Scholarlink

<p>The rapid proliferation of Artificial Intelligence (AI) and Electric Vehicles (EVs) is reshaping electricity demand patterns, with important implications for grid stability and long-term energy planning. Conventional forecasting approaches, largely based on historical trends and macroeconomic indicators, have limited ability to represent structurally distinct and technology-driven demand growth. This study proposes a modular electricity demand forecasting framework that explicitly separates baseline demand from AI- and EV-driven components. Within this structure, the ARIMAX model is used to estimate baseline demand from macroeconomic variables, while independent modules capture additional demand associated with AI data centers and EV deployment. Using the United States as a case study, annual data from 2015 to 2050 were analyzed. ARIMA, ARIMAX, and Prophet models were evaluated through observed-data validation and long-term scenario consistency analysis relative to publicly available outlook trajectories. In addition, a diffusion-based probabilistic module was implemented to characterize long-horizon demand uncertainty. Results indicate that incorporating AI and EV components improves scenario consistency relative to baseline configurations. The proposed framework provides a transparent and adaptable structure for integrating emerging demand drivers into medium- to long-term electricity planning under technological disruption.</p> Graphical abstract <p>Modular forecasting framework integrating economic and demographic drivers with technology-induced demand drivers (AI data centers, EV) using ARIMAX, Prophet, and Diffusion models.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Framework for Electricity Demand Forecasting under AI and EV Expansion: Comparative Analysis of ARIMAX, Prophet, and Diffusion Models

  • 期刊论文

  • 发表日期：

    2026年4月15日

查看全文
  • Wonyoung Jeong,
  • Dongil Shin

摘要

The rapid proliferation of Artificial Intelligence (AI) and Electric Vehicles (EVs) is reshaping electricity demand patterns, with important implications for grid stability and long-term energy planning. Conventional forecasting approaches, largely based on historical trends and macroeconomic indicators, have limited ability to represent structurally distinct and technology-driven demand growth. This study proposes a modular electricity demand forecasting framework that explicitly separates baseline demand from AI- and EV-driven components. Within this structure, the ARIMAX model is used to estimate baseline demand from macroeconomic variables, while independent modules capture additional demand associated with AI data centers and EV deployment. Using the United States as a case study, annual data from 2015 to 2050 were analyzed. ARIMA, ARIMAX, and Prophet models were evaluated through observed-data validation and long-term scenario consistency analysis relative to publicly available outlook trajectories. In addition, a diffusion-based probabilistic module was implemented to characterize long-horizon demand uncertainty. Results indicate that incorporating AI and EV components improves scenario consistency relative to baseline configurations. The proposed framework provides a transparent and adaptable structure for integrating emerging demand drivers into medium- to long-term electricity planning under technological disruption.

Graphical abstract

Modular forecasting framework integrating economic and demographic drivers with technology-induced demand drivers (AI data centers, EV) using ARIMAX, Prophet, and Diffusion models.