Adaptive Support Vector Machine Based for Optimal Sizing of Renewable Generation and Battery Storage for Standalone Households

This paper proposes a novel adaptive support vector machine based demand-side management strategy for the optimal sizing of battery storage (BS), wind turbine (WT), and photovoltaic (PV) in a standalone household. The proposed strategy uses day-ahead forecasts of wind speed and solar irradiance and the battery state of charge level to reduce electricity consumption when renewable generation is expected to be low. Additionally, it minimizes the levelized cost of electricity (LCOE) by minimizing the system's net present cost (NPC) over the project lifespan. The NPC includes replacement present value, capital present value, salvage value of the system components, and maintenance present value. To evaluate the efficiency of the proposed method, a typical case study from South Australia with three different configurations of BS, PV, and WT is considered. The results demonstrate that a system including BS, PV, and WT is the best configuration that achieves the lowest LCOE and eliminates load curtailment. Furthermore, the proposed method enhances system reliability by eliminating load curtailment (from 6 days to zero) and extends battery lifetime by 3 times through intelligent battery management.