SBIR Phase I: Real-Time Decision Making Software for Wastewater Treatment Operators

这个小企业创新研究（SBIR）项目的更广泛的影响/商业潜力是开发新一代机器学习/人工智能工具，以提高废水处理系统的效率和有效性。在美国，超过1600亿美元用于下水道费用，对于许多用户来说，每年的成本增加超过5%。即使在废水处理上花费了大量资源，每年仍有30亿至100亿加仑未经处理的污水从美国废水处理厂释放出来。开发利用机器学习和人工智能来更好地管理复杂的生物和化学过程的技术，不仅会对910亿美元的废水处理系统控制市场产生重大影响，还会对其他依赖生物化学的行业产生重大影响。除了减少与废水处理相关的社会财政负担外，这项技术还将提高基础设施的可持续性，并将限制人类活动对环境的影响。污水处理运营商将能够利用该技术的实时决策软件来大幅降低市政设施的运营成本，并减少因违规和溢出而造成的环境污染。SBIR第一期项目旨在开发实时软件，以帮助污水处理运营商进行决策，从而提高效率和效果。现有的用于控制和监测的商业模拟解决方案不能准确地反映实际的处理厂行为，不能对生物过程进行建模，不需要使用大量的配置，并且不能快速响应工厂性能的变化。该项目通过使用机器学习技术将废水处理厂的生物成分与历史数据联系起来，改进了当前的方法。第一阶段的研究将集中于开发：1）允许准确模型输入的进水流量/组成模型; 2）结合物理过程和机器学习生物过程模块的全尺寸混合模型，能够准确预测工厂流出物流量和质量; 3）管理模型过程的软件平台。该技术方法将侧重于平衡将大规模基因组数据作为非线性处理过程的一部分的复杂性，同时确保高准确性和保持模型稳定性。预期的技术成果将为废物运营商提供一个软件产品，该产品利用历史和当前工厂数据实时开发全尺寸处理厂模型，从而显著改善运营决策。该奖项反映了NSF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) project is the development of a new generation of machine learning/artificial intelligence tools for improving the efficiency and effectiveness of wastewater treatment systems. Over $160 billion is spent on sewer fees in the U.S., with year-to-year costs increasing by over 5% for many users. Even with the significant resources spent on wastewater treatment, 3 to 10 billion gallons of untreated sewage are still released from US wastewater treatment plants each year. Development of technologies leveraging machine learning and artificial intelligence to better manage complex biological and chemical processes will not only have a major impact on the $91 billion wastewater treatment system control market, but on other biochemical-dependent industries as well. In addition to reducing the societal financial burden associated with wastewater treatment, this technology will improve the sustainability of the infrastructure and will limit the environmental impact of human activities. Wastewater operators will be able to utilize this technology's real-time decision-making software to significantly reduce their municipal facility operating costs and decrease environmental pollution caused by non-compliance and overflows.This SBIR Phase I project proposes to develop real-time software for assisting wastewater treatment operators with decision-making for improved efficiency and effectiveness. Existing commercial simulation solutions for control and monitoring do not accurately reflect actual treatment plant behavior, do not model biological processes, do not require extensive configuration to be used, and do not respond rapidly to changes in plant performance. This project improves upon current approaches by linking biological components of the wastewater treatment plant with historical data using machine learning techniques. Phase I research will focus on development of: 1) an influent flow/composition model allowing accurate model inputs; 2) a full-scale hybrid model combining physical process and machine-learning bioprocess modules able to accurately predict plant effluent flow and quality, and; 3) a software platform to manage the model processes. The technical approach will focus on balancing the complexity of incorporating large-scale genomic data as part of a non-linear treatment process while ensuring high accuracy and maintaining model stability. Anticipated technical results will provide waste operators a software product that develops full-scale treatment plant models in real-time, using historical and current plant data, enabling significantly improved operational decision-making.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.