The Effect of Changing Temperature Extremes on Renewable Energy Production Risk

As is often noted, gradual shifts in mean temperatures due to climate change cause dramatic shifts in the frequency of temperature extremes. While the relationship between extreme temperatures and power grid demand is relatively straight-forward, the effect of changing temperature distributions on renewable energy generation (and the associated risks to profitability or grid reliability) is less studied.

Like all mechanical systems, wind turbines and photovoltaic systems are designed to operate optimally within certain ranges of weather conditions. Wind turbines compliant with IEC 61400 standards, for example, are designed to have a “normal” and “extreme” operating ranges of -10C to 40C and -20C to 50C while solar inverters in PV systems degrade above 45C.

In this talk, we use stochastic climate simulations to analyze the effect of climate change on extreme temperature distributions and the downstream ramifications to wind and solar production. Taking a sample of 1000 simulation scenarios, we first examine the changing raw frequencies of temperature extremes with respect to the thresholds above. We then use the Pickands-Balkema-De Haan theorem to fit time-varying generalized Pareto distributions to these raw tail frequencies and obtain continuous and analytic extreme temperature distributions. We multiply these distributions with mechanical deterioration profiles and modeled electricity price distributions to quantify expected losses due to extreme temperature events in both energy and dollar terms.

Finally, we note that weather extremes and electricity prices cannot be considered independently, and that to obtain accurate estimates of expected losses one must integrate over the joint distribution of weather extremes and prices. That is, extreme heat and extreme cold events are highly correlated with periods of extreme grid stress–which are in turn associated with high energy prices and reliability risks. We illustrate this phenomenon by incorporating a simple electric price model to consider joint distribution losses.