A Recommendation System for Prosumers Based on Large Language Models

Area of Science:

• Energy Management
• Artificial Intelligence
• Smart Homes

Background:

• Increasing integration of smart home technologies and sensors.
• Homeowners acting as prosumers with access to detailed energy data and local energy market (LEM) information.
• Need for decision support systems to manage complex energy data for cost reduction and comfort maintenance.

Purpose of the Study:

• To propose a recommendation system for prosumers to optimize energy consumption and costs.
• To provide personalized advice for load adjustment and LEM transactions.
• To evaluate the system's performance using specific prosumer scenarios.

Main Methods:

• Development of a recommendation system powered by large language models (LLMs), Scikit-llm, and zero-shot classifiers.
• Evaluation of two prosumer scenarios (5.9 kW) with candidate labels: Decrease, Increase, Sell, and Buy.
• Comparison with a content-based filtering system using relevant performance metrics.

Main Results:

• The LLM-powered system offers tailored advice for prosumers based on real-time data.
• Demonstrated potential for optimizing energy consumption and LEM engagement.
• Comparative analysis highlights the system's effectiveness against traditional methods.

Conclusions:

• LLM-based recommendation systems can effectively support prosumers in managing home energy.
• Personalized recommendations enhance cost savings and comfort in smart homes.
• The proposed system shows promise for future energy management solutions.