基于趋势分析的间歇过程异常工况超早期报警研究

为了实现化工间歇过程阶跃变化中异常工况的早期报警，重点监测过程参数的变化趋势，提出基于最小二乘法的拟合 微分 再微分的化工间歇过程趋势分析方法。采用最小二乘法对历史数据进行拟合，依据过程参数的变化趋势，将间歇过程分为3个阶段。对数据微分求导计算，根据得出的过程参数一阶变化率的取值区间进行间歇过程多时段工况识别，针对危险性较大的阶跃变化过程利用再微分计算过程参数的二阶变化率取值区间，结合滑动窗算法，实现连续的间歇过程异常工况早期报警监测。在聚丙烯装置异常工况早期预警案例分析中，以聚合釜升温过程中上温为目标参数，实时监测上温变化趋势，识别间歇过程工况，并且在上温升温速率过快但温度未超出分布式控制系统(Distributed control system, DCS)阈值时发出警报。结果表明，所提出方法能够在参数状态出现异常的早期发出警报，相比于3σ报警阈值的方法提前24 min 34 s报警。

Study on Abnormal Situation Ultra-Early Warning of Batch Process Based on Trend Analysis

The batch process parameters present a cyclical step change with the change of an operation process. In order to realize the early warning of the abnormal situation in the step change of a chemical batch process, the trend of the process parameters was monitored. The method of fitting differentiating re differentiating based on the least square method was proposed. The least squares method was used to fit the historical data, and the batch process was divided into three stages according to the trend of the process parameters, ie, the rising stage, reacting stage and declining stage. In order to ensure the effectiveness of the method, historical data was selected as many as possible to making differential calculation to get the first order derivation control thresholds. Then, the second order differential derivative was applied to history data, and the second order derivation control thresholds were obtained. During the continuous monitoring of the batch process parameters, the differential method was used for the real time data, and then, according to the threshold of the first order differential of the process parameters, the batch process multi period working conditions were identified. For the most dangerous step change part, rising stageand declining stage, the second order differential range of the process parameter was calculated by using the re differential method. Combined with the sliding window algorithm, continuous batch process abnormal situation of early warning monitoring was achieved. In the early warning case study of abnormal situation of a polypropylene plant, the upper temperature of the polymerizer was used as the target parameter. The fluctuated trend of the upper temperature was monitored in real time, the batch process conditions were identified, and the alarm would be triggered when the temperature changed a little faster but did not exceed the distributed control system(DCS)threshold. The results showed that the method proposed in this paper was capable of issuing an alarm in the early abnormal state of the parameter, and compared to the Three Sigma method, it can alarm in the early 24 minutes 34 seconds in advance.