A study is conducted on the power prediction method of distributed photovoltaic(PV)power generation under differ-ent weather conditions,and a power prediction method for distributed PV stations based on output characteristics and long and short term memory(LSTM)network is proposed.Firstly,the basic principle,grid-connection structure and power characteristics of PV pow-er generation in optical storage power generation system are introduced.On this basis,the output characteristics of distributed PV ar-rays are analyzed.Then,LSTM's advantages of state characterization and feature extraction in time series are utilized,and feature ex-traction is carried out for the feature values before and after the optical power time point.Finally,based on the output characteristics of distributed PV array and LSTM,a new power prediction model of PV power generation is constructed,which can realize the accurate power prediction of PV power generation under different weather conditions.Based on the actual parameters of a PV power generation system in Baoding City,the effectiveness of the proposed model is verified by simulation experiments.The results show that the pro-posed power prediction method of distributed PV stations can accurately predict the power of distributed PV stations under various weather conditions,and the root-mean-square error and mean absolute error are the lowest on sunny days,which are 4.35％and 5.74％respectively.
维普
万方数据