Journal of Electrical Engineering, Electronics, Control and Computer Science – JEEECCS, Volume 7, Issue 23, pages 21-32, 2021 Electrical Load Identification for Household Appliances Ritesh Purrahoo 1 , Bhimsen Rajkumarsingh 2 Department of Electrical and Electronic Engineering, University of Mauritius, Mauritius 1 riteshpurrahoo@gmail.com 2 b.rajkumarsingh@uom.ac.mu Abstract — Electrical energy is one of the leading sources of production and of crucial significance in nowadays life. Energy saving is a key element and hence, efficient management of energy in buildings is pivotal to reduce electricity consumption. For this motive, it is essential to provide individual appliance energy consumption data to homeowners with the use of efficient assets and thus, the overall energy consumption obtained from the house main circuits must be disaggregated into separate device. Electrical load identification helps determine the type of load, operating conditions and electricity consumption of electrical appliances. This work examines different identification techniques based on power signature of household electrical appliances obtained from Reference Energy Disaggregation Dataset (REDD) system which uses a single electricity sensor connected to a building's main circuit to measure aggregated energy consumption. Each distinct algorithm extracts dissimilar features for analysis and classification. Various sets of samples were generated for simulation purposes to evaluate the proposed methods. From results obtained, we were able to identify the appliances chosen for use within a certain level of inaccuracy. The Load Switching Transient (LST), Mean Steady State (MSS) and Discrete Fourier Transform (DFT) algorithms have been determined to have overall accuracy of 99.8%, 97.1% and 100% respectively based on the samples generated for simulations. Finally, the DFT method was deemed to be unsuitable for use in practise to its limitations with the other two method preferred despite their drawbacks and lower precision percentages. Keywords-Appliance States; Load monitoring; Non-Intrusive Load Monitoring (NILM); Power Signatures; REDD I. INTRODUCTION Electrical energy is the most important and convenient source of energy related to the economy and people. In the last few decades, the electrical consumption has kept on rising at an alarming rate. With more technologies at our disposal to facilitate our daily tasks and the world population growing annually, the demand of electricity is increasing rendering it more and more challenging to meet such requirements. A significant percentage of energy is used in household for different purposes: heating and cooling of premises, lighting and other electrical appliances. In 2017 in the European Union (EU), 27.2 % of total energy consumption was represented by households [1]. Saving energy has transitioned from an option to a necessity, due to limited amount of available resources, so as to meet current and future demand of energy consumption of the population. This may be achieved through efficient use of energy. A reduction in energy usage would result into the better protection of the globe since a decrease in demand requires less production of energy from non-renewable sources of energy. Therefore, it will lessen its negative impact on climate change and greenhouse effect on the environment. It is impossible to deduce which electrical devices have an important contribution in the consumption of the total electricity bill due to the increasing number of household electrical appliances. Thus, people are unable to diminish their consumption as they are unaware of the factors impacting on their total electricity demand. Studies have shown that providing consumers with real-time power consumption information, at the aggregate level, helps them to change their behaviour and save 10-15% on power costs [2-4]. Electrical load Identification is an approach used in order to disaggregate the total energy consumed in households during a certain period of time to detailed ones. It will help customers determine the kind of appliance, operating modes and consumption details of electrical devices. Normally, sensors acquire data on usage details of appliances such as power consumed, time and frequency of use for a limited period of time. The data are then transferred to a processing system for processing and storage from where it can be accessed and displayed anytime. It is more practical and less costly to use a single sensor at the main meter instead of a sensor for each appliance for load identification. As a result, a simpler hardware network is obtained nevertheless requiring more complex algorithms for load identification. This approach is referred to as Non-Intrusive Load Monitoring and research on the latter was started in the 1980 by George W. Hart, Fred Schweppe and Ed Kern from Massachusetts Institute of Technology (MIT) for the purpose of measuring voltage and current data to determine the discrete states of electrical appliances [5]. For households, power consumption of appliances is the parameter used to record the usage of devices. This data however, can only reveal the overall energy consumption of the