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

Claudio Battiloro,Gianluca Guidi,Falco J. Bargagli-Stoffi,Francesca Dominici

2024-10-12

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.

Signal Processing,Systems and Control,Applications

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the issue of ignoring health impacts in current power system optimization methods. Specifically:

1. **Limitations of traditional optimization methods**:
- Current power system optimization mainly focuses on reducing carbon emissions (emission - based optimization) or improving energy efficiency (efficiency - based optimization), but these methods do not directly consider the harmful health impacts.
- Although reducing greenhouse gas emissions is beneficial to the environment, it is not equivalent to reducing the adverse health consequences caused by harmful air pollutants (HAPs).

2. **Increased demand in the era of artificial intelligence**:
- Driven by the artificial intelligence (AI) revolution, the power demand of data centers has increased significantly, which has put great pressure on power resources and grid infrastructure.
- The International Energy Agency (IEA) estimates that data centers consumed 240 - 340 terawatt - hours (TWh) of electricity in 2022 and are expected to double by 2026.

3. **Proposed new paradigm**:
- The paper proposes a new health - based power system optimization paradigm, aiming to directly minimize adverse health consequences while meeting energy efficiency and emission limitations.
- This new method not only considers carbon dioxide (CO₂) emissions but also pays special attention to the negative impacts of fine particulate matter (PM₂.₅) and other harmful air pollutants (HAPs) on human health.

4. **Specific objectives**:
- Minimize the average number of hospitalizations caused by exposure to harmful air pollutants.
- Set constraints on the average emissions during the optimization process to ensure that the set CO₂ and HAPs emission thresholds are not exceeded.

By introducing health factors, this research hopes to provide a more comprehensive power system optimization framework to promote truly sustainable technological progress, which is in line with global climate goals and gives priority to public health.

### Overview of the mathematical model

To achieve the above - mentioned goals, the paper proposes a dynamic fuel - mix allocation optimization problem, which is formulated as the following optimization problem:

\[
\min_{\{w_i(t)\}_{i = 1}^S} \lim_{t\rightarrow\infty} \frac{1}{t} \sum_{\tau = 0}^{t - 1} E\left(\sum_{i = 1}^S h_i(\tau)\right)
\]

Subject to the following constraints:

\[
\begin{aligned}
&\text{C1:} \quad \lim_{t\rightarrow\infty} \frac{1}{t} \sum_{\tau = 0}^{t - 1} E\left(\sum_{i = 1}^S e_{\text{CO2}}(i, w_i(\tau), P_i(\tau))\right) \leq C_{\text{CO2}} \\
&\text{C2:} \quad \lim_{t\rightarrow\infty} \frac{1}{t} \sum_{\tau = 0}^{t - 1} E\left(\sum_{i = 1}^S e_{\text{HAP}}(i, w_i(\tau), \{W_i(\tau)\}_{i = 1}^S, P_i(\tau))\right) \leq C_{\text{HAP}} \\
&\text{C3:} \quad 0\leq w_{i,n}(t)\leq 1, \quad i = 1,\ldots, S, \quad n = 1,\ldots, N \\
&\text{C4:} \quad w_{i,n}(t) P_i(t)\leq p_n(i, t), \quad i = 1,\ldots, S, \quad n = 1,\ldots, N \\
&\text{C5:} \quad \sum_{n = 1}^N w_{i,n}(t)= 1, \quad i = 1,\ldots, S
\end{aligned}
\]

where:
- \( w_i(t)=[w_{i,1}(t)