SBIR Phase I: Smart Control Automation and Learning for Energy

这个小型企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是开发一个商业建筑管理系统，该系统使用模型预测控制为中小型商业建筑业主，以帮助他们灵活地管理日益复杂的能源法规和价格。该技术使用机器学习来自动化高级建筑控制的昂贵方面，并消除复杂性，挫折感，以及那些试图节省资金，满足代码，减少碳足迹并适应快速变化的能源价格的人员精简的建筑业主的费用。所提出的方法显着减少了设置时间，训练数据量以及该技术使用建筑热动力学收敛于准确模型和预测所需的计算时间。这些改进在不牺牲精度的情况下降低了控制器成本。该技术将简化从K-12学校开始的楼宇自动化、效率和基于模型的控制市场中代表性不足的细分市场的设置和实施过程。这种简化具有几个明显的社会和环境效益，包括：增加能源和需求费用节省，提高能源效率，改善环境足迹，增加建筑控制技术人员的就业机会，提高弹性，并为K-12家庭和社区提供额外的教育机会。SBIR第一阶段项目开发了一种能够进行建筑效率控制的技术。 该创新采用了一种基于受限深度学习工具的混合方法，该工具建立在建筑系统和建筑的物理知识基础上，从而利用采样数据，同时在建模和控制预测中产生物理一致的准确性。具体来说，项目团队希望集中在一个架构上，与最先进的控制方法相比，该架构可以更可靠，更准确地管理能源使用和居住者舒适度。该项目团队还旨在证明在目标建筑物中显著降低供暖、通风和空调（HVAC）驱动的高峰系统需求，同时将每栋建筑的仪器、劳动力和数据成本保持在目标市场可承受的成本。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to develop a commercial building management system that uses model predictive control for small to mid-sized commercial building owners to help them flexibly manage increasingly complex energy codes and prices. This technology uses machine learning to automate costly aspects of advanced building control, and eliminating complexity, frustration, and expense for leanly staffed building owners who are attempting to save money, meet code, reduce carbon footprint and adapt to rapidly changing energy prices. The proposed approach significantly reduces the setup time, the amount of training data, and the compute time needed for the technology to converge on accurate models and predictions using building thermal dynamics. These improvements reduce the controller costs without sacrificing accuracy. This technology will simplify the setup and implementation process for under-represented segments in the building automation, efficiency and model-based controls market starting with K-12 schools. This simplification has several distinct societal and environmental benefits including: increased energy and demand charge savings, increased energy efficiency, improved environmental footprint, increased job creation for building controls technicians, improved resiliency, and additional educational opportunities for K-12 families and communities.This SBIR Phase I project develops a technology capable of building efficiency control. The innovation employs a hybrid approach based on constrained deep learning tools that build on physical knowledge of building systems and architecture, thereby making use of sampling data while producing physics-consistent accuracy in modeling and control predictions. Specifically, the project team hopes to converge on an architecture that can more reliably and accurately manage energy use and occupant comfort compared to state of the art control approaches. The project team also aims to demonstrate a significant reduction in heating, ventilation and air-conditioning (HVAC)-driven peak system demand in target buildings while keeping instrumentation, labor, and data costs per building to an affordable cost for the target market.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.