Supply curves in electricity markets: a framework for dynamic modeling and Monte Carlo forecasting

Official URL: https://dx.doi.org/10.1109/TPWRS.2022.3208765

We introduce a new time series model for supply curves in a short-term electricity market. The model accounts for the contribution of different resources to the aggregate supply curve, as well as the impact of external factors on the price and amount of supplies provided by each of those resources. We equip the proposed model with a unified Monte Carlo methodology for tracking the latent variables, forecasting, and hyperparameter estimation. Specifically, we present a sequential Markov chain Monte Carlo (S-MCMC) algorithm for tracking the latent variables of the model, which in turn enables forecasting day-ahead supply curves. We present two stochastic variants of the expectation-maximization (EM) algorithm for estimating the hyperparameters of the proposed model. Both variants of EM employ S-MCMC in their expectation steps. We apply the proposed framework to the Turkish electricity market and show its performance on a real dataset from that market.