| The increasing use of traditional energy in social production and life has brought about environmental problems that can not be underestimated.Under the premise of advocating green sustainable development around the world,new energy power generation technologies,especially photovoltaic(PV)power generation technologies,plays an important role in the field of power energy.However,PV power has certain characteristics such as randomness,fluctuation and uncertainty,which may make the power grid unable to operate safely and stably.Therefore,the forecast of PV power is particularly important for the development of PV power generation.In this paper,the fluctuation characteristics of PV power,the identification of abnormal data of PV power,short-term PV power combination forecast and error analysis of PV power forecast are studied.Firstly,after studying the distribution characteristics of PV output power fluctuation,a probability distribution description method of PV power random component fluctuation characteristics based on the mixed t Location-Scale distribution model is proposed.The random components of PV power at different sampling intervals at home and abroad are described respectively,which provides a the basis for PV power to identify abnormal data.Secondly,the current recognition method for PV power abnormal data is complex and tedious,and the recognition effect is not ideal.In addition,the method has poor generality and can not effectively identify the random and fluctuation PV power data.As well as the poor universality of the method,it is unable to effectively identify the PV power data with strong randomness and fluctuation.According to the mixed t Location-Scale probability distribution model,a class 3 Sigma criterion model is proposed to identify the abnormal PV power data.This method can effectively eliminate abnormal data values in PV power.Then,in view of the limitations of each single forecasting method,the traditional combination forecasting methods,such as the average weight combination method and the fixed weight combination method,can not guarantee the minimum forecasting error at each moment.Aiming at the determination of the weight of the combined prediction method,a short-term PV power dynamic weighted combination forecasting method based on least squares method is proposed.Taking the PV power plant in Ashland area of the United States as an example,it is verified that the dynamic weighted combination forecasting method can effectively reduce the PV power forecasting error compared with the single forecasting method and the traditional combination forecasting method.Finally,when studying the distribution characteristics of PV power forecasting errors,firstanalyzed the different distribution characteristics of PV power forecasting errors with different forecasting methods,and then analyzed the distribution characteristics of PV power forecasting errors under different weather types.The single distribution model and the mixed distribution model are used to fit the probability density of the forecasting error. |
