Interpretable data-driven building energy analytics

在过去的二十年中，机器学习在建筑研究中得到了发展和测试，得益于数据可用性的增加、强大且价格合理的计算资源和先进的算法，并证明了其提高建筑性能的潜力。所使用的方法将解决新建筑和改造中的运营和隐含能源和碳。这个博士项目的主要目标是为数据驱动的能源建模创建一个可解释的“数字孪生”方法。“数字孪生”是物理对象、过程或服务的数字副本，可以克服传统基于仿真的工程方法的局限性。虽然模拟和数字孪生都是对象的虚拟表示，但数字孪生可以验证物理对象、流程或服务在实时和真实世界条件下的执行情况。此外，可以设计可解释的数据驱动方法，以“人在循环方法”将人类和机器智能结合起来，其基本目标是加速向建筑库存的净零过渡。另一个主要目标是开发“数字孪生”方法，使其可以应用于建筑生命周期的各个阶段，从早期设计到运营，并且它是可扩展的，从整个建筑分析向下到单个建筑技术，向上到建筑物集群。

Over the last two decades, machine learning has been developed and tested in building research, aided by increased data availability, powerful and affordable computing resources, and advanced algorithms, and has demonstrated its potential to improve building performance. The methods used will address both operational and embodied energy and carbon in new construction and retrofit. The primary goal of this PhD project will be to create an interpretable "digital twin" approach for data-driven energy modelling. A "digital twin" is a digital replica of a physical object, process, or service that can overcome the limitations of traditional simulation-based engineering approaches. While simulations and digital twins are both virtual representations of objects, digital twins can verify how a physical object, process or service performs in real time and in real world conditions. Furthermore, interpretable data-driven methods can be designed to combine human and machine intelligence in a "human-in-the-loop approach" whose fundamental goal is to accelerate the transition to Net Zero of the building stock. The other primary goal will be to develop the "digital twin" approach so that it can be applied at various stages of the building life cycle, from early design to operation, and that it is scalable, from whole-building analysis downwards to individual building technologies and upward to clusters of buildings.