Short-term prediction of photovoltaic power based on multi-parameter predictability

With the rapid development of photovoltaic power generation, its volatility has brought great challenges to the safe and stable operation of the power system. In order to efficiently absorb photovoltaic power generation, the power system needs to accurately understand the volatility characteristics of photovoltaic power generation. This paper proposes a prediction method based on the standard deviation and volatility variation of photovoltaic power generation power. By analysing the coupling time series characteristics between different parameters, a joint prediction study of multi-parameter sets is carried out. The experimental results show that through the analysis of serial autocorrelation, it can be confirmed that the multi-parameter set has strong short-term predictability. Furthermore, the multi-task 1dCNN model proposed in this paper performs better in the prediction accuracy of each parameter than the single-task ANN, single-task LSTM and single-task 1dCNN models. This model transforms the multi-parameter prediction problem into a multi-task prediction mode, combines the local receptive field technology of 1dCNN for multi-channel attributes, and reasonably uses the shared module and the task module to complete the extraction of temporal correlation information and differentiated modelling between parameters, which effectively improves the prediction accuracy.