A Decision Support Tool for Energy Storage Optimization

We examine the problem faced by owners of gas storage contracts of how to manage the injection/withdrawal schedule of natural gas, given historical price behavior and a predictive model of future prices. We describe a hybrid price prediction model that combines a support vector machine technique with a stochastic model for short and long-term gas prices, respectively. We present a theoretical result regarding the complexity of the scheduling problem and propose a formulation for an integer program given capacity, injectability and deliverability constraints. We also outline how this was incorporated in a decision support tool that is used by energy traders at a gas utility company.