Embed Size (px)
Citation preview
32 risk.net March 2018
Quant ideas: Renewable energy
The rapid deployment of renewable energy is fundamentally transforming energy supply and power markets, resulting in lower average prices but also more volatile prices. These new market dynamics are causing pain for inflexible genera-
tion resources, but also creating significant revenue-enhancing opportuni-ties for flexible resources, particularly in the ancillary services and real-time markets.
Under this new paradigm, the shift to renewable generation has created new value chains and opportunities, while providing a threat to more entrenched players. From a risk- and decision-analysis perspective, traditional decision-support tools fall short of capturing the dichotomy between monthly price volatility and the new volatility in the hourly and five-minute markets. Power market participants that fail to adapt to these structurally changing market conditions will likely suffer the same fate as Kodak, Nokia and other companies that failed to respond to struc-tural changes in their industries.
In this article, we will explore some of the key implications for decision analysis.
Structural change in power marketsStructural change in a market is a fundamental shift in the basic ways that a market operates. Power markets are at a turning point due to disruptive technologies on both the supply and demand sides. The shift in supply fundamentals of base energy from coal to renewable generation, and the shift in demand due to the rise of distributed generation and batteries, rep-resents a structural change in the traditional models that electric utilities have operated in recent decades.
Fortunately, the speed of change in the utilities business model is slower than in other industries, such as cellphone technology or data storage, allow-ing for established companies to adapt to changing market fundamentals.
Some of the changes in power prices that have been brought by increased renewable penetration are consistent with developments in European power markets that have gone through a similar transformation: lower forward prices; substantial changes in the shape and dynamic of hourly price and implied heat rate profiles; and higher price volatility; particularly in hourly and sub-hourly markets.
As a result of these changes, the traditional approach of measuring and managing ‘average’ market exposures will result in suboptimal decisions and inaccurate risk projections. A key consideration moving forward is that uncertainty matters as much as the average and restructuring the port-folio to take advantage of hourly and sub-hourly price behaviour will be critical to capturing value in power markets.
Changing load patternsOne of the new market dynamics introduced by higher reliance on renewables is the changing load patterns. A clear example is the state of Hawaii, which has been working proactively to meet its goal of generating 100% clean energy by 2045. Policymakers have been aligning government regulations and policies with clean energy goals, facilitating the processes for developing renewable energy and deploying the renewable generation and grid infrastructure.
Figure 1 shows the system demand over the same week in June in the years 2006, 2009, 2012, 2014 and 2017. We can observe how system load levels and profiles have changed dramatically in 10 years. Loads for on-peak hours are down more than 20% and large increases in distributed generation pho-tovoltaics have moved the peak load period to occur later in the day.
Additionally, system ramp patterns have changed, with increased frequency of changes in ramp direction. The erosion of demand is being embraced by the state of Hawaii: capital is being deployed, retail customers are installing rooftop solar, and large hotels have installed co-generation facilities, which fits the business model and strategic objective.
Solar generation and CAISOOn the US mainland, the impact of growing renewable penetration is particularly critical for CAISO (California Independent System Operator) and Ercot (Electric Reliability Council of Texas), driven primarily by solar and wind power, respectively. There are important lessons to be learned from the changes in market dynamics observed in recent years. In this arti-cle, we focus on CAISO.
Supply and demand fundamentals, CAISOIn CAISO, utility scale renewable generation is being added at a record pace. The California Energy Commission estimates 30% of 2017 retail electricity sales in California were served by renewable energy facilities, which is a sig-nificant increase from the 20% contribution level in 2014. The increasing supply of low variable cost wind and solar resources can be seen in changes in the supply stack, which represents the order in which units are dispatched to meet demand, stacked from the lowest cost to the least efficient. Renewables are displacing traditional base load resources and pushing the supply stack
Utilities need to adopt new decision-making tools in order to compete in the “new normal” environment of renewable energy supply. By Gary Dorris and Carlos Blanco
Navigating the new energy market dynamics
Power markets are at a turning point due to disruptive technologies on both the supply and demand sides
ER0218_Technical.indd 2 23/02/2018 11:01
risk.net 33
Quant ideas: Renewable energy
outward, and this trend is projected to continue as renewables displace tradi-tional baseload resources. However, while lower generation costs are already adding pressure on coal units and other traditional baseload resources to retire, the intermittency from renewables are at times creating higher prices and introducing new opportunities from additional market volatility.
Th e changes in the dynamics of hourly prices, as measured through hourly implied market heat rates (power price divided by natural gas price), follow the solar generation profi les advancing in California between 2014 and 2016, as shown in Figure 2.
Figure 2 shows the system-wide hourly solar generation in gigawatts (left Y-axis) as well as the hourly implied heat rates (right Y-axis) in Califor-nia for 2014, 2015 and 2016. Average super-peak prices have dropped from $47/MWh to $36/MWh in the 2014–16 period and the implied heat rates during those hours have decreased from 10MMBtu/MWh to 8MMBtu/MWh. Th e decline in heat rates during those hours is due to the fact that there is signifi cantly more solar generation.
While we see the decrease in prices and implied heat rates during the super-peak hours, the implied heat rates during shoulder hours increases due to high demand and low solar generation levels in the morning and early evening hours. Hours 18–21 had implied heat rates of about 14 in 2014 and, over the course of two years, the implied heat rates increased by 20% to a level close to 17MMBtu/MWh with similar conditions occurring in the early morning hours. Th ese changes suggest a greater oppor-tunity for highly fl exible generation that can counterbalance the natural vicissitudes of renewable generation.
Th e inherent intermittency induced through increasing levels of renewable generation has become directly manifest in the volatility of market prices. Th e eff ect of renewable generation on market prices are demonstrated through the three diff erent charts in Figure 3 of the CAISO node SP-15 (Southern California) in the 2014–16 period. Th e top chart shows with contribution of renewables increasing from 10% to 22% over the three-year period (top chart). Over the same period, the volatility in the day-ahead market prices correspondingly increased from 42% to 60% (middle
chart) while the real-time (RT) price volatility increased from 50% to 100%.
As wind and solar generation displaces conven-tional sources of base energy, it is expected that volatility in locational marginal prices will continue increasing.
Another important dynamic is the increase in frequency and magnitude of real-time price spikes – both upwards and downwards, but principally upwards. Markets with high renewable penetra-tion are characterised by extremely volatile real-time prices related to the inherent intermittency of renewables.
Th e nature of spot price dynamics are shown in Figure 4 through a series of frequency plots that show the number of times the fi ve-minute RT prices exceeds $100/MWh using 2014 to 2016 data for SP15.
Understanding the relationship between the amount of renewable generation and the probabil-ity of price spikes can be used by owners of fl exible resources to make optimal bidding decisions
between energy and ancillary services markets.At the node level, price-volatility dynamics are not just a function of the
hour of the day and month of the year, but also the node location. Figure 6 shows a volatility map of locational marginal prices, measured by their stand-ard deviation. We can observe that volatility in CAISO is not homogeneous across nodes. As expected, the intermittency of renewables is one of the key drivers of volatility. Congestion appears greatest in the Central Valley, but vola-tility is also considerably larger as a result of the greater renewable penetration.
Decision-analysis tools in the ‘new normal’ environmentStructural change has arrived in power markets and it is here to stay. Management teams of market participants need to look forward and identify where opportunities and risks lie in this new environment.
In the presence of wild price swings caused by the intermittency of renewables, decision-support tools need to incorporate weather-driven uncertainty and can no longer be based on weather normalised conditions or average prices and deterministic views.
1 Hawaiian Electric’s changing load patterns, 2006–17
Source: Hawaiian Electric
0
200
400
600
800
1,000
1,200
1,400 06/05/2017 06/09/2014 06/11/2012 06/08/2009 06/12/2006
Mon
00:
00
Mon
04:
50
Mon
09:
40
Mon
14:
30
Mon
19:
20
Tue
00:1
0
Tue
05:0
0
Tue
09:5
0
Tue
14:4
0
Tue
19:3
0
Wed
00:
20
Wed
05:
10
Wed
10:
00
Wed
14:
50
Wed
19:
40
Thu
00:3
0
Thu
05:2
0
Thu
10:1
0
Thu
15:0
0
Thu
19:5
0
Fri 0
0:40
Fri 0
5:30
Fri 1
0:20
Fri 1
5:10
Fri 2
0:00
Sat 0
0:50
Sat 0
5:40
Sat 1
0:30
Sat 1
5:20
Sat 2
0:10
Sun
01:0
0
Sun
05:5
0
Sun
10:4
0
Sun
15:3
0
Sun
20:2
0
Source: Ascend Analytics
2 Implied heat rate versus solar generation in California
7
9
11
13
15
17
19
0
1
2
3
4
5
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Impl
ied
heat
rate
(MM
btu/
MW
h)
Sola
r (G
W)
Hour ending
2016 Solar output 2015 Solar output 2014 Solar output
Declining average prices
Increasing heat rates during morning ramp
Increasing heat rates during evening ramp
Super peak
Decreasing heat rates midday
2016 Implied heat rate 2015 Implied heat rate 2014 Implied heat rate
$47/MWh
$39/MWh
$36/MWh
ER0218_Technical.indd 3 23/02/2018 11:01
34 risk.net March 2018
With a high concentration in renewables, weather has become the new fuel driving the price dynamics of power markets with unprec-edented levels of volatility. Weather now has a pronounced infl uence on co-determining position and cashfl ow exposures. As a result, energy position exposures have become signif-icantly divorced from fi nancial position expo-sures. Th e question utilities face is not whether there are enough resources to meet the average load, but whether there are enough fl exible resources in the portfolio that can isolate cash-fl ows against the extreme market variability.
From a decision-analysis perspective, valua-tion and risk models that do not incorporate the relationship between weather and the supply and demand of power will likely lead to suboptimal decisions. As higher weather variability drives higher power price volatility, more accurate weather forecasting will have increasing value in the new market dynamics.
In addition, it is critical to ensure that models incorporate real-time price dynamics to provide greater insights into actual fi nan-cial exposures. Under the new market dynam-ics, modelling day-ahead prices and loads is not suffi cient because of the inherent natural exposure nearly all portfolios contain to large positive or negative price spikes and their
Quant ideas: Renewable energy
4 Frequency of real-time price spikes in SP-15, by month
Source: Ascend Analytics
Percentage of fi ve-min intervals where Caiso price > $100/MWh, 2014–16
January February March April
May June July August
September October November December10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20Hour of day Hour of day Hour of day Hour of day
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20
10
5
0
Price
spi
kes
> $
100/
MW
h
0 10 20Hour of day Hour of day Hour of day Hour of day
Hour of day Hour of day Hour of day Hour of day
3 SP-15 Market Analysis. Renewable penetration and price volatility (2014–16)
0%
5%
10%
15%
20%
25%
Jan 2014 Jul 2014 Jan 2015 Jul 2015 Jan 2016 Jul 2016
CAISO renewable penetration as proportion of load by month (2014–16)
Rene
wab
le p
enet
ratio
n 22%
7%
0.3
0.4
0.5
0.6
0.7
0.8
60%
42%
CAISO day ahead (DA) price volatility by month (2014–16)
DA v
olat
ility
Jan 2014 Jul 2014 Jan 2015 Jul 2015 Jan 2016 Jul 2016
0 0.5
1 1.5
2 2.5
3 3.5
RT v
olat
ility
100%50%
CAISO real time (RT) price volatility by month (2014–16)
Jan 2014 Jul 2014 Jan 2015 Jul 2015 Jan 2016 Jul 2016
Source: Ascend Analytics
ER0218_Technical.indd 4 23/02/2018 11:01
risk.net 35
material fi nancial implications. Any risk metrics based on average prices and average loads are likely to provide a false sense of security and misalignment of hedges to adequately mitigate uncertainty in future cashfl ows.
In this new environment, fl exibility and timing is critical, even though energy is worth less on average. For example, when valuing renewable projects, it is essential to model hourly and sub-hourly price dynamics, because renewables are settled against the real-time price. Proactive management teams at leading utilities are already repositioning their port-folios to be more fl exible, to capture volatility and short-duration price spikes. Highly fl exible resources such as short-duration and high-capacity batteries or internal combustion engine generation units can provide isolation as well as take advantage of those price dynamics.
For example, let us assume a utility in California owns a fl exible genera-tion resource. Th e interaction of the fl exible resource to market prices follows from fi gure 6, where the day-ahead energy, day-ahead generation off er price, and real-time (fi ve-min-ute) settlement price are shown for a typical day in September.
For the hours when the day-ahead hourly energy price is above the gen-eration off er price – shown in red – the utility would operate in the day-ahead market and eff ectively commit the generation to serve load. How-ever, an additional source of revenue for the utility would come from price spikes in the real-time market in the time intervals where the real-time price exceeds marginal cost of
generation. Another alternative is to collect rents in the ancil-lary services market when the likelihood of price spikes is low. Predictive analytics could help the utility determine the likeli-hood of a price spike in the real-time market based on existing grid conditions and expected renewable generation output for that time interval, and maximise net revenues between the real-time and the ancillary services market.
Markets with high renewable penetration are also seeing a large spread between day-ahead and real-time prices. Th e imbal-ance between buyers and sellers means sellers require a premium to sell in the day-ahead market given the increased uncertainty in the real time market. Th is again spells opportunity and risk for asset holders to retain customers by designing hedging pro-grammes to take advantage of these new market dynamics.
SummaryAs a result of unprecedented technological evolution, renewa-bles are entering the market at an inexorable rate and causing a pronounced transformation of price formation, which is the foundation for structural change. Utilities can draw upon the many lessons from companies that have faced structural or regime changes in other industries, and develop an action plan to succeed in the new environment.
Management teams should keep the big picture in sight and incorporate the impact of structural change on price levels and volatility in the overall business strategy. In the new paradigm,
uncertainty counts as much as averages to capture some of advantages of market structure and dynamics. Flexibility will also be a key source of value, while infl exibility will become a source of pain.
In order to navigate these new market dynamics, tools to support plan-ning, portfolio and risk management decisions need to be improved to adequately capture the structural changes brought about by the increasing role of weather on power markets. ■
The authors would like to thank Lisa Giang, Josh Papernik and Mehdi Shahriari for assistance with the fi gures used in the article.Gary Dorris is the CEO and co-founder of Ascend Analytics. Carlos Blanco is managing director of Analytic Solutions. Emails: [email protected]; [email protected]
Quant ideas: Renewable energy
5 Spatial heterogeneity of market price volatility
Source: Ascend Analytics
6 Day ahead and real-time energy prices for SP-15 and generation offer price
DA and RT energy prices ($/MWh)September 2, 2017
1,000
800
600
400
200
0
Ener
gy P
rice
($/M
Wh)
Hour of day
Additional revenuefrom RT price spikes
Additional revenuefrom RT price spikes
DA priceRT priceDA generation offer price
Serve native load
0 5 10 15 20
Source: Ascend Analytics
ER0218_Technical.indd 5 23/02/2018 11:01
