Electricity Storage and It’s Economic Potential in Deregulated Markets

- An analysis from a trading perspective

1

Xiabing (Harbing) Lou

Harvard University

Apr-03-2016

Outline

• Electricity market, energy storage and trading

• Potential profit of arbitraging the electricity market

• Demonstration of the trading strategies

2

Part 1: An overview

3

• Electricity market, energy storage and trading • Deregulated electricity market

• Trading opportunities for private investors

• Flow battery as a new energy storage

• Potential profit of arbitraging the electricity market

• Demonstration of trading strategies.

Regulated and Deregulated Electricity Market

4

• Traditional regulated market • Fixed wholesale and retail price

Utilities Generators End users

• Deregulated market

• Bidding to set whole sale price

Electricity bidding mechanism

5

Nuclear, Coal, Hydro

Large scale gas, oil plant

Small scale combustion plant

• Marginal price paid for all electricity • Lower bidding price encouraged, but volatility gets higher

Compare Germany and PJM electricity source

6

Solar 21%

Wind 25%

Gas 14%

Oil 2%

Coal 25%

Nuclear 6%

Biomass 4%

Hydro 3%

Germany Market Solar 0%

Wind 1%

Gas 35%

Oil 5% Coal

34%

Nuclear 19%

Biomass 1%

Hydro 5%

PJM Market

46% unstable source (solar & wind) Almost no unstable source

7

Ger

man

y

PJM

Price valley at noon due to solar input

No valley at noon.

Negative price occasionally

Spikes at 7pm due to household usage

Summer electricity price curve

Can we store the electricity to arbitrage and reduce the volatility?

Energy storage technologies

8

Pump Hydro Li ion battery Flywheel Flow battery

Pros Large capacity, low cost Easy setup Fast charging Low cost

Cons Restricted location Expensive Expensive Occupy large volume

Cost $/kWh

30-50 350 ~1,200 ~80 (traditional)

Flow battery may be a good choice of arbitraging electricity market

Organic flow battery

9

By replacing Vanadium with organic compound Cost reduced to $27/kWh

Summary

• Electricity price is volatile in a deregulated market • Price fluctuates with daily with periodic pattern

• Unstable generators (solar and wind) may significantly change the price pattern of a day.

• Flow battery may be a good choice for arbitraging the electricity price.

10

Part 2: Profit limit

11

• Electricity market, energy storage and trading

• Potential profit of arbitraging the electricity market • Optimized profit with linear optimization model

• Storage parameter design: efficiency and capacity

• Economic comparison between Li battery and flow battery

• Demonstration of trading strategies.

Optimized profit: buy low and sell high

• Assumptions: • Full knowledge of future price curve

• Instant turn on and off

• Uniform charge and discharge

• Approach: • Linear optimization algorithm: iterate until highest profit is achieved

• Setting boundary conditions: starting with empty storage, charging and discharging speed limited by the power.

12

Boundary conditions

13

total profit = 𝑃𝑡 𝐷𝑡 − 𝐶𝑡 ; (1)

𝑇

𝑡=1

𝑆𝑡 = 𝑆𝑡−1 + 𝐷𝑡 +ω ∗ 𝐶𝑡 (2)

𝑆𝑡 ∈ 0, ℎ ∗ 𝑘 (3)

𝐷𝑡 , 𝐶𝑡 ∈ 0, 𝑘 (4)

Maximize total profit with Linear Optimization function in Scipy

Example: a 1MW storage profitability in PJM

14

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecR

ela

tive

Sta

nd

ard

Dev

iati

on

Pro

fit

($)

PJM profit and relative standard deviation

Profit relative deviation

7MWh storage capacity, 1 MW power, 80% round trip efficiency

• 2016 Total profit: $27,700 • Profit increases with standard deviation linearly

y = 0.0023x + 2.853 R² = 0.945

4

6

8

10

12

14

16

500 1500 2500 3500 4500

Stan

dar

d D

evia

tio

n (

$)

Monthly Profit

PJM Montly Profit VS Standard Deviation

Design of a storage device

• Two major components of a storage device: • Power component (inverter, converter, transformer). Cost ∝ power (MW)

• Capacity component (battery, chemical tank, reservoir, etc). Cost ∝ capacity (MWh)

• Use one parameter to depict both power and capacity: capacity factor • Capacity factor(h)= Capacity(discharge) / Power

• Depict how many hours does it take to deplete a storage device

15

Assume 1MW power, use capacity factor and efficiency as variables

Capacity Factor and Efficiency

16

0

1000

2000

3000

4000

5000

6000

0 5 10 15

Pro

fit

$

Capacity Factor (h)

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

40% 50% 60% 70% 80% 90% 100% 110%

Pro

fit

$

Round Trip Efficiency

Li battery Flywheel

Flow battery Pump hydro

Assuming 1MW power • Profit increase with efficiency linearly but render much higher cost • Marginal benefit of increasing capacity factor decreases beyond 5h

Capacity factor vs Profit Efficiency vs Profit

Estimating payback periods of flow battery(FB) and Li battery(LB)

• Capacity cost: $27,000/MWh (FB), $350,000/MWh (LB)

• Power components cost roughly $200,000/MW for both system.

• Annual profit: ~$25,000(FB), ~28,500(LB)

17

Capacity cost (5MWh)

Power cost (1MW) Total cost Annual profit $ Payback period

Flow battery 135,000 200,000 335,000 25,000 15.4y

Li battery 1,750,000 200,000 1,950,000 28,500 68y

Both coal fire plant and pump hydro have payback period around 20y

Flow battery seems to be economically feasible to arbitrage electricity market

Summary

• A medium size 1MW(5h) storage device can achieve an annual profit of $25,000 in PJM 2016.

• Marginal profit of capacity factor decreases beyond 5h.

• Although 90% storage render higher profit, the higher cost does not work with this business model.

• Payback period indicates flow battery is commercially feasible

18

Part 3: on going

19

• Electricity market, energy storage and trading

• Potential profit of arbitraging the electricity market

• Demonstration of trading strategies. • Dumb approach

• Machine learning approach

Optimized trading strategy: from linear optimization

20

Mon Tue Wed Thu Fri Sat Sun

Price in a week

Trading activities in a week

Charging : ~0:00 to 6:00; Discharging: 10:00-14:00 and 19:00-20:00

Dumb approach: fixed trading every day

• Charge during night 0:00 to 6:15 (assume 80% efficiency)

• Discharge at noon and evening: 10:00-14:00 and 19:00 to 20:00

21

Recovers 52% profit compare with optimized strategy

Mon Tue Wed Thu Fri Sat Sun

Machine learning approach

22

Training Set: (date, hour,

weather, etc.)

Label: price

Feature Vectors Machine

Learning Algorithm

New set: (date, hour, etc)

Feature Vector

Predicted price Predictive model

Supervised Learning

Trading strategy based on machine learning

23

Mon Tue Wed Thu Fri Sat Sun

Predicted and actual price

Trading behavior based on predicted price

Trading behavior based on actual price

Recovers 90% profit

Summary

• Simple dumb trading approach cannot effectively recover the profit.

• Machine learning can effectively predict the future electricity price.

• 90% of theoretical profit can be recovered based on the predicted price.

24

Conclusions

• Electricity price in deregulated market is volatile due to demand and supply fluctuation thus provides an opportunity for arbitrage.

• The profit of a electricity storage device can be evaluated with historical data and optimization algorithm.

• Flow battery is economically the best choice.

• A SVR machine learning trading strategy can effectively recover the theoretical profit.

25

26 Thank you very much

Monthly profit vs standard deviation

27

y = 0.0023x + 2.853 R² = 0.945

0

2

4

6

8

10

12

14

16

0 1000 2000 3000 4000 5000

Stan

dar

d D

evia

tio

n $

Monthly Profit

PJM Montly Profit VS Standard Deviation

y = 0.0048x - 0.6772 R² = 0.9364

0

5

10

15

20

25

0 500 1000 1500 2000 2500 3000 3500 4000 4500

Stan

dar

d D

evia

tio

n $

Monthly Profit

German Monly profit vs Standard Deviation

References

• http://www.omnesenergy.com/product/

28

29

R² = 0.9756

20%

25%

30%

35%

40%

45%

50%

800 1300 1800 2300 2800 3300 3800 4300 4800

Re

lati

ve S

tan

dar

d D

evia

tio

n

Monthly Profit Adjusted $

Standard deviation difference over the years

30

-40.00%

-20.00%

0.00%

20.00%

40.00%

60.00%

80.00%

100.00%

120.00%

0

10

20

30

40

50

60

70

80

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

German_Price PJM_Price German Relative std PJM Relative std

Both markets has about 40% average fluctuation in electricity price

Winter electricity curve

31

Dumb strategy 2

32

33

34

35