Sustainable Energy, Grids and Networks 37 (2024) 101246 Available online 11 December 2023 2352-4677/© 2023 Elsevier Ltd. All rights reserved. Contents lists available at ScienceDirect Sustainable Energy, Grids and Networks journal homepage: www.elsevier.com/locate/segan STNILM: Switch Transformer based Non-Intrusive Load Monitoring for short and long duration appliances L.N. Sastry Varanasi ∗ , Sri Phani Krishna Karri National Institute of Technology, Andhra Pradesh (NIT AP), Tadepalligudem, India ARTICLE INFO Keywords: Non-Intrusive Load Monitoring (NILM) Energy disaggregation Transformer model Deep learning algorithm Sequence-to-sequence learning ABSTRACT Non-Intrusive Load Monitoring (NILM) is a technique used by contemporary energy management systems to predict and optimize appliance load distribution in real time. The real-time reduction of energy consumption and improvement of electricity efficiency are two major benefits of energy disaggregation. Transformer models have made NILM far better at forecasting device power values. Due to the absence of inductive bias in the local context, transformers may not be able to capture local signal patterns in sequence-to-point settings. In this work, we present a Switch Transformer based Non-Intrusive Load Monitoring (STNILM). STNILM utilizes switching and routing layers by replacing the vanilla transformer final layers to accurately estimate the power signals of short and long duration domestic appliances. It also uses self attention mechanisms to extract global dependencies between the aggregate and the domestic appliance signals. STNILM works with minimal dataset pre-processing and unbalanced. With extensive experiments and quantitative analysis, we demonstrate the efficiency and effectiveness of the proposed STNILM with considerable improvements in terms of accuracy and F1-score compared to state-of-the-art baselines. 1. Introduction Sensor data is used by intelligent energy management systems to optimize use for effective domestic energy consumption [1,2]. Smart meter data analytics are applied to forecast, manage the load [3], and identify the demand response in energy management systems [4]. Such sensor-based power analysis gives an estimate on electric power usage, which promotes the sustainable energy eco system [5]. Bidgely [6] establishes that NILM based energy efficiency service can reduce house- hold energy consumption by 14%. Smart sensors offer increasing promise for sustainable energy management through artificial intel- ligence and the internet of things [1,7]. Sensor based systems can be utilized to detect appliance state and estimate power consump- tion values [5]. Early versions of smart metres and sensor systems are made to measure power used by electrical appliances [8,9]. For energy disaggregation and event-based analysis, intelligent sensor sys- tems with integrated hardware and software architecture have been proposed [10]. In the development of smart grids and smart cities, sensor-based energy management has recently been proposed to lower total consumption and maximize operational efficiency [1,11]. Deploy- ing sensor based systems is costly, intrusive and involves proprietary communication protocols [12,13]. Large-scale smart meter rollouts have recently escalated interest towards efficient Non-Intrusive Load Monitoring (NILM) methods. ∗ Corresponding author. E-mail addresses: sastry.sclr@nitandhra.ac.in (L.N.S. Varanasi), sriphani@nitandhra.ac.in (S.P.K. Karri). Non-intrusive load monitoring (NILM) is a technique used to iden- tify how much power is used by specific home equipment. Smart metres give information on a building’s total energy usage, but they could not give enough specifics to affect consumer behavior. Data gathering, feature extraction, event detection, load identification, and energy separation are all steps in the NILM process. During steady- state and transient situations, machine learning algorithms are used to recognize appliances and extract features. This technique makes it possible to assess appliance performance over a long period of time, which helps manufacturers improve energy efficiency. Consumers may receive suggestions encouraging them to use portable appliances less frequently or later in order to save energy [14]. Building energy use can be tracked so that waste can be avoided and consumers can take the appropriate action. Smart metres track overall energy use and often update users, which has led to a 3% drop in energy consumption. Furthermore, quick feedback on energy use at the building level can result in savings of up to 9% [15]. Because customers are likely to modify their usage habits in response to feedback, consumer behavior has a substantial impact on efficient energy use. The NILM system is a useful tool for tracking energy use and detecting high-energy equipment, giving both users and manufacturers insightful data to increase energy efficiency [1]. https://doi.org/10.1016/j.segan.2023.101246 Received 14 February 2023; Received in revised form 1 December 2023; Accepted 3 December 2023