Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact...

17
Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi * , Soroush Shaee, Logan Rakai, Andrew M. Knight, Hamidreza Zareipour Schulich School of Engineering, University of Calgary, Alberta, Canada article info Article history: Received 4 August 2016 Received in revised form 11 February 2017 Accepted 19 February 2017 Available online 23 February 2017 Keywords: Electricity market Energy storage system Price impact abstract This paper analyzes the impact of an independently-operated large-scale energy storage system on the electricity prices of a fully competitive pool-based electricity market. From a consumer's perspective, the impact of storage operation on energy cost is investigated. The changes in the revenue of generation units and entities caused due to the arbitrage operation of storage facility are also explored. In the current study, we have considered an optimization-based scheme to schedule the operation of a large- scale storage system. Also, a number of ad-hoc strategies to participate in the energy market are examined for the sake of comparison. Modeling the impacts of a large-scale energy storage system can inform planners and operators of the potential effects of storage on the rest of the system and help them to use storage most effectively. These analyses can also inform regulators and policy-makers of the potential societal costs and benets to energy storage deployments that are not necessarily monetizable by the energy storage owner. © 2017 Elsevier Ltd. All rights reserved. 1. Introduction 1.1. Background and motivation The electrical power industry is entering a transition towards sustainable, reliable and clean solutions as a policy. It is a quiet but continuous revolution trending to a large-scale expansion of re- newables in power systems. There have been, however, serious concerns over reliable and satisfactory operation of the power systems. Energy storage facilities are increasingly being used to help integrate renewable energy resources into the grid. These systems are capable of absorbing and delivering power with sub- second response times and therefore can be used to compensate the high variability and uncertainty of renewables. These emerging technologies are likely to have an economically- important effect on the dynamics of electricity prices. This is a concern to different sections of electricity markets including power plant owners, policy maker, and end users. Participation of a large- scale Energy Storage System (ESS) may challenge the revenue suf- ciency of conventional generation units. On the other hand, it may lead to higher cost of energy for end-user consumers in an elec- tricity market. At the same time, in some competitive electricity markets it is yet to be claried if ESS should be treated as genera- tors, consumers, or classied as another category of market participant. Thus, it is essential to quantify and measure the impact on wholesale electricity prices of an upcoming large-scale ESS. 1.2. Literature review and contribution Many of previous papers, e.g. Refs. [1e 19], published on the area of storage systems, have considered storage as a facilitator to in- crease the penetration of renewable energy resources and assumed that the operation of ESS is governed by system operator. Other works consider that ESS is co-located and combined with a wind farm and is integrated to maximize the total operation prot [20e24]. On the other hand, another category of papers, consider the storage system as an individual participant in electricity markets [25e32]. These papers aim at selecting optimal operation sched- uling for ESS to achieve the optimum operation prot. In these papers, it is assumed that ESS is price-taker, i.e., it is not large enough to have an impact on electricity prices. However, large- scale ESS is identied as one of the priority areas to build a smart grid [33,34]. With large-scale merchant ESS, it is now time to * Corresponding author. E-mail address: [email protected] (P. Zamani-Dehkordi). Contents lists available at ScienceDirect Energy journal homepage: www.elsevier.com/locate/energy http://dx.doi.org/10.1016/j.energy.2017.02.107 0360-5442/© 2017 Elsevier Ltd. All rights reserved. Energy 125 (2017) 27e43

Transcript of Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact...

Page 1: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

lable at ScienceDirect

Energy 125 (2017) 27e43

Contents lists avai

Energy

journal homepage: www.elsevier .com/locate/energy

Price impact assessment for large-scale merchant energy storagefacilities

Payam Zamani-Dehkordi*, Soroush Shafiee, Logan Rakai, Andrew M. Knight,Hamidreza ZareipourSchulich School of Engineering, University of Calgary, Alberta, Canada

a r t i c l e i n f o

Article history:Received 4 August 2016Received in revised form11 February 2017Accepted 19 February 2017Available online 23 February 2017

Keywords:Electricity marketEnergy storage systemPrice impact

* Corresponding author.E-mail address: [email protected] (P. Zamani

http://dx.doi.org/10.1016/j.energy.2017.02.1070360-5442/© 2017 Elsevier Ltd. All rights reserved.

a b s t r a c t

This paper analyzes the impact of an independently-operated large-scale energy storage system on theelectricity prices of a fully competitive pool-based electricity market. From a consumer's perspective, theimpact of storage operation on energy cost is investigated. The changes in the revenue of generationunits and entities caused due to the arbitrage operation of storage facility are also explored. In thecurrent study, we have considered an optimization-based scheme to schedule the operation of a large-scale storage system. Also, a number of ad-hoc strategies to participate in the energy market areexamined for the sake of comparison. Modeling the impacts of a large-scale energy storage system caninform planners and operators of the potential effects of storage on the rest of the system and help themto use storage most effectively. These analyses can also inform regulators and policy-makers of thepotential societal costs and benefits to energy storage deployments that are not necessarily monetizableby the energy storage owner.

© 2017 Elsevier Ltd. All rights reserved.

1. Introduction

1.1. Background and motivation

The electrical power industry is entering a transition towardssustainable, reliable and clean solutions as a policy. It is a quiet butcontinuous revolution trending to a large-scale expansion of re-newables in power systems. There have been, however, seriousconcerns over reliable and satisfactory operation of the powersystems. Energy storage facilities are increasingly being used tohelp integrate renewable energy resources into the grid. Thesesystems are capable of absorbing and delivering power with sub-second response times and therefore can be used to compensatethe high variability and uncertainty of renewables.

These emerging technologies are likely to have an economically-important effect on the dynamics of electricity prices. This is aconcern to different sections of electricity markets including powerplant owners, policy maker, and end users. Participation of a large-scale Energy Storage System (ESS) may challenge the revenue suf-ficiency of conventional generation units. On the other hand, it may

-Dehkordi).

lead to higher cost of energy for end-user consumers in an elec-tricity market. At the same time, in some competitive electricitymarkets it is yet to be clarified if ESS should be treated as genera-tors, consumers, or classified as another category of marketparticipant. Thus, it is essential to quantify and measure the impacton wholesale electricity prices of an upcoming large-scale ESS.

1.2. Literature review and contribution

Many of previous papers, e.g. Refs. [1e19], published on the areaof storage systems, have considered storage as a facilitator to in-crease the penetration of renewable energy resources and assumedthat the operation of ESS is governed by system operator. Otherworks consider that ESS is co-located and combined with a windfarm and is integrated to maximize the total operation profit[20e24].

On the other hand, another category of papers, consider thestorage system as an individual participant in electricity markets[25e32]. These papers aim at selecting optimal operation sched-uling for ESS to achieve the optimum operation profit. In thesepapers, it is assumed that ESS is price-taker, i.e., it is not largeenough to have an impact on electricity prices. However, large-scale ESS is identified as one of the priority areas to build a smartgrid [33,34]. With large-scale merchant ESS, it is now time to

Page 2: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

Notation

Indicest Index for operation intervals running from 1 to Nt

s Index for the steps of discharging price quota curvesfrom 1 to Nd

s0 Index for the steps of charging price quota curves from1 to Nd

Parametersqc Rated power of charging [MW]

qd Rated power of discharging [MW]Smin Minimum level of energy storage [MWh]Smax Maximum level of energy storage [MWh]h Roundtrip storage efficiency [%]VOMc Variable operation and maintenance cost of charging

[$/MWh]VOMd Variable operation and maintenance cost of

discharging [$/MWh]Pdt;s Electricity price corresponding to step number s of the

discharging price quota curve at time interval t[$/MWh]

Pct;s0 Electricity price corresponding to step number s’ of thecharging price quota curve at time interval t [$/MWh]

qds Summation of power blocks from step 1 to step s-1 ofthe discharging price quota curve at time interval t[MW]

qcs0 Summation of power blocks from step 1 to step s’-1 ofthe charging price quota curve at time interval t [MW]

bds Size of step s of the discharging price quota curve attime interval t [MW]

bcs0 Size of step s' of the charging price quota curve at timeinterval t [MW]

FOM Fixed operation and maintenance cost [$/MW-year]CC Overnight capital cost of the storage facility [$]CRF Capital Recovery Factor of the storage facility [%]

Real variablesst Stored energy at time interval t [MWh]qct Charged energy at time interval t [MWh]qdt Discharged energy at time interval t [MWh]bdt;s The fractional value of the power block corresponding

to step s of the discharging price quota curve at timeinterval t [MW]

bct;s0 The fractional value of the power block correspondingto step s’ of the charging price quota curve at timeinterval t [MW]

Binary variablesudt Unit status indicator in discharging modeuct Unit status indicator in charging modeudt;s Binary variable that is equal to 1 if step s of the

discharging price quota curve is the last stepuct;s0 Binary variable that is equal to 1 if step s' of the

charging price quota curve is the last step

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4328

answer the following question:What is the impact of a price-makerESS on the price of electricity?

The literature on the large-scale price-maker ESS is scarce. Tothe best of our knowledge, there are a limited number of papers[35e39] with focus on the optimal scheduling and market partic-ipation of large-scale and price-maker storage units that areinvestor-owned and independently-operated. Authors in Ref. [35]present the economic ramifications of a growing storage marketshare on the spot market. The basic analysis is done by applyingpower plant dispatch and a load levelling algorithm to adjust theoriginal hourly demand curve. It is concluded that by introducingadditional storage capacity to the system, the initial price spread isnearly half-cut due to the smoothing effect on the residual demand.In Ref. [36], the problem is formulated as a mathematical pro-gramming with equilibrium constraints. It is assumed that thebidding strategy of other market participants is known. Then, anoptimization problem is solved to find the optimal operation ofstorage system to maximize its profit. A main concern about thepresented study in Ref. [36] is that solving the formulated optimi-zation problem in real electricity markets with large number ofparticipants can be laborious if not impossible. Moreover, thisassumption that each market's participants other than ESS has thesame bidding strategy for all the conditions in the market is notvalid based on our experience from Alberta market.

Other papers [37e39] have developed their models by incor-porating the relationship between electricity prices and net de-mand in the electricity market. Authors in Ref. [37] estimate a linearrelationship between load and electricity prices to model the priceimpact of a storage facility. In Ref. [38], one representative supplycurve is considered to model the relationship between net demandand electricity prices. In another paper [39], authors have modeledthe residual demand by an approximated sigmoid function to

formulate the operation scheduling of a price-maker pumpedhydro storage. It is uncertain if the representative functionsconsidered by previous studies [37e39] to express the relationshipbetween demand and price, are capable of modeling real supplycurves in the market.

Our contribution in the present research, is a data analyticsapproach to investigate the impact of a large-scale ESS on the dy-namics of electricity price. We rely on the real data from electricitymarket to estimate the impact of an upcoming price-maker storagesystem on the energy cost of end-user consumers and also therevenue of different generation units. Our experience from theAlberta electricity market reveals that market participants changetheir bidding strategies significantly from time to time. Thus, weconstruct real supply curves at each hour based on the observedsubmitted offers by the generation units. Next, an optimization-based scheme and a group of ad-hoc bidding strategies for theoperation of the ESS in the energy market are developed to ensurethe profitability of its arbitrage operation. For each strategy, theoperation of ESS is considered in a study period of five years and itsimpact on wholesale electricity prices, energy cost, and revenue ofother market participants is analyzed. The proposed methodologyhas considered a general model for the large-scale ESS and providesample flexibility to estimate the price impact of storage systemswith different sizes and different bidding strategies.

1.3. Paper organization

The rest of this paper is organized as follows. Section 2 describesour case study and the database used for the simulations. Section 3explains the simulation procedure to characterize the price impactof additional supply or demand. The operational basics of large-scale ESSs are presented in Section 4. Section 5 provides

Page 3: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

Accumulated Quantity of Submitted Offers (MW)0 5000 10000 15000

Pric

e ($

/MW

h)

0

200

400

600

800

1000

Fig. 1. Sample supply curve for hour ending 1, December 31, 2014.

Accumulated Quantity of Submitted Offers (MW)0 2000 4000 6000 8000 10000 12000

Pric

e ($

/MW

h)0

200

400

600

800

1000

Fig. 2. Supply curves for each of the 24 h on August 1, 2014.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 29

economic analyses for different operational strategies of the storagefacility. Finally, section 6 closes the paper, providing some relevantconclusions.

2. Database

In this paper, we consider the pool-based electricity market ofAlberta as a case study. In order to model the potential impacts ofan ESS facility, a first necessary step is to build a historical databaseof market operations over the period of interest. The data requiredis publicly available, published by the Alberta Electric SystemOperator (AESO) through its online data-publishing portal [40].Data preparation requires the identification of relevant data,retrieving the data from the AESO website and processing into areadily useable database format. The database for the study in-cludes hourly generator offers and pool price data for the period ofJanuary 1, 2010 to December 31, 2014.

2.1. Generators hourly offer data

All generators with a capacity of more than 5 MW must offertheir capacity to sell energy in the market to the AESO through theEnergy Trading System (ETS). Generators submit their offers in theform of quantity and price pairs. Each unit may submit up to sevenprice-quantity blocks for each hour. The price for each block mayvary between $0/MWh to $999.99/MWh; the sum of the quantityfor the blocks must equal the total capacity of the generator.1

The full list of price-quantity data for each of the 43,824 h in thestudy period of five years is downloaded and stored in a database.Sorting the price-quantity offers from the lowest-priced to thehighest-priced, a supply curve, also referred to as a merit ordercurve, is constructed for each hour. The resulting supply curve forHour Ending 1, December 31, 2014 is shown in Fig. 1 Two importantcaveats should be noted when considering the historical database:(1) Wind generators were not required to submit offers to theAESO; this requirement changed on April 1, 2015. (2) Generatorswith capacities smaller than 5 MW are exempt from submittingoffer data to the AESO. The supply curves used in the study onlyconsider units greater than 5 MW, that are dispatchable, and thathave regularly submitted offer data to the market.2

As stated earlier, the study uses the actual supply curve data foreach of the 43,824 h in the study period. The importance of usingactual hourly data, rather than a representative curve, can be seenby considering Fig. 2 which plots the supply curves for each of the24 h in a single day, August 1 2014. This plot demonstrates signif-icant variability in the supply curves. This is an indication thatmarket participants submit their offers strategically depending onmarket conditions. For instance, minimum run-times of thermalpower plants could be a reason for different offers at differenthours. Thus, using a single supply curve for all hours of a long studyperiod may not fully capture the realities of market participants'offering strategies.

Similar to supply curves, one can build a demand curve for eachhour of the market if the demand-side also submits price-quantitybid blocks to the market. To build a demand curve, the bids aresorted from highest-priced values to the lowest-priced ones. Intheory, the intersection of the supply and demand curves for eachperiod is the market clearing price. In Alberta, while the demand-side market participants are technically allowed to submit hourlybids for purchasing energy, they are inelastic for the most part and

1 https://www.aeso.ca/aeso/training/guide-to-understanding-albertas-electricity-market/.

2 http://ets.aeso.ca/.

typically do no actively submit bids to the market. In this study, wehave not considered any elastic demand bids from marketparticipants.

2.2. Hourly Alberta electricity market price data

Electricity market price data in Alberta is available by downloadof historical pool prices. The pool price is the hourly average of theminute-by-minute system marginal price (SMP) and is used forsettlement purposes. The SMP values are determined as the offerprice of the marginal generator, i.e., the generator on the meritorder that supplies that last 5 MW to the market. There are occa-sions in the market that the AESO makes adjustments to the supplycurve in order to establish the SMP. For example, in case a must-rununit must be dispatched, market rules specify how exactly themarket price needs to be determined. The circumstances of suchoccasions are outlined in the AESO Rules. The Pool Price reportprovides data on the hourly pool price and the Alberta Internal Loadat each hour.3

2.3. Market equivalent demand

As mentioned above, the Alberta Internal Load (AIL) is reportedby the AESO on a hourly basis. However, the AIL data does notprovide sufficient information to determine price from a supplycurve. The supply curves in the database only consider dispatchablegeneration above 5 MW. To determine equivalent system load on asupply curve such as that developed in the study, the pool price iscross-referenced against the supply curve, with the interceptproviding the load supplied by the dispatchable generation in the

3 http://ets.aeso.ca/.

Page 4: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

Market Equivalent Demand (MW)0 2000 4000 6000 8000 10000 12000

Pric

e ($

/MW

h)

0

200

400

600

800

1,000

Supply CurveInitial Demand Bid

Fig. 3. Initial supply offer and demand bid curves.

Market Equivalent Demand (MW)8800 8900 9000 9100 9200 9300 9400

Pric

e ($

/MW

h)0

10

20

30

40

50

60

Supply Curve80 MW @ $30/MWh bid130 MW @ $35/MWh bid170 MW @ $45/MWh bid

Fig. 4. Illustration of possible outcomes from three bids. Initial pool price is $33.28/MWh, corresponding to an existing market equivalent demand of 9110 MW.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4330

merit order. We refer to this value as theMarket Equivalent Demand.

3. Price impact of additional demand or supply

Prior to investigating the impact of an energy storage facility, itis important to model the impact that additional demand andsupply may have on pool prices, without the constraints of storageoperations. This section evaluates how perturbation of the supplyor demand curves, in general, translate into price fluctuations in theAlberta wholesale electricity market.

A merchant ESS facility that is in the market to make a profitthrough energy arbitrage, buys energy during low-price hours andsells the energy back to the grid during high-price hours. The fa-cility can be thought of as being in one of three operating modes (1)chargingmode, acting similar to a dispatchable load (2) dischargingmode, when it is technically a generator (3) standby, neithercharging or discharging. When acting like a dispatchable load, thedecision to charge will impact the demand curve in the market.Similarly, the decisions to discharge energy will change the supplycurve.

Each case of adding a new block of demand bid or a new block ofsupply offer to the market can be considered independently.Assuming all other historical supply offers and demand bids areunchanged, it is possible to estimate the change to the pool pricethat results at any given hour, for a given range of supply offers ordemand bids. These estimates are applicable to adding any demandor supply to the market and are not constrained with the operationlogic of a ESS facility. The findings of this section are used later tocalculate the price impact of a ESS facility with specificcharacteristics.

In the remainder of this section, first, the price impact of addinga load to the market is estimated. Secondly, the price impact ofadding a new supply offer to the market is estimated. For bothcases, the statistical analysis of resulting price impacts for a widerange of single-block supply offers and demand bids are presented.

3.1. Price increases due to new dispatchable load bids

A data-mining algorithm has been developed in order to esti-mate how much the price would increase at a given hour if a newdemand bid was added to the market. The steps of the algorithmare:

1. Identify historical equivalent market demand,2. Add a new demand bid to the market and create the new de-

mand curve,3. Calculate intercept between supply and demand curves, and

determine the resulting price,4. Calculate the price impact, i.e., the increase in price due to an

additional demand bid.

3.1.1. Constructing new demand curves and estimating marketprices

The process of constructing new demand curves is bestdescribed with the illustrative case shown in Figs. 3 and 4. For thesupply curve shown in Fig. 3, begin with the case where the poolprice is $33.28/MWh. This corresponds to 9110 MW, which is thedefined asmarket equivalent demand for that hour. This is assumedto be a single block demand bid of 9110 MW at $999.99/MWh, andresults in the demand bid curve shown in Fig. 3. The point ($33.28/MWh, 9110 MWh) is the intercept of the two curves.

Fig. 4 shows the cases for 3 additional bids added to the demandcurve. Each case illustrates a different outcome on themarket price.In all cases, the price is set by the intercept of the supply and

demand curves. The first case is a bid of 80MWat $30/MWh. As theexisting market price is greater than $30, this bid is not accepted,and has no impact on the market price. The second case is a bid of130 MW at $35/MWh. This bid price is above the existing marketprice, however, a part of the block is below the supply offers. Thesystem operator can dispatch up to 9209 MW below $35/MWh, butadditional supply is not available until the price reaches $36.30/MWh. In this case, the bid will be partially accepted, with 109 MWaccepted, setting the market price to $35/MWh, i.e., the intercept ofthe supply and demand curves. The third case is a bid of 170 MWat$45/MWh. The entire block is above the supply curve. This demandbid is fully accepted, with the price set at the intersection of thesupply and demand curves, i.e., $39.35/MWh.

3.1.2. Statistics of price increases resulting from increased load inthe market

For every hour in the study period, the market prices for all bidblocks are determined. Statistics are compiled for all cases wherethe demand bid block was fully or partially accepted. These sta-tistics reflect the impact of bids made irrespective of whether thebid timing may make sense for the ESS operation. As such, thesestatistics should be considered together with the remainder of thepaper, rather than as stand-alone information.

For a sample bid block of 160 MW at $40/MWh, Fig. 5 plots thefrequency distribution of price increases over the five-year period.The most likely price increases are in the range of $2-$4/MWh. Dueto the long-tailed distribution, the mean and median price impactsare significantly different. Fig. 6 plots the mean increase in poolprice as a function of demand bid price and demand bid quantity.Fig. 7 plots the median increase in pool price as a function of de-mand bid price and demand bid quantity.

Page 5: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

Increase in Elecricity Price ($/MWh)0 5 10 15 20

Perc

enta

ge o

f Hou

rs (%

)

0

2

4

6

8

10

12

Fig. 5. Distribution of price increases for accepted demand bid of 160 MW at $40/MWh, all hours.

200

Bid Quantity (MW)

100020

30Bid Price ($/MWh)

40

0

5

10

50Mea

n Pr

ice

Incr

ease

($/M

Wh)

Fig. 6. Mean change in price (increase) when discharging, due to different demand bidquantity price pairs.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 31

3.2. Price decreases due to new supply offers

Similar to additional demand, an algorithm is used to estimatehow much prices would decrease if a new supplier was to makeoffers to the market. The steps are similar to the work on newdemand. In this case, a new supply curve is constructed, adding inan additional supply offers block. Again, a key assumption is thatother historical offers remain the same. The steps involved inestimating the price decreases are further explained next.

3.2.1. Constructing new supply curves and estimating market pricesThe first step of the algorithm is to create new supply curves, for

different offer price-quantity combinations, at every hour in the

200

Bid Quantity (MW)

100020

30Bid Price ($/MWh)

40

3

4

0

1

2

50

Med

ian

Pric

e In

crea

se ($

/MW

h)

Fig. 7. Median change in price (increase) when discharging, due to different demandbid quantity price pairs.

database. The offers range from 10 MW to 160 MW quantity with a10 MW increment, and price range from $50/MWh to $800/MWh.Fig. 8, illustrates the case of a supply offer of 160MWat $155/MWh;as a result of the new offer, the supply curve is extended to theright.

The second step of the algorithm is to determine the equivalentmarket demand for each hour, based on the historical supply curvesand pool prices.

The impact of the new supply offer can be seen as the differencebetween the original and modified supply curves, at the equivalentmarket demand. Again, three cases may occur. In this first case, thenew offer price is less than the actual market price and is such thatit is fully accepted in the market. For example, with 10,900 MWmarket equivalent demand for the particular hour presented inFig. 8, the original market price is $930/MWh, and becomes $500/MWh after the new offer block is inserted.

In the second case, the newoffer block sets the price for the hourand it is partially accepted. For example, at 10,250 MW, the originalprice is $171, the new price is $155/MWh, set by the new offer; but,only 90 MW of the 160 MW offer is accepted.

The final case occurs when the original market price is lowerthan the new offer block and as such, is out of the market and is notaccepted. In this case the price does not change.

3.2.2. Statistics of price decreases resulting from adding new supplyoffer blocks to the market

Similar to the demand bid blocks, statistics of market priceimpact can be calculated for all accepted offer blocks, for every hourin the study period. As with the price increases due to new demandbids, the price decrease resulting from a new offer block is highlydependent upon the price and quantity of the offer block. However,in general, the prices decrease more severely as a result of addingnew supply offers compared to price increases resulting fromadding new demand bids.

Fig. 9 plots the mean change in price after an accepted supplyoffer over a range of offer price-quantity pairs. Fig. 10 plots themedian price decreases for the same cases. Note that the verticalaxis plots the magnitude of the change in price (e.g. $300 repre-sents a $300/MWh reduction in pool price) and that the horizontalprice axes are the offer price-the actual pool price may be higherthan or equal to the offer price. From these figures, it can be notedthat for offer prices close to $100/MWh, the impact on prices is lesscompared to those for offer prices ranging between $300/MWh to$650/MWh. As expected, the impact is more significant when theoffer quantities are large.

There are relatively few supply offers in the range $300-$600/MWh and thus, the supply curve has a steep gradient in this pricerange. This implies that a new submitted offer inserted into this

Market Equivalent Demand (MW)10100 10300 10500 10700 10900 11100

Pric

e ($

/MW

h)

0

200

400

600

800

1000

Original CurveNew OfferShifted Blocks

Fig. 8. Inserting a new supply offer 160 MW at $155/MWh to the supply curve.

Page 6: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

200

Offer Quantity (MW)

10000Offer Price ($/MWh)

500

200

300

0

100

1000Mea

n Pr

ice

Dec

reas

e ($

/MW

h)

Fig. 9. Mean change in price (decrease) due to different supply offer quantity pricepairs.

200

Offer Quantity (MW)

10000Offer Price ($/MWh)

500

0

200

300

100

1000

Med

ian

Pric

e D

ecre

ase

($/M

Wh)

Fig. 10. Median change in price (decrease) due to different supply offer quantity pricepairs.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4332

steep part of supply curve can decrease the electricity pricesubstantially which also explains the convex surface seen inFig. 10.

4. Operational basics of large-scale ESS

In this section, a number of bidding strategies for a merchantESS facility are developed. We consider two strategies to schedulethe operation of the price-maker energy storage system. First, anoptimization platform is developed in Section 4.1 to maximize thearbitrage revenue of the storage system by taking into account itsimpact on electricity prices. Limitations are placed due to thephysical constraints of the energy storage facilities. For the sake ofcomparison, we also consider another alternative in Section 4.2 thatthe ESS has ad-hoc strategies with decisions to buy and sell atcertain pre-defined market prices. Pumped Hydro Storage (PHS)and Compressed Air Energy Storage (CAES) as the two primarytechnologies for bulk storage of electric energy (hundreds of MW-hours) are considered to be integrated to the Alberta electricitymarket.

4.1. Self-scheduling of the price-maker energy storage

In this section, a general optimization-based formulation forthe operation scheduling of an ESS is presented. The energy isstored during off-peak hours where prices are low and injectedback to the gird during peak periods. The goal of the storage plantis to maximize its profit through energy arbitrage. The objectivefunction and constraints for the self-scheduling optimization areas follows:

maxqct ;q

dt ;u

ct ;u

dt

Xt¼1

Nt

qdt�Pdt

�qdt

�� VOMd

�� qct

�Pct�qct�þ VOMc� (1)

Subject to:

0 � qct � qc:uct ct2Nt (2)

0 � qdt � qd:udt ct2Nt (3)

Smin � St � Smax ct2Nt (4)

Stþ1 ¼ St þ h:qct � qdt ct2Nt (5)

S0 ¼ Sint (6)

uct þ udt � 1 ct2Nt (7)

The objective function is presented in (1), which consists of twoterms. The first term is the revenue from electricity sales to themarket by discharging the stored energy. The second term is thecost of purchasing the electricity from the market. It is noted thatthe electricity price cleared in the market is a function of charging/discharging quantity in the case of a price-maker ESS. The charging/discharging power of the ESS is constrained by its nominal power asexpressed in (2) and (3). Equation (4) models the minimum andmaximum energy limits of the storage facility, while (5) indicatesthat the storage state of charge dynamically changes based on theamount of charging and discharging power by considering theround-trip efficiency. The initial stored energy in the system isshown in (6). The constraint presented in (7) indicates that the ESScannot operate in both charging and discharging mode at a timeinstance.

It could be observed that the objective function presented in (1)is a non-linear equation. To be able to solve the developed opti-mization platform by utilizing the commercial software, we need tofind the equivalent linear formulation. To do so, we employ theproposed price-quota-curve methodology in Ref. [41]. An alterna-tive equivalent formulation of problem (1) that is linear is providedas follows:

maxqct ;q

dt ;u

ct ;u

dt

Xt¼1

Nt"Xs¼1

Nd �bdt;s þ udt;sq

ds

��Pdt;s � VOMd

��

Xs0¼1

Nc �bct;s0

þ uct;s0qcs0

��Pct;s0 þ VOMc

�#(8)

Subject to: (2)e(7)

qdt ¼Xs¼1

Nd �bdt;s þ udt;sq

ds

�ct2Nt (9)

0 � bdt;s � bds ct2Nt (10)

Xs¼1

Nd

udt;s ¼ udt ct2Nt (11)

qct ¼Xs0¼1

Nc �bct;s0 þ uct;s0q

cs0

�ct2Nt (12)

Page 7: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

Pt,1c

Pt,2c

Pt,3c

Pt,4c

Pt,5c

Elec

trici

ty P

rice

ut,1c =0

ut,2c =0

ut,4c =0

ut,5c =0

b3c

ut,3c =1

b3,tcq3

c

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 33

0 � bct;s0 � bcs0 ct2Nt (13)

Xs0¼1

Nc

uct;s0 ¼ uct ct2Nt (14)

The objective function (8) expresses the profit of the price-maker ESS. To express the total revenue as a linear function, step-wise price quota curves are built as illustrated in Figs. 11 and 12.It is observed that the real variables, i.e., bdt;s and bct;s0 , and binaryvariables, i.e., udt;s and uct;s0 , are used to relate the electricity price tothe storage system's charging/discharging quantity.

Charging Quantity

Fig. 12. Charging price quota curve.

Table 1ESS parameters.

Parameter CAES PHS

qd (MW) 100 150

qc (MW) 100 150Smin (MWh) 100 0Smax (MWh) 500 1500h (%) 55 [43] 75 [44]VOMc ($/MWh) 1.5 [43] 0.9 [44]

VOMd ($/MWh) 1.5 [43] 0.6 [44]

4.2. Ado-hoc operation strategies

In this study, we assume that ESS has an unchanging biddingstrategy to participate in the electricity market. To do so, the ESSoperators submits determined demand bids and supply offers tothe market at preset prices. On the charging side, demand bids aresubmitted with constant price of P1, which implies that electricityis bought when the price is lower than or equal to P1. To sell thestored energy back to the market, ESS submits supply offers tomarket with price of P2 during discharging periods to ensure thateach MWh generated energy is paid more than the threshold of P2.

ESS owners may be able to make profit and take advantage ofthe arbitrage operation if the values of P1 and P2 are chosenreasonably. Let us assume that PC ½$=MWh� is the price of electricityduring charge period. Then, the total cost of generating 1 MWhpower by using the energy stored in the ESS is the sum of amountpaid for the energy during charging and the operation and main-tenance cost. It should be noted that the amount of energy thatneeds to be bought from the market to generate 1 MWh electricityis calculated based on the round-trip efficiency of the ESS as 1/hMWh. In order tomake profit by generating electricity in a period oftime, the price of electricity during discharge, i.e., PD½$=MWh�,should be higher than the cost of operation:

PD > PC=hþ VOMc=hþ VOMd (15)

Two options of PHS and CAES for the large-scale ESS areconsidered in this study. Parameters for the two type of ESS arepresented in Table 1. We obtain the related numbers based on areport published by International Renewable Energy Agency(IRENA) on different electricity storage technologies [42]. Equation(15) suggests a logic for the relationship between P1 and P2 asfollows:

Discharging Quantity

Pt,5d

Pt,4d

Pt,3d

Pt,2d

Pt,1d

Elec

trici

ty P

rice

ut,1d =0

ut,2d =0

ut,4d =0

ut,5d =0

q3d

ut,3d =1

b3d

bt,3d

Fig. 11. Discharging price quota curve.

P2 > P1=hþ VOMc=hþ VOMd (16)

We assume that the ESS has an unvarying strategy to submitdemand bids to the market. Our analysis from the historical data ofthe Alberta electricity market demonstrates that for more than 70%of hours between years 2010e2014, electricity prices have beenmore than $40/MWh. Thus, by submitting demand bids with un-changing price of $40/MWh, the ESS would have enough oppor-tunity to buy electricity from the pool-basedmarket to store energyand then sell it back to the market during peak hours to capturearbitrage revenue.

By considering P1 ¼ $40/MWh, then we can conclude based on(16) and the parameters in Table 1 that the CAES and PHS facilitiesshould submit their supply offers to the market with prices greaterthan $79.77/MWh and $62/MWh, respectively, to ascertain theprofitability of their arbitrage operation. Thus, we develop a seriesof rational ad-hoc strategies for operation scheduling of the ESS. Allthree strategies use the same demand bid, i.e., $40/MWh. Thestrategies differ in the price used for the supply offers. A number ofstrategies are suggested in which one single block of supply offer issubmitted to the pool-based electricity market:

STG1. Single supply offer block at $80/MWh

STG2. Single supply offer block at $150/MWh

STG3. Single supply offer block at $300/MWh

Some multi-block pre-determined supply offers are also inves-tigated to explorewhether there is an advantage to breaking supplycapacity into multiple blocks, reserving some capacity for the sit-uation when high prices occur. Again, the demand bid is submittedas a single block $40/MWh; differences between the strategies arelimited to supply offer blocks, as follows:

STG4. Two offer blocks: equal quantities at $80/MWh and $300/MWh

Page 8: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

Table 2Costing inputs for assessing the economics of CAES [43].

Parameter Value

FOM of expander ($/kW-yr) 8Capital Cost of expander ($/kW) 440FOM of compressor ($/kW-yr) 8Capital Cost of compressor ($/kW) 415Capital Cost of cavern ($/kWh) 0.15Life time (year) 30

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4334

STG5. Two offer blocks: equal quantities at $100/MWh and $500/MWh

STG6. Two offer blocks: equal quantities at $150/MWh and $900/MWh

STG7. Three offer blocks: equal quantities at $80/MWh, $300/MWh, and $900/MWh

STG8. Four offer blocks: equal quantities at $100/MWh, $300/MWh, $600/MWh, and $900/MWh.

The quantity of submitted bids and offers are considered to beequal to the rated charging and discharging power as long as theconstraint on the minimum and maximum capacity of the ESS isnot violated. If the stored energy is not sufficient to submit offerswith maximum power rating, then the available energy remainedin the ESS determines the quantity of submitted supply offer.Conversely, there may be some hours that the ESS does not havesufficient spare capacity to charge with the maximum power. Then,the free capacity of ESS is used to determine the quantity of sub-mitted demand bid. In summary, the quantities of submitted bidsand offers is defined as follows:

qc;subt ¼�

qc st � Smax � hqc

ðSmax � stÞ=h otherwise(17)

qd;subt ¼�

qd st � qd þ Sminst � Smin otherwise

(18)

where qC;subt and qD;subt are the quantity of submitted demand bidsand supply offers at each instance t, respectively. The results ofapplying the eight operation strategies are discussed in thefollowing section.

Table 3Costing inputs for assessing the economics of PHS [44].

Parameter Value

Capital Cost of charge device ($/kW) 1230Capital Cost of discharge device ($/kW) 1270Capital Cost of reservoir ($/kWh) 50FOM ($/kW-yr) 9Life time (year) 50

4.3. Calculating the net profit

After calculating the operational revenue of storage facilities,their yearly net profit is estimated by incorporating their annual-ized cost. In finance, Equivalent Annual Cost (EAC) is the cost peryear of owning and operating an asset over its entire lifespan and iscalculated as follows [45]:

EAC ¼ CC� CRFþ FOM (19)

where CC, i.e., overnight capital cost, is measured in dollars, FOM, isthe Fixed Operation and Maintenance costs in dollars per kilowatt-year ($/kW-yr), and CRF, i.e., capital recovery factor, is a fractioncalculated based on the interest rate, i, and life time, n, as follows:

CRF ¼ ið1þ iÞn��ð1þ iÞn� 1 (20)

It should be noted that the variable operation and maintenancecost is not included in (19) as it is already incorporated in the arbi-trage operation of the ESS in (1). For both storage technologies ofCAES and PHS, the EAC is calculated to investigate the profitability oftheir arbitrage operation. Compressor, cavern, and expander are thethree main components of the CAES [43]. We obtain the requirednumbers for calculating the EACof CAES based on a report publishedby the Electric Power Research Institute (EPRI) on the cost evalua-tion of CAES [43]. The associated values for the CAES technology arepresented in Table 2. The PHS facility composed of pumping stationfor charging, hydro turbines and generators for discharging, andreservoirs to store the water [44]. The costing inputs for assessingthe economics of PHS are presented in Table 3. These numbers arefrom a report published by British Columbia Hydro on the costestimation of a PHS at Mica generating station [44].

In this paper, we consider the value of interest rate to be equal to7% as it is suggested by the Open Energy Information (OEI) [46].After estimating the EAC, the operational profit is calculated as thedifference between the yearly net revenue and EAC.

5. Economic analysis

Each of the operational strategies is applied to the database ofhistorical market data for the five-year study period. The operatingschedules and hourly supply curves are used to determine marketprice impacts and the potential operating revenue of ESS. Potentialimpacts on energy consumers, wind producers and other types ofgenerators are also investigated. Again, these results are based onan assumption that other market participants behavior is un-changed. Results only consider profits from arbitrage operation ofthe storage facility in the energy market; potential revenue streamsfrom participation in other markets (such as ancillary services,fixed load serving contracts or arbitrage with neighboring markets)are excluded.

5.1. Operating profit

The first aspect of the economic analysis is to calculate the totalobtained operational profit. To do so, the arbitrage revenue of eachstorage facilities is evaluated when different offering strategies areadopted. Then, their net profit is estimated by incorporating theirannualized cost to investigate the profitability of their arbitrageoperation.

Figs. 13 and 14 depict the charging cost, discharging income,and net revenue of the CAES and PHS, respectively, for differentstrategies. As expected, charging cost and discharging income forthe PHS facility are higher compared to the CAES due to its higherpower ratings and also better round-trip efficiency. Consequently,the total operational revenue of the PHS facility is on averagetwice the earning of CAES. Observe from Fig. 13 that for the CAES,among different bidding strategies, STG3 reports the highestoperational revenue since it has the lowest charging cost. Thisstems from the fact that CAES facility charges less frequently byadopting STG3 since the supply offers are submitted with higherprices and thus, accepted for lower number of hours. It is thesame condition for STG6 and it reveals a relatively low chargingcost compared to other strategies. For the CAES facility andamong the single offer block bidding strategies, STG3 offers thehighest discharging income. It should be noted that at some

Page 9: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-20

0

20

40

60

80$

Mill

ion

Charging CostDischarging IncomeNet Revenue

Fig. 13. Charging cost, discharging income, and net revenue of the CAES for differentstrategies.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-20

0

50

100

150

$ M

illio

n

Charging CostDischarging IncomeNet Revenue

Fig. 14. Charging cost, discharging income, and net revenue of the PHS for differentstrategies.

2010 2011 2012 2013 2014Year

0

50

100

150

200

Elec

trici

ty P

rice

($/M

Wh)

0

0.5

1

1.5

2

Rev

enue

($)

107

Revenue-OPTRevenue-STG3Ave. of PricesStd. of Prices

Fig. 15. Average and standard deviation of electricity prices in different years versusyearly net revenue of the CAES by adopting STG3.

2010 2011 2012 2013 2014Year

0

50

100

150

200

Elec

trici

ty P

rice

($/M

Wh)

0

1

2

3

4

Rev

enue

($)

107

Ave. of PricesStd. of PricesRevenue-OPTRevenue-STG5

Fig. 16. Average and standard deviation of electricity prices in different years versusyearly net revenue of the PHS by adopting STG5.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 35

discharging instances the storage facility is the marginal unit inthe market and sets the electricity price. In these cases, STG3results the biggest discharging income as it submits supply of-fers with higher prices compared to other strategies. Conse-quently, it is able to earn higher operational revenue byemploying STG3 as its operational strategy. On the other hand, asit is clear from Fig. 14, PHS facility obtains the highest revenue byadopting STG5. This stems from the higher storage capacity androundtrip efficiency of the PHS facility compared to CAES. PHSasset would be able to discharge for higher number of hours perMWh of charging quantity due to its higher roundtrip efficiency.It can also capture electricity price spikes more effectively since ithas higher discharge power rating.

Our results indicate that the revenue of storage facilities varysignificantly among different years. For the two assets, the results ofyearly net revenues for the particular strategies that offer thehighest arbitrage values are displayed in Figs. 15 and 16. Theaverage and standard deviation of electricity prices in differentyears versus yearly net revenue of the CAES for STG3 is presented inFig. 15. Yearly net revenue of the PHS for STG5 is also depicted inFig. 16. These figures imply that the yearly operational revenue ofboth storage facilities is proportional to average and variance ofobserved electricity prices. Higher average of electricity pricesoriginates form frequent spikes in the market which in turn willincrease the variance as well. Higher standard deviation of elec-tricity prices in a year will provides the ESS facilities with more

opportunities for arbitrage operation. For the CAES facility, theoperational revenue in years from 2011 to 2013 is more than threetime larger than the corresponding number in years 2010 and 2014.The difference between yearly revenues of PHS is more significantdue to its higher power and capacity ratings. The operational rev-enue of the PHS asset in year 2011 is reported to be more than $25million. This number shrinks to less than $5 million for years 2010and 2014.

After calculating the operational revenue of storage facilities,their yearly net profit is estimated by incorporating their annual-ized cost as it is presented in Section 4.3. Fig. 17 represent the yearlyoperational profit of the CAES facility by adopting different strate-gies. The values of net profit in years 2010 and 2014 for all theoperating strategies have been negative. The reason is that elec-tricity prices in these years are lower and thus, there is less op-portunity to capture arbitrage revenue. The highest net profit of$4.2 million is achieved in year 2011 by adopting STG3. For year2012, the mean pool price is $64.31 which is 20% lower than thecorresponding value in year 2011. However, the reduction in netprofit form year 2011e2012 is significantly higher and is estimatedto be more than 50% for all the operation strategies. Strategy STG1results the lowest obtained profit among different applied strate-gies where it outputs positive net profit solely for year 2011 in thefive-year study period.

Fig. 18 illustrates the yearly operational profit of the PHS facilityby adopting different strategies. Observe that for all the years,

Page 10: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-5

0

5

10Y

early

Net

Pro

fit ($

Mill

ion)

20102011201220132014

Fig. 17. Yearly operational profit of the CAES facility by adopting different strategies.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-30

-20

-10

0

-10

-5

0

Yea

rly N

et P

rofit

($ M

illio

n)

20102011201220132014

Fig. 18. Yearly operational profit of the PHS facility by adopting different strategies.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

0

5

10

15

Perc

enta

ge o

f Tim

e (%

)

Charging<50%Charging>50%Discharging<50%Discharging>50%

Fig. 19. Dispatch characteristic of the CAES facility by adopting different strategies.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

0

5

10

15

20

Perc

enta

ge o

f Tim

e (%

)

Charging<50%Charging>50%Discharging<50%Discharging>50%

Fig. 20. Dispatch characteristic of the PHS facility by adopting different strategies.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4336

different bidding strategies outcome negative net profits. Thisstems from the fact that capital cost the PHS asset is notably higherthan CAES. The EAC of the PHS facility is calculated as $33.75millionwhich is four times bigger than $8.45million that is the EACof CAES. The highest yearly operational revenue of PHS facility isobtained in year 2013 by adopting STG5 which is $25.77 million.Thus, the upshot of year 2013 is a deficit of $7.98 million. It shouldbe noted that in this study, we have only considered the arbitrageoperation of the ESS facilities in energy market. Incorporating otherrevenue streams such as participating in ancillary service marketmay result in higher operational profit and is a subject of interest tobe considered by future studies.

5.2. Impact of each strategy on energy cost

Another aspect of the economic analysis is to explore the impactof ESS operation on energy cost. Figs. 19 and 20 depict the dispatchcharacteristic of the CAES and PHS facilities, respectively, based ondifferent strategies. PHS operates for a higher number of hours thanthe CAES facility due to its larger storage capacity. By adoptingSTG1, ESS facilities submit a single block of supply offer with lowestoffer price of $80 and thus, ESS discharges for higher number ofhours compared to other strategies. Accordingly, the number ofcharging hours increases when STG1 is applied. On the other hand,by adopting STG3, ESS facilities operate for less than 8% of the time.Since only for 4.36% of the hours electricity prices have been higherthan the supply offer prices of $300. It also should be noted that the

number of charging hours are generally higher than discharginghours because of the roundtrip efficiency of storage units. Thenumber of charging hours for the two of storage facilities areapproximately equal as they adopt the same strategy to submitdemand bids. However, for the PHS facility, the number of dis-charging hours are relatively higher than CAES due to its higherroundtrip efficiency. It can also be inferred from this figure that thestorage facilities often operate with more than 50% of their ratedpowers both during charging and discharging periods.

Figs. 21 and 22, respectively, represent the impact of CAES andPHS facilities on electricity prices during charging, dischargingperiods, and also the overall impact. All ad-hoc strategies have thesame demand bid, and as a result, the price increase when chargingis very similar for them. The percentage increase in electricityprices for the PHS facility when charging is almost 50% higher thanCAES since its rated charging power is 1.5 times bigger. The ad-hocstrategy with the lowest priced supply offer (STG1) results in thehighest impact on prices when discharging. The impact ofSTG6eSTG8 on prices when discharging is substantially lower. Thisis to be expected as a $900/MWh supply offer can reduce the priceby at most $99.99/MWh. The price reduction during discharging issignificantly higher than the price impact throughout chargingperiods. This stems from the fact that supply curves have extremelyhigher slopes for periods of peak demand that storage unitsdischarge. Thus, a small decrease in the net demand due to gen-eration of the ESS facility, may lead to an extreme reduction in thelevel of electricity prices. The average percentage decrease in

Page 11: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-40

-30

-20

-10

0

10

20C

hang

e in

Ele

ctric

ity P

rices

(%)

ChargingDischargingOverall

Fig. 21. Price impact of CAES facility during charging, discharging, and also the overallimpact.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-50

-40

-30

-20

-10

0

10

20

Cha

nge

in E

lect

ricity

Pric

es (%

)

ChargingDischargingOverall

Fig. 22. Price impact of PHS facility during charging, discharging, and also the overallimpact.

Fig. 23. Locations of wind farms in the Alberta electric system- Wind map curtesy ofAlberta Electric System Operator.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 37

electricity prices during discharge when ESS units adopt STG1 isreported to be 30%. In spite of substantial price reductions duringdischarge, the overall impact on electricity prices are relativelysmall. This is expected as the number of charging hours are higherbased on Figs. 19 and 20 and thus, price increase during chargeoffsets the price reduction of discharge. The operation of CAES fa-cility by adopting STG7 and STG8 leads to an overall increase in themean pool price. While, PHS decreases the final average of elec-tricity prices for all the strategies since its negative price impactduring discharge is higher compared to the CAES facility.

5.3. Impact on the revenue of different types of generation

In this section, we will investigate the impact of the PHS andCAES facilities on the operation and revenue of different type ofgeneration units in the Alberta electricity market. First, must-takegeneration units are considered that are given priority by theAESO to be dispatched in the market. Wind-power in Alberta wasconsidered as must-take generation before the new legislationenacted in April 2015 that allows the participation of wind farms inthe electricity market. Thus, as the first analysis, the impact on therevenue of different wind farms as a result of different strategiesadopted by the ESS facilities is studied. Then, other generation unitsthat actively participate in the market and submit supply offers tosell their available capacities are explored. The impact of ESS on thequantity of their generation and also their revenue is investigated.

The output power of individual wind farms in the Albertaelectricity market is gathered for year 2014. There were 18 windfarms commissioned in the market by end of this year which wereconsidered as price-taker must-take generation units. Impact of theESS operation on wind generation is estimated by integrating thereported hourly mean wind generation with the predicted whole-sale market prices. The results show that the net impact of ESSoperation on wind farm revenue varies, sometimes significantly,depending on the wind speed pattern of wind farms. Fig. 23 depictsthe locations of wind farms in the Alberta electric system. Ouranalysis indicated that three distinct wind speed regimes areidentifiable in the province as it is presented in the map. Fig. 24represents the correlation between wind-power outputs ofdifferent farms. This figure shows that wind farms that belong tothe same cluster demonstrate a high level of correlation. While,wind farms in different clusters show a relatively low cross-correlation between their wind generation outputs.

We investigate the impact of large-scale ESS facilities on therevenue of wind farms in distinct regions for the year 2014. Figs. 25and 26 illustrates the percentage change in the revenue of windfarms belonging to each cluster due to the operation of the CAESand PHS facilities, respectively. Observe that CAES has a relativelymore positive impact on the revenue of wind farms compared tothe PHS facility. This can be justified with reference to Figs. 21 and22 that show the price impact of storage units. It can be concludedthat PHS decreases electricity price more significantly duringdischarge periods and thus, it substantially reduces the revenue ofwind farms during peak demand hours. On the other hand, thechange in pool prices due to operation CAES during discharge is

Page 12: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

3 6 9 12 15 18

3

6

9

12

15

18 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1Region 3Region 1 Region 2

Fig. 24. Correlation between wind farms.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-2

-1

0

1

2

3

4

5

Cha

nge

in th

e re

venu

e of

win

d fa

rms (

%)

Region 1Region 2Region 3

Fig. 25. Percentage change in the revenue of wind farms belonging to each cluster inthe year 2014 due to the operation of the CAES facility.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-3

-2

-1

0

1

2

Cha

nge

in th

e re

venu

e of

win

d fa

rms (

%)

Region 1Region 2Region 3

Fig. 26. Percentage change in the revenue of wind farms belonging to each cluster inthe year 2014 due to the operation of the PHS facility.

Coa

l

Cog

ener

atio

n

Bio

mas

s

Com

bine

d C

ycle

Sim

ple

Cyc

le

Hyd

ro

0

20

40

60

80

100

Perc

enta

ge o

f Cap

acity

(%) BP<40

40<BP<8080<BP<100100<BP<150150<BP<300300<BP<500500<BP<600600<BP<900900<BP

Fig. 27. Distribution of offer prices submitted by different type of generation units inthe Alberta market.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4338

lower since it has lower power rating. Hence, CAES in general has amore positive impact on the revenue of wind generation units.STG7 and STG8 represent the most positive impact on the revenueof wind farms. This stems from the fact that these strategies havethe lowest impact on prices during discharge and consequently, do

not dilute the revenue of wind farms during peak demand hours. Itcan be also inferred from our results that the percentage change inthe revenue of wind farms strongly depends on the cluster towhichthey belong. Wind farms in region three benefit more notably fromthe operation of ESS facilities compared to other regions. WhenCAES facility adopts STG1 and STG4, wind farms in region three arethe only group of wind generation units that experience an increasein their revenue. Similarly, by applying the strategy STG4 by PHS,wind farms in region three gain a higher revenue while othersundergo a decline in their revenue. This implies that to investigatethe impact of ESS operation onwind generation unit, wind farms indifferent regions that possess dissimilar wind speed regimesshould be studied distinctly.

In order to scrutinize the impact of ESS operation on the revenueof other types of generation units, we should analyze the biddingstrategies adopted by each of them. Fig. 27 represents the distri-bution of offer prices submitted by different type of generationunits in the Alberta electricity market. Observe that coal andcogeneration are base load generation units that submit more than90% of their generation capacity with the bid price lower than $40/MWh. Coal and cogeneration assets together include more than70% on the total installed generation capacity in the Alberta elec-tricity market. Biomass and combined cycle generation units whichjointly encompass about 8% of total installed generation, respec-tively, provide the intermediate level of electricity demand in theAlberta electricity market. Finally, hydro units and simple cycle gasturbines supply the peak demand in the market. Supply offerssubmitted by the simple cycle and hydro generation units to theAlberta market in years 2010e2015 had the average bid price of$403.52/MWh and $671.10/MWh, respectively.

Figs. 28 and 29 depict the change in the energy output quantityof different types of generation units due to the operation of theCAES and PHS facilities, respectively. Observe that coal andcogeneration units that provide the base-load experience an in-crease in the their output generated power. Strategy STG1 results inthe highest increase in the generation quantity of coal and cogen-eration power plants since it has higher number of charging hourscompared to other strategies. Both of biomass and combined cycleplants supply the intermediate load, but the operation of ESS hasdifferent impact on their generation quantity. This stems from thefact that biomass units submit a big portion of their available ca-pacity with the price floor of $0/MWh and thus, ESS do not changetheir participation in the market during charging period. However,a slight decrease in the generation quantity of biomass units isobserved when ESS discharges during peak hours. Consequently,

Page 13: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-20

0

20

40

60

80

100

120

Cha

nge

in th

e ge

nera

tion

quan

tity

(GW

h)CoalCogenerationBiomassCombined CycleSimple CycleHydro

Fig. 28. Change in the generation quantity of different types of generation units due tothe operation of the CAES facility.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-40

-20

0

20

40

60

80

100

120

Cha

nge

in th

e ge

nera

tion

quan

tity

(TW

h)

CoalCogenerationBiomassCombined CycleSimple CycleHydro

Fig. 29. Change in the generation quantity of different types of generation units due tothe operation of the PHS facility.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-8

-6

-4

-2

0

2

4

Cha

nge

in th

e re

venu

e (%

)

CoalCogenerationBiomassCombined CycleSimple CycleHydro

Fig. 30. Percentage change in the revenue of different types of generation units due tothe operation of the CAES facility.

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-30

-25

-20

-15

-10

-5

0

5

Cha

nge

in th

e re

venu

e (%

)CoalCogenerationBiomassCombined CycleSimple CycleHydro

Fig. 31. Percentage change in the revenue of different types of generation units due tothe operation of the PHS facility.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 39

biomass power plants undergo an overall reduction in their outputenergy due to the operation of ESS facilities. But, combined cyclegeneration units enjoy an increase in their output energy as ESSfacilities increase the net demand in off-peak periods when operatein the charging mode. On the other hand, simple cycle and hydrogeneration plants that mainly supply the peak load in the Albertaelectricity market experience a drop in their output energy. This isdue to the reduction in net demand during peak load hours as aresult of ESS discharging.

The percentage change in the revenue of different types ofgeneration units due to the operation of the CAES facility, is pre-sented in Fig. 30. As it is clear, operation of CAES during discharginghours leads to a sharp decline in electricity prices and in turn causesa decrease in the revenue of peak load generation units. Off-peakand intermediate load power plants encounter a lower impact ontheir revenues since the increase in their generation quantity off-sets the reduction in electricity prices. Coal and cogeneration unitsenjoy an increase in their aggregate revenue when CAES adoptsstrategies STG3, STG6, STG7, and STG8 for its operation. StrategySTG7 results in the highest positive change in the revenue of alltype of generation units. This can be justified with respect to Fig. 21which implies that STG7 has the lowest impact on themagnitude ofelectricity prices during discharge. Thus, STG7 does not dilute therevenue of generation units during peak hours and also increasetheir revenue during charging instances by raising the net demandand electricity prices. Fig. 31 also depicts the percentage change in

the revenue of different types of generation units due to theoperation of the PHS facility. It can be concluded that PHS facility inaverage decreases the revenue of generation units more notablycompared to CAES. The reason is that PHS facility cause a biggerdecrease in the level of electricity prices when discharges as it has ahigher power rating and drops the level of net demand moresignificantly. Thus, for all the adopted strategies adopted by PHSexcept for STG3, the revenue of different generation technologiesdiminishes.

5.4. Impact on the revenue of different generation entities

Total capacity of commissioned generation in the Albertaelectricity market was reported to be 15,200 MW by end of year2014. Six generation entities encompassed more than 65% of thetotal installed capacity by end of that year. In this work, we willlabel these entities as ENT1, ENT2,…, ENT6 to keep their identityanonymous. Fig. 32 represents the share of each entity indifferent types of generation in the market by the end of year2014. These six entities own more than 55% of the total installedwind power in the province. While, ENT5 has the highest sharewith 180 MW of installed wind power. Four firms of ENT2, ENT3,ENT4, and ENT6 possess 97% of the available coal generationcapacity. These six generation companies own 34%, 92%, 14%, and72% of the installed capacity of cogeneration, hydro, combinedcycle, and simple cycle units, respectively. Observe from Fig. 32

Page 14: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

Win

d

Coa

l

Cog

ener

atio

n

Hyd

ro

Com

bine

d C

ycle

Sim

ple

Cyc

le

Bio

mas

s0

20

40

60

80

100Sh

are

(%)

ENT1ENT2ENT3ENT4ENT5ENT6Others

Fig. 32. Share of main entities in different types of generation in the Alberta electricitymarket by the end of year 2014.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4340

that none of these entities own any biomass generation unit inthe Alberta electricity market.

Fig. 33 shows the generation mix of the entities in the market bythe end of year 2014. It is clear from this figure that these firms havedifferent mix of installed generation. ENT1 has a total installedcapacity of 83 MW that cogeneration and simple cycle have sharesof 55% and 45%, respectively. ENT2 owns about 1700 MW ofinstalled coal and cogeneration capacity and thus is a base loadsupplier in the Alberta electricity market. It has also hydro andsimple cycle power plants in operation with total capacities of32 MW and 198 MW, respectively. ENT3 owns the second biggestwind farm in the Alberta with nameplate capacity of 150 MW. Thiswind farm is located in central Alberta, i.e., Region 3, as its indicatedin Fig. 23. Coal and simple cycle power plants with capacities of2050 MW and 250 MW, respectively, have the highest share ingenerationmix of ENT3. The generation mix of ENT4 entity covers awide range electricity demand in the market with 1200 MW ofinstalled coal to provide base load and 120 MW and 144 MW ofcombined cycle and simple cycle, respectively, to supply interme-diate and peak load. ENT5 with 489 MW of commissioned windpower in the south of province and 790 MW hydro generation hasthe highest share of renewable energy resources in the Albertamarket. ENT6 as a main base load supplier in the market, possess2130 MW and 438 MW of installed coal and cogeneration capac-ities, respectively.

ENT1

ENT2

ENT3

ENT4

ENT5

ENT6

0

500

1000

1500

2000

2500

3000

Inst

alle

d C

apac

ity (M

W)

WindCoalCogenerationHydroCombined CycleSimple Cycle

Fig. 33. Generation mix of the main generation entities in the Alberta market by theend of year 2014.

As it can be inferred from Figs. 32 and 33, the generation entitiesof ENT1, ENT2, ENT3, ENT4, ENT5, and ENT6 have significantlydifferent generation mixes. It was concluded in Section 5.3 thatoperation of the ESS facilities has different impacts on the revenueof various type of generation units. In this section, we investigatethe effect of ESS operation on the income of different generationentities with dissimilar generation mix. To do so, the impact on therevenue of individual power plants due to the dispatch of ESS unitsis analyzed to calculate the overall influence on the earning ofdifferent generation firms. Fig. 34, respectively, depict the per-centage change in the revenue of main generation entities due tothe operation of the CAES and PHS facilities in the year 2014. Bycomparing these figures it can be concluded that in general, gen-eration entities endure amore significant decrease in their earningsdue to the operation of PHS facility than CAES. The reason is thatPHS facility has a higher impact on pool prices during dischargingas it has a higher power rating compared to CAES. This in turn di-lutes the revenue of peak generation units as it was discussed inSection 5.3 and implemented in Fig. 31. It can be understood fromFigs. 34 and 35 that ENT1 company encounters the biggest loss inits revenue compared to other entities. This stems from the fact thatENT1 owns simple cycle and small cogeneration units that act aspeak load generators and thus, undergo a major loss in their rev-enue during discharge operation of the ESS facilities. ENT2, ENT3,and ENT4 with a high share of simple cycle in their generation mixalso face a substantial shortfall in their revenue. ENT5 encountersthe lowest decline in revenue among all the entities in spite of itshigher share of hydro generation as a peak demand supplier. This isbecause of high portion of wind power in the generation mix ofENT5 which generally experiences a mild change in its earningwhen ESS facilities operate in the market. ENT6 possess coal andcogeneration units that supply the base load in the Alberta elec-tricity market. As it is depicted in Figs. 30 and 31, coal and cogen-eration units receive higher revenues when CAES adopts strategiesSTG3, STG6, STG7 and STG8 for its operation and also PHS operatesbased on the ad-hoc strategy of STG3. Consequently, the ENT6companymeets a higher incomewhen storage facilities adopt thesespecific strategies for their operation scheduling.

5.5. Sensitivity analysis

In this section, we evaluate the impact of power rating and ca-pacity of the ESS on its economic performance. It is assumed that theESS participates as an independent facility in the energymarket andtakes advantage of price arbitrage. We have considered the

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-6

-4

-2

0

2

4

Cha

nge

in re

venu

e (%

)

ENT1ENT2ENT3ENT4ENT5ENT6

Fig. 34. Percentage change in the revenue of main generation entities due to theoperation of the CAES facility in the year 2014.

Page 15: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

STG

1

STG

2

STG

3

STG

4

STG

5

STG

6

STG

7

STG

8

OPT

Strategy

-20

-15

-10

-5

0

5C

hang

e in

reve

nue

(%)

ENT1ENT2ENT3ENT4ENT5ENT6

Fig. 35. Percentage change in the revenue of main generation entities due to theoperation of the PHS facility in the year 2014.

4 hr 7 hr 10 hr 13 hr 16 hrReservior Capacity

-30

-20

-10

0

-5

0

-30

Yea

rly N

et P

rofit

($ M

illio

n)

20102011201220132014

Fig. 37. Yearly operational profit of the PHS facilities with the power rating of 150 MWand different reservoir capacities-optimization-based strategy.

40 M

W70

MW

100 M

W

130 M

W

160 M

W

Power Rating

-10

0

10

20

Yea

rly N

et P

rofit

($ M

illio

n) 20102011201220132014

Fig. 38. Yearly operational profit of the CAES facilities with the cavern capacity of 5 hand different power ratings-optimization-based strategy.

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 41

optimization-based strategy to schedule the operation of storagefacilities. Thus, price quota curves are constructed and fed into thedeveloped optimization platform in Section 4.1 to maximize thearbitrage revenue of the ESS. Then, the EAC is calculated from (19)and (20). The cost parameters for CAES and PHS are presented inTables 2 and 3, respectively. After calculating the EAC for the specificpower rating and capacity of ESS, the yearly net profit is determinedas the difference between yearly arbitrage revenue and EAC.

Fig. 36 shows the yearly operational profit of the CAES facilitieswith the power rating of 100 MWand different cavern capacities. Itcould be inferred from this figure that higher capacity of cavernincreases the operational profit of the CAES monotonously. Thisstems from the fact that capital cost of cavern is significantly lowercompared to other components and scaling up the capacity of CAESdoes not change EAC drastically. On the other hand, the arbitragegrows substantially by enlarging the capacity of CAES and thus, thefinal result is an overall improvement in the yearly net profit.

The yearly operational profit of the PHS facilities with the powerrating of 150 MW and different reservoir capacities is presented inFig. 37. In comparison to CAES, the capital cost of reservoir for thePHS is notably higher and thus, larger reservoirs increase the EACremarkably. Consequently, the increase in the arbitrage revenuecaused by higher capacity is not sufficient to offset the change inEAC and yearly net profit reduces in total. The decrease in net profitfor the years 2010 and 2014 is more evident, since the level ofelectricity prices and arbitrage revenue have been relatively lowerin these years.

1 hr 3 hr 5 hr 7 hr 9 hrCavern Capacity

-10

-5

0

5

10

15

Yea

rly N

et P

rofit

($ M

illio

n) 20102011201220132014

Fig. 36. Yearly operational profit of the CAES facilities with the power rating of100 MW and different cavern capacities-optimization-based strategy.

Figs. 38 and 39, respectively, display yearly operational profit ofthe CAES and PHS facilities with a constant cavern capacity anddifferent power ratings. A general conclusion could be drawn thatfor those years that level of electricity prices is high and there issufficient opportunity to capture arbitrage revenue, higher powerrating improves the yearly net profit. However, in those years withlow arbitrage income, the increase in the capital cost of storage

50 M

W

100 M

W

150 M

W

200 M

W

250 M

W

Power Rating

-40

-30

-20

-10

0

10

Yea

rly N

et P

rofit

($ M

illio

n)

20102011201220132014

Fig. 39. Yearly operational profit of the PHS facilities with the reservoir capacity of 10 hand different power ratings-optimization-based strategy.

Page 16: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e4342

facilities is significantly higher than the corresponding improve-ment in the revenue. Thus, the overall yearly profit is negativelyinfluenced by increasing the power rating of the ESS. For instance, itis clear from Fig. 38 that higher power rating has improved theyearly net profit in years 2011, 2012, and 2012, while in years 2010and 2014 the all-inclusive impact on the net profit is negative.

6. Discussion

It is crucial for an energy storage system to take advantage ofrevenue streams other than energy arbitrage by participation inancillary services. At this moment, there is a technical discourse inthe Alberta electricity market, which is aimed at defining the reg-ulatory framework for operation of storage systems in reserve andregulation markets. Having access to various revenue sourcesmakes the operation of independent investor-owned storage sys-tems more captivating. This can ensure the potential investors thereturn of their capital investment and make the storage systemsspeculation economically viable.

However, if a storage system is intended to operate solely in theenergy market, then there is a high probability that arbitrage rev-enue would not be sufficient to recover capital cost as we observedin our results from the Alberta market in years 2010 and 2014.Nevertheless, the storage owner would be able to maximize thearbitrage profit be developing an optimization platform to schedulethe operation of storage by incorporating its impact on wholesaleprices. Doing that, having access to reliable forecasts of electricityprice is a crucial requirement. The input to the optimization plat-form is not only the point forecast of price, but also the price-demand sensitivity curves, which we labeled as ”Price QuotaCurves” in this paper. An interesting direction for future researchcould be to investigate the utilization of different data mining toolsto develop a price forecasting engine to predict price quota curves.The better the accuracy of price forecasts, the higher the revenue ofstorage system form energy arbitrage.

This analysis disregards the collateral effect of strategic offering.But, the reality of a wholesale electricity is a much more compli-cated case in which existing generation units will adjust theirbidding strategies according to new market conditions after theintegration of a large-scale energy storage system. Thus, the priceimpact on revenue of generation entities or companies needs to bemodeled by incorporating their reactions to the new player in themarket. This suggests that an upcoming energy storage systemcannot optimize its profits in isolation. It must also consider thereaction of other market participants including suppliers andconsumers. This interacting optimization problem is called a non-cooperative game. The solution of such a game represents a mar-ket equilibrium under imperfect competition and is called a Nashequilibrium [47]. It should be noted that modeling the imperfectcompetition in the electricity market is beyond the scope of thiswork and is left as a compelling direction for future research.

Another aspect to improve the accuracy of market modeling is toincorporate the impact of fuel price alterations onwhole sale pricesby evaluating the observed correlations in historical data. Authorsdid not have access to the price of natural gas and coal as the maintype of fuels of generation units in the Alberta electricity market.However, it is suggested as an interesting direction for futureresearch to investigate the relationship between fuel prices andelectricity prices to incorporate this factor in the models developedfor competitive electricity markets.

7. Conclusion

This report summarizes the findings of a study into the priceimpact and potential operating profit of merchant ESS facilities

operating in Alberta. As the first step, a database of hourly historicalprice supply and demand curves for the five year period 2010e2014is created. The hourly pool-based electricity prices and equivalentnet demand values are also collected to be included in the database.Prior to investigating the impact of an energy storage facility, theimpact that additional demand and supplymay have onpool prices,without the constraints of storage operations is analyzed.

After that, a number of bidding strategies for a merchant ESSfacility are developed. An optimization-based scheme is consideredto schedule the operation of a large-scale storage system. Also, anumber of ad-hoc strategies to participate in the energy market areexamined for the sake of comparison. Limitations are placed due tothe physical constraints of the energy storage facilities. PumpedHydro Storage (PHS) and Compressed Air Energy Storage (CAES) asthe two primary technologies for bulk storage of electric energy(hundreds of MW-hours) are considered to be integrated to theAlberta electricity market.

Each of the operational strategies is applied to the database ofhistorical market data for the five-year study period. The operatingschedule and prices are used to determine market price impactsand potential operating revenue. Potential impacts on energyconsumers, wind producers and other types of generators are alsoinvestigated. Again, these results are based on an assumption othermarket participants would not have adjusted their operating stra-tegies. Results only consider profits from arbitrage operation of thestorage facility in the energy market; potential revenue streamsfrom participation in other markets are excluded.

The study results indicate that the operating profit of ESS variesfrom one strategy to another one. Strategies that submits supplyoffers with higher prices result in bigger arbitrage revenuecompared to other strategies. The reason is that in by adoptingthese strategies, electricity prices are less affected during dis-charging periods and thus the revenue of ESS from selling itsavailable energy is not diluted. It should be also noted that theyearly net revenue of arbitrage operation strongly depends on thedistribution of electricity prices in a year. Our analyses suggest thatyears with higher average and standard deviation of observedelectricity prices provide better arbitrage opportunities.

Our results also demonstrate the merchant ESS operations inAlberta had the potential to significantly reduce total wholesaleelectricity prices in the 5-year study period. This stems from thefact that ESS charges during off-peak hours that supply curve has amoderate slope and thus, increase in electricity price is not signif-icant. But, discharging instances coincide with peak-demand hoursthat result in a considerable decrease in the pool prices due to thehigher slope of supply curves. This in turn leads to a reduction inenergy costs of electricity end-users that is on the other hand at theexpense of reduced revenue for generators. Investigations indicatethat the impact of merchant ESS facilities on the revenue of indi-vidual wind farms depends on their wind speed regime andchanges from one region to another one. While some wind gener-ation facilities may benefit higher revenues due to the operation ofESS, others will experience a decline in their income.

After exploring the impact of ESS onmust-takewind generation,we analyze other generation units that actively participate in themarket and submit supply offers to sell their available capacities.Our results indicate that coal and cogeneration units that providethe base-load, experience an increase in their generation quantity.The reason is that ESS increases the level of demand during off-peak hours whenever it is in charging mode. However, it shouldbe noted that overall impact of ESS on electricity prices is negativeand this would offset the increase in their generation quantity.Overall, the revenue of base-load generation units may either in-crease or decrease based on the strategy adopted for the operationof ESS facilities. Simple cycle and hydro generation plants that

Page 17: Price impact assessment for large-scale merchant energy storage … · 2017-03-13 · Price impact assessment for large-scale merchant energy storage facilities Payam Zamani-Dehkordi*,

P. Zamani-Dehkordi et al. / Energy 125 (2017) 27e43 43

mainly supply the peak load in the Alberta electricity marketexperience a drop in their output energy. This is due to thereduction in net demand during peak load hours as a result of ESSdischarging. Combined with the reduction in wholesale electricityprices, peak units face a sharp decline in their income from pool-based electricity market.

It should be noted that the present study in this paper isintended to investigate impact of a large-scale merchant energystorage system on the dynamics of electricity prices in a competi-tive electricity market. In this paper we have solely considered thearbitrage operation of ESS in the energy market and potentialrevenue streams from participation in ancillary service market areexcluded. A future study may focus on optimizing the joint-operation of ESS in energy and reserve markets to investigate itsprofitability and also explore its impact on competitive prices ineach market.

References

[1] Rodrigues E, Godina R, Santos S, Bizuayehu A, Contreras J, Catal~ao J. Energystorage systems supporting increased penetration of renewables in islandedsystems. Energy 2014;75:265e80.

[2] Edmunds R, Cockerill T, Foxon T, Ingham D, Pourkashanian M. Technicalbenefits of energy storage and electricity interconnections in future Britishpower systems. Energy 2014;70:577e87.

[3] Li N, Hedman KW. Economic assessment of energy storage in systems withhigh levels of renewable resources. IEEE Trans Sustain Energy 2015;6(3):1103e11.

[4] Hill CA, Such MC, Chen D, Gonzalez J, Grady WM. Battery energy storage forenabling integration of distributed solar power generation. IEEE Trans SmartGrid 2012;3(2):850e7.

[5] Denholm P, Ela E, Kirby B, Milligan M. The role of energy storage withrenewable electricity generation. National Renewable Energy Laboratory;2010.

[6] Divya K, Østergaard J. Battery energy storage technology for power system: anoverview. Electr Power Syst Res 2009;79(4):511e20.

[7] Beaudin M, Zareipour H, Schellenberglabe A, Rosehart W. Energy storage formitigating the variability of renewable electricity sources: an updated review.Energy Sustain Dev 2010;14(4):302e14.

[8] Bueno C, Carta JA. Wind powered pumped hydro storage systems, a means ofincreasing the penetration of renewable energy in the canary islands. RenewSustain Energy Rev 2006;10(4):312e40.

[9] Black M, Strbac G. Value of bulk energy storage for managing wind powerfluctuations. IEEE Trans Energy Convers 2007;22(1):197e205.

[10] Barton JP, Infield DG. Energy storage and its use with intermittent renewableenergy. IEEE Trans Energy Convers 2004;19(2):441e8.

[11] Ibrahim H, Ilinca A, Perron J. Energy storage systems characteristics andcomparisons. Renew Sustain Energy Rev 2008;12(5):1221e50.

[12] Hadjipaschalis I, Poullikkas A, Efthimiou V. Overview of current and futureenergy storage technologies for electric power applications. Renew SustainEnergy Rev 2009;13(6):1513e22.

[13] Deshmukh M, Deshmukh S. Modeling of hybrid renewable energy systems.Renew Sustain Energy Rev 2008;12(1):235e49.

[14] Hozouri MA, Abbaspour A, Fotuhi-Firuzabad M, Moeini-Aghtaie M. On the useof pumped storage for wind energy maximization in transmission-constrained power systems. IEEE Trans Power Syst 2015;30(2):1017e25.

[15] Salgi G, Lund H. System behaviour of compressed-air energy-storage inDenmark with a high penetration of renewable energy sources. Appl Energy2008;85(4):182e9.

[16] Nair N-KC, Garimella N. Battery energy storage systems: assessment for small-scale renewable energy integration. Energy Build 2010;42(11):2124e30.

[17] Lund H, Salgi G. The role of Compressed Air Energy Storage (CAES) in futuresustainable energy systems. Energy Convers Manag 2009;50(5):1172e9.

[18] Vazquez S, Lukic SM, Galvan E, Franquelo LG, Carrasco JM. Energy storagesystems for transport and grid applications. IEEE Trans Ind. Electron2010;57(12):3881e95.

[19] Celik A. A simplified model for estimating the monthly performance ofautonomous wind energy systems with battery storage. Renew Energy

2003;28(4):561e72.[20] Khatod DK, Pant V, Sharma J. Optimized daily scheduling of wind-pumped

hydro plants for a day-ahead electricity market system. In: Internationalconference on power systems (ICPS’09); 2009. p. 1e6.

[21] Costa LM, Bourry F, Juban J, Kariniotakis G. Management of energy storagecoordinated with wind power under electricity market conditions. In: Inter-national conference on probabilistic methods applied to power systems(PMAPS’08); 2008. p. 1e8.

[22] Dicorato M, Forte G, Pisani M, Trovato M. Planning and operating combinedwind-storage system in electricity market. IEEE Trans Sustain Energy2012;3(2):209e17.

[23] Garcia-Gonzalez J, la Muela D, Ruiz RM, Santos LM, Gonz�alez AM. Stochasticjoint optimization of wind generation and pumped-storage units in an elec-tricity market. IEEE Trans Power Syst 2008;23(2):460e8.

[24] Abbaspour M, Satkin M, Mohammadi-Ivatloo B, Lotfi FH, Noorollahi Y.Optimal operation scheduling of wind power integrated with Compressed AirEnergy Storage (CAES). Renew Energy 2013;51:53e9.

[25] Kanakasabapathy P, Swarup KS. Evolutionary tristate PSO for strategic biddingof pumped-storage hydroelectric plant. IEEE Trans Syst Man Cybern2010;40(4):460e71.

[26] Figueiredo FC, Flynn PC. Using diurnal power price to configure pumpedstorage. IEEE Trans Energy Convers 2006;21(3):804e9.

[27] Lu N, Chow JH, Desrochers AA. Pumped-storage hydro-turbine bidding stra-tegies in a competitive electricity market. IEEE Trans Power Syst 2004;19(2):834e41.

[28] Tsai C-C, Cheng Y, Liang S, Lee W-J. The co-optimal bidding strategy ofpumped-storage unit in ercot energy market. In: North American powersymposium (NAPS); 2009. p. 1e6.

[29] Akhavan-Hejazi H, Mohsenian-Rad H. Optimal operation of independentstorage systems in energy and reserve markets with high wind penetration.IEEE Trans Smart Grid 2014;5(2):1088e97.

[30] Mohsenian-Rad H. Optimal bidding, scheduling, and deployment of batterysystems in California day-ahead energy market. IEEE Trans Power Syst PP2015;99:1e12.

[31] Walawalkar R, Apt J, Mancini R. Economics of electric energy storage for en-ergy arbitrage and regulation in New York. Energy Policy 2007;35(4):2558e68.

[32] Sioshansi R, Denholm P, Jenkin T. A comparative analysis of the value of pureand hybrid electricity storage. Energy Econ 2011;33(1):56e66.

[33] Department of Energy, The smart grid: An introduction.[34] Locke G, Gallagher PD. Nist framework and roadmap for smart grid interop-

erability standards. National Institute of Standards and Technology; 2010.p. 33.

[35] Kondziella H, Bruckner T. Economic analysis of electricity storage applicationsin the German spot market for 2020 and 2030. In: 7th conference on energyeconomics and technology; 2012. p. 1e13.

[36] Mohsenian-Rad H. Coordinated price-maker operation of large energy storageunits in nodal energy markets. IEEE Trans Power Syst 2016:786e97.

[37] Sioshansi R, Denholm P, Jenkin T, Weiss J. Estimating the value of electricitystorage in PJM: arbitrage and some welfare effects. Energy Econ 2009;31(2):269e77.

[38] Mannan P, Baden G, Olein L, Brandon C, Scorfield B, Naini N, et al. Techno-economics of energy storage. Alberta Innovates-Technology Futures; 2014.p. 3e11.

[39] Sousa JA, Teixeira F, Faias S. Impact of a price-maker pumped storage hydrounit on the integration of wind energy in power systems. Energy 2014;69:3e11.

[40] AESO, Current supply demand report.[41] de la Torre S, Arroyo JM, Conejo AJ, Contreras J. Price maker self-scheduling in

a pool-based electricity market: a mixed-integer LP approach. IEEE TransPower Syst 2002;17(4):1037e42.

[42] Simbolotti G, Kempener R. Electricity storage: technology brief. The Interna-tional Renewable Energy Agency (IRENA); 1997.

[43] Wright S. Reference design description and cost evaluation for compressed airenergy storage systems. Palo Alto (CA): Electric Power Research Institute(EPRI); 2011. p. 104.

[44] BC Hydro, Pumped storage at mica generating station, preliminary costestimate.

[45] Copeland TE, Weston JF, Shastri K, Education P. Financial theory and corporatepolicy, vol. 3. Reading, MA: Addison-Wesley; 1983.

[46] Open Energy Information, Transparent cost database.[47] Kirschen DS, Strbac G. Fundamentals of power system economics. John Wiley

& Sons; 2004.