Abstract: The rapid growth of electricity demand in residential and commercial sectors, combined with the increasing penetration of renewable energy sources, necessitates intelligent, real-time energy management solutions. This paper presents an AI-Based Smart Energy Monitoring and Consumption Optimization System with Appliance-Level Control, which integrates Internet of Things (IoT) sensing infrastructure, Non-Intrusive Load Monitoring (NILM) and reinforcement learning-based optimization to achieve granular control over individual household appliances. The proposed system architecture comprises five functional layers: a sensing layer using smart meters and IoT-connected plugs, a NILM disaggregation engine based on LSTM and Transformer models, an ML-based forecasting module, a Multi-Objective Deep Q-Network (MO-DQN) optimization agent and an IoT actuator control layer with a mobile user interface. Experimental evaluations demonstrate appliance-level energy classification accuracy exceeding 95%, a forecasting accuracy of 99.7% for hourly consumption using LSTM and electricity cost reductions of up to 18% through demand response integration. The system further supports dynamic peak clipping, valley filling and load shifting strategies under real-time pricing signals. This work contributes a scalable, privacy-aware and user-adaptive framework for next-generation smart home energy management

Keywords: Smart Energy Monitoring, Appliance-Level Control, Non-Intrusive Load Monitoring (NILM), Reinforcement Learning, Home Energy Management System (HEMS), IoT, Demand Response, Deep Q-Network, LSTM, Smart Grid.

Downloads: | DOI: 10.17148/IJIREEICE.2026.14529