基于K均值小波神经网络的二阶段空调负荷预测  

A two-stage prediction for air-conditioning load base on K-means wavelet neural network

作　　者：赵超 郑守锦 ZHAO Chao;ZHENG Shoujin(College of Chemical Engineering,Fuzhou University,Fuzhou,Fujian 350116,China)

机构地区：[1]福州大学石油化工学院,福建福州350116

出　　处：《福州大学学报（自然科学版）》2018年第3期416-421,共6页Journal of Fuzhou University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(6080402;61374133);高校博士点专项科研基金资助项目(20133314120004)

摘　　要：结合聚类分析和小波神经网络模型,提出一种二阶段空调负荷建模方法,以提高空调负荷预测精度.首先利用K均值聚类算法将原始负荷样本数据依据其统计分布特性划分为若干簇类,以降低数据相关性对建模精度的影响;然后基于对每个划分簇类所属的样本数据建立相应小波神经网络空调负荷预测模型.最后基于De ST平台模拟数据,将构造的小波神经网络预测模型运用于福建某办公大楼的逐时空调负荷预测.通过对比均方根误差(RMSE)和平均绝对误差(MAPE),结果表明该模型的预测精度明显优于传统单一的小波神经网络和BP神经网络模型.In order to improve the accuracy of air conditioning load prediction,a two-stage predictive model based on K-means clustering and wavelet neural networks（ WNN） was proposed. Aiming at the strongly coupling nonlinear characteristics of the air conditioning load data,K-means clustering method was employed to divide the historical load data into several clusters which could reduce the interference between samples and eliminate the noise in load sample data. Then,the wavelet neural network model was constructed with the training samples of the identified cluster. Based on the simulated data from the De ST platform,the two-stage WNN model was used to predict the hourly air-conditioning load of an office building in South China. Experiment results shown that the proposed model performed significantly higher prediction accuracy than the traditional single WNN model and BP model in terms of the root mean square error（ RMSE） and the mean absolute percentage error（ MAPE）.

关 键 词：空调负荷 预测 K均值聚类算法 小波神经网络 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程]