To address the issue of uncertain load conditions in photovoltaic power generation systems, this paper proposes an adaptive robust optimization scheduling model and microgrid operation control strategy. The method first uses GPR adaptive generation to obtain the mean and variance of the day-ahead power output forecast values, and introduces key data features from the forecast stage to reduce the error in the robust optimization uncertainty set. Subsequently, the proposed multi-state ant colony-bacterial foraging algorithm can achieve maximum power point tracking (MPPT) for the photovoltaic system under PSC conditions. In the case study analysis, the total operating cost of the proposed model is 309,200 yuan lower than that of the classical two-stage robust optimization model, validating that the proposed adaptive robust optimization model better balances the operational economic advantages during microgrid optimization scheduling. Additionally, the maximum power value tracked by the algorithm under the given conditions is 732.6 W, with an error of only 0.01 W compared to the actual maximum power. This verifies that the proposed algorithm has the advantages of fast optimization speed and very small system steady-state oscillations.
