Uncertainty quantification in energy management of microgrid: Extended state observer surrogate model approach

With the wide application of renewable energy sources in microgrids, the uncertainty of photovoltaic power has become a key factor affecting the stability and operational efficiency of microgrids. To address this challenge, this paper proposes extended state observer (ESO) surrogate model for quantifying photovoltaic uncertainty and optimizing real-time dispatch strategies for microgrid. First, the mechanism model is constructed to explore the dynamic relationship between solar irradiance and power. Then, the ESO surrogate model is designed to successfully estimate the trend of photovoltaic power by capturing the irradiance fluctuation uncertainty in real-time. The weighted reconstruction method is further introduced to fuse the results of the mechanism model with those of the surrogate model, which significantly reduces the prediction error and provides smoother power prediction, especially under sudden or rapidly changing solar irradiance conditions. The experimental results show that the ESO surrogate model achieves high accuracy in one-step prediction and is capable of adapting to different weather conditions and regions, but the prediction error increases significantly when irradiance fluctuates rapidly. In contrast, the weighted reconstruction model demonstrates enhanced robustness in handling large uncertainties, with improved prediction performance indexes. Compared to other uncertainty modeling methods, the weighted reconstruction ESO model results in the lowest operational costs. This paper provides an innovative solution for the energy management of microgrids and an effective technical path to cope with complex uncertainties. The successful application of this method demonstrates its great potential in enhancing the reliability and economy of microgrids.