Abstract: For problem of incorrect modeling method which takes environmental temperature instead of photovoltaic array temperature in existing simulation of solar photovoltaic arrays, the paper pointed out environmental temperature and actual working temperature of photovoltaic array should be distinguished in modelling of photovoltaic arrays. It analyzed relationship between temperature of photovoltaic array and environmental temperature as well as output power, and proposed a temperature prediction method of photovoltaic array based on BP neural network. At last, it compared predicting result with the measured one and got following conclusions: the method can predict temperature of photovoltaic array effectively and get better effect by use of data of the last day and the last three days. According to the conclusion, temperature of photovoltaic array of the day can be predicted by use of data of the last day in actual application.