基于浅层和深度机器学习算法的区域建筑电力负荷预测研究  

Research on Regional Building Power Load Prediction Based on Shallow and Deep Machine Learning Algorithms

作　　者：赵泽坤 高岩[1] 安晶晶 王者 ZHAO Zekun;GAO Yan;AN Jingjing;WANG Zhe(Beijing Energy Conservation&Sustainable Urban and Rural Development Provincial and Ministry Co-construction Collaboration Innovation Center,Beijing University of Civil Engineering and Architecture,Beijing 100044,China;Beijing Key Laboratory of Heating and Gas Supply Ventilation and Air Conditioning Engineering,Beijing University of Civil Engineering and Architecture,Beijing 100044,China;The Hong Kong Polytechnic University,Department of Civil&Environmental Engineering,Hong Kong 100872,China)

机构地区：[1]北京建筑大学北京节能减排与城乡可持续发展省部共建国家协同创新中心,北京100044 [2]北京建筑大学供热供燃气通风空调工程北京市重点实验室,北京100044 [3]香港科技大学,土木与环境工程学院,中国香港100872

出　　处：《建筑科学》2025年第2期229-236,共8页Building Science

基　　金：国家重点研发计划项目“零碳建筑及社区可再生能源应用关键技术研究”(2022YFE0134000);北京市教委科技计划重点项目(KZ202110016022)。

摘　　要：建筑电力负荷预测技术可用于辅助区域建筑能源规划、能耗及碳排放计算等。本文探索了机器学习算法在区域电力负荷短期预测方面的应用,使用了5种机器学习算法对北京某区域建筑的日电力负荷进行了预测,结果显示XGBoost(极限梯度提升)和LSTM(长短期记忆)2种先进的学习算法可以提供较为精准的预测结果。进一步,考虑到天气预报的不确定性,我们选取这2种模型讨论了输入不确定性对电力负荷预测结果的影响,结果表明使用LSTM模型可以有效降低由于天气预报不稳定性引起的预测误差,更适用于开展电力负荷短期预测。The building power load forecasting technology can be used to assist in regional building energy planning,energy consumption,and carbon emission calculations.This paper explored the application of machine learning algorithms in short-term prediction of regional power loads.Five machine learning algorithms were used to predict the daily power loads of a building in Beijing.The results showed that XGBoost(Extreme Gradient Boosting)and LSTM(Long Short-Term Memory)advanced learning algorithms can provide more accurate prediction results.Furthermore,considering the uncertainty of weather forecasts,we selected these two models to discuss the impacts of input uncertainty on the results of power load forecasting.The results indicated that the LSTM model can effectively reduce prediction errors caused by weather forecast instability and is more suitable for short-term power load forecasting.

关 键 词：电力负荷预测 机器学习 长短期记忆(LSTM) 极限梯度提升(XGBoost) 不确定性 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程]