With the continuous development of power system, power system load Load forecasting comes into play increasingly important role. Compared with traditional prediction methods, machine learning shows better performance in dealing with the temporal and nonlinear characteristics of load, but the traditional LSTM cannot make past data and future data more closely linked. To solve this problem, In this paper, a hybrid model short-term load prediction method based on a combination of CNN, BiLSTM and attention mechanism to reduce memory loss and improve prediction accuracy. Finally, taking the load data in a certain area of Australia as an example, the method is compared with other commonly used models. It can be found from the experimental results that this method has high prediction accuracy.