Short-term PV generation forecasting based on weather type clustering and improved GPR model

A combined prediction method based on weather type index and Gaussian Process Regression(GPR) improved by linear weight reduction PSO algorithm was proposed to deal with the problem of forecasting of short-term PV generation. Firstly, the samples, according to the weather type index, was built, and then three types were classified based on daily mean temperature and humanity; the different GPR model, considering different highly correction factors, which measuring combined covariance functions and taking LinW-PSO to optimize the hyperparameters of model, were built; The corresponding computer code was programmed in MATLAB. Seen from simulation results of forecasting of the PV station in XinJiang, the given classification method and LinW-PSO-GPR combination predicting method enable the rate of convergence and the accuracy of results improved.