The present paper summarizes an article titled “An alternative optimal strategy for stochastic model predictive control of a residential battery energy management system with solar photovoltaic” by Meer, Wang, and Munkhammar. The article describes a study conducted in 2021, which proposed an alternative scenario-based stochastic model predictive control (MPC) algorithm. This algorithm includes an innovative constraint that improves the outcomes of all scenarios during the first prediction step. The new approach was tested on a villa with a battery and a photovoltaic (PV) in Borås, Sweden.
The researchers used the data on electricity production, electricity prices, and weather forecasts to create an innovative MPC. The data was analyzed using gradient boosted regression trees (GBRTs), quantile regression (QR), multivariate scenario generation, and stochastic model predictive control. The results of the performance of MPC were validated by comparing the results of the scenarios with the actual outcomes using relative percentage differences. The probabilistic forecasts were ranked using a continuously ranked probability score, while the multivariate scenarios were validated using a pre rank framework.
The results of the research revealed that while the probabilistic models were not perfectly calibrated, the calibration was satisfactory to produce reliable results. The calibration was improved by postprocessing the GBRT forecasts using QR outcomes. In general, the results provided sufficient evidence that finding a strategy that is optimal across all scenario forecasts helps to operate the system optimally. Additionally, the model was beneficial for decreasing the variance of the power exchange with the grid. However, the results were limited by the length of the test period and the size of the battery. The authors of the article suggest that future research should include heating, ventilating, and air conditioning systems.