Volume 69

An Explainable Artificial Intelligence-Driven Framework for Automated System Dynamics Modeling in Zero-Carbon Industrial Parks Yanchen Wu, Boyu Guo, Liang Zhao

Abstract

Zero-carbon industrial parks are important carriers for the transition toward carbon neutrality, and system dynamics can support the analysis of energy-carbon emission evolution, scenario simulation, and governance pathway design. However, conventional modeling relies heavily on expert knowledge and faces problems such as long development cycles, poor reusability, difficult structural review, and limited software compatibility. This study proposes an explainable AI-driven framework for automated system dynamics modeling. The framework adopts a six-layer architecture, uses SD-IR as the core representation, and establishes a three-role collaborative mechanism. Through structured output, rule-based validation, Vensim-compatible export, and a human review loop, it generates verifiable, compilable, and simulatable model drafts. The results show that the framework can generate multi-system, multi-variable, and multi-scenario models. After repair, the model passes unit checking and produces scenario rankings consistent with the manual model, thereby improving modeling efficiency and supporting calibration and decision-making.

Keywords zero-carbon industrial park, system dynamics, explainable artificial intelligence, automated modeling, large language model