Sensitivity to forecast errors in energy storage arbitrage for residential consumers

With the massive deployment of distributed energy resources, there has been an increase in the number of end consumers that own photovoltaic panels and storage systems. The optimal use of such storage when facing Time of Use (ToU) prices is directly related to the quality of the load and generation forecasts as well as the algorithm that controls the battery. The sensitivity of such control to different forecast techniques is studied in this paper. It is shown that good and bad forecasts can result in losses in, particularly bad days. Nevertheless, it is observed that performing Model Predictive Control (MPC) with a simple forecast that is representative of the pasts can be profitable under different price and battery scenarios. We observe that performing MPC at a faster sampling time with a receding optimization horizon makes arbitrage less sensitive to uncertainties in forecasting. We use real data from Pecan Street and ToU price levels with different buying and selling price for the numerical experiments.