Optimal charging / discharging strategies for batteries in smart energy grids

The intermittent nature of the ever increasing renewable electricity production demands an ever increasing flexibility in the electrical grid. With the price of the most flexible device dropping rapidly, batteries are close to large-scale implementation. In this thesis, we start by formulating a Stochastic Dynamic Programming (SDP) algorithm to control a stand-alone battery that is acting as a price taker in an hour-ahead electricity market. By using a simulation with realistic data we find that our algorithm significantly outperforms two lower bound methods and gets within 92% of the theoretical upper bound. We then extend the SDP algorithm to minimize the costs of a household with a home battery. Through simulation we show that this algorithm still outperforms both lower bound methods and this time gets within 93% of the theoretical upper bound. By applying this algorithm to a situation where the battery affects the price without being aware of it, we show the robustness of our algorithm. Finally, by adapting the algorithm to include a scenario where the battery is aware of its influence on the price, we show that the choice of price function significantly influences the actions of the battery and with it the effectiveness of its peak shaving.