ULTRA-SHORT-TERM PROBABILITY PREDICTION OF PV OUTPUT BASED ON MULTI-WEATHER-TYPE-WEIGHTED AND IMPROVED GAUSSIAN MIXTURE MODEL
This paper proposes a PV output ultra-short term probability prediction method based on multi-weather-type-weighted and improved gaussian mixture model,which can realize ultra-short-term probability prediction of photovoltaic output 15 minutes ahead.Firstly,the historical data is divided into several meteorological types according to the meteorological characteristics.And then,the improved Gaussian mixture model is constructed to obtain the probability distribution of each meteorological type.Thirdly,the membership function is constructed to quantify the similarity degree of the meteorological characteristics to each meteorological type with the time to be predicted.Finally,the probability distribution of each meteorological type is weighted according to the membership.Taking the actual PV power station data as an example,the results show that compared with methods with a single meteorological type,the weighted model of multiple meteorological types has an average reduction of 16.87%in MAPE,an average increase of 10.45%in ACD,and an average decrease of 2.49%in AW.
photovoltaic power outputprobabilistic forecastultra-short-termmulti-weather-type-weightedimproved Gaussian mixture model
NETL
NSTL
万方数据
