Short-term PV Power Prediction Based on TPE With TCN-BiGRU-Attention

Short-term power prediction is of great significance to improve the stability of power system, for the short-term power prediction of photovoltaic power plant in the meteorological conditions, time features and power cyclic features of the integrated extraction and fusion of the problem of insufficient, put forward a combination of prediction method based on the time convolutional network (TCN), bidirectional gated recurrent unit (BiGRU) and Attention mechanism (Attention). Firstly, multi-scale features are fused with meteorological features, temporal features and hidden periodic features, secondly, temporal and spatial features are extracted using Time Convolution Network and Bi-directional Gated Recurrent Unit respectively, and then features are fused using Attention model, and finally TPE optimization model based on Bayesian Networks is used to optimize the network parameters, and finally the prediction results are obtained. The PV power data from a PV power plant in Ningxia region is used for validation, and the results show that the proposed method has higher computational efficiency and higher prediction accuracy compared with the methods that have been proposed so far.