摘要
为解决BP神经网络预测速度慢不适于建筑能耗短期预测的问题,采用Levenberg-Marquardt算法改进BP神经网络建立了基于LMBP神经网络的建筑能耗短期预测模型。通过某建筑物1个月的电量,对模型进行训练和测试,结果表明基于LMBP神经网络的预测模型预测速度显著提高,预测精度满足实际需要,适用于建筑能耗短期预测。
BP neural network isn＇t suitable for short-term prediction of building energy consumption because of its slow prediction speed. Therefore, Levenberg-Marquardt algorithm is adopted to establish short-term prediction model based on LMBP neural network instead of BP neural network. Then the power load data of a building for one month are used to train and test the improved prediction model. Results show that the efficiency of the advanced model is greatly increased. Besides, the prediction accuracy is in the allowed range. It indicates the prediction model based on LMBP neural network is applicable to shortterm prediction of building energy consumption.
出处
《建筑节能》
CAS
2014年第11期79-81,100,共4页
BUILDING ENERGY EFFICIENCY
基金
上海市教委学科专业建设资助项目(XKCZ1212)
