论文部分内容阅读
建筑能耗影响因素复杂,研究新的能耗预测方法可简化预测过程,提高预测精度。首先对一栋高校建筑的能耗样本进行主成分分析(PCA),去除信息冗余,消除输入变量之间的相关性。把经过PCA提取的主成分作为Elman神经网络的输入,隐含层和输入层均采用tansig函数,在训练过程中不断对权值和偏差进行修正,最终建立基于PCA-Elman的建筑能耗预测模型。采用测试样本对模型精度进行验证,实例表明,基于PCA-Elman的建筑能耗预测模型相对误差为5.49%,优于单一Elman神经网络预测结果。本方法简单易行,可用于建筑能耗预测和建筑能耗监测系统的报警阈值设置。
The influencing factors of building energy consumption are complex. Researching new energy consumption prediction methods can simplify the prediction process and improve the prediction accuracy. First, a principal component analysis (PCA) is performed on energy consumption samples of a university building to remove information redundancy and eliminate the correlation between input variables. The principal component extracted by PCA is regarded as the input of Elman neural network. The tansig function is used for both the hidden layer and the input layer. The weights and deviations are continuously corrected in the training process. Finally, a PCA-Elman-based prediction model of building energy consumption . The test samples are used to verify the accuracy of the model. The results show that the relative error of PCA-Elman model is 5.49%, which is better than the single Elman neural network prediction. The method is simple and easy to use and can be used for the alarm threshold setting of building energy consumption prediction and building energy consumption monitoring system.