摘要
危险气体探测器是炼化装置安全的重要保护层,合理的选址可大幅提高检测效率,降低泄漏事故损失。本文从削减气体泄漏风险的角度提出了一种危险气体探测器选址的定量优化方法。基于多种影响危险气体泄漏扩散的外界因素,确定了危险气体泄漏场景发生概率、泄漏后果以及泄漏风险的定量表征方法。以累积泄漏风险最小化为优化目标,针对气体探测器的工作失效模式和冗余结构等内部因素,建立了考虑气体探测器不可用性、表决逻辑和条件风险值约束的选址优化模型,并提出求解优化方案的粒子群算法。以某柴油加氢装置为例进行算例研究,结果表明优化方案可有效削减气体泄漏风险,且效果明显优于原布设方案。后继还将开展优化模型的验证试验及多目标优化研究,以进一步提高优化模型的实用性。
As an important protective layer of petroleum refinery installations,gas detectors can greatly improve the detection efficiency and reduce the loss if they are placed in a reasonable position.From the perspective of minimizing leakage risks,a quantitative optimization method of gas detectors placement was proposed.Several external factors of leakage diffusion process were quantified to calculate the scenarios occurrence probability,leakage consequence and risks.According to the failure mode and redundancy structure,a minimal risk p-median problems(MRPMP-UVCVaR)optimization model was demonstrated,as well as unavailability,voting effects and Conditional-Value-at-Risk(CVaR)was considered.Particle swarm optimization(PSO)was used to solve the model.The applicability of the methodology was shown in a case study and the results were compared with the original scheme,improving the design of detection systems on a risk-oriented basis.Future work including the research of the experimental validation and multi-objective optimization will be taken to improve the practicability of the optimization model.
作者
章博
慕超
王志刚
王彦富
ZHANG Bo;MU Chao;WANG Zhigang;WANG Yanfu(College of Mechanical and Electronic Engineering in China University of Petroleum,Qingdao 266580,Shandong,China;Center for Offshore Equipment and Safety Technology in China University of Petroleum,Qingdao 266580,Shandong,China;Guangxi Natural Gas Pipeline Co.,Ltd.,Nanning 530000,Guangxi,China)
出处
《化工进展》
EI
CAS
CSCD
北大核心
2017年第8期2809-2815,共7页
Chemical Industry and Engineering Progress
基金
山东省自然科学基金(ZR2016EEM27)
国家自然科学基金(51409260)项目
关键词
气体探测器
炼化装置
泄漏风险
优化模型
求解算法
gas detectors
petroleum refinery installations
leakage risks
optimization model
solution algorithm
