Development of predictive emission monitoring system (PEMS)

预测排放监测系统(PEMS)根据已知的操作参数并结合关键工艺条件(例如，燃料消耗、环境条件等)来预测工业污染源的空气污染物排放量。拟议研究的目标是通过将最先进的机器学习技术，特别是深度学习体系结构应用于环境监测，监测燃烧来源的NOx、废气流量和温度，旨在将具有成本效益的环境监测技术引入加拿大工业。PEMS的资本成本估计为可比的连续排放监测系统(CEMS)的50%；此外，PEMS的运营和维护(O&M)成本预计约为CEMS成本的10%至20%。开发的产品将部署在赞助商在艾伯塔省的新空气监测设施中。尽管美国和欧洲的商业PEMS安装数量自1990年以来大幅增加，但PEMS在加拿大，特别是在艾伯塔省的实施和利用率仍然很低。主要原因和挑战包括1)行业意识2)软件许可，3)PEMS安装和应用的监管审批过程缺乏明确性，4)缺乏从一个设施到另一个设施的可转移性，以及5)有限的预测参数(变量)。本研究将(1)利用开源软件为PEMS开发新一代PEMS模型，以降低软件许可成本；(2)增加PEMS的可移植性。该模型可以直接用于多个设施，而不是特定于设施的模型；(3)包括所有监管要求的参数作为模型输出；因此，CEMS单元可以完全更换；(4)获得省(艾伯塔省)和联邦监管机构的批准，为在加拿大的行业中更广泛地安装PEMS铺平道路。

Predictive Emission Monitoring Systems (PEMS) predict the air pollutant emissions from industrial sources based on the known operating parameters in combination with key process conditions (e.g., fuel consumption, ambient conditions, etc.). The objective of the proposed research is to monitor NOx, exhaust flow, temperature from combustion sources by applying the state-of-the-art machine learning technologies, particularly deep learning architecture, to environmental monitoring and aims to bring cost-effective environmental monitoring technologies to Canadian industries. The capital cost for PEMS is estimated to be 50% of a comparable Continuous Emission Monitoring System (CEMS); furthermore, the operations & maintenance (O&M) costs for PEMS are expected to be approximately 10% to 20% of the cost of CEMS. The developed product will be deployed in the sponsor's new facility for air monitoring in Alberta. It can also be included as an option to third-party licensees of the sponsor's Hydrofraction technology.Although the number of commercial PEMS installations in the US and Europe has significantly increased since 1990, PEMS implementation and utilization in Canada, particularly in Alberta, remains low. The major reasons and challenges include 1) Industry awareness 2) Software licensing, 3) lack of clarities of the regulatory approval process for PEMS installation and application, 4) lack of transferability from one facility to another, and 5) limited predicted parameters (variables). The research will (1) Develop next generation PEMS model using open source software for PEMS to reduce the software licensing cost; (2) Increase PEMS' transferability. The model can be used on a number of facilities directly, instead of facility-specific model; (3) Include all regulatory-required parameters as model outputs; therefore, CEMS unit can be completely replaced; (4) Obtain provincial (Alberta) and federal regulatory agencies' approval to pave a way for a broader installation of PEMS in Canadian industries.