Electricity Sources Fossil Fuels Fossil Fuels From Deep Within.
The RES-induced Switching E ect Across Fossil Fuels: An Analysis … · The RES-induced Switching E...
Transcript of The RES-induced Switching E ect Across Fossil Fuels: An Analysis … · The RES-induced Switching E...
-
The RES-induced Switching Effect Across Fossil Fuels:An Analysis of the Italian
Day-ahead and Balancing Prices
Angelica Gianfreda, Lucia Parisio, Matteo Pelagatti
Department of Economics, Management and Statistics,University of Milano–Bicocca, Milan, Italy
19th Jan 2017, FEEM-IEFE Joint Seminar
1 / 38
-
Introduction
Aim of the Paper
The intermittent and unpredictable nature of wind and solarproduction has made the real-time balancing activities more complexand relevant for the continuous matching of supply and demand.
We show how RES have affected the fuels-electricity nexus in Italy,considering the relationship between fuel prices and between fuels andelectricity prices (DAM & BAMs).
We analyze how the massive introduction of RES has influencedbalancing activities and we calculate the incurred costs for balancingneeds across hours, technologies and market purposes.
2 / 38
-
Introduction
Main focus on balancing sessions
High RES shares modify the shape of the aggregate supply function inDAM, misplacing gas-fired units.
BAMs are dominated by conventional technologies (thermal, hydroand pumping) which have the required degree of flexibility and enjoya higher degree of market power with respect to the DAM.
In this scenario, we expect two distinct dynamics of thefuels-electricity nexus induced by the growth of RES (less relationshipin DAM and a stronger nexus in BAMs).
We also expect that the new results documented for DAM sessionmay have influenced prices and quantities in real time sessions.
3 / 38
-
Literature
Relevant literature
Papers about long run dynamics among fuels and fuels-electricityprices (mainly on day-ahead):
Erdös (2012) using VECM estimates shows that US natural gas priceshave decoupled from European gas and crude oil prices since 2009.Bosco et al. (2010) found strong evidence of a common long-termdynamics between electricity prices and gas prices for the major EUpower exchanges. This long run common dynamics is one of the keyfactors explaining the almost strong integration among price series ofthe different power exchanges.More recently, this relationship appears to be weakened Gianfreda et al.(2016b), so that the introduction of RES appears to have obstacled thelong run convergence of EU prices.
4 / 38
-
Literature
Relevant literature 2
Papers studying the relationship between RES-E and electricity prices(Texas, Australia, Spain, Denmark, Norway, United Kingdom, TheNetherlands and Germany):
Woo et al. (2011), Ketterer (2014), Mulder and Scholtens (2013),Mauritzen (2013), Gelabert et al. (2011), and Cruz et al. (2011).However, these recent contributions are mainly devoted to the analysisof day-ahead prices and not on balancing and fuel prices.Hirth and Ziegenhagen (2015) provide a clear description of the mainissues regarding balancing activities and relate them to therequirements imposed by the increasing share of variable RESproduction. They describe the German market data and, surprisingly,notice that while German wind capacity has tripled since 2008,balancing reserves have been reduced by 15% and balancing costs by50%.
5 / 38
-
Literature
Relevant literature 3
Papers considering structure and rules for the functioning ofbalancing markets.
Papers studying conditions for participation of RES units in thebalancing market: Fernandes et al. (2016).
Papers studying the relationships among spot, adjustment andregulation prices: empirical evidence that the intra-daily sessions arewell-functioning and low-cost market tools to ease the introduction ofa high share of RES:
Gianfreda et al (2016), MI sessions in ItalyChaves-Avila and Fernandes (2015), Spain
Both papers conclude that market design leaves room to possiblestrategic behavior across day-ahead and intra-day markets, giving riseto higher system costs.
6 / 38
-
Background
Evolution of the Italian generation mixIdentification of Two Scenarios: “low” (06-08) and “high” (13-15) RES
Italian shares by technology generation (on the left), and RES penetrationtogether with Demand levels in TW (on the right)
7 / 38
-
Background
RES generation in ItalySelection of the Northern Zone
Hydro (left), solar PV (middle) and wind (right) generation
In Northern Italy, there is the majority of hydro and solar PV. Whereas,most wind power is generated in Southern Italy.However, there are only few observations in Southern BAMs.
8 / 38
-
Background
Inspection of Intra-daily ProfilesSelection of Hours: 3–9–11–13–19–21
1200
014
000
1600
018
000
2000
022
000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2012 2013 2014
Load Intra-daily Profile in Northern Italy
010
0020
0030
00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2012 2013 2014
Solar Intra-Daily Profile in Northern Italy
2000
2500
3000
3500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2012 2013 2014
Hydro Intra-Daily Profile in Northern Italy
1214
1618
2022
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2012 2013 2014
Wind Intra-Daily Profile in Northern Italy
9 / 38
-
Background
Inspection of Intra-daily Profiles
2040
6080
100
120
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2006 2007 20082013 2014 2015
Intra-daily Mean Zonal Prices for Northern Italy
Spread between peak and off-peak: in 2008 peak price was three times theoff-peak, whereas in 2015 peak price was only 50% higher.
10 / 38
-
Background
Marginal technology index, MGP - North zone
020
4060
8010
00
2040
6080
100
2006 2007 2008 2013 2014 2015 2006 2007 2008 2013 2014 2015 2006 2007 2008 2013 2014 2015
2006 2007 2008 2013 2014 2015 2006 2007 2008 2013 2014 2015 2006 2007 2008 2013 2014 2015
3 9 11
13 19 21
CCGT & TG Other Coal
Hydro Run Hydro Basin Hydro Pump
Natural Gas Oil Oil-Coal
Oil-Gas Foreign VZ & MC FER
Marginal Technology Index - North (MGP)
11 / 38
-
Background
ITM - comments
Decreasing role of gas
Coal maintains or even increases its role (see in particular h3)
Foreign zones are marginal with high frequency
RES start to be the marginal technology even if with very lowfrequency
12 / 38
-
Description of rules for balancing markets
Real time markets
Ancillary services markets have a scheduling sub-stage (ex-ante MSDwith 4 sessions) and a balancing market (MB) with 5 sessions.
MSD is the marketplace where the Italian TSO, Terna, negotiates allresources necessary to guarantee the system security, includingdispatching services useful for resolving intra-zonal congestions, theestablishment of an adequate reserve and real time balancing.
During MB sessions, Terna accepts energy demand bids and supplyoffers in order to provide secondary control and to balance energyinjections into and withdrawals from the grid in real time.
The ex-ante MSD and MB are based on the pay-as-bid pricingmechanism (a reference price usually calculated as the weightedaverage of all accepted bids, both for purchases and for sales).
Italian suppliers of balancing power are obliged to deliver energyunder fixed technical conditions, like time of response, ramp rates andduration.
13 / 38
-
Description of rules for balancing markets
Timing of transactions in different market sessions
Bids submitted in MB sessions can only contain better economicconditions with respect to MSD bids, otherwise ex-ante MSD bids remainvalid.
14 / 38
-
Description of rules for balancing markets
Balancing products
Balancing products can be divided into two main categories:
1 balancing capacity, not committed in other markets2 balancing energy, which refers to the actual variation of generation (or
consumption) with the purpose of reestablishing the balance betweengeneration and demand in real time
Market purpose: ‘upward’ reserve (for balancing capacity/energyprocured to compensate a negative imbalance) and ‘downward’reserve (for balancing capacity/energy procured to compensate apositive imbalance)
Participants are obliged to comply with the production/consumptionprogram established in the day-ahead and in the intra-day marketsand they are financially responsible for any deviations with respect totheir market schedules.
15 / 38
-
Description of rules for balancing markets
Participants to balancing sessions
Balancing sessions are more concentrated than DAM session.
Thermal
Pumping units
Hydro units
In recent years we notice a reduction of capacity entitled to bid intobalancing session, expecially in the thermal segment (-5,7%).
0 20000 40000 60000 80000 100000 120000
2012
2013
2014
MW
MSD Thermal MSD Power Tot Power
16 / 38
-
Description of rules for balancing markets
Balance of TERNA operations
Negative balance of Terna’s operations (cost for the system coveredby the so-called uplift component). Its value was 3.82e/MWh in2009, but it almost doubled in 2014 (being equal to 6.25e/MWh).
The main cost components are represented by:
1 ‘the planning of services’ (approvvigionamento servizi) concerningactivities in the ex-ante MSD sessions, which was mainly stable aroundone billione across years;
2 the ‘energy component’ (componente energia) taking into account allrealized imbalances (a cost of e459 M in 2014);
3 contracts to secure upward reserves (stable across years)4 Start-up and status change cost (gettone di avviamento) introduced in
2014 (e82 M in 2014)
We concentrate on the two first components.
17 / 38
-
Empirical Analysis Data
Data Description and Providers
Two samples: 2006–2008 and 2013–2015
Zonal day-ahead electricity prices (GME)
Balancing prices as weighted averages of awarded quantities under the‘pay-as-bid’ rule (on both MSD & MB), and at disaggregated level(GME)
Oil, Coal and ICE UK Natural Gas prices (Datastream)
Actual Load as proxy for Demand (ENTSO-E for Italy & Terna forNorth zone, but only from 2010)
18 / 38
-
Empirical Analysis Methods
Methods: VECM
We decided to keep all the time series at their original (daily)frequency and treat the seasonal components with a datapre-processing.
All time series of electricity, coal and gas prices were tested for a unitroot using the ADF test
Johansen’s test: for each considered hour and for each subsample, wetested for the presence of cointegration among the logarithms ofelectricity and fuels prices.
We estimated a vector error correction model (VECM) for each hour,coherently with the number of cointegrating relations found byJohansen’s test.
19 / 38
-
Empirical Analysis Methods
Methods: FEVD
In the VECM, the best way to assess the role thatfuel prices play in influencing electricity prices in the long-run is bythe forecast error variance decomposition, (FEVD), which allows todetermine how much of the forecast error variance of each variablescan be explained by exogenous shocks to the other variables
The relationship among fuel prices (oil, gas and coal) is firstly tested
Then, the influence of fuel prices on electricity prices is considered atboth the day-ahead and balancing levels
20 / 38
-
Empirical Analysis Dynamics of Fuel Prices
Forecast Error Variance Decomposition: OIL
Oil prices became largely independent from shocks affecting other fuels
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_oil
l_gasl_coal
l_oil
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_oil
l_gasl_coal
l_oil
2006-2008 (left) and 2013–2015 (right)
21 / 38
-
Empirical Analysis Dynamics of Fuel Prices
Forecast Error Variance Decomposition: GAS
The role of OIL in explaining the long-run dynamics of gas prices largelydecreased (decoupling)
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_gas
l_gasl_coal
l_oil
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_gas
l_gasl_coal
l_oil
2006-2008 (left) and 2013–2015 (right)
22 / 38
-
Empirical Analysis Dynamics of Fuel Prices
Forecast Error Variance Decomposition: COAL
The role of OIL in explaining the long-run dynamics of coal prices largelyreduced
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_coal
l_gasl_coal
l_oil
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_coal
l_gasl_coal
l_oil
2006-2008 (left) and 2013–2015 (right)
23 / 38
-
Empirical Analysis Day-Ahead & Balancing Prices
Forecast Error Variance Decomposition: H3DA (left column), BA (right column), 1st sample (top row), 2nd sample (bottom row)
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_nord3adj
l_gasl_coal
l_oill_nord3adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap3adj
l_gasl_coal
l_oill_bap3adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_nord3adj
l_gasl_coal
l_oill_nord3adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap3adj
l_gasl_coal
l_oill_bap3adj
24 / 38
-
Empirical Analysis Day-Ahead & Balancing Prices
Forecast Error Variance Decomposition: H9DA (left column), BA (right column), 1st sample (top row), 2nd sample (bottom row)
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_nord9adj
l_gasl_coal
l_oill_nord9adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap9adj
l_gasl_coal
l_oill_bap9adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_nord9adj
l_gasl_coal
l_oill_nord9adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap9adj
l_gasl_coal
l_oill_bap9adj
25 / 38
-
Empirical Analysis Day-Ahead & Balancing Prices
Forecast Error Variance Decomposition: H13DA (left column), BA (right column), 1st sample (top row), 2nd sample (bottom row)
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for nord13adj
l_gasl_coal
l_oilnord13adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap13adj
l_gasl_coal
l_oill_bap13adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for nord13adj
l_gasl_coal
l_oilnord13adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap13adj
l_gasl_coal
l_oill_bap13adj
26 / 38
-
Empirical Analysis Day-Ahead & Balancing Prices
Forecast Error Variance Decomposition: H21DA (left column), BA (right column), 1st sample (top row), 2nd sample (bottom row)
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for nord21adj
l_gasl_coal
l_oilnord21adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap21adj
l_gasl_coal
l_oill_bap21adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_nord21adj
l_gasl_coal
l_oill_nord21adj
0
20
40
60
80
100
0 50 100 150 200 250 300 350
days
forecast variance decomposition for l_bap21adj
l_gasl_coal
l_oill_bap21adj
27 / 38
-
Empirical Analysis Balancing Costs
Computations
We compute the actual balancing costs1 multiplying the awardedprices for corresponding awarded quantities at unit level
then, we aggregate the information across technologies, hours, yearsand market ‘purpose’
“sales” are situations in which Terna buys quantities incurring in ‘costs’for the system (represented with negative values) – “up-regulation”
⇒ general increasing yearly mean prices across the two samples
whereas “purchases” are situations in which Terna sells quantitiesobtaining instead ‘profits’ (depicted with positive values) –“down-regulation”
⇒ decreasing yearly mean prices across the two samples
1Focusing only on two components of the uplift: the first one is the planning ofservices, which concerns the ex-ante MSD sessions, and the second one is the energycomponent which takes into account all the realized imbalances.
28 / 38
-
Empirical Analysis Balancing Costs
Balancing Quantities in the “ex-ante MSD”Yearly Sum of Awarded Purchased (on the first row) and Offered or “Sold” (on thesecond row) Quantities across hours and technologies
0
50
100
150
200
250
300
350
400
450
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
3 9 11 13 19 21
Pumping Hydro Thermal
GW
Awarded BIDs on MSD
0
50
100
150
200
250
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
3 9 11 13 19 21
Pumping Hydro Thermal
GW
Awarded OFFs on MSD
29 / 38
-
Empirical Analysis Balancing Costs
Balancing Quantities in “MB”Yearly Sum of Awarded Purchased (on the first row) and Offered or “Sold” (on thesecond row) Quantities across hours and technologies
0
50
100
150
200
250
300
350
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
3 9 11 13 19 21
Pumping Hydro Thermal
GW
Awarded BIDs on MB
0
50
100
150
200
250
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
3 9 11 13 19 21
Pumping Hydro Thermal
GW
Awarded OFFs on MB
30 / 38
-
Empirical Analysis Balancing Costs
Price variations across the two samples in MSD and MB
Hydro Water Pumping ThermalMax Mean Max Mean Max Mean
Hour MSD MB MSD MB MSD MB MSD MB MSD MB MSD MB3 ↓ 20 ↑ 111 ↓ 3 ↑ 8 ↑ 19 ↑ 67 ↑ 36 ↑ 63 ↑ 148 ↑ 884 ↓ 3 ↑ 319 ↓ 54 ↑ 176 ↓ 33 ↓ 31 ↑ 19 ↑ 57 ↑ 11 ↑ 37 ↑ 48 ↑ 30 ↓ 28 ↑ 4511 ↓ 12 ↑ 1422 ↓ 44 ↓ 20 ↑ 34 ↑ 55 ↑ 15 ↑ 34 ↑ 38 ↑ 25 ↓ 34 ↑ 2113 ↓ 46 ↑ 13 ↓ 28 ↓ 31 ↑ 25 ↑ 39 ←→ ↑ 28 ↑ 35 ↑ 1717 ↓ 34 ↑ 1719 ↑ 22 ↑ 1689 ↓ 22 ↓ 24 ↑ 48 ↑ 60 ↑ 35 ↑ 40 ↓ 11 ↑ 903 ↓ 33 ↑ 1821 ↓ 41 ↑ 1922 ↓ 28 ↓ 23 ↑ 43 ↑ 55 ↑ 36 ↑ 42 ↓ 50 ↑ 379 ↓ 34 ↑ 18
Dynamics across samples for the average Maximum and Mean Pricesawarded for “Sales” on MSD and MB across hours and technologies,where ↑, ↓ and ↔ represent an average increment, decrement or nochanges across the two samples measured by the corresponding amountsexpressed in e/MWh.
31 / 38
-
Empirical Analysis Balancing Costs
Evolution of balancing costs across technologiesThermal Costs (in thousands of e)
-35000
-25000
-15000
-5000
5000
15000
25000
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
3 9 11 13 19 21
MB MSDTHERMAL
32 / 38
-
Empirical Analysis Balancing Costs
Evolution of balancing costs across technologiesHydro Costs (in thousands of e)
-5000
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
20
06
20
07
20
08
20
13
20
14
20
15
20
06
20
07
20
08
20
13
20
14
20
15
20
06
20
07
20
08
20
13
20
14
20
15
20
06
20
07
20
08
20
13
20
14
20
15
20
06
20
07
20
08
20
13
20
14
20
15
20
06
20
07
20
08
20
13
20
14
20
15
3 9 11 13 19 21
MB MSDHYDRO
33 / 38
-
Empirical Analysis Balancing Costs
Evolution of balancing costs across technologiesWater Pumping Costs (in thousands of e)
-12500
-11000
-9500
-8000
-6500
-5000
-3500
-2000
-500
1000
2500
4000
5500
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
3 9 11 13 19 21
MB MSDWATER PUMPING
34 / 38
-
Empirical Analysis Balancing Costs
Overall Balance (in thousands of e)as the difference between profits and costs, faced by the Italian TSO for the Northern zone
We quantify the overall profits/costs as sum across technologies on bothmarket sessions within a year. Clearly the activities of planning resourcesand dispatching balancing power are highly costly, and increasing acrosssamples for all hours but H19 & H21
-8282-11158
-20244
-21728
-19134
-25694
-19935
-26537
-20842
-16133
-19200
-13907
-35000
-30000
-25000
-20000
-15000
-10000
-5000
0
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
2006
2007
2008
2013
2014
2015
3 9 11 13 19 21
Balance
35 / 38
-
Conclusions
Conclusions
We documented a decoupling between oil and gas prices in oursecond sample (2013-15) with respect to the first sample (2006-08)
We documented a switching effect among fuels in influencingelectricity prices
the switching effect is remarkable in the day-ahead marketthe same effect is observed in balancing prices but with a reduced size
Balancing costs are generally higher in the second sample
The planning activity executed in MSD is actually a substantial partof computed costs and a migration towards a “capacity market” maybe of help for the system
36 / 38
-
Appendix
Transactions in North zone at h11 as an example
05
01
00
15
02
00
/M
Wh
-10
00
0-8
00
0-6
00
0-4
00
0-2
00
00
MW
1/1/2012 1/1/2013 1/1/2014 1/1/2015
Net Trades MGP price
MGP Market Session
37 / 38
-
Appendix
Results about MI market sessions
05
01
00
15
02
00
25
0/
MW
h
-10
00
-50
00
50
01
00
01
50
0M
W
1/1/2012 1/1/2013 1/1/2014 1/1/2015
Net Trades MGP price MI1 price
MI1 Market Session
38 / 38
-
Appendix
Results about MSD market sessions
020
040
060
080
010
00/M
Wh
-100
00
1000
2000
3000
MW
1/1/2012 1/1/2013 1/1/2014 1/1/2015
Net Trades MGP price MSD price
MSD Market Session
39 / 38
-
Appendix
Results about MB market sessions
020
040
060
080
010
00/M
Wh
-300
0-2
000
-100
00
1000
2000
MW
1/1/2012 1/1/2013 1/1/2014 1/1/2015
Net Trades MGP price MB price
MB Market Session
40 / 38
IntroductionLiteratureBackgroundDescription of rules for balancing marketsEmpirical AnalysisDataMethodsDynamics of Fuel PricesDay-Ahead & Balancing PricesBalancing Costs
ConclusionsAppendix