negative wholesale power prices: why they occur and what to do about them

Negative Wholesale Power Prices: Why They Occur

and What to Do about Them

A Study of the German Power Market

Maria Woodman, StudentEconomics Department, New York University

Research Motivation:The German Power Market and Negative Wholesale Power Prices

• Market Anomaly– Electricity doesn’t obey traditional commodity price

behaviors – Negative wholesale prices can result

• Three Causes– Increases in wind infeed when demand is low– Government incentives to favor wind producers – Flat rate prices cause customers to consume

electricity regardless of the value of each MW

Gately

I don't understand this point -- is there a clearer way to explain this? The idea is that consumers don't pay the hourly prices that wholesale buyers pay. Consumers pay a single price for all hours regardless of what is happening to the hourly variation in wholesale prices.

mwoodman

Is this clearer? I plan on explaining further in depth in the presentation.

Wind Behavior

01.01.2009

03.01.2009

05.01.2009

07.01.2009

09.01.2009

11.01.2009

13.01.2009

15.01.2009

17.01.2009

19.01.2009

21.01.2009

23.01.2009

25.01.2009

27.01.2009

29.01.2009

31.01.20090

2000

4000

6000

8000

10000

12000

14000

16000

45000

50000

55000

60000

65000Daily Variation in Wind at 12 PM - January

Wind Infeed

Total Load

Win

d In

feed

MW T

otal Load

MW

Wind generation is uncorrelated with demand, meaning it can cause erratic price swings

0.15 0.2 0.25 0.3 0.35 0.4 0.45

-200.00

-150.00

-100.00

-50.00

0.00

50.00

100.00

f(x) = − 930.633692798745 x + 230.635050673888R² = 0.724892998688939

High Wind Infeed, Weekend

Wind % of Total Generation

Pric

e

0.15 0.2 0.25 0.3 0.35 0.4 0.45

-200.00

-150.00

-100.00

-50.00

0.00

50.00

100.00

f(x) = − 325.97474887977 x + 115.715956658648R² = 0.443294399655772

High Wind Infeed, Weekday


Pric

e 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

-200.00

-150.00

-100.00

-50.00

0.00

50.00

100.00

f(x) = − 392.397917461131 x + 39.826474793298R² = 0.229910999865855

Low Wind Infeed, Weekday


Pric

e

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

-200.00

-150.00

-100.00

-50.00

0.00

50.00

100.00

f(x) = − 517.312095396445 x + 58.0461812081028R² = 0.136458462960702

Low Wind Infeed, Weekend


Pric

e

Power Prices and Wind Generation

Gately

new graphs - still have to change to percentagesNot CRUCIAL now that you've labeled the axes.

Gately

this slide is MUCH better now.

Power Generation Supply (without Wind)

Hydro Nuclear Lignite Coal

Bituminous Coal

Gas Turbine

Wholesale clearing price

MW of Capacity

M

argi

nal

Cost

Demand

Supply

Merit Order Curve

Combined Cycle Gas Turbine

mwoodman

my thought was to show this as the curve with existing supply types and then use the following curve to illustrate what happens when wind enters the curve

Gately

If this is Without Wind then you'll need a follow-up graph With Wind. I'd suggest you combine into one graph.Also, I think this is really about MWhours not MW. We're talking about using existing capacity, and it's the lowest MC per MWHr that gets used first, isn't it?It looks like your focus is on how to add new capacity, not on the order in which to use existing capacity.

Nuclear

Lignite Coal

Bituminous Coal

Wholesale clearing price

MW of Capacity

M

argi

nal C

ost

Demand

Supply – with Wind Shifted Merit Order Curve

Combined Cycle Gas Turbine

WIND0

Power Generation Supply (with Wind)

Supply – No Wind

mwoodman

Does this make sense?

Increased Wind Infeed

10000 12000 14000 16000 18000 20000 22000 24000 26000 28000-200

-150

-100

-50

0

50

100

150

200

Volume of MW

Pric

e pe

r MW

h

An influx of wind power shifts that merit order curve rightward, which drives prices down

Gately

don't abbreviate MW or MWhrWind was omitted from Merit Order curve in previous slide. How can change in wind affect it? Not obvious.Don't use such huge vertical scale. Make min about -200. Max about 200.

mwoodman

makes sense. this granular view makes it easier to understand

Gately

OK, but you still need to make the axes agree. One is in MW and the other is P/MWHr.

Flat Rate Retail Prices

• Retail prices don’t represent a consumers true willingness to pay

• In the case of negative wholesale prices, they grossly overpay in the retail market.

D

Retail Market

S

L

P

Wholesale Market

0

P

L

SD D

Gately

why is retail D flat? It should be sloped.

Gately

this second bullet is very confusing. we're talking about negative wholesale prices.

mwoodman

yes, negative wholesale prices. but the price is fixed in retail market (in Germany in 2009 it was ~12.33euro/MW), so customers are paying much more in the retail market then the electricity is valued in the wholesale marketthis bullet is perhaps not clear. i'll think of a concise way to revise

Gately

with flat rate pricing, isn't demand flat? since at each quantity of MW, people are subject to the same constant priceNo, D isn't flat. With fixed price, the demanders still have a sloped D curve but it's S that is flat (if either is flat). Consumers appear to face a flat S curve at the fixed P; they can buy as little or as much as they want, without affecting the P.

Gately

still needs to be explained better

Is Dynamic Pricing the Answer?

• What is dynamic (“time-of-day”) pricing?– Allows retail prices to match wholesale prices in real

time• Stimulates a demand side price response

• How can it impact negative prices?– The solution isn’t simple• An estimated retail demand curve isn’t defined when

prices are negative

Gately

is this the same (or close to) time-of-day pricing?

Gately

But your graphs show that D for negative prices is defined, and it's vertical.

mwoodman

Yes, dynamic pricing is the catch all term for all types of time-of-use pricing - real time pricing, time of use, critical peak pricing - to name a few.

mwoodman

Wholesale demand curves are defined that way based on the bids by utilities into the auction. The retail demand curve estimates that have generally been used in studies of consumer price response have assumed a constant elasticity curve, so I do the same.

Gately

I've used constant elas. D curves as well, but I don't believe that literally.And in your case, if you have const.elas.D it's IMPOSSIBLE to have negative prices. So I'd prefer that you say that you have low elasticities estimated for the range of prices used historically. Also, you could estimate low elas. for a linear D curve, but the elas. would change as you have lower or higher prices.

Existing Studies of Dynamic Pricing

• Studies evaluating the results of implemented programs have returned varied ranges of end-user price response• Long Run Elasticity – 0.3-0.5 (Borenstein, 2005)• Short Run Industrial End-User Elasticity – 0.01-0.27 (Boisvert, 2007;

Neenan, 2004; Braithwait and Sheasy, 2002; Patrick and Wolak, 1997)

• Using these ranges, the studies focused on the use of dynamic pricing to reduce peak price and load

Method for Analyzing RTP

• Construct wholesale supply and demand curves for a set of hours representing different combinations of demand and wind infeed

• Construct demand curves representing different levels of price response using differing price elasticities

• Induce an increase in wind power by shifting the supply curve

• Solve for the new equilibrium points given the new supply and demand curves

Gately

this is not clear without a graph to explain it.

Gately

I don't understand this

Gately

I'm not sure I understand this

mwoodman

Is this wording clearer? Dynamic pricing has been shown to increase consumer price elasticity, so studies have used different levels of elasticty to measure consumer response.

mwoodman

I changed the wording. Is this clearer?

Gately

It's not dynamic pricing that's generating the level of price elasticity, it's higher prices (or lower P) with a D curve that does not have constant elasticity. For example, with linear D curve, elas. in abs.value is lowest at the lowest P levels. As you move up the linear D curve, the elas. gets bigger in absolute value, and becomes 1 (abs.value) at midpoint of curve and elastic (>1 abs.value) above the midpoint). THAT is what's happening when you change the prices that you're talking about.

Allocation of Hours• It was found that a necessary condition for negative

prices appeared to be either high wind in-feed ( >12 GW) coupled with moderate system demand (40-50 GW) or low system demand ( <40 GW) coupled with moderate wind in-feed (5-10 GW) (Genoese, 2010).

• Using these metrics, I was able to disaggregate the hours into their respective buckets for analysis

High Wind Infeed Low Wind Infeed

High Demand

Jan 12 - 7 PM - Weekday Oct 7 - 8 PM - Weekday

Wind MW: 12594.75 Wind MW: 2361.5

Total Demand: 72826 Total Demand: 65847

Low Demand

Mar 8 – 9 AM - Weekend May 17 – 5 AM - WeekendWind MW: 9914.25 Wind MW: 2276Total Demand 43358 Total Demand: 33394

Hours of focus

Constructing the Model

13000 13500 14000 14500 15000 15500 16000 16500 17000-20

-10

0

10

20

30

40

Supply

Retail Demand e=-0.01



Increased Wind Infeed

€/M

Wh

Gately

is this a low-demand time? Weekday?Put labels on graphs, not in legends. Use fewer demand elasticity cases; this is too cluttered.Price is per MWhr but Q is MW ??

Gately

I don't understand why you're using constant-elasticity demand curves. I'd suggest using linear demand "curves". Yours suggest that at low enought prices the Q demanded expands very rapidly. I'd think that linear response is much more realistic.

mwoodman

I used constant elasticity curves because in all of the literature I read that was dealing with consumer response/dyamic pricing, demand curves with constant elasticity were used.

mwoodman

is this what you were thinking in terms of labels?

Gately

aren't these MWHours not MW?On both axes.

Preliminary Results

• For the hour type of low demand and high wind infeed, on average, a price elasticity of at least -0.14 was needed to have a market clearing price of €0.00– For the case of increased wind generation

• Following the same logic, a price elasticity of approximately -0.44 was necessary to raise the price to equal the existing retail flat rate price. – The elasticity value is unrealistic given previous estimates of consumer

price response

Occurrence of Negative

PricesWeekday Weekend Grand Total

Early Morning 14.08% 36.62% 50.70%

Mid Day 2.82% 1.41% 4.23%

Night 21.13% 23.94% 45.07%

Grand Total 38.03% 61.97% 100.00%

Negative Prices: Hours of Occurrence

Of the 1% of hours that were affected in 2009

The vast majority fell during early morning and weekend hours

Conclusions• RTP may not have a significant effect and in some cases

might even be a hindrance to the market.• Other demand side management techniques may be

more effective in mitigating the market inefficiency of negative prices

• Additional R&D in electric vehicles, smart grid technology and implementation, and smart appliances could aid in making demand side management viable

Gately

this is a consequence of your assumption that you have constant elasticity demand curves. It is not a conclusion.

mwoodman

ok, makes sense. i'll work on the conclusion section

Thank You!

Questions?

Maria Woodman, New York UniversityEmail: [email protected]

negative wholesale power prices: why they occur and what to do about them

Documents