Research on PV Power Forecasting Based on Wavelet Decomposition and Temporal Convolutional Networks

Abstract

Photovoltaic power generation generally exhibits periodicity and volatility. Large-scale integration of photovoltaic power generation into the grid will bring challenges to the stability and security of the grid system, and inaccurate output power prediction will bring great impact to the grid. Therefore, accurate photovoltaic power generation prediction is very important for power system dispatch. This paper proposes a temporal convolutional networks model based on wavelet decomposition to predict short-term photovoltaic power generation. First, in order to extract the time-frequency information of the input feature, the input feature is decomposed into several component sequences using wavelet decomposition. Then, a case study was carried out using photovoltaic power generation data from a certain region in South China, and the feasibility of the temporal convolutional networks model based on wavelet decomposition proposed in this paper was tested. The research results show that the temporal convolutional networks after wavelet decomposition has slightly higher prediction accuracy than the long short-term memory networks, but the operating efficiency of this method is greatly improved.