ULTRA-SHORT-TERM PV POWER DYNAMIC PREDICTION METHOD BASED ON SFLA AND MSISSA-ANFIS
PV power output is characterized by volatility and randomness,and accurate power prediction has important application value for safe grid operation in the process of achieving grid connection.In this paper,an intra-day dynamic prediction model of ultra-short-term PV power based on SFLA,MSISSA and ANFIS is proposed.Firstly,MSISSA is used to optimize the ANFIS model for its drawback of being influenced by the membership function,and the power prediction model based on SFLA and MSISSA-ANFIS is constructed by combining the SFLA method of selecting similar days.Then a feature vector is constructed to predict the PV power for the next 4 h based on the set of power,meteorological features and similar days with high correlation.Finally,the real-time updated future meteorological data obtained from small weather stations are stored in the database and predicted every 15 min to realize the intra-day dynamic prediction of PV power.The results show that the proposed method improves the accuracy of ultra-short-term PV prediction.
photovoltaic power predictiontime seriesadaptive neuro-fuzzy inference systemalgorithm optimizationsimilar day selection
万方数据
维普
