Short-term Output Prediction Of Photovoltaic Power Generation Based On Weather Clustering And Improved Neural Network

Photovoltaic power generation technology directly converts light energy into electric energy.Renewable and pollution-free are its remarkable advantages.With the rapid development of photovoltaic power generation in China,the stability and reliability of grid-connected photovoltaic power generation has become a key issue in the process of the development of independent photovoltaic power generation to grid-connected system.Because of its intermittent,volatility,randomness and other characteristics,photovoltaic output power will have a negative impact on grid-connected,and brings great challenges to the scheduling work.For this reason,accurate short-term prediction of photovoltaic power is the key to ensure stable and reliable operation of the grid in the process of photovoltaic grid-connected.Based on weather clustering and improved neural network,the short-term output forecasting method of PV power generation is studied in this paper.Considering that some of the data used for prediction may be missing,a two-dimensional sequential filling method is used to complete missing data,and subsequent modeling is carried out on the basis of complete and effective data.Firstly,the influencing factors of PV output are analyzed in detail.Based on this,improved Kohonen model and S-Kohonen model are used to realize weather clustering and weather type identification of the day to be predicted.At the same time,the optimal similarity theory is used to select the similar days of the day to be predicted,and the results are compared.Then,two kinds of improved neural network models are established: RBF neural network optimized by improved thought evolutionary algorithm and generalized regression neural network optimized by Drosophila species.Then,the forecast of PV power output is completed by using the data of corresponding historical days and the meteorological data of the day to be predicted as the input of the model.Finally,the above models are simulated in Matlab,and the prediction results are compared and analyzed.The results show that the forecasting accuracy is effectively improved by the above methods,which provides a feasible way to achieve accurate short-term output forecasting of PV power generation.