The Arctic blast has shown how challenging is to hedge renewable merchant risk
ERCOT (i.e., Texas) is home to many wind projects under a merchant model, and some of them had to hedge their energy price to secure financing. However, it is far more complicated to hedge wind power generation than dispatchable energy. For instance, the Arctic blast that threw Texas in deep freeze led some wind producers to massive losses because of their hedging program. Without the hedges, they would have triple their cumulated 12-month margin.
Two reasons make hedging wind power generation challenging in ERCOT. One is the “covariance risk” associated with wind, i.e., a negative relationship between its generation and price. This risk is present in many markets where intermittent renewable energy represents a significant share of the generating capacity. Second, ERCOT is an energy-only market, i.e., there is no organized market mechanism to secure long-term capacity. Therefore, the value of capacity shows up in energy price spikes — the more often the price spikes, the higher the incentive to build generating capacity. It is beyond the scope of this article to discuss the pros and cons related to this approach.
We use the ERCOT West zone and the recent events in ERCOT to illustrate the challenge of hedging wind power generation.
What is the covariance risk?
The following graphic illustrates the spot price distribution as a function of wind power generation in ERCOT West. We have juxtaposed a hypothetical hedge consisting of P80 (i.e., 80% chance on average that our wind power generation exceeds that volume) baseload sold at $35/MWh in the ERCOT West hub. It is possible to sell into other hubs that are more liquid (e.g., North or South) but at the expense of a basis risk that further exacerbates the overall risk exposure.
We can easily see on average that prices decrease as wind power generation increases.
As illustrated here, this hedging strategy represents more a speculative position than a synthetic hedge as we bet on limited price spikes when wind power generation is below P80. It is not always the case in ERCOT because wind power generation represents a considerable share of the generating capacity. Therefore, when there is little wind, price spikes are more likely to occur. The Arctic blast is a clear illustration of this phenomenon, as depicted in the following graphic.
What has been the financial impact of the Arctic blast on the assumed hedge?
Our hypothetical hedge combined with the Arctic blast has led to massive losses bringing the average margin down to $-8.37/MWh for the 12-month horizon ended by the blast. The average margin was $21.73/MWh just before the blast. The following graphic summarizes the impact of the Arctic blast on our supposed hedge.
There are three observations to note:
1. Even before the blast, our hedge did not perform well as it has closed only 1/3 of the revenue gap. We could have hedged more than P80, but that will have increased the covariance risk. For instance, if we had sold P20 baseload, the hedge would have closed 100% of the gap just before the blast. However, the blast would have created a negative margin of $-105/MWh.
2. Day-ahead prices are usually more favorable to intermittent energy as they do not directly include the cost of balancing the system. The day-ahead/real-time arbitrage consists of offering P50 in the day-ahead market, and we settle the imbalances in the real-time market. We can improve this simple strategy by varying the day-ahead offer as a function of the anticipated prices (e.g., we can use deep learning techniques to determine the optimal day-ahead offers).
3. The covariance risk shows up as a substantial discount for wind power generation revenue relative to baseload price. In our example, the discount is at least 30% in a normal price regime and up to 70% when we include extreme price spikes.
How should we hedge then?
The very first step is to determine the discount associated with the covariance risk and what will drive this discount in the future. Unfortunately, a task that requires heavy analytics as the intricacies related to intermittent renewable energy are complex to assess, even more in a nodal market approach.
We can use the assessed discount to figure out the amount of capital we are willing to invest into a static hedging program (a kind of a “put option”) to protect us against low prices. We can also use this capital to set-up a commercial function that will dynamically hedge our exposure like any sophisticated trading house will do.
In some cases, the anticipated discount may trigger an investment in energy storage that will (partially) offset the renewable merchant risk.