ILB: Graph Neural Network Enabled Emergency Demand Response Program For Electricity

TL;DR

This paper introduces ILB, a graph neural network-based emergency demand response program that efficiently manages electricity demand during crises by incentivizing flexible households, validated through extensive experiments in multiple states.

Contribution

The paper presents a novel GNN-based framework for participant selection in emergency demand response, improving demand management during crises.

Findings

ILB effectively reduces demand during emergencies.

GNN-based participant selection outperforms traditional methods.

Successful deployment demonstrated in California, Michigan, and Texas.

Abstract

Demand Response (DR) programs have become a crucial component of smart electricity grids as they shift the flexibility of electricity consumption from supply to demand in response to the ever-growing demand for electricity. In particular, in times of crisis, an emergency DR program is required to manage unexpected spikes in energy demand. In this paper, we propose the Incentive-Driven Load Balancer (ILB), a program designed to efficiently manage demand and response during crisis situations. By offering incentives to flexible households likely to reduce demand, the ILB facilitates effective demand reduction and prepares them for unexpected events. To enable ILB, we introduce a two-step machine learning-based framework for participant selection, which employs a graph-based approach to identify households capable of easily adjusting their electricity consumption. This framework utilizes two Graph Neural Networks (GNNs): one for pattern recognition and another for household selection. Through extensive experiments on household-level electricity consumption in California, Michigan, and Texas, we demonstrate the ILB program's significant effectiveness in supporting communities during emergencies.