科研之友
微信
新浪微博
Facebook
分享链接
摘要
Urban energy management nowadays has put more focus on residential houses energy consumption. Lots of machine learning based data-driven approaches have the abilities to characterize and forecast total energy consumption of commercial data. However, a paucity of research applying data-driven methods have been tested on the hour ahead energy consumption forecast for typical single family houses in the US. With the advances in smart metering, sub meter usages forecast in household-level is also getting more and more popular on smart building control and demand response program. The situation here inspires us to develop a hybrid model to address the problem of residential hour and day ahead load forecasting through the integration of data-driven techniques with a physics-based model. In this article, we report on the evaluations of five different machine learning algorithms, artificial neural network (ANN), support vector regression (SVR), least-square support vector machine (LS-SVM), Gaussian process regression (GPR) and Gaussian mixture model (GMM), applied to four residential data set that contains smart meters. Both total and non-air conditioning (AC) power consumption are forecasted for hour ahead and 24-h ahead. The variation of patterns captured from non-AC part is further input as internal heat gain to a physics-based model. The model uses a 2R-1C thermal network and an AC regression model. By utilizing this hybrid approach, we get AC load prediction and non-AC energy forecast simultaneously. Total energy consumption is further produced by summing up the two sub meters forecast for hybrid model. The results from new modeling are compared with those from pure data-driven techniques. The final result showing improvements of coefficient of variance between the best data-driven model and hybrid model are 6-10% and 2-15% for hour ahead and 24-h ahead, respectively.
- 出版日期2016-4-1
