Long sequence prediction method for photovoltaic power generation based on CNN-LSTM-KAN modeling

In recent years with the development of cluster photovoltaic power plants, a large amount of photovoltaic power generation is integrated into the power grid affecting the frequency and voltage stability of the power grid, effective prediction of photovoltaic power generation can help the grid operator to grasp the output of photovoltaic power generation in advance, to maintain the load balance of the grid, the current existing prediction methods have a high degree of time-space complexity as well as algorithmic complexity and other problems. Therefore, this paper proposes an improved LSTM-CNN-KAN model for long sequence PV power prediction based on the deep learning model LSTM. The method extracts spatial features, reduces feature dimensions, and captures multi-scale features in the time series through the convolutional neural network CNN, uses LSTM to effectively deal with long-term dependencies in the time series data and utilizes the KAN neural network to select the focus of attention through an adaptive approach to improve the prediction accuracy. Through simulation experiments using publicly available datasets and comparisons with related prediction methods in terms of RMSE and MAE, the experimental results show that the method proposed in this paper has high prediction accuracy.