Fuel Poverty Risk Scoring at Scale: A Multi-Signal Weighted Approach Using Half-Hourly Smart Meter Data

Fuel poverty remains a critical challenge affecting millions of households and commercial premises across the United Kingdom, yet systematic, data-driven methods for identifying at-risk properties at portfolio scale remain underdeveloped. This paper presents the Fuel Poverty Risk Score (FPRS), a novel multi-signal weighted scoring framework applied to 1,112 commercial and residential buildings using half-hourly smart meter data from the Building Data Genome Project 2 (BDG2). The FPRS integrates five behavioural signals derived entirely from consumption patterns: safe heating threshold violations, extended low-consumption periods, high variability in usage, anomalous consumption spikes, and LSTM autoencoder reconstruction error. Applied across the full portfolio, the FPRS produces a mean score of 23.8 (SD=20.0), identifying 160 buildings (14.4%) as HIGH or CRITICAL risk.