Deep Transfer Learning from Data for Operational Excellence in Refineries

化工流程行业正在采用使用人工智能的数字技术来提高效率。新的大数据分析时代(由使用智能数字测量设备的数据爆炸式增长、云计算提高的数据存储容量、强大的硬件和软件技术以及通信平台推动)正在成为将海量数据转化为更好的运营和商业决策的新征程。大多数制造设施，包括对加拿大经济福祉至关重要的石油和天然气行业，目前正在经历数字化转型，以保持全球竞争力。为了通过这场数字革命达到下一个里程碑，政府、学术界和行业需要共同关注创造价值的创新，以及培养训练有素的具有数字专业知识的劳动力。 使用数据科学，可以智能、可靠、准确地关联大量过程数据。机器学习和深度学习算法能够自动从数据中收集见解并做出预测，并提供以极高的精度查明过程扰动的根本原因的方法，并在过程不稳定和故障有机会影响生产之前预测它们。西部大学和帝国石油公司的这一合作项目旨在应对数字技术对人员、工艺安全、实施和具有相关技能的训练有素的人员的可用性带来的挑战。拟议的研究计划将在开发创新的数据驱动建模方法方面带来新的见解，以便更好地控制和了解(A)炼油厂的脱盐操作及其优化操作，以及(B)加氢处理操作单元，目的是优化整个炼油厂单元。结果将使帝国石油公司能够有效地改善他们的炼油厂运营，以实现可持续性、安全性和盈利能力。该项目的关键成果是开发新的知识、工具和训练有素的合格人员，以推动数字技术和催化经济增长。

Chemical Process industries are adopting digital technologies using artificial intelligence for improved efficiency. The new 'big data analytics' era (driven by the explosion of data using smart digital measurement devices, improved data storage capacity due to cloud computing, powerful hardware and software technology and communication platforms) is emerging as the new journey to turn voluminous data into insights for better operational and business decisions. Most manufacturing facilities, including the Oil and Gas industry, which is paramount to the well-being of the Canadian economy, are currently experiencing a digital transformation to stay globally competitive. To reach the next milestone with this digital revolution, government, academia, and industry need to collaboratively focus on innovation for value generation and development of well-trained workforce with digital expertise. Using data science, vast process data can be intelligently correlated reliably and accurately. Machine learning and deep learning algorithms are capable of automatically gathering insights from data and making predictions and provide means to pinpoint the root cause of process disturbances with extreme accuracy, and predict process instabilities and failures before they have the chance to affect production. This collaborative project between Western University and Imperial Oil aims to address the challenges of digital technology implications on personnel, process safety, implementation and availability of trained personnel with relevant skills. The proposed research program will lead to new insights in developing innovative data-driven modeling approaches for better control and understanding of (a) de-salter operation at refinery and in optimizing its operation, and (b) hydro-processing operating units with the intention of optimization of the entire refinery unit. Results will allow Imperial Oil to effectively improve their refinery operations for sustainability, safety, and profitability. The key deliverable of the project are development of new knowledge, tools, and highly trained, qualified personnel to advance digital technologies and catalyze economic growth.