基于BP-Adaboost算法的建筑能耗预测研究  

Prediction for Building Energy Consumption Based on BP-Adaboost Algorithm

作　　者：方涛涛[1] 马小军[1] 陈冲[1] Fang Taotao;Ma Xiaojun;Chen Chong(College of Electrical Engineering and Control Science,Nanjing Tech University,Nanjing 211816,China)

机构地区：[1]南京工业大学电气工程与控制科学学院,江苏南京211816

出　　处：《科技通报》2017年第7期162-166,共5页Bulletin of Science and Technology

摘　　要：针对传统BP神经网络存在收敛速度慢和容易陷入局部极小值缺点,影响建筑能耗预测精度的问题,引入Adaboost算法对BP神经网络进行改良,提出一种基于BP-Adaboost算法的建筑能耗预测方法。该方法充分结合BP神经网络良好学习能力和Adaboost算法预测精度高的优点,通过Adaboost算法将BP神经网络训练所得的弱预测器组合集成为BP-Adaboost强预测器,完成对建筑能耗的预测。案例仿真结果表明:与传统BP神经网络预测比较,该算法预测速度快、预测精度较高,其预测结果可为建筑节能方案的实施提供参考依据。In view of the traditional BP neural network easy to fall into local minimum and slowconvergence speed value.It can affect building energy consumption prediction problem.BP neuralnetwork was improved by Adaboost algorithm.The method combines the advantages of BP neuralnetwork learning ability with the higher prediction accuracy of Adaboost algorithm.It integrate themultiple weak predictors of BP neural network into strong BP-Adaboost predictor to predicte buildingenergy consumption.The result show that the BP-Adaboost algorithm prediction is higher than BPneural network.It can provide the reference for building energy efficiency.

关 键 词：BP神经网络 ADABOOST算法 建筑能耗 预测 

分 类 号：TU12[建筑科学—建筑理论]