【摘要】 研究了RBF神经网络在城市燃气日负荷预测中的应用及输入变量的选择问题,提出了基于逐步回归的输入变量选取方法。对实例进行了预测,对不同输入变量方案进行了对比分析。以逐步回归选取的输入变量为基础,增加日期类型、前一天平均气温两项数据作为输入变量,完全满足神经网络用于城市燃气日负荷预测精度的要求,且合理可行。更多还原

【Abstract】 The application of RBF neural network to daily city gas load forecast and the selection of input variables are studied.The selection method of input variables based on stepwise regression is put forward.A case forecast is carried out,and a comparison among schemes with different input variables is made.Based on the input variables selected by stepwise regression,adding date type and average air temperature of previous day as input variables can meet the accuracy requirement of neural network for daily city gas load forecast,and it is reasonable and feasible.更多还原