Generative models for energy modelling and management: a structured framework for technology adoption in energy systems
针对现代能源系统复杂性，提出GM-EMM框架，通过分类生成方法并评估六种核心技术，为能源建模与管理提供结构化技术采纳路线图。
The growing complexity of modern energy systems demands intelligent, adaptable, and theoretically grounded approaches to modelling and management. Generative Artificial Intelligence (Gen-AI) and Deep Learning Models (DLMs) configured for generative tasks offer significant potential for scenario simulation and rare-event modelling; however, their adoption remains constrained by conceptual ambiguity and the absence of structured, theory-informed frameworks. To address this gap, this study proposes the Generative Models for Energy Modelling and Management (GM-EMM) framework. By establishing a functional taxonomy of generative approaches, the research addresses critical analytical challenges, including counterfactual reasoning and data scarcity in decentralised energy systems. A multi-method research design was employed, integrating a scoping review, expert elicitation via the Fuzzy Delphi method, and non-parametric statistical prioritisation using Friedman and Nemenyi tests to evaluate six core technologies (RNNs, GANs, GNNs, TNNs, VAEs, and CNNs). Based on expert consensus and statistical differentiation, technologies were positioned across four analytically defined phases: Deployment, Validation, Exploration, and Monitoring. To enhance practical relevance, an interpretive validation stage with industry experts confirmed the roadmap’s coherence and decision-support value.