Influence of “abnormal” consumers on fire hazardous situation in electric networks with utility consumers

Abstract

The dynamics in changes in electricity consumption by multi-apartment residential buildings in recent years creates uncertainty in determining the correspondence of the actual load with the standardized values. Identification of this uncertainty is possible using machine learning (neural network models) based on the local anomaly coefficient (LOF) method which makes it possible to identify “abnormal” consumers which are the objects that create an excess voltage unbalance value due to excess power consumption. In turn, they are sources that create a fire hazard in urban electrical distribution networks with household utility consumers. To identify these consumers at the initial stage, by using the developed computer model, the values of electrical loads were simulated taking into account the climate and meteorological factors. The adequacy of the computer model is confirmed by high convergence with the data given in the SP 256.1325800.2016. To determine the correspondence between the normalized electrical load values and the actual load, “abnormal” consumers are identified based on monthly power consumption data using machine learning. The results obtained show discrepancies in the range from -58 to +155% of the normalized values. By using the developed software application “Simulation model of the thermal process of an insulated conductor” for “abnormal” consumers with 60% asymmetry, the changes in the heating temperature of the wire cores were studied. The results obtained in the developed software application showed that with this asymmetry, the short-term steady-state temperature of the conductor was 225°C which on average was 3 times higher than the permissible values. This accelerates the aging process of the conductor insulation, contributes to the appearance of micro-cracks due to its drying out, and in general contributes to the emergence of a fire hazard.