Introducing A New Method To Detect Technical And Non-technical Losses In The Smart Grid Using Machine Learning Algorithms

International audienceA smart grid is an electrical network enabling a two-way flow of both data and electricity. During electrical power generation in the Smart Grid, a vast amount of electricity is lost. Mainly in transmission and distribution lines, where 17% of electricity is lost in transmission lines, and 50% is lost in distribution lines. This is called Transmission and Distribution Losses (T&D), and it describes when the amount of electricity generated is greater than the amount of electricity delivered to end-users. There are two types of T&D losses. First are Technical Losses (TL), which are ordinarily 22.5%, and directly depend on the network characteristics and mode of operation. Second is Non-technical Losses (NTL) at 16.6% and related to meter reading, defective meter and error in meter reading, billing of customer energy consumption, lack of administration, financial constraints, and estimating the unmetered supply of energy as well as energy thefts. Previously proposed solutions for detecting TL and NTL depend on monitoring user behavior, analysis, and feature extraction. In contrast, new proposed solutions for detecting TL and NTL made use of Machine Learning (ML) Algorithms and Synthesis data, differing from each other in applying various ML approaches (Supervised, Unsupervised, Semi-Supervised) and diverse Data processing techniques. As the rising need to detect TL and NTL in the Smart Grid, the industry is not equipped with an applicable solution to overcome the problem. The overall deliverables of this study were: 1. multi-layers system model, 2. Machine learning model, 3. new methodology. In principle, the model uses a new method to detect TL and NTL and Real-time data from Smart Grid