Abstract:Based on the standard back propagation neural network, this paper proposes a method combining principal component analysis and improved error back propagation neural network to predict the long-term power load of the power grid. Firstly, principal component analysis is used to extract the factors affecting the power load, effectively reduce the dimension of the data sample, eliminate the redundancy and linear information of the data, and retain the main components as the input data of the model. The momentum term and the steepness factor are then introduced in the back propagation of the standard neural network. The combination of the two methods effectively solves the problem of slow network convergence and easy to fall into local minimum. Applying this method to medium and long-term electric load forecasting in Jiyuan City, the experimental results show that the method based on principal component analysis combined with improved back-propagation neural network is more common than the standard back-propagation neural network, multivariate-based time series network and time series network have higher computational efficiency and prediction accuracy, which proves that the proposed prediction model is effective in power load forecasting.