A Machine Learning-Assisted Decomposition Method for Efficient Unit Commitment Solving in ESS System

In this study, we propose a novel decomposition method that decomposes the model by predicting the energy values of storage units at each time step. We also introduce a feasibility detection method to ensure problem feasibility and enable warm-starting. Experimental results demonstrate that our algorithm outperforms traditional methods in rapidly exploring the feasible region and significantly accelerating solution times, thereby enhancing optimization efficiency for energy system scheduling.