标题： BuildEvo：通过LLM驱动的进化设计建筑能耗预测启发式方法 显示英文标题

标题： BuildEvo: Designing Building Energy Consumption Forecasting Heuristics via LLM-driven Evolution

摘要： 准确的建筑能耗预测至关重要，但传统的启发式方法往往缺乏精度，而先进的模型可能不透明，并且通过忽视物理原理而在泛化上遇到困难。 本文介绍了 BuildEvo，一种新颖的框架，利用大型语言模型（LLMs）自动设计有效且可解释的能耗预测启发式方法。 在进化过程中，BuildEvo通过系统地结合来自建筑特征和运行数据（例如，来自建筑数据基因组项目 2）的物理洞察力，指导LLMs构建和改进启发式方法。 评估结果显示，BuildEvo在基准测试中达到了最先进的性能，提供了更好的泛化能力和透明的预测逻辑。 这项工作推进了稳健、基于物理的启发式方法的自动化设计，促进了复杂能源系统中可信模型的发展。 显示英文摘要

摘要： Accurate building energy forecasting is essential, yet traditional heuristics often lack precision, while advanced models can be opaque and struggle with generalization by neglecting physical principles. This paper introduces BuildEvo, a novel framework that uses Large Language Models (LLMs) to automatically design effective and interpretable energy prediction heuristics. Within an evolutionary process, BuildEvo guides LLMs to construct and enhance heuristics by systematically incorporating physical insights from building characteristics and operational data (e.g., from the Building Data Genome Project 2). Evaluations show BuildEvo achieves state-of-the-art performance on benchmarks, offering improved generalization and transparent prediction logic. This work advances the automated design of robust, physically grounded heuristics, promoting trustworthy models for complex energy systems.