改进用影响因素最为输入变量的BP神经网络进行短期负荷预测的传统方法,讨论基于BP神经网络的两种天然气负荷预测方法,分析其稳定性和误差分布及平均误差大小,最后得出用前4年的数据作为输入变量,后一年的数据作为目标矢量的网络训练方法,误差分布均匀且平均误差非常小,可达到天然气负荷预测的要求。 The traditional method of short-term load forecasting based on BP neural network with the most input variables is improved. Two kinds of natural gas load forecasting methods based on BP neural network are discussed. The stability, error distribution and average error of BP neural network are discussed. Finally, Year data as input variables and the data of the following year as the network training method of target vector, the error distribution is uniform and the average error is very small, which can meet the requirements of natural gas load forecasting.