Intelligent categorization and interactive mechanism for smart demand side management of residential consumers

Abstract

Power Grids are becoming resilient to the erupting complications of imbalanced supply and demand through intelligent technologies that analyze and allow communication between utility and consumers. This paper proposes a strategy for smart energy management among residential consumers using machine learning algorithm that intelligently classifies high power consumers contributing to peak demand in every quarter of a day as Category A, Category B, Category C and Category D. Classification is performed on daily load profile data of 29,546 residential consumers using different supervised learning techniques. The raw data of consumers are preprocessed with statistical analysis to label peak consumers in different time slots and highest testing accuracy of 92.23% has been achieved with Random Forest classifier. The peak loads of classified consumers are then individually clipped to curtail the load of every consumer within baseline power of every quarter. Adoption of this methodology will facilitate to provide recommendations to every classified consumer of different categories for load curtailment and imposition of penalty based on the baseline limit. The power deficits during different quarters of a day estimated as 0.968%, 0.740%, 0.993% and 0.164% of load have been effectively adjusted by the classified consumers with per capita load adjustment of 3.641 kW, 1.164 kW, 1.628 kW and 0.984 kW respectively. Thus, overall power deficit, 0.73% of daily total load (12,265.7 kW) has been mitigated through intelligent distribution of load with minimal stress on consumers, which proves the efficiency of proposed Demand Side Management algorithm in sustaining growing demands within limited resources.