A TPA-TCN Prediction Model Applied In Photovoltaic Power Generation Field

Abstract

To solve the problem of large fluctuation and instability of photovoltaic power generation, a deep learning prediction model (TPA-TCN) based on temporal pattern attention mechanism (TPA) and temporal convolutional network (TCN) is proposed, and then applied to photovoltaic power generation. First of all, the k-means clustering algorithm is used to cluster historical data to obtain three typical weather types, and the model is trained by dividing test sets according to the clustering results. After TPA is introduced into the TCN model, which can capture the influence of each variable on the predicted series of the model, help the model pay better attention to the key features in the time series, improve the model’s ability to understand the data, and thus efficiently and accurately predict the short-term photovoltaic power. Combined with the measured data, the experiment results show that the TPA-TCN model has good generalization ability and high precision in different weather types.