Day-ahead energy management of grid-connected multi-source complementary microgrid combined with PV forecasting

This paper proposes a multi-source microgrid day-ahead energy management considering the uncertainty of renewable energy output. First of all, to alleviate the impact of the uncertainty of photovoltaic output on the management of microgrids, an improved deep learning model (QPSO-LSTM) based on the quantum particle swarm optimization algorithm long short-term memory network was developed to predict photovoltaic power output. Secondly, according to the forecast results, a day-ahead load optimization dispatching model for grid-connected microgrid based on non-dominated sequencing genetic algorithm and elite strategy (NSGA-II) algorithm is proposed. The microgrid includes photovoltaic systems, energy storage systems and diesel generators. We use data sets collected in the real world to test the established models. The final results show that the performance of the QPSO-LSTM model is better than the PSO-LSTM model and the LSTM model. Actual data have also verified the feasibility and efficiency of the energy management model.