Risk Management in Wholesale Electricity Markets: A Signal Processing Approach

Medium-term load forecasting is a critical tool utilized by power market system operators for system planning and load serving entities for procuring power supply contracts. In this paper, we employ techniques from the field of signal processing to model seasonal load in New York wholesale electricity market (NYISO). We begin by using mathematical filtration techniques to smooth raw NYISO load and price data that spans 2006 to 2018. The resulting filtered data series express load and prices as percentage deviations from their annual moving averages. Next, we develop a nonlinear, almost-periodic model that uses the 90-day moving average of the filtered price data and purely timedependent periodic terms to model the cyclicality of the 90-day moving average of the filtered load. In a final modeling step, we model the noisy model residuals using autoregressive (AR) regressions. We conclude by performing out-of-sample forecasts of our model using test data from 2019 and 2020. The forecasting results reveal that our model can predict seasonal load with a high degree of accuracy and potentially assist market participants with their medium-term planning objectives.