Short-term photovoltaic power prediction based on coyote algorithm optimized long-short-term memory network

Enhancing the accuracy of PV power prediction is crucial for guaranteeing secure scheduling and steady power system operation. This research proposes a coyote algorithm (COA) to optimize the prediction model of the long-short-term memory network (LSTM). Taking into full consideration of the five factors constraining the output power of PV, and taking PV power generation as the research object, the power generation efficiency under different weather is analyzed, and COA is used to optimize the parameters of the LSTM fully-connected layer, and establish a COA-LSTM combination model to predict the PV power, which has a better convergence speed and solving efficiency, and it can also avoid the local optimal solution effectively. Finally, based on the real-time data of a photovoltaic power station in Xinjiang, simulation is carried out, and the experimental results show that the COA-LSTM is more accurate in predicting the photovoltaic power than the LSTM.