A short-term power prediction of photovoltaic based on SSA-DBN

A single prediction model is not a sufficient solution to the problems arising from the stochasticity and high volatility of PV power output. To improve prediction accuracy and mitigate the negative impact on the grid, this study proposes a short-term PV power prediction technique based on SSA-DBN. Firstly, we performed Pearson's method correlation analysis and feature screening for each meteorological feature quantity, then we used SSA to optimise the parameter settings of the DBN, and selected the optimal hyperparameters to give full play to the model performance, so as to overcome the limitations of the traditional DBN model, and ultimately compared it with a single Long-Short-Term Memory (LSTM) and Deep Belief Network (DBN) model. The experimental findings demonstrate that the methodology proposed in this paper exhibits superior predictive precision.