SBIR Phase I: Human-Centered, Augmented Intelligence Software for Water and Wastewater

这项小企业创新研究（SBIR）第一阶段项目的更广泛影响/商业潜力将来自增强智能软件的开发，该软件可以改善水和废水系统的规划和运营决策。对于水务公司来说，最具挑战性的问题包括解决老化的基础设施，使复杂的系统适应不断变化的法规，以及解决环境变化的影响。它们源于与建筑和自然环境相互连接的基础设施网络。目前针对水的人工智能和机器学习解决方案都是针对特定用例进行狭义定义的。拟议的智能软件将通过来自模拟器和数据驱动模型的混合模型的无缝组合来实现水管理的革命性变化，以克服数据和信息孤岛，使决策者能够将数据集成到系统模型中，从而在降低客户成本的同时提高弹性。这些改进可以显著减少每年约47亿美元的水/废水能源支出，500亿美元的联合下水道系统项目，以及高达1万亿美元的老化基础设施需求。该小企业创新研究（SBIR）第一阶段项目将开发多种方法，将多保真仿真模型和数据驱动模型相结合，以支持水和废水系统的长期规划需求和实时操作决策支持。将元建模技术嵌入物理系统的理解从高保真物理模拟器到低保真模型将进行评估。精度和运行时的权衡将评估多个降阶方法（如线性和非线性方程，基于投影的方法），以实现更有效的优化大型解决方案空间。领域应用包括一维流动的圣维南方程的降阶版本，以及二级废水处理中生物，物理和化学过程的分析解决方案。该项目将评估多种机器学习方法，包括深度神经网络、强化学习、随机森林、支持向量机和增强学习算法，以检测观察数据中的模式，为操作实时决策提供近期预测能力。专家启发技术将用于量化主观决策标准的人类专业知识，将有价值的隐性人类知识整合到决策过程中。将评估可选混合建模工作流的元分析，以确定优化复杂规划挑战的计算有效途径。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识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 will result from development of augmented intelligence software improving planning and operational decisions for water and wastewater systems. The most challenging issues for water utilities include addressing aging infrastructure, adapting complex systems to changing regulations, and addressing the impacts of environmental change. They arise from interconnected infrastructure networks that interface with both the built and natural environment. Current artificial intelligence and machine learning solutions for water are narrowly defined for specific use cases. The proposed intelligence software will enable transformative changes in water management by seamless composition of hybrid models from simulators and data-driven models to overcome data and information silos, enabling decision-makers to integrate data in a system model that increases resilience at reduced customer costs. These improvements can lead to significant reductions in the roughly $4.7 B annual energy spend for water/wastewater, $50 B in combined sewer system programs, and up to $1 T in aging infrastructure needs.This Small Business Innovation Research (SBIR) Phase I project will develop methods for combining multi-fidelity simulation models and data-driven models to support decision-making for both long-term planning needs and real-time operational decision support for water and wastewater systems. Meta-modeling techniques for embedding physical system understanding from high-fidelity physics-based simulators to low-fidelity models will be evaluated. Accuracy and runtime tradeoffs will be evaluated for multiple reduced-order methods (e.g. linear and non-linear equations, projection-based methods) to enable more efficient optimization of large solution spaces. Domain applications include reduced-order versions of the St Venant equations for one-dimensional flow, and analytical solutions of biological, physical, and chemical processes in secondary wastewater treatment. The project will evaluate multiple machine learning methods, including deep neural networks, reinforcement learning, random forest, support vector machines, and boosted learning algorithms, to detect patterns in observed data for near-term predictive power toward operational real-time decisions. Expert elicitation techniques will be used to quantify human expertise for subjective decision criteria, integrating valuable tacit human knowledge into the decision process. Meta-analysis of alternative hybrid modeling workflows will be evaluated to identify computationally efficient pathways to optimize complex planning challenges.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.