This research investigates how LLMs can function as autonomous coordinators for home energy systems. A hierarchical architecture with one orchestrator and three specialist agents using the ReAct reasoning pattern was developed. Testing across three open-source models with Austrian electricity pricing data showed that Llama-3.3-70B successfully coordinates all appliances across all scenarios to match cost-optimal benchmarks.

This paper presents a novel optimization algorithm that integrates user behavior into day-ahead load scheduling for residential demand response by utilizing data from the Harmonized European Time Use Survey (HETUS). The algorithm schedules household appliances based on dynamic pricing and user preferences, achieving significant cost savings of up to 50% while minimizing disruption to user behavior through buffer zones and alternative scheduling options.

This paper presents a reinforcement learning approach using Soft Actor-Critic for optimizing household battery charging and discharging under real-time pricing scenarios. The agent operates without requiring future price predictions, relying only on current conditions and historical data, demonstrating effective cost minimization and adaptability across European markets including Germany, Austria, and Italy.

This research proposes a home energy management system utilizing genetic algorithms to optimize household appliance scheduling. The system incorporates dynamic electricity pricing, solar energy utilization, and user comfort considerations. Using actual consumption data from a Moroccan household, three optimization scenarios were evaluated, demonstrating cost reductions up to 63.48% compared to baseline consumption.