Photovoltaic power output forecasting based on similar day analysis and sensitive pruning extreme learning machine

A prediction model (SDA-SP-ELM) based on the combination of similar day analysis (SDA) and sensitive pruning extreme learning machine (SP-ELM) is proposed to predict the hourly output power of PV plants for the problem of low accuracy of PV output prediction. Firstly, samples with similarities to the prediction date are selected as model training data by the Pearson coefficient method, avoiding the input of redundant information. Secondly, the optimal number of nodes in the hidden layer is determined from the use of a sensitive pruning extreme learning machine, which greatly improves the generalization ability of the model and avoids the problem of increasing the complexity of the model. And the simulation results show that the proposed model can accurately predict the hourly PV plant output power, and it is significantly better than SDA-ELM and similar day analysis back-propagation neural network (SDA-BPNN) in terms of accuracy and stability of PV output prediction, with a difference of about 1.2516 kW in MAE value and 85.99% in nRMSE error compared to SDA-ELM.