基于EMD与机器学习算法的近零能耗建筑负荷预测方法
Load forecasting method of nearly zero energy buildings based on EMD and machine learning algorithm
摘要：
采用皮尔逊相关系数法分析了不同特征变量与冷热负荷的相关性,确定了预测模型的输入特征变量。采用经验模态分解(EMD)对逐日冷热负荷按频分解,然后采用机器学习算法,即反向传播神经网络(BPNN)、随机森林(RF)和支持向量机(SVM),分别对不同频率的负荷量进行了训练、验证,最后重构得到了近零能耗建筑预测负荷。基于上述方法,以北京市某近零能耗居住建筑为研究对象,比较了不同算法预测结果的精确度。结果表明:采用EMD与RF算法相结合对近零能耗建筑冷热负荷的预测精确度较高。进一步采用穷举搜索法对模型初设参数进行了优化,冷热负荷预测结果精确度提高,冷负荷预测结果的决定系数R2、平均绝对百分比误差MAPE分别为0.996、1.32%,热负荷预测结果的R2、MAPE分别为0.997、0.79%。
Abstract：
The Pearson correlation coefficient method is used to analyse the correlation between different feature variables and cooling and heating loads,and the input feature variables of the prediction model are determined.The empirical mode decomposition (EMD) is used to decompose the daily cooling and heating loads according to frequency,and the machine learning algorithms,namely,back propagation neural network (BPNN),random forest (RF) and support vector machine (SVM),are used to train and verify the loads at different frequencies.Finally the predicted load of nearly zero energy buildings is obtained by reconstruction.Based on the above methods,this paper takes a nearly zero energy residential building in Beijing as the research object,and compares the accuracy of the prediction results of different algorithms.The results show that the method combining EMD with RF algorithm has a high accuracy in predicting the cooling and heating loads of nearly zero energy buildings.Furthermore,the exhaustive search method is used to optimize the preliminary parameters of the model,and the accuracy of the cooling and heating load forecasting results is improved.The determination coefficient (R2) and the average absolute percentage error (MAPE) of the cooling load forecasting results are 0.996 and 1.32%,respectively,and the R2and MAPE of the heating load forecasting results are 0.997 and 0.79%,respectively.
Keywords:nearly zero energy building; load forecasting; empirical mode decomposition; machine learning algorithm; back propagation neural network (BPNN); random forest (RF); support vector machine (SVM); exhaustive search method