Evolving Radial Basis Function Neural Networks for One-Day-Ahead Hourly Forecasting of PV Power Generation

摘要: This paper proposes a novel method to predict one-day-ahead hourly photovoltaic (PV) power generation. The proposed comprises three stages: data classification, training and forecasting. In the first stage, fuzzy k-means algorithm is used classify historical for daily PV generation into various weather types. second five models are established, according verbal forecast of Taiwan Central Weather Bureau (TCWB). Each model constructed using radial basis function neural network (RBFNN), which parameters each RBFNN, including position RBF centers, width RBFs weights between hidden output layers, optimized harmony search (HSA). To select an adequate forecasting from trained models, inference in stage. approach tested on practical system. results show that provides better than existing methods over one-year testing data.