Two-Stage Robust Optimization for Microgrid Incorporating Renewable Energy Uncertainties via Bayesian Inference with Gaussian Processes

This study contributes a data-driven two-stage robust optimization framework to account for the uncertainties in renewable energy sources within microgrid systems. The framework integrates Bayesian inference with Gaussian Process regression (GPR) analysis to evaluate and model the output uncertainties of photovoltaic (PV) and Wind turbine (WT) power sources. Furthermore, the correlation between these sources is established at various uncertainty levels. The two-stage optimization approach incorporates the day-ahead dispatch stage and real-time adjusting under economic factors in the second stage. So, the framework controls day-ahead dispatch strategies to optimize microgrids in the prospect of varying PV-WT output and electricity demand. Subsequently, it involves real-time optimization under the worst-case scenario, enhancing the microgrid's economic viability and operational efficiency. Finally, a column-and-constraint generation (C&CG) method is employed to address the robust optimization model by incorporating different levels of uncertainty. The proposed method is established on an IEEE 33-bus distribution network, ensuring its applicability and reliability.