Predicting Public Health Impacts of Electricity Usage

TL;DR

The paper tackles how to predict and mitigate the public health impacts of electricity usage by linking demand-side dynamics to emissions, dispersion, and health costs through an end-to-end AI pipeline called HealthPredictor.HealthPredictor comprises a Transformer-based Fuel Mix Predictor, an Air Quality Converter for emissions and dispersion, and a Health Impacter that monetizes health changes, trained with a health-aware loss.Across three U.S. regions and an EV charging case study, the approach yields lower health-impact errors than fuel-mix baselines and demonstrates substantial potential health benefits from health-informed energy management.The work emphasizes data fusion, end-to-end optimization, and actionable signals for individuals and system operators, while releasing datasets and code to support reproducibility.

Abstract

The electric power sector is a leading source of air pollutant emissions, impacting the public health of nearly every community. Although regulatory measures have reduced air pollutants, fossil fuels remain a significant component of the energy supply, highlighting the need for more advanced demand-side approaches to reduce the public health impacts. To enable health-informed demand-side management, we introduce HealthPredictor, a domain-specific AI model that provides an end-to-end pipeline linking electricity use to public health outcomes. The model comprises three components: a fuel mix predictor that estimates the contribution of different generation sources, an air quality converter that models pollutant emissions and atmospheric dispersion, and a health impact assessor that translates resulting pollutant changes into monetized health damages. Across multiple regions in the United States, our health-driven optimization framework yields substantially lower prediction errors in terms of public health impacts than fuel mix-driven baselines. A case study on electric vehicle charging schedules illustrates the public health gains enabled by our method and the actionable guidance it can offer for health-informed energy management. Overall, this work shows how AI models can be explicitly designed to enable health-informed energy management for advancing public health and broader societal well-being. Our datasets and code are released at: https://github.com/Ren-Research/Health-Impact-Predictor.

Figures (5)

• Figure 1: Total public health costs of electricity generation and on-road emissions in the contiguous U.S. in 2023 and 2028 Health_COBRA_EPA_Website. The error bars represent high and low estimates provided by COBRA using two different exposure-response models.
• Figure 2: Overview of our end-to-end HealthPredictor. The pipeline begins with energy contribution, $E'_t$, from various sources (e.g. gas, coal). It then models pollutant dispersion (e.g. SO2, PM2.5) to receptors. Finally, it quantifies the resulting health impacts by monetary cost metrics ($/MWh).
• Figure 3: Trade-off between health impact prediction and fuel mix prediction accuracy across CISO, ERCO, and PJM regions. NAME refers to Normalized Mean Absolute Error. The top-left points of each curve correspond to the Fuel-mix-driven Opt, while the bottom-right points represent the Health-driven Opt.
• Figure 4: Simulation results of using different EV charging strategies based on health impact predictions in CISO, PJM and ERCO regions. With the provided prediction signals from the HealthPredictor, EV users can choose the optimal hours to charge their vehicles, achieving the greatest adverse health outcomes reduction compared to other charging strategies.
• Figure 5: Distribution of the energy generation mix by different fuel types in CISO, PJM, and ERCO.