An extensible framework for energy community embedding business models and AI-generated code snippets

This paper presents a modular and extensible framework for simulating the operation of Energy Communities (ECs) and their participation in Local Energy Markets (LEMs). The framework integrates agent-based modelling, weather-driven asset simulation and flexible market-clearing mechanisms to support high-resolution techno-economic analysis. It introduces three main contributions: (i) a heuristic, data-driven algorithm for estimating household consumption under partial observability, (ii) the Dynamically Adjusted Uniform Price (DAUP), a new pricing mechanism that increases traded volumes through iterative bid adjustment and (iii) GenECodes , a generative module powered by Large Language Models that automatically produces simulation-ready asset and market models using a three-layer architecture. Simulations based on a Romanian EC show high accuracy of the asset models, significant improvements in traded energy and self-consumption using DAUP, and seamless integration of new assets such as heat pumps and electric vehicles via GenECodes . The framework enables rapid prototyping of emerging business models and supports the digital transition of decentralized energy systems. • Modular agent-based framework for simulating communities in local markets • Data-driven algorithm estimates household loads under partial observability • Pricing mechanism boosts self-consumption and reduces grid dependence • GenECodes uses LLMs to auto-generate simulation-ready models • Simulations show improved accuracy, market gains and seamless asset integration