Dynamic decision making in energy systems with storage and renewable energy sources

Abstract

We model an energy system with a storage device, a renewable energy source and with market access as a Markov decision process. We have identified four classes of pure policies (PFAs, CFAs, VFAs and lookaheads), each of which may work best depending on the characteristics of the system (volatility of prices, stationarity, accuracy of forecasts). We demonstrate that each of the four classes can work best on a particular instance of the problem. We describe the problem characteristics that bring out the best of each policy.