Fourth-type region photovoltaic power prediction method based on prediction point data recombination-WaveNet

The invention discloses a power prediction method for photovoltaic power stations in four types of regions based on prediction point data recombination-WaveNet. The method comprises the following steps: preparing historical data information of the photovoltaic power stations; data preprocessing and feature selection; data recombination; a prediction point classification self-recognition model is constructed through the data recombination blocks and an improved WaveNet model; the method comprises the following steps: preprocessing historical data, screening original data features by using a maximum information coefficient and a Pearson correlation coefficient, then re-classifying the data by using a data recombination module to obtain a data set of a corresponding category prediction model (an improved WaveNet model), then constructing an improved WaveNet model consisting of two WaveNet models and Self-Attention, and carrying out prediction on the improved WaveNet model by using the improved WaveNet model and