地源热泵系统能耗预测及实例验证研究  

Study on Energy Consumption Prediction and Case Verification for Ground-source Heat Pump System

作　　者：钟寒露 朱尧星 徐成良[1] 陈焕新[1] ZHONG Hanlu;ZHU Xiaoxing;XU Chengliang;CHEN Huanxin(School of Energy and Power Engineering,Huazhong University of Science and Technology,Wuhan 430074,Hubei,China)

机构地区：[1]华中科技大学能源与动力工程学院,湖北武汉430074

出　　处：《制冷技术》2020年第1期54-58,共5页Chinese Journal of Refrigeration Technology

基　　金：国家自然科学基金(No.51876070,No.51576074)。

摘　　要：本文以某办公大楼的地下水源热泵系统为研究对象,在夏季对该系统的运行功率及其他运行数据进行监控和数据统计,将此数据作为能耗预测的样本数据。针对传统预测模型无法兼顾负荷数据的时序和非线性的问题,提出了基于长短期记忆神经网络(LSTM)能耗预测模型,并与神经网络模型进行预测性能的比较。结果表明,LSTM模型预测值的两个误差评价指标,即均方根误差(RMSE)和平均绝对误差(MSE)分别为13.6292和6.3105;神经网络模型的RMSE和MSE分别为34.1411和21.6430;因此LSTM模型的预测性能优于神经网络。In this paper,the groundwater-source heat pump system of an office building is taken as the research object,and the operating power and other operational data of the system are monitored and statistically analyzed during summer,using the data as sample data for energy consumption prediction.Aiming at the problem that the traditional prediction model cannot balance the timing and nonlinearity of the load data,an energy consumption prediction model based on Long Short-term Memory(LSTM)neural network is proposed,and the prediction performance is compared with the neural network model.The results show that,the two error evaluation indicators RMSE and MSE of the LSTM model predictions are 13.62292 and 6.3105,respectively,while the RMSE and MSE of the neural network model are 34.1411 and 21.6430,respectively.So that the prediction performance of the LSTM model is better than that of the neural network.

关 键 词：地下水源热泵系统 LSTM神经网络 神经网络 能耗预测 

分 类 号：TQ051.5[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]