Design and Implementation of Non-Intrusive Load Monitoring using Machine Learning Algorithm for Appliance Monitoring

Abstract

Listen

Translate article

Energy Conservation and management is gaining popularity in research area due to the increase in energy demand. It also plays a vital role in industrial/commercial/domestic/power sectors for reducing the carbon emission, the energy bill. Thus, increase in conservation would change the economy of the world energy crisis. To conserve energy, it is very important to have a better monitoring and identification system. The existing monitoring technique has conventional energy meter or smart energy meter that gives total energy consumption only. To enhance a better quality of services in existing monitoring technique, there is a need to monitor energy consumption of individual appliances and hence one meter/sensor for each appliance are necessary. Due to more sensors and its associated installation cost, this technique is not a cost effective in nature. To overcome Non-intrusive Load monitoring technique was introduced to disaggregate the total energy consumption from a single meter using Machine learning disaggregation algorithm. Thus, to identify the appliances malfunctioning Non-intrusive load monitoring (NILM) technique can be used as a Real time Monitoring technique. In this paper, it is proposed to use single energy meter for the set of appliances to monitor the status of the individual appliances. Non-intrusive Load Monitoring technique using machine learning algorithms has been discussed for appliances identification and monitoring for energy conservation. The MATLAB/Simulink Software has been used for designing and mathematical modeling of each appliance. The NILM technique mainly involves the three stages via; Data acquisition, feature extraction and training of data under different classification algorithm for appliance identification. Data acquisition used for acquiring the voltage and current from a single phase system. Using the features extracted like active, reactive power the different load patterns of individual appliances can be studied. Training of data under DT and K-NN which are supervised learning techniques are used as disaggregation algorithm. Moreover, the algorithms are compared using the Confusion matrix and ROC curve for the prediction of accuracy. The result shows that the K-NN algorithm is having a better accuracy of performance compared with DT algorithm.