A Literature Review of AI-Based Smart Energy Management and Consumption Prediction Systems

Abstract

Electricity use in homes, offices, and institutions keeps rising, yet most users cannot easily see which appliances waste energy or how much they will consume next. Research on this problem is spread across several strands: machine-learning (ML) forecasting of consumption, Internet of Things (IoT) energy monitors, appliance-level analysis, home energy management systems (HEMS), recommendation systems, and occupancy-based control. This review analyzes twenty-two publications and one technical report from these strands that are relevant to the AI-Based Smart Energy Management and Consumption Prediction System, a proposed system that combines an energy/current sensor, a presence sensor, a Wi-Fi-enabled microcontroller with a relay module, an ML prediction model, and a web dashboard. Through thematic and comparative analysis we identify four research gaps: forecasting evidence is mostly dataset-based, and the few hardware-linked systems depend on cloud services or heavier edge hardware; prediction-driven control and presence-based control are studied separately; quantified evidence for presence-based control comes mostly from building-scale lighting and HVAC; and recommendation research targets load shifting rather than idle or unnecessary appliance use. We then outline how the proposed system’s integrated design is intended to address these gaps. The prototype is under development, so this paper reports no experimental results.