In order to ensure the safety and stability of the power grid and the efficient operation of the power market, grid dispatchers and power market participants have put forward higher requirements for the accuracy of power load forecasting. However, the distributed power and intermittent loads increase the difficulty of predicting loads accurately. In order to solve the problem that the current load forecasting method cannot simultaneously model the change law of the load itself and its influencing factors, load forecasting method based on long short-term memory(LSTM) is proposed. LSTM is used to construct recurrent neural network(RNN), and comprehensively historical load and various load influence factors are considered to establish load forecasting model. The method utilizes the feature extraction ability of the neural network and the memory ability of the LSTM to identify the internal variation law of the load and the nonlinear influence of various influencing factors on the load in a longer historical time range. The actual load datas are used to verify the prediction performance of different historical time windows and different network architectures. Meanwhile, compare with other load prediction algorithms. Experimental results show that the model can improve the accuracy of load forecasting.