论文部分内容阅读
针对传统BP神经网络的建筑能耗预测中不变的预测影响因素难以保证预测的准确性和人工确定网络结构耗时长的问题,本文提出一种基于GA-BP的自优化的建筑能耗预测方法。该方法利用遗传算法对建筑能耗BP神经网络预测模型的输入因素和网络结构进行自动寻优确定,有效地减少了最佳预测模型的设计时间,节省了人工实验成本。利用该方法建设的建筑能耗预测系统已应用在某建筑群的能耗预测中,有效地减少了建筑能源浪费。
Aiming at the problem that the invariable forecasting factors in the prediction of building energy consumption of traditional BP neural network can not guarantee the accuracy of prediction and the time-consuming determination of network structure manually, this paper proposes a GA-BP based self-optimizing method of forecasting building energy consumption . The method uses genetic algorithm to automatically determine the input factors and network structure of BP neural network prediction model of building energy consumption, which effectively reduces the design time of the best prediction model and saves the artificial experiment cost. The building energy consumption forecasting system constructed by this method has been applied to forecast the energy consumption of a building complex, which effectively reduces the building energy waste.