AI-BASED MONITORING AND FORECASTING OF ELECTRICITY CONSUMPTION IN MULTI-APARTMENT RESIDENTIAL BUILDINGS
DOI: 10.31673/2412-4338.2025.038703
Abstract
The subject of this study is the process of forecasting and managing electricity consumption in multiapartment residential buildings during emergency outages or operation on backup generators. The problem addressed concerns the optimal application of intelligent load management methods such as predictive control, reinforcement learning algorithms, fuzzy logic, and multi-agent systems.
The article explores promising practices for improving the resilience and efficiency of power supply in apartment buildings in cases of diesel generator operation or limited energy availability. The results of the study include a comprehensive analysis of modern load forecasting methods using neural networks. Special attention is given to load prioritization techniques under backup power scenarios. Intelligent energy management systems based on multi-agent
technologies are examined. The effectiveness of hybrid control strategies is also evaluated. The study found that reinforcement learning (RL) methods, when combined with prioritization rules and fuzzy logic, contribute to greater stability and efficiency during backup power operation.
The implementation of multi-agent systems enables flexible load distribution among apartments, taking into account individual priorities and real-time consumption conditions. Hybrid approaches have shown superior performance in scenarios with unpredictable load changes. The proposed trust and reputation mechanism improves the reliability of decision-making regarding load limitation for consumers with a history of non-compliant consumption behavior.
The application of intelligent technologies for managing electricity consumption in multi-apartment residential buildings during backup generator operation helps enhance system resilience, reduce the risk of emergency shutdowns, and optimize the use of available capacity.
Keywords: load management; fuzzy logic; neural networks; reinforcement learning; multi-agent systems; load forecasting; backup power supply; load shedding.
