Non-intrusive energy estimation using random forest based multi-label classification and integer linear programming

• DOI: 10.1016/j.egyr.2021.08.045
• Publication Date: 2021-11-25T07:19:11Z
• Authors: Yu Liu; Congxiao Liu; Yiwen Shen; Xin Zhao; Shan Gao; Xueliang Huang
• Keywords: Identification

Home energy management system is proposed to reduce the influences caused by high ratio penetration of renewable generation, through managing and dispatching residential power consumption in demand side. Being aware how electric consumed a key step this system. Non-intrusive Load Monitoring regarded as most potential method address problem, which aims separate individual appliances households decomposing total consumption. In recent years, NILM framed multi-label classification problem many researches has been investigated field. paper, non-intrusive can identify usage information from thoroughly investigated. Firstly, random k-labelset algorithm enhanced introducing forest base classifier. Then, grid search cross validation are integrated determine optimal paraments set. This used achieve identification. Finally, based on identification result, integer linear programming employed for estimation each appliance, especially multi-state appliances. Experimental results low voltage networks simulator demonstrate that accuracy compared with traditional methods other classifiers, it capable identifying usages different accurately. The desirable performance broadened applications machine learning monitoring.