Stochastic Model Predictive Control for Solar Homes with Battery Energy Storage

Predictive control of solar homes with battery energy storage systems requires forecasts of key inputs including PV generation; residential load; and, under some tariff structures, wholesale electricity market price. Stochastic model predictive control (SMPC) is a control strategy that incorporates uncertainty in forecasts to optimise system operation under a number of possible forecast scenarios. This study implements SMPC using a constrained approach where optimisation is performed simultaneously across all forecast scenarios rather than simply averaging all forecast scenarios. It also proposes a flexible scenario generation method that models the forecast errors in a Markov chain Monte Carlo simulation which allows for any forecast to be used. This study quantifies the benefits of SMPC over deterministic MPC for six representative solar homes in Australia found using a clustering-based methodology. It finds that under all studied tariff structures, monthly electricity bill savings were in the range of $3-$5 with only a small increase in computation time making it worthwhile for implementation within home energy management systems or virtual power plants.