基于粒子群优化算法的建筑能耗管理系统  

Building energy consumption management system based on particle swarm algorithm optimization

作　　者：潘旭 刘清惓[1,2,3] 邹永奇 王柯 PAN Xu;LIU Qingquan;ZOU Yongqi;WANG Ke(School of Electronics and Information Engineering,Nanjing University of Information Technology,Nanjing 210044,China;Jiangsu Key Laboratory of Meteorological Detection and Information Processing(Nanjing University of Information Science and Technology),Nanjing 210044,China;Jiangsu Meteorological Sensing Network Technology Engineering Center,Nanjing 210044,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,南京210044 [2]江苏省气象探测与信息处理重点实验室(南京信息工程大学),南京210044 [3]江苏省气象传感网技术工程中心,南京210044

出　　处：《中国科技论文》2024年第12期1335-1344,共10页China Sciencepaper

基　　金：国家自然科学基金资助项目(42275143)。

摘　　要：针对传统BP神经网络预测建筑能耗时容易陷入局部最优解的情况,提出一种粒子群优化(particle swarm optimization,PSO)算法优化BP神经网络预测建筑能耗。首先利用建筑热环境设计模拟工具包(designer’s simulation toolkit, DeST)构建建筑维护结构并仿真出不同气象参数条件下的空调能耗输出,将仿真所得样本数据分为训练集样本和测试集样本;然后使用PSO-BP神经网络算法对训练集进行训练学习,获得一个能耗预测经验方程,为便于算法的实际应用,用Java语言将方程封装在后端代码中;最后将测试集样本投入该方程中,对方程的准确性进行验证。结果表明,PSO-BP能耗预测结果与仿真结果相比,误差百分比介于[-1.110%,1.167%]。As a promising countermeasure to the problem that traditional BP neural networks are prone to getting stuck in local optima when predicting building energy consumption,an improved PSO(particle swarm optimization)-BP neural network was proposed for the prediction of building energy consumption management system.Firstly,DeST(designer’s simulation toolki)was used to construct the building maintenance structure and simulate the air conditioning energy consumption output under different meteorological parameter conditions.The simulated sample data was divided into training set samples and testing set samples.Then the PSO-BP neural network algorithm was used to train and learn the training set and obtain an empirical formula for energy con⁃sumption prediction.To facilitate the practical application of the algorithm,the formula was encapsulated in the backend code in Java.Finally,the test set samples were input into the formula to verify the accuracy of the equation.The results show that,com⁃pared with the simulation results,the error percentage of PSO-BP energy consumption prediction is between[-1.110%,1.167%].

关 键 词：建筑能耗 DeST仿真 PSO-BP神经网络 JAVA语言 

分 类 号：TN919-34[电子电信—通信与信息系统] TP312[电子电信—信息与通信工程]