Algorithm for Operational Detection of Abnormally Low Electricity Consumption in Distribution

Currently commercial losses of electric power are an actual global problem for electric grid companies and final consumers, since a high level of losses is the reason for a significant increase in electricity tariffs and a decrease in the quality of supplied energy. The solution to this problem is the creation of monitoring systems for operational search and detection of abnormal electricity consumption using elements of artificial intelligence. The algorithm of operational intelligent search for non-technical electricity losses (NTL) in 0.4 kV distribution networks is presented and applied in the study. The peculiarity of the developed algorithm is that the identification of energy consumers with abnormally low consumption is carried out without a pre-marked sample on the presence/absence of NTL. This significantly increases the value of the proposed algorithm, since in developing energy systems such accounting is not always carried out in digital form. Statistical analysis and machine learning methods were applied during the research. An experiment was also conducted on real energy consumption data of a sample site with 126 final consumers, all of which are fully equipped with advanced metering infrastructure (AMI). As a result of the experiment, the effectiveness of the algorithm was demonstrated as a basis for a monitoring system and management decision support in minimizing losses in the energy system.