Speeding Ticket: Unveiling the Energy and Emission Burden of AI-Accelerated Distributed and Decentralized Power Dispatch Models

TL;DR

This work tackles the environmental burden of AI-accelerated power dispatch in modern grids by comparing centralized, distributed, and decentralized neural-network surrogates on the IEEE 33-bus system. It replaces traditional iterative solvers with neural networks to achieve rapid online decisions, and uses Code Carbon to quantify energy use and carbon emissions over a week of inferences. Key findings show that distributed and decentralized configurations incur higher per-inference energy and emissions than centralized ones, with model size and system scalability affecting energy costs differently across setups. The study provides practical guidance for designing AI-enabled energy systems that improve operational efficiency without compromising ecological integrity, informing practitioners and policymakers on when to favor speed versus sustainability in grid dispatch.

Abstract

As the modern electrical grid shifts towards distributed systems, there is an increasing need for rapid decision-making tools. Artificial Intelligence (AI) and Machine Learning (ML) technologies are now pivotal in enhancing the efficiency of power dispatch operations, effectively overcoming the constraints of traditional optimization solvers with long computation times. However, this increased efficiency comes at a high environmental cost, escalating energy consumption and carbon emissions from computationally intensive AI/ML models. Despite their potential to transform power systems management, the environmental impact of these technologies often remains an overlooked aspect. This paper introduces the first comparison of energy demands across centralized, distributed, and decentralized ML-driven power dispatch models. We provide a detailed analysis of the energy and carbon footprint required for continuous operations on an IEEE 33 bus system, highlighting the critical trade-offs between operational efficiency and environmental sustainability. This study aims to guide future AI implementations in energy systems, ensuring they enhance not only efficiency but also prioritize ecological integrity.

Figures (4)

• Figure 1: Different power dispatch setups. In centralized decision-making, a central coordinator handles all computations and decisions, directing unidirectional communication to all nodes. Distributed decision-making involves local controllers coordinating with a central entity through bidirectional communication, while decentralized decision-making operates without a central coordinator, relying on peer-to-peer communication for independent, cooperative decisions.
• Figure 2: Energy consumption and carbon emission with varying model sizes, characterized by the total number of parameters of all agents' neural networks.
• Figure 3: Comparison of energy consumption and carbon emissions based on system scalability, indicated by the number of decision-making units or generators. Distributed and decentralized models exhibit less sensitivity to scalability changes in energy usage compared to centralized models.
• Figure 4: Model size increase as the system scalability increases. The rate of increase in model size for the distributed and decentralized models is slower than for the centralized model as system scalability increases.