基于机理模型的精馏塔DCS报警优化研究与应用

发布时间：2018-03-05 00:26

本文选题：报警管理　切入点：小波分析　出处：《青岛科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：报警系统构成了大型现代化工装置操作界面的一个重要组成部分,在预防、控制和减轻异常情况方面具有重要的作用。因此,报警的优化以及报警处理策略对化工装置的正常运行具有十分重要的意义。精馏塔是化工企业生产过程中应用广泛的传质设备,在化工行业有着很重要的地位。精馏塔的安全、平稳运行己经成为了化工生产的重要环节,也直接关系到化工企业的经济效益。本文针对精馏塔DCS报警信息冗余问题,进行了报警优化以及报警根原因分析研究。本文首先对报警系统关键性能指标进行了统计分析,并对报警系统的性能进行了评估,为后续报警系统优化做好了准备。针对重复报警,在考虑死区的基础上,引入了基于数据过滤的报警限优化设计。数据过滤的引入有效地解决了过程数据受噪声干扰的不足,增强了鲁棒性。死区和数据过滤的结合,有效地解决了重复报警的问题。但是,仅根据测量信号进行报警优化是十分有限的,只有找到报警根源才能彻底解决报警的问题。故障诊断策略也是处理报警的一种有效方法,只要将故障消除,报警也会随之消失。但是,直接将化工机理模型用于故障诊断还存在计算量大、收敛性不好的问题。因此,本文研究了基于双层法模型的故障检测和诊断方法。使用非线性模型监控精馏过程和识别大的测量值偏差时的异常情况,通过基于线性模型的线性最小二乘法估计精馏塔内部的故障参数。为了进一步缩小模型计算范围,还运用希梅尔布劳算法对精馏系统进行分解,将高维的、不易求解的问题分解成若干低维的子问题。然后对分解后的小系统采用基于中值滤波和提升小波分析去噪的方法对数据进行去噪处理,增强了报警优化的鲁棒性。最后,将上述报警优化策略应用到TEP仿真流程的汽提塔中。案例研究结果表明,双层诊断结构有效地获得了故障参数的变化情况,比单纯的非线性模型结构更有效。还给出了因物流C的压差推动力发生阶跃下降导致的故障7时的塔底进料损失参数的变化情况。结果表明,该方法不仅可以准确地给出故障原因还可以大大缩短报警诊断的时间。
[Abstract]:The alarm system constitutes an important part of the operating interface of large-scale modern chemical plants and plays an important role in preventing, controlling and mitigating anomalies. The optimization of alarm and the strategy of alarm processing are very important to the normal operation of chemical plant. Distillation tower is a mass transfer equipment widely used in the production process of chemical enterprises and plays an important role in the chemical industry. Smooth operation has become an important part of chemical production, and also directly related to the economic benefits of chemical enterprises. This paper aims at the redundancy of DCS alarm information in distillation tower. The alarm optimization and the cause analysis of the alarm root are carried out. Firstly, the key performance indexes of the alarm system are statistically analyzed, and the performance of the alarm system is evaluated. For the repeated alarm, considering the dead zone, the optimal design of alarm limit based on data filtering is introduced. The introduction of data filtering can effectively solve the problem of process data being disturbed by noise. The combination of dead zone and data filtering can effectively solve the problem of repeated alarm. However, it is very limited to optimize the alarm only according to the measured signal. Only by finding the root cause of the alarm can the problem of alarm be solved completely. The fault diagnosis strategy is also an effective way to deal with the alarm. As long as the fault is eliminated, the alarm will disappear. However, The direct application of chemical mechanism model to fault diagnosis still has the problems of large computation and poor convergence. In this paper, the fault detection and diagnosis method based on the two-layer model is studied. The nonlinear model is used to monitor the distillation process and identify the abnormal situation of the large deviation of the measured value. The linear least square method based on the linear model is used to estimate the internal fault parameters of the distillation tower. In order to further reduce the calculation range of the model, the Simmel Braugh algorithm is also used to decompose the distillation system. The problem that is difficult to solve is decomposed into some sub-problems of low dimension. Then, the decomposed small system is de-noised based on median filter and lifting wavelet analysis, which enhances the robustness of alarm optimization. The above alarm optimization strategy is applied to the stripper of TEP simulation flow. The case study shows that the two-layer diagnosis structure can effectively obtain the variation of fault parameters. It is more effective than the simple nonlinear model structure. The variation of the parameters of the bottom feed loss caused by the step drop of the pressure differential driving force of the logistics C at 7:00 is also given. The results show that, This method can not only give the cause of fault accurately, but also shorten the time of alarm diagnosis.
【学位授予单位】：青岛科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TQ053.5

【参考文献】