Physics-informed Machine Learning Modelling for Multi-scale Building Energy Systems with Enhanced Accuracy and Interpretability

具有更高准确性和可解释性的多尺度建筑能源系统的基于物理的机器学习建模

• 批准号: 2725680
• 金额: --
• 依托单位: University College London
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2725680
• 关键词: Physics; informed; Machine; Learning; Modelling

Buildings significantly contribute to global energy consumption and carbon emissions and play an important role in accelerating the transformation of a low-carbon energy system. Accurate and transparent modelling is essential for developing energy-flexible and resilient buildings, characterising building demand flexibility, and comprehensively assessing different control strategies before implementation. In practice, physics-based building modelling requires many inputs, some uncertain; this often leads to overly simplistic or inaccurate approaches, especially in large-scale multi-building contexts. Pure data-driven modelling approaches are attractive but lack interpretability, and model predictions are not fully explainable or trustable.

建筑物是全球能源消耗和碳排放的重要组成部分，在加快低碳能源体系转型中发挥着重要作用。准确和透明的建模对于开发能源灵活性和弹性建筑，描述建筑需求灵活性以及在实施前全面评估不同的控制策略至关重要。在实践中，基于物理的建筑建模需要许多输入，其中一些是不确定的;这通常会导致过于简单或不准确的方法，特别是在大规模的多建筑环境中。纯数据驱动的建模方法很有吸引力，但缺乏可解释性，模型预测不完全可解释或可信。