Calibration of a Stochastic Price Model for American Electricity Markets

Abstract

This thesis discusses models for electricity spot prices from the Midwestern American and Manitoba market. The models are based on experiences in European markets and rely on a superposition model with several jump components. The methodology of Bayesian Inference solved with a Markov chain Monte Carlo algorithm has been applied to find estimators for the processes of the model. The specific Markov chain Monte Carlo algorithm applied a Random Walk Metropolis combined with a Gibbs sampler. The different estimators of the models are evaluated with the posterior predictive value and simulations of the electricity spot prices. We have modified this methodology to apply to the US market.