基于径向基神经网络的建筑物空调负荷预测模型

本文建立了一种基于径向基（RBF）神经网络的建筑物空调负荷预测模型。对广州市某办公楼在夏季不同月份的逐时冷负荷，分别用RBF神经网络模型和BP神经网络模型进行训练和预测计算，发现RBF神经网络模型预测的均方根误差ΔRMSE和平均相对误差ΔMRE都仅是BP神经网络方法的64％左右。仿真结果表明径向基（RBF）神经网络具有更高的预测精度及更好的泛化能力，是建筑物空调负荷预测的一种有效方法。

Building cooling load prediction model based on radial-based function neural network

On the basis of radial-based function (RBF) neural network, a kind of building cooling load prediction model is established. Back-propagation (BP) neural network model and RBF model are both used for the summer hourly cooling load prediction of an office building in Guangzhou. It is found that the mean square root error and the mean relative error from RBF model are about 64 percent of those from BP network model. The simulation results show that the RBF method can achieve better accuracy and generalization, and is effective for building cooling load prediction.