Indonesian Journal of Electrical Engineering and Computer Science Vol. 30, No. 3, June 2023, pp. 1678~1688 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v30.i3.pp1678-1688  1678 Journal homepage: http://ijeecs.iaescore.com Detection of occupancy status from internet connectivity for non-intrusive load monitoring Manjula Wickramathilaka 1,2 , Md Pauzi Abdullah 1 , Mohammad Yusri Hassan 1 , Hayati Abdullah 3 1 School of Electrical Engineering, Faculty of Engineering, Universiti Teknologi Malaysia, Johor Bahru, Malaysia 2 Division of Electrical, Electronics and Telecommunication Engineering Technology, Institute of Technology, Universiti of Moratuwa, Homgama, Sri Lanka 3 School of Mechanical Engineering, Faculty of Engineering, Universiti Teknologi Malaysia, Johor Bahru, Malaysia Article Info ABSTRACT Article history: Received Aug 11, 2022 Revised Dec 17, 2022 Accepted Jan 9, 2023 Non-intrusive load monitoring (NILM) methods are widely used for appliance level energy disaggregation in residential buildings. These methods mostly depend on electrical features, and they have not been much successful in applying for commercial buildings. However, recent research has indicated that the accuracy of existing NILM methods can be improved by associating with occupancy data. Therefore, in this paper a novel occupancy detection algorithm is proposed which can detect occupancy status of individuals using the connectivity of their information technology (IT) devices to the local area network of the building. The model is validated using data collected at a university building, with mean errors of 01:23 and 04:02 minutes for the detection of arrival and departure. The occupancy profiles developed by the proposed model can be used to disaggregate energy consumption in a commercial building to appliance and occupant level. Keywords: Commercial buildings Energy disaggregation Internet connectivity Non-intrusive load monitoring Occupancy detection This is an open access article under the CC BY-SA license. Corresponding Author: Md Pauzi Abdullah School of Electrical Engineering, Faculty of Engineering, Universiti Teknologi Malaysia 81310 Johor Bahru, Johor, Malaysia Email: mpauzi@utm.my 1. INTRODUCTION Global energy consumption has increased by more than 50% in the last twenty years, and it is estimated to increase annually by 2% in the next 30 years [1]. Commercial and residential buildings share more than 40% of the global energy consumption [2] and are also responsible for a similar portion of green house gas (GHG) emissions [3]. In Malaysia, contribution of the entire building sector to the total energy consumption is 48% [4]. Out of the total electricity consumption in Malaysia commercial buildings (including universities) have consumed 43,724 GWh (almost 30%) in 2017 [5]. Load monitoring plays a vital role in energy management. Out of the two load monitoring methods; intrusive and non-intrusive, non-intrusive load monitoring (NILM) has become a largely researched topic [6] due to its competitive advantages. The very first NILM model was introduced by Hart et al. [7] in 1986. As opposed to intrusive load monitoring (ILM), the interesting feature of NILM was its ability to disaggregate energy consumption down to the sub-circuit level while having an energy meter installed only at the main circuit breaker of the interested building or section. After the introduction of first model, Hart himself as well as many others have proposed a variety of NILM solutions [6], [8]-[10]. In the literature, classification of NILM approaches has been done in many different manners [9], [11], [12]. Even though NILM approaches have shown a significant success in residential building sector, it has not been successful so far commercial buildings due to various reasons like large number of identical appliances and occupants [11], [13]-[15]. A detailed review of NILM applications in commercial buildings has been done in our previous work [16].