机器学习方法在建筑能耗预测中的应用研究

Research on the application of machine learning methods in building energy consumption prediction

为了减少建筑对能源的过多消耗,通过对建筑能耗数据进行数据挖掘,提出了使用机器学习的方法对建筑能耗数据进行预测分析。首先对建筑能耗数据中的自变因素和从变因素进行相关性分析;然后进行数据归一化;接下来使用KNN、决策树、AdaBoost和随机森林这四种机器学习方法进行能耗预测。在实验过程中,使用交叉验证的技术,分别对四种算法计算了MAE、MRE、MSE、RMSE四种评价函数。最后将四种方法所得的结果做了比较。实验结果显示随机森林算法表现出最好的效果。

In order to reduce building energy consumption of excessive energy consumption,building energy consumption data for data mining can provide a theoretical basis for decision‑making of building energy‑saving program.In this paper,the machine learning methods were used to analyze and predict data of building energy consumption.The main works are as follows.First,make correlation analysis of building energy consumption data from independent and dependent variables.After data normalization,KNN(k‑nearest neighbor algorithm),decision trees,AdaBoost and Random Forests(RF)are used to predict the energy consumption.During the experiments,cross‑validation technique was used in this paper.Four functions are used to evaluate by mean absolute error(MAE),the mean relative error(MRE),the mean square error(MSE)and the root mean square error(RMSE)respectively.The result of the experiments indicated that the best performance was made by RF.