Abstract: To improve the prediction accuracy of PV plant’s power output, a hybrid grey relational analysis（GRA） and long-short term memory（LSTM）neural network model was established in this study. Firstly, main meteorological factors affecting the photovoltaic output were identified and the similar days of the days to be predicted were determined by GRA method. Next, BP neural network model and LSTM neural network model were trained by using the meteorological parameters and actual power generation of similar days. Finally,the intelligent prediction model based on GRA method was developed, which was applied in the photovoltaic power station in Yunnan.Compared with traditional single prediction model and the combined model of BP and GRA, the accuracy of LSTM prediction model considering similar days is significantly improved, which meets the relevant requirements and owns good application prospect.