面向工业大系统安全高效运行的报警设计与消除方法及应用

Alarm systems play a vital role for safe operation and high efficiency of modern large-scale industrial systems. However, there are several major drawbacks in the existing alarm systems, namely, configuration of alarm variables lacks scientific basis, alarm trippoint design is not associated with related variables, nuisance alarms are overwhelming and difficult to identify; and alarms may propagate to many places so that it is difficult to find their root causes. As a result, existing alarm systems often generate more alarms than they can be handled promptly and face great difficulties in the management of alarms, which often prevents them from timely reporting deviations from normal operating regions through alarms, and removing abnormal situations associated with alarms, resulting in low production efficiency, numerous safety issues and even severe disasters. Aiming at large-scale industrial processes such as those in power and petrochemical industries and targeting at the entire lifetime of alarm monitoring in terms of alarm configuration, alarm generation, and alarm removal, we shall study in this project configuration methods of alarm variables, alarm design methods, alarm removal methods, and experimental application of such alarm design and removal theories. Our goal is to solve some fundamental key scientific problems, including the dynamic modeling for high-dimensional normal/efficient operating zones and the design of dynamic alarm trippoints associated with related process variables. Thus, fundamental solutions and key techniques can be provided to overcome the major drawbacks of existing alarm systems and build a theoretical framework of alarm design and removal for safe and efficient operation of large-scale industrial systems.

报警系统对保障现代化工业大系统的安全生产与高效运行发挥着至关重要的作用，然而由于报警变量配置缺乏科学性、报警阈值设计缺乏与相关变量的关联、干扰报警多且判断难、报警传播广且判断其根源难度大等原因，使得现有报警系统普遍存在“报警过多、处理困难”的突出问题，经常导致无法通过报警及时发现生产过程偏离正常工作范围，无法及时排除异常状况、消除报警，不仅严重影响生产效益，更带来大量生产安全问题甚至灾难性事故。本项目以大规模工业生产过程为对象（如电力和石化），面向报警配置、生成、消除的完整生命周期，研究报警变量配置方法、报警器设计方法与报警消除方法，开展报警设计与消除理论的应用验证研究，解决研究中遇到的正常/高效工作区的高维动态空间建模及动态关联报警阈值设计等若干深层次关键科学问题，为解决现有报警系统的突出问题提供基础性的解决方案与关键技术，初步建立面向工业大系统安全与高效运行的报警设计与消除理论的框架。

报警系统对保障现代化工业大系统的安全生产与高效运行发挥着至关重要的作用；然而，现有报警系统普遍存在“报警过多、处理困难”的突出问题，经常导致无法通过报警及时发现生产过程偏离正常工作范围和消除报警，不仅严重影响生产效益，更带来大量生产安全问题甚至灾难性事故。.本项目解决了正常/高效工作区的高维动态空间建模及动态关联报警阈值设计等若干深层次关键科学问题，提出了报警设计与消除等多项关键技术，在电力与石化企业的生产过程中开展了研究结果的应用验证研究。项目取得的主要结果是：.1. 提出了综合工艺知识和统计规律的监控变量运行状况推断方法，划分了处于正常或异常状况的数据段，突破了工业现场长期依靠人工识别正确报警与干扰报警的难题。.2. 建立了基于正常/高效工作区的高维空间模型，设计了当前工作点下过程变量的动态关联报警阈值，增加了生产过程处于高效安全运行状态的时间。.3. 丰富了报警死区、延迟器等多种单变量报警系统优化设计方法，优化了误报警率、漏报警率等性能指标，大幅度减少了多种类型干扰报警的出现次数。.4. 独创了报警泛滥的检测、相似性分析和预测方法，为运行人员在严重紧急异常状况下及时发现、分析和处理报警泛滥现象提供了技术手段。.5. 研发了具有自主知识产权的火电机组智能报警系统等工业应用系统，在大型火电机组、石化装置进行了应用，大幅度提高了报警监控的有效性和准确性。.项目成果主要包括SCI论文69篇、授权发明专利15项、学术专著2部、EI会议论文19篇、中文核心期刊论文15篇。培养博士后5人、博士生20人、硕士生37人。在IFAC2017等国际会议组织了5个相关研究方向的国际学术专题研讨会（workshop）。获得2018年中国自动化学会科技进步一等奖1次、中国过程控制会议张钟俊优秀论文奖1次和优秀学生论文奖1次等科研奖励。