基于PCA与RBF的建筑能耗预测建模
A prediction model for energy consumption of building based on PCA and RBF

由于建筑能耗因子间存在非线性和高度冗余特性，传统预测方法很难消除数据之间冗余和捕捉非线性特征，导致预测精度较低. 为了提高建筑能耗预测精度，提出一种将主成分分析(principal component analysis，PCA)和径向基函数(radial basic function，RBF)神经网络相结合的建筑能耗预测方法(PCA-RBF). 利用PCA消除建筑能耗高维变量数据的相关性，并按累积贡献率提取主成分，将主成分作为RBF神经网络的输入进行训练学习. 通过PCA避免了模型过多的输入导致的训练耗时长及预测精度较低的不足. 通过将PCA-RBF模型方法应用于某办公建筑能耗的预测中，并与RBF神经网络及BP神经网络模型相比，实验结果表明PCA-RBF模型方法能有效提高建筑能耗预测精度.
There are highly redundant features in affecting factors of building energy consumption，and the traditional method has low predictive accuracy. In order to improve the accuracy of building energy consumption forecasting，a model method for energy consumption based on principal component analysis (PCA) and radial basic function (RBF) neural network is proposed，which combines the abilities of PCA to de-correlate the variables and reduce the dimensionality of the data with that of neural network to approximate any complex nonlinear function. The PCA-RBF model is applied to the energy consumption prediction for an office building，and the simulated results show that the PCA-RBF has better accuracy compared with RBF neural network model and BP neural network model，which is considered that the PCA-RBF is effective for building energy consumption prediction