EVENT DOSSIER
The researchers proposed a hybrid framework that combines large language models (LLMs) with calibrated Irish dairy farm solar photovoltaic adoption agents (ABMs). This framework enhances the original technical and economic adoption mechanisms through bounded behavioral norms and structured scenario norms. Experimental results show that this method produces stable and economically reasonable behaviors under various policy settings, keeping photovoltaic adoption results within bounded and monotonic ranges. Compared with the corresponding logistic case, using LLM-assisted norms resulted in a maximum increase in dairy farm solar photovoltaic adoption rate of approximately 13%, without producing unstable or false saturation dynamics. The study confirms that LLM-assisted norms can be controlled, reproducible, and integrated into the calibrated energy ABM in a policy-relevant manner.
本文提出一种混合框架，将大语言模型（LLMs）与校准后的爱尔兰奶牛场太阳能光伏采用代理模型（ABM）结合。该框架保留原有技术经济采用机制，通过有界行为规范和有结构场景规范进行增强。实验结果显示，该方法在多种政策设置下产生稳定且经济合理的行为，使采用结果保持在有界和单调范围内，相比对应逻辑斯谛案例，行为采用率最高提升约 13%，且不产生不稳定或虚假的饱和动态。研究证实，LLM 辅助规范可受控、可复现且与政策相关地集成到校准能源 ABM 中。