EnergyBridge: Benchmarking Household Energy Management, User Participation, and Grid Flexibility

2026-08-09

AI Summary

This study addresses a critical gap in existing residential virtual power plant (VPP) benchmarks, which overlook user authorization behavior at the event level, thereby hindering the reliable conversion of physical flexibility into grid capacity. To bridge this gap, the work proposes the first end-to-end evaluation framework that integrates event-level user authorization, high-fidelity physical simulation, and human behavior modeling. The framework combines region-specific EnergyPlus environments for Tianjin and Berlin with a large language model–driven user participation simulator. By incorporating human-centric factors into VPP benchmarking for the first time, it introduces a novel dataset of real human judgments. Experimental results from both cities demonstrate superior performance, achieving the highest authorization rates, lowest event-window energy consumption, and most reliable capacity commitments, with an average absolute error of only 5.3 points in user acceptance prediction—significantly outperforming current approaches.

Abstract

Residential virtual power plants (VPPs) can provide grid flexibility by shifting household demand, but physical flexibility becomes dependable capacity only when residents authorize a plan and the promised response is delivered. Existing benchmarks evaluate control but omit event-specific authorization. We present EnergyBridge, a benchmark and agent framework connecting capacity reporting, household authorization, and physical execution. It combines region-specific EnergyPlus environments for Tianjin and Berlin with an LLM-based User Participation Simulator. Against 584 persona- and event-matched human role-play judgments, the LLM-based User Participation Simulator preserves method ordering with a 5.3-point mean absolute acceptance error. Across conventional controllers and agent baselines, EnergyBridge achieves the highest simulated authorization, lowest event-window energy, and the most reliable capacity commitment in both regions. We release human data and codes for reproducible human-centered grid-flexibility research: https://github.com/Agentic-Intelligence-Lab/EnergyBridge.

Problem

• virtual power plant
• user authorization
• grid flexibility
• residential energy management
• benchmarking

Innovation

• virtual power plant
• user participation simulation
• LLM-based agent
• grid flexibility benchmarking
• EnergyPlus co-simulation