Towards a Health-Based Power Grid Optimization in the Artificial Intelligence Era

TL;DR

The paper argues that health impacts from power-plant emissions, notably $PM_{2.5}$ and $HAP$, should be integrated into grid optimization alongside $CO_2$ and energy efficiency. It proposes a formal dynamic optimization framework that minimizes long-run hospitalizations due to $HAP$ under average $CO_2$ and $HAP$ emission caps across spatial regions and time. Solution approaches discussed include Lyapunov optimization and model-based reinforcement learning, with an inner dispatch loop to allocate energy production across plants. A toy example demonstrates that emission minimization and health minimization are not equivalent, motivating a health-based paradigm to support policymaking under carbon caps while protecting public health.

Abstract

The electric power sector is one of the largest contributors to greenhouse gas emissions in the world. In recent years, there has been an unprecedented increase in electricity demand driven by the so-called Artificial Intelligence (AI) revolution. Although AI has and will continue to have a transformative impact, its environmental and health impacts are often overlooked. The standard approach to power grid optimization aims to minimize CO$_2$ emissions. In this paper, we propose a new holistic paradigm. Our proposed optimization directly targets the minimization of adverse health outcomes under energy efficiency and emission constraints. We show the first example of an optimal fuel mix allocation problem aiming to minimize the average number of adverse health effects resulting from exposure to hazardous air pollutants with constraints on the average and marginal emissions. We argue that this new health-based power grid optimization is essential to promote truly sustainable technological advances that align both with global climate goals and public health priorities.

Figures (1)

• Figure 1: A toy example showing that minimizing emissions is not the same as minimizing adverse health outcomes, and the need for health-based power grid optimization. Assume a grid world where each square represents an area with either a power plant (clean, hybrid, or dirty) or a population. Clean, hybrid, and dirty plants can produce, with the available supply, 20, 40, and 40 units of energy, respectively. Assume the energy demand is 100, and the weather-dependent emission transport directions are as depicted by the arrows. In A, energy demand is allocated to power plants by minimizing emissions, therefore clean and hybrid plants are used. However, emissions reach populated areas, causing significant adverse health outcomes. In B, energy demand is allocated to power plants to minimize health impacts, therefore clean, hybrid, and dirty plants are used while avoiding emissions in populated areas. However, the overall emissions are high. In C, our framework, energy is allocated to power plants to minimize adverse health outcomes while constraining emissions, achieving lower emissions than B and lower adverse health outcomes than A.