Portfolio Optimisation Science

The Australian National Electricity Market (NEM) is a dynamic and often unpredictable environment, where organisations with market exposures must make complex decisions amongst uncertain market conditions. Traditional risk management tools can often fall short in capturing the true challenge facing financially exposed businesses.

We have adapted Cumulative Prospect Theory (CPT), which is an extension of Prospect Theory developed by Daniel Kahneman and Amos Tversky. In 2002, Daniel Kahneman was awarded the Nobel Memorial Prize in Economic Sciences for his pioneering work in behavioral economics. Prospect Theory fundamentally changed how economists understand decision-making under risk and uncertainty. CPT builds on this foundation, and after our adaption provides a more nuanced and realistic framework for evaluating the risk-and-reward trade-offs in a complex market like the NEM.

Understanding Risk-Adjusting

The fundamental premise of CPT is that it recognises that people value gains and losses differently:

• Losses loom larger than gains as a 1 million loss is felt more acutely than the satisfaction from a 1 million gain.
• The relationship is non-linear where the pain of a 2 million loss is valued more than a simple doubling of a painful 1 million loss. Conversely, a 2 million gain delivers less satisfaction than twice the satisfaction from a 1 million gain.
• This is captured through the concept of economic utility, which is central to CPT.

In the wholesale electricity market, these behavioural insights are crucial. Participants are not just striving to maximise expected profits, but also managing risk, reflecting their aversion to losses.

Calibrating Risk Aversion: Our Experience

One of the most challenging aspects of applying CPT in the wholesale electricity market is quantifying risk aversion behaviour. In practice, measuring risk aversion behaviour is as much an art, as it is a science.

The same individual may change their risk appetite from day to day, influenced by recent experiences, market news, or even personal mood. Within the same organisation, perspectives can vary widely. A Chief Financial Officer may prioritise capital preservation and be highly risk-averse, while a Trading Manager might be more comfortable with volatility in pursuit of higher returns. Meanwhile, a Risk Manager may focus on minimising tail risks and ensuring compliance with internal risk limits.

This diversity of views means that the “correct” risk aversion parameters are often a moving target, shaped by internal debates, external pressures, and evolving market conditions.

The way we quantify the market’s risk aversion is by:

• Simulating Spot Prices:
  We build detailed probabilistic spot price simulations, capturing a wide range of possible market outcomes. These simulations account for weather fluctuations reflecting different temperature outcomes, wind speeds and solar irradiance; and then we add many generation strategies, outages, transmission constraints, curtailment, potential input cost changes to natural gas or coal prices, planned retirements, new generation, etc. The final melting pot of simulations has a wide and varied set of possible outcomes, some previously seen and many unseen in the market’s history.
• Observing Forward Markets:
  We then analyse forward financial market trading where the traded prices reflect the collective perception of risk aversion and profit opportunity of buyers and sellers in the market; however, this is a moving target.
• Matching Economic Utilities:
  Buyers and sellers look at the market through a different lens given the spot price distribution is skewed, and the risk of holding a long position is not the same as holding a short one. By adjusting the parameters of the CPT mathematical functions to represent a buyer and seller, we can find the point where the economic utility of a seller matches that of a buyer, evidenced by trading activity. This means we can empirically measure risk aversion behaviour. Quite an elegant outcome.
• Using the “Do-Nothing” Scenario as Reference:
  All strategies are evaluated relative to a baseline “do-nothing” scenario where the participant takes no further hedging or trading action. This provides a clear reference point for measuring gains and losses in utility terms.

Applying Risk-Adjusting

With market calibrated risk aversion behaviour measured and an adapted CPT theory for the NEM, we can combine with our probabilistic spot forecasts to:

• Derive Optimal Hedging and Trading Strategies:
  By comparing risk-adjusted returns across regions and trading instruments, risk-adjusting guides us to where capital is best deployed. The optimal strategy is often complex as it not only pursues the trading strategy with the best use of capital, but measures the risk-adjusted reward and may include a counter position to protect the overall robustness of the portfolio. It is fascinating to observe this dynamic when managing trading portfolios.

Below is an example of a Portfolio's expected quarterly trading profit trading across the NEM using about $600,000 of working capital. The left side of the figure shows the expected profit distribution of the Existing portfolio, and then a group of small trades are tested to show the Hypothetical Impact. The end of quarter Risk Adjusted Average Gain improved from $264,000 to $292,000 which is a favourable movement of $28,000. This risk adjusted gain is after taking account of weighting the increased potential losses more heavily then the potential gains.

On the right of the figure below shows the impact of adding an additional trade to the Hypothetical trades that not only increased the expected profit, but reversed all the unfavourable hypothetical simulations. Such a trade does not always exist and depends upon both the portfolio and the market pricing levels at the time; however, these trades do occur reasonably often. In this particular case, the additional trade would improve the Risk Adjusted Average Gain by almost $80,000; a significant improvement with less risk.

• Value Renewable PPAs Accurately:
  For run-of-plant Power Purchase Agreements (PPAs) with wind, solar, or hybrid farms, risk adjusting provides a robust framework for capturing the unique risks faced by owners and off-takers. There are two key risks to consider:
  • Timing Risk: The value of generation depends heavily on when it occurs.
    Not all megawatt-hours are created equal. Producing 1 extra MW of power at midday, when solar output is high and spot prices are typically low, does not have the same value as producing 1 extra MW during the 6:00pm peak when demand is high, prices are higher and more prone to spike. This timing mismatch is a critical risk for PPA valuations and must be accounted for in any robust valuation or hedging strategy.
  • Quantity Risk: The actual amount of generation is uncertain and variable. High generation periods tend to coincide with lower spot prices, while low generation periods often align with higher spot prices. This negative correlation means that PPA off-takers are more likely to receive large volumes of energy when it is least valuable, and lower volumes when it is most valuable. This dynamic reduces the economic utility of the PPA.

By incorporating both timing and quantity PPA risks into the risk-adjusted framework that encompasses our probabilistic spot forecasts and calibrated risk aversion parameters, we can more accurately assess the risk-adjusted fair value of renewable PPAs and design complementary hedging strategies.

Furthermore, it is possible to measure the incremental impact of any supplementary trade or hedge on a portfolio, whether long-term or short-term, to determine its overall impact. This insight of the portfolio impact is crucial when assessing trading and hedging opportunities.

Why Risk-Adjusting Makes a Difference

The non-linear, asymmetric nature of risk-adjusting means that the optimal trading and hedging strategies are highly specific. Any deviation from the risk-adjusted optimal strategy can significantly impact the risk-adjusted value of the portfolio because the weighting across Regions and/or instruments, can become out of balance. This approach provides a more realistic and effective framework for players exposed to the spot and forward financial markets, enabling them to make decisions that truly reflect the opportunistic prize, risk preferences and the realities of the NEM.

To apply this risk-adjusting process into businesses we use our watt.if portfolio optimisation model that measures the impact of long-term transactions, or shorter term hedging and trading strategies. It has a wide range of applications to drive trading and hedging decision making.

If you are interested to have a demonstration of our short-term watt.if portfolio optimisation in action or wish to understand our long-term watt.if, just let us know by contacting me at carl.daley@zawee.com.au