WO2017088207A1 - 一种基于变量时段分解的间歇过程无模型在线滚动优化方法 - Google Patents

一种基于变量时段分解的间歇过程无模型在线滚动优化方法 Download PDF

Info

Publication number
WO2017088207A1
WO2017088207A1 PCT/CN2015/096372 CN2015096372W WO2017088207A1 WO 2017088207 A1 WO2017088207 A1 WO 2017088207A1 CN 2015096372 W CN2015096372 W CN 2015096372W WO 2017088207 A1 WO2017088207 A1 WO 2017088207A1
Authority
WO
WIPO (PCT)
Prior art keywords
variable
optimization
batch
time period
period
Prior art date
Application number
PCT/CN2015/096372
Other languages
English (en)
French (fr)
Inventor
栾小丽
王志国
刘飞
Original Assignee
江南大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江南大学 filed Critical 江南大学
Priority to US15/736,578 priority Critical patent/US20190094838A1/en
Publication of WO2017088207A1 publication Critical patent/WO2017088207A1/zh
Priority to US16/455,679 priority patent/US10739758B2/en

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/0205—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system
    • G05B13/021—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system in which a variable is automatically adjusted to optimise the performance
    • G05B13/022—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric not using a model or a simulator of the controlled system in which a variable is automatically adjusted to optimise the performance using a perturbation of the variable
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
    • G05B19/41865—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by job scheduling, process planning, material flow
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/18—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form
    • G05B19/4155—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form characterised by programme execution, i.e. part programme or machine function execution, e.g. selection of a programme
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
    • G05B19/4188—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by CIM planning or realisation
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/30—Nc systems
    • G05B2219/32—Operator till task planning
    • G05B2219/32015—Optimize, process management, optimize production line
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/30—Nc systems
    • G05B2219/32—Operator till task planning
    • G05B2219/32077—Batch control system
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/30—Nc systems
    • G05B2219/32—Operator till task planning
    • G05B2219/32252—Scheduling production, machining, job shop
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/30—Nc systems
    • G05B2219/32—Operator till task planning
    • G05B2219/32287—Medical, chemical, biological laboratory
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Definitions

  • the invention belongs to the field of chemical process manufacturing, and relates to an on-line rolling optimization method for operating trajectory variable period decomposition of batch process, which is suitable for including batch reactor, batch distillation tower, batch drying, batch fermentation, batch crystallization and other adoptions.
  • the process of batch mode operation and the optimal operation of the system are optimized online.
  • the most common method for obtaining the best operating curve is the model-based offline optimization method, which is based on the process model to solve the optimization problem offline.
  • off-line optimization is only applicable to the ideal model.
  • the uncertain factors and disturbances in the process model have an impact on the real-time operation of the system, the obtained trajectory will no longer be optimal.
  • the operation strategy and operating conditions of the process must be updated in real time. Therefore, the online real-time optimization method and technology for studying intermittent processes is an important issue in the process industry.
  • the invention relates to a variable time period decomposition model-free online scrolling optimization method for a batch process.
  • the variable operation data closely related to the product quality is collected, and the data-driven method is used to divide the process on the time domain to integrate the optimization action on each subset and form a global optimization strategy. On this basis, the online rolling minimum error correction of the optimization strategy is implemented.
  • a data-driven online scrolling optimization method for variable-time decomposition of batch processes based entirely on operational data of the production process, without prior knowledge and mechanism model of the process mechanism.
  • the steps of the invention are divided into two parts.
  • the first part offline data collection and establishment of basic optimization strategies
  • the second Points online rolling error correction implementation method.
  • Step 1 Collect the variables to be optimized and the final quality or yield indicators in batches for the complete batch process.
  • the data collection interval may be an equal time interval or an unequal time interval. During one time interval, the variables to be optimized of the process do not change significantly, or may not have a significant impact on the final quality or yield indicator. 30-50 batches of valid data are generally required.
  • Step 2 Perform the principal component analysis on the collected data according to the batch and eliminate the singular points in the pivot mode, so that all data points are within a certain degree of confidence.
  • Step 3 The remaining data after the singular points are removed are equally divided or unequal intervals on the time axis.
  • Step 4 Express each batch of data contained in each interval as a continuous variable. These variables are called time-separated variables after decomposition. The value of the time period variable consists of the individual batch data of the variable to be optimized in a specific time interval.
  • Step 5 Each batch quality or yield indicator corresponding to step 4 is referred to as an indicator variable.
  • the value of the indicator variable is a continuous variable formed by the final mass or yield of each batch.
  • Step 6 Combine the time period variable and the index variable formed in steps 4 and 5 to form a joint data matrix of the time period variable and the index variable.
  • Step 7 Perform a principal component analysis on the joint matrix to form a principal component load map.
  • Step 8 Classify the direction and magnitude of the target variable by the time period variable according to the principal component load map in step 7. It is divided into three categories: positive action, reactionary action and no (micro) action.
  • Step 9 Calculate the optimization strategy for each period variable according to the following perturbation formula:
  • J(i), M(i) and ⁇ (i) are the optimization target values, mean and standard deviation of the i-th period variable, respectively; sign(i) is the i-th period variable and indicator variable.
  • Angle cosine symbol When the angle is less than 90 degrees, it is +1, when the angle is greater than 90 degrees, it is -1, and when the angle is 90 degrees, it is 0.
  • Step 11 Generally, the above optimization variable curve needs to be digitally filtered, so that the new optimization curve is relatively smooth and easy to track and control.
  • the basic optimization control variable trajectory obtained through the above steps is subjected to online rolling error correction for the basic optimization strategy in each time period when it is put into practical application.
  • Step 13 On the offline basic optimization strategy, form a new optimization target value for the next period:
  • an error sequence can be formed using errors over a period of time in the past, digital filtering is applied to the error sequence, and the filtered prediction values are applied to the optimization strategy for the current time period.
  • the invention collects the variable operation data closely related to the product quality, and uses the data driven method to integrate the optimization action amount on each subset and form a global optimization strategy on the basis of the time domain variable division of the process. On this basis, the online rolling minimum error correction of the optimization strategy is implemented.
  • the method of the present invention forms an online optimization strategy based entirely on the operational data of the batch process, without prior knowledge and models of the mechanism itself. At the same time, the use of the online rolling correction strategy makes the optimized operating trajectory more adaptable and better meets the requirements of actual industrial production anti-interference.
  • Figure 1 is an example of a temperature profile for a batch process.
  • Figure 2 is a principal mode diagram of an intermittent process temperature as an optimized variable.
  • Figure 3 is a block diagram of the time period variable.
  • Figure 4 is a plot of the principal component load for the time period variable and the indicator variable.
  • Figure 5 is a classification diagram of the effect of time period variables on indicator variables.
  • Figure 6 is a comparison of the optimized temperature profile and the original temperature profile for a batch process.
  • FIG. 7 is a generation diagram of an online scrolling error correction strategy.
  • FIG. 8 is a block diagram showing the steps of implementing the present invention.
  • Figure 9 shows the optimized curve and the original optimization curve after moving average filtering.
  • Figure 10 is an optimization (partial) diagram of a batch crystallization process.
  • the implementation method is divided into four parts.
  • the first part is data collection and preprocessing.
  • the second part is the construction of the joint data matrix.
  • the third part is the calculation of the basic optimization strategy.
  • the fourth part is to establish an online optimization strategy for rolling error correction.
  • Step 1 For the operation of the complete batch crystallization process, select the operating temperature closely related to the product yield as the variable to be optimized, and collect 50 sets of temperature variables and final yield indicator data by batch. The data collection interval is 1 minute.
  • Figure 1 is an example of temperature profile data collection for a batch crystallization process. For the sake of clarity, only two batches of temperature profiles are plotted.
  • Step 2 For all 50 batches of temperature data collected, perform a principal component analysis of the temperature variables by batch and eliminate the singular points in the pivot mode, so that all data points are within a certain degree of confidence.
  • Figure 2 is a principal mode diagram of the batch process temperature as the optimization variable. It can be seen from the figure that there is a gap between the data on the right side and the overall data pattern. The temperature data of this batch should be eliminated.
  • Step 3 The temperature data of the remaining 49 batches are equally divided into 300 time periods on the time axis to form 300 time period variables C 1 , C 2 ... C 300 .
  • Figure 3 shows the period variables for C 40 to C 70 .
  • Step 4 The yield indicator data of each batch corresponding to step 3 is formed to form the indicator variable Q.
  • Step 5 Combine the 300 time-varying variables C 1 , C 2 ... C 300 formed in steps 3 and 4 with an index variable Q to generate a joint data matrix L of 49 ⁇ 301 dimensions.
  • Step 6 Perform a principal component analysis on the joint matrix L to form a principal component load map.
  • Figure 4 shows an example of a Principal Load Diagram resulting from the combination of C 36 to C 60 25 time periods and the indicator variable Q.
  • Step 7 Classify the direction and magnitude of the target variable by the time-variant variable according to the principal component load map in step 6.
  • Figure 5 is a classification example. It can be seen from Figure 5 that C 154 , C 155 , C 156 and C 273 have the greatest effect on the index variable Q, where C 154 , C 155 , C 156 are reaction and C 273 is positive. On the other hand, C 66 and C 111 which are in the direction of about 90 degrees from the index variable Q have little effect on the index variable Q.
  • Step 8 Calculate the mean and standard deviation of each period variable separately.
  • the average value of C 154 which is counterproductive to the indicator variable Q, is 134.58 degrees Celsius and the standard deviation is 6.08 degrees Celsius.
  • the optimization target value of the i-th period variable is obtained by the following perturbation calculation formula:
  • J(i), M(i) and ⁇ (i) are the optimization target values, mean and standard deviation of the i-th period variable, respectively; sign(i) is the i-th period variable and indicator variable.
  • Angle cosine symbol As shown in the classification diagram of Fig. 5, when the angle is less than 90 degrees, it is +1, when the angle is greater than 90 degrees, it is -1, and when the angle is 90 degrees, it is 0.
  • Step 11 The above-mentioned basic optimization curve is subjected to moving average filtering, so that the filtered optimization curve is relatively smooth, which is convenient for the later tracking control design.
  • Figure 6 is a comparison of the optimized temperature profile and the original temperature profile
  • Figure 9 is the optimization curve and the original optimization curve after the moving average filtering. It can be seen from Figure 9 that the filtered optimization curve is smoother and easier to track the implementation of the controller.
  • Step 12 When the basic optimized control track obtained in the above series of steps is used online, the rolling error correction is performed in each time period:
  • FIG. 7 is a schematic diagram of generation calculation of an online scrolling error correction strategy.
  • Figure 10 is an illustration of an optimization result of a batch crystallization process. From the results shown in the figure, the unoptimized yield is 90.25%. Under the rolling correction optimization strategy, the actual operating yield is 94.88%, which is close to the theoretical optimal yield of 95.51%. This result shows the effectiveness and utility of the method of the invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Manufacturing & Machinery (AREA)
  • Quality & Reliability (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Human Computer Interaction (AREA)
  • Feedback Control In General (AREA)
  • Control Of Metal Rolling (AREA)
  • General Factory Administration (AREA)

Abstract

一种基于变量时段分解的间歇过程无模型在线滚动优化方法，通过采集与产品质量密切相关的变量运行数据，利用数据驱动方法在对过程做时间域上变量划分的基础上，整合各个子集上的优化作用量并形成全局优化策略。在此基础上，实施在线滚动误差修正优化策略。上述方法完全基于间歇过程的操作数据形成在线优化策略，不需要过程本身机理的先验知识和模型。同时在线滚动修正策略的使用，使得所优化的操作轨线具有更好适应性，更好地满足实际工业生产抗干扰的要求。

Description

一种基于变量时段分解的间歇过程无模型在线滚动优化方法 技术领域
本发明属于化工流程制造业领域，涉及针对间歇过程的操作轨线变量时段分解无模型在线滚动优化方法，适用于包括间歇反应器、间歇精馏塔、间歇干燥、间歇发酵，间歇结晶以及其它采用间歇方式操作的过程和系统的最佳操作轨线在线优化。
背景技术
间歇过程在工业实际生产中，操作人员通常是从各种控制指标出发，根据长时间积累的经验摸索出一条针对某个具体间歇过程的操作曲线。如此依靠经验寻找操作曲线的方法费时费力并且难以标准化和推广。因此推行简捷有效的间歇过程优化方法，获取更满意的经济指标成为必要。间歇过程的优化问题，通常以提高产品的质量或产率为目标，获取最优操作轨线。因此对获取间歇过程的最优操作曲线的方法研究是解决问题的关键。
获取最佳操作曲线最普通的方法是基于模型的离线优化方法，即基于过程模型离线求解优化问题。但离线优化只适用于理想模型，当过程模型中不确定因素及干扰对系统实时运行产生影响时，已求得的轨线将不再是最优的。同时由于操作过程中进料的变化、产品和原料的切换、生产过程的开/停车，都会要求过程的操作策略和操作条件必须进行实时更新。因此研究间歇过程的在线实时优化方法和技术是流程工业中一个重要课题。
基于连续过程的实时优化已有许多成功的工业应用案例，而针对间歇过程的在线实时优化技术，目前仍缺乏比较通用的适合工业应用的有效方法。因此提出一个比较通用的，能解决实际工业问题的间歇过程在线实时优化策略和实施框架，促进间歇过程在线实时优化的工业化进程，为解决实际生产领域的控制难题提供新的方法迫在眉睫。
发明内容
本发明涉及一种针对间歇过程的变量时段分解无模型在线滚动优化方法。采集与产品质量密切相关的变量运行数据，利用数据驱动方法在对过程做时间域上变量划分的基础上，整合各个子集上的优化作用量并形成全局优化策略。在此基础上，实施优化策略的在线滚动最小误差修正。
本发明为实现上述目的，采用如下技术方案：
一种针对间歇过程的变量时段分解的数据驱动在线滚动优化方法，完全基于生产过程的操作数据，不需要过程机理的先验知识和机理模型。
本发明步骤分为两个部分。第一部分，离线数据收集和建立基础优化策略；第二部 分，在线滚动误差校正实施方法。
离线数据收集和基础优化策略步骤如下：
步骤一：针对操作完整的间歇过程，按批次收集待优化变量和最终质量或产率指标。数据的收集时间间隔可以是等时间间隔或不等时间间隔，在一个时间间隔内，过程的待优化变量没有显著变化，或不会对最终质量或产率指标有显著影响。一般要求30-50个批次有效数据。
步骤二：对所采集的数据，按批次为变量进行主元分析并在主元模式图中剔除奇异点，使得所有数据点在一个可信度之内。
步骤三：将剔除奇异点后的剩余数据在时间轴上进行等间隔划分或不等间隔划分。
步骤四：将每一个间隔所包含的各批次数据表达为一个连续变量，这些变量被称为分解后的时段变量。时段变量的值由待优化变量在一个特定时间区间的各个批次数据所组成。
步骤五：将步骤四中所对应的每个批次质量或收率指标，称为指标变量。指标变量的值是由各个批次最终质量或收率形成的连续变量。
步骤六：将步骤四和步骤五中形成的时段变量和指标变量进行合并，形成时段变量和指标变量的联合数据矩阵。
步骤七：对上述联合矩阵做主元分析，形成主元载荷图。
步骤八：对步骤七中的主元载荷图按时段变量对指标变量的作用方向和大小进行分类。分为正作用，反作用及无(微)作用三类。
步骤九：按以下摄动量公式计算各时段变量的优化策略：
J(i)＝M(i)+sign(i)×3σ(i)
此处的J(i)，M(i)和σ(i)分别是第i个时段变量的优化目标值，均值和标准差；sign(i)是第i个时段变量和指标变量所形成的夹角余弦符号。夹角小于90度时为+1，夹角大于90度时为-1，夹角为90度时为0。
步骤十：将步骤九中所得各时段的优化目标值，按时段顺序i＝1,2,…N组成一条针对整个批次过程的基础优化变量曲线。
步骤十一：一般地需要将上述优化变量曲线进行数字滤波，使得新的优化曲线比较光滑，易于跟踪控制。
为了克服动态控制偏差和不可控的随机干扰，经过上述步骤得到的基础优化控制变量轨迹在投入实际应用时，在每一个时间段都对基础优化策略加以在线滚动误差修正。
在线滚动误差校正步骤如下：
步骤十二：在第i-1时间段，计算离线基础优化目标值J(i-1)和实际测量值RV(i-1)的误差：E(i-1)＝J(i-1)-RV(i-1)。
步骤十三：在离线基础优化策略上，构成下一时段新的优化目标值：
Jo(i)＝J(i)+E(i-1)。
将步骤十二以及步骤十三按时段顺序i＝1,2,…N依次计算并施加到过程中，直至整个批次过程操作结束。
更一般地，在步骤十二中，可以使用过去若干时段的误差形成误差序列，对此误差序列施加数字滤波，将滤波预测值施加到当前时段的优化策略中。
本发明通过采集与产品质量密切相关的变量运行数据，利用数据驱动方法在对过程做时间域上变量划分的基础上，整合各个子集上的优化作用量并形成全局优化策略。在此基础上，实施优化策略的在线滚动最小误差修正。本发明方法完全基于间歇过程的操作数据形成在线优化策略，不需要过程本身机理的先验知识和模型。同时在线滚动修正策略的使用，使得所优化的操作轨线具有更好适应性，更好地满足实际工业生产抗干扰的要求。
附图说明
图1为间歇过程的温度曲线示例。
图2为一个间歇过程温度为优化变量的主元模式图。
图3为时段变量的构成图。
图4为时段变量和指标变量主元载荷图。
图5为时段变量对指标变量作用分类图。
图6为一个间歇过程的优化温度曲线和原始温度曲线比较图。
图7为在线滚动误差修正策略的生成图。
图8为本发明实施步骤框图。
图9为滑动平均滤波后的优化曲线和原优化曲线。
图10为一个间歇结晶过程的优化结果(局部)图。
具体实施方式
本例以一个间歇结晶过程为例，所述方法不构成对本发明的范围限制。
本实施方法分为四个部分。第一部分为数据收集和预处理。第二部分是联合数据矩阵的构造。第三部分是基础优化策略的计算。第四部分是建立滚动误差修正在线优化策略。
本方法实施步骤框图如图8所示，具体实施步骤和算法如下：
步骤1：针对操作完整的间歇结晶过程，选择与产品收率密切相关的操作温度作为待优化变量，按批次收集50组温度变量以及最终收率指标数据。数据的收集时间间隔为1分钟。图1是一个间歇结晶过程的温度曲线数据收集示例，为清晰起见，图中仅画出2个批次的温度曲线。
步骤2：对采集的所有50批次的温度数据，按批次对温度变量进行主元分析并在主元模式图中剔除奇异点，使得所有数据点在一个可信度之内。图2是一个间歇过程温度为优化变量的主元模式图，从图中可以看出，右边有一个批次的数据与整体数据模式差距太大，应剔除这个批次的温度数据。
步骤3：将剩余49个批次的温度数据在时间轴上进行等间隔划分为300个时段，以此构成300个时段变量C1,C2…C300。为清晰起见，图3给出了C40到C70的时段变量。
步骤4：将步骤3所对应的每个批次的收率指标数据，形成所述的指标变量Q。
步骤5：将步骤3以及步骤4中形成的300个时段变量C1,C2…C300和一个指标变量Q进行合并，生成49×301维的联合数据矩阵L。
步骤6：对上述联合矩阵L做主元分析，形成主元载荷图。为清晰起见，图4给出了C36到C6025个时段变量和指标变量Q合并产生的主元载荷图示例。
步骤7：对步骤6中的主元载荷图按时段变量对指标变量的作用方向和大小进行分类。图5是一个分类示例，从图5可以看出C154、C155、C156以及C273对指标变量Q的作用最大，其中C154、C155、C156是反作用，C273是正作用。而与指标变量Q夹角90度左右方向的C66、C111等对指标变量Q几乎不起作用。
步骤8：分别计算出每一个时段变量的均值和标准差。例如对指标变量Q起反作用的C154的均值为134.58摄氏度，标准差为6.08摄氏度。
按以下摄动量计算公式获取第i个时段变量的优化目标值：
J(i)＝M(i)+sign(i)×3σ(i)
此处的J(i)，M(i)和σ(i)分别是第i个时段变量的优化目标值，均值和标准差；sign(i)是第i个时段变量和指标变量所形成的夹角余弦符号。如图5分类图上，夹角小于90度时为+1，夹角大于90度时为-1，夹角为90度时为0。
步骤10：将步骤9中所得各时段的优化目标值，按时段顺序i＝1,2,…300构成一个基础优化变量曲线。
步骤11：将上述基础优化曲线进行滑动平均滤波，使得滤波后的优化曲线比较光滑，便于后期的跟踪控制设计。图6是优化的温度曲线和原始温度曲线的比较，图9是采用滑动平均滤波后的优化曲线和原优化曲线。从图9中可以看出经过滤波后的优化曲线更加平滑，便于跟踪控制器的实施。
步骤12：对上述系列步骤得到的基础优化控轨迹在线使用时，在每一个时间段加以滚动误差修正：
(1)对于第i-1时间段，计算离线基础优化目标值J(i-1)和实际测量值RV(i-1)的误差：
E(i-1)＝J(i-1)-RV(i-1)
(2)在离线基础优化策略上，构成下一时段新的优化目标值：
Jo(i)＝J(i)+E(i-1)
将步骤12按时段顺序i＝1,2,…300依次推进计算，直至整个批次过程操作完毕。图7是在线滚动误差修正策略的生成计算示意图。
图10是一个间歇结晶过程的优化结果示例。从图中结果看出，无优化收率为90.25％，在滚动修正优化策略下，实际运行收率为94.88％，基本接近理论最优收率95.51％。此结果显示出本发明方法的有效性和实用性。

Claims (6)

  1. 一种基于变量时段分解的间歇过程无模型在线滚动优化方法，其特征在于包括下述步骤：
    (1)针对操作完整的间歇过程，按批次收集待优化变量和最终质量或产率指标；
    (2)对所述步骤(1)中采集的数据，以批次为变量进行主元分析并在主元模式图中剔除奇异点，使得所有数据点在一个可信度之内；
    (3)将剔除奇异点后的剩余数据在时间轴上进行间隔划分；将每一个间隔所包含的各批次数据表达为一个连续变量，这些变量被称为分解后的时段变量；所述时段变量值由待优化变量在一个特定时间区间的各个批次数据所组成；
    (4)将步骤(3)中所对应的每个批次质量或收率指标，称为指标变量；所述指标变量的值是由各个批次质量或收率形成的连续变量；
    (5)将步骤(3)和步骤(4)中形成的时段变量和指标变量进行合并，形成时段变量和指标变量的联合数据矩阵；并对所述联合数据矩阵做主元分析，形成主元载荷图；
    (6)对步骤(5)中的主元载荷图按时段变量对指标变量的作用方向和大小进行分类；
    (7)按以下摄动量公式计算各时段变量的优化策略：
    J(i)＝M(i)+sign(i)×3σ(i)
    此处的J(i)，M(i)和σ(i)分别是第i个时段变量的优化目标值，均值和标准差；sign(i)是第i个时段变量和指标变量所形成的夹角余弦符号；
    (8)将步骤(7)中所得各时段的优化目标值，按时段顺序组成一条针对整个批次过程的基础优化变量曲线；
    (9)在第i-1时间段，计算离线基础优化目标值J(i-1)和实际测量值RV(i-1)的误差：E(i-1)＝J(i-1)-RV(i-1)；
    (10)在离线基础优化策略上，构成下一时段新的优化目标值：
    Jo(i)＝J(i)+E(i-1)；
    (11)将步骤(9)以及步骤(10)按时段顺序i＝1,2,…N依次计算并施加到过程中，直至整个批次过程操作结束。
  2. 根据权利要求1所述的基于变量时段分解的间歇过程无模型在线滚动优化方法，其特征在于：所述步骤(1)中间歇过程数据收集的时间间隔相等或不相等。
  3. 根据权利要求1所述的基于变量时段分解的间歇过程无模型在线滚动优化方法， 其特征在于：所述步骤(3)中间隔划分为等间隔划分或不等间隔划分。
  4. 根据权利要求1所述的基于变量时段分解的间歇过程无模型在线滚动优化方法，其特征在于：所述步骤(6)中分类为正作用、反作用和无/微作用三类。
  5. 根据权利要求1所述的基于变量时段分解的间歇过程无模型在线滚动优化方法，其特征在于：所述步骤(7)中夹角余弦符号sign(i)的取值为：夹角小于90度时sign(i)＝+1，夹角大于90度时sign(i)＝-1，夹角为90度时sign(i)＝0。
  6. 根据权利要求1所述的基于变量时段分解的间歇过程无模型在线滚动优化方法，其特征在于：所述步骤(8)中对优化变量曲线进行数字滤波，使得新的优化变量曲线光滑。
PCT/CN2015/096372 2015-11-26 2015-12-04 一种基于变量时段分解的间歇过程无模型在线滚动优化方法 WO2017088207A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US15/736,578 US20190094838A1 (en) 2015-11-26 2015-12-04 Model-Free Online Recursive Optimization Method for Batch Process Based on Variable Period Decomposition
US16/455,679 US10739758B2 (en) 2015-11-26 2019-06-27 Model-free online recursive optimization method for batch process based on variable period decomposition

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510834481.0A CN105373098B (zh) 2015-11-26 2015-11-26 一种基于变量时段分解的间歇过程无模型在线滚动优化方法
CN201510834481.0 2015-11-26

Related Child Applications (2)

Application Number Title Priority Date Filing Date
US15/736,578 A-371-Of-International US20190094838A1 (en) 2015-11-26 2015-12-04 Model-Free Online Recursive Optimization Method for Batch Process Based on Variable Period Decomposition
US16/455,679 Division US10739758B2 (en) 2015-11-26 2019-06-27 Model-free online recursive optimization method for batch process based on variable period decomposition

Publications (1)

Publication Number Publication Date
WO2017088207A1 true WO2017088207A1 (zh) 2017-06-01

Family

ID=55375381

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/096372 WO2017088207A1 (zh) 2015-11-26 2015-12-04 一种基于变量时段分解的间歇过程无模型在线滚动优化方法

Country Status (3)

Country Link
US (2) US20190094838A1 (zh)
CN (1) CN105373098B (zh)
WO (1) WO2017088207A1 (zh)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109254531B (zh) * 2017-11-29 2021-10-22 辽宁石油化工大学 具有时滞和干扰的多阶段间歇过程的最优成本控制方法
CN108375908B (zh) * 2018-04-11 2020-07-03 江南大学 一种基于系统运行模式的双酚a结晶过程滚动优化方法
CN109635465B (zh) * 2018-12-18 2020-06-05 江南大学 基于tpls模型的批次内操作轨迹调整方法
US11119456B2 (en) * 2019-03-14 2021-09-14 Kaneka Corporation Production system, information processing method, and production method
CN110579970B (zh) * 2019-10-24 2023-02-03 海南师范大学 一种2d滚动优化下间歇过程终端约束预测控制方法
CN114200834B (zh) * 2021-11-30 2023-06-30 辽宁石油化工大学 丢包环境下批次过程无模型离轨策略最优跟踪控制方法
CN116861224B (zh) * 2023-09-04 2023-12-01 鲁东大学 基于间歇过程软测量建模方法的间歇过程软测量建模系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101620416A (zh) * 2009-06-12 2010-01-06 浙江工业大学 一种流程工业企业生产过程的生产调度智能优化方法
CN101872444A (zh) * 2010-05-21 2010-10-27 杭州电子科技大学 一种结合中期修正策略的间歇过程批到批优化方法
WO2010128354A1 (en) * 2009-05-06 2010-11-11 Abb Research Ltd A method and a system for on-line optimization of a batch pulp digester
CN103092074A (zh) * 2012-12-30 2013-05-08 重庆邮电大学 半导体先进过程控制的参数优化控制方法
CN103092078A (zh) * 2013-01-07 2013-05-08 北京中医药大学 多阶段间歇生产过程的全程优化方法

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100561383C (zh) * 2006-10-25 2009-11-18 上海梅山钢铁股份有限公司 板坯连铸二次冷却及动态轻压下离线仿真系统
US8374709B2 (en) * 2008-03-03 2013-02-12 Alstom Technology Ltd Control and optimization system
US20090287320A1 (en) * 2008-05-13 2009-11-19 Macgregor John System and Method for the Model Predictive Control of Batch Processes using Latent Variable Dynamic Models
CN101587328B (zh) * 2008-05-23 2011-12-14 朱豫才 非线性过程动态模型辨识方法
US9098093B2 (en) * 2008-09-30 2015-08-04 Rockwell Automation Technologies, Inc. Model predictive control of biofuel denaturant blending
US9134711B2 (en) * 2010-05-04 2015-09-15 Frito-Lay North America, Inc. Advanced batch control
US9110462B2 (en) * 2010-05-04 2015-08-18 Frito-Lay North America, Inc. Batch control using bang-bang control
US9046882B2 (en) * 2010-06-30 2015-06-02 Rockwell Automation Technologies, Inc. Nonlinear model predictive control of a batch reaction system
WO2012003324A1 (en) * 2010-06-30 2012-01-05 Xcellerex, Inc. Batch authoring tool and bioreactor control system
CN102183699B (zh) * 2011-01-30 2012-12-26 浙江大学 化工过程多变量预测控制系统模型失配检测与定位方法
FR2978577B1 (fr) * 2011-07-26 2020-04-03 Ip Leanware Dispositif et procede de diagnostic d'un processus industriel evolutif
ES2644303T3 (es) * 2012-05-11 2017-11-28 Bp Corporation North America Inc. Control por lotes automatizado de un coquizador retardado

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010128354A1 (en) * 2009-05-06 2010-11-11 Abb Research Ltd A method and a system for on-line optimization of a batch pulp digester
CN101620416A (zh) * 2009-06-12 2010-01-06 浙江工业大学 一种流程工业企业生产过程的生产调度智能优化方法
CN101872444A (zh) * 2010-05-21 2010-10-27 杭州电子科技大学 一种结合中期修正策略的间歇过程批到批优化方法
CN103092074A (zh) * 2012-12-30 2013-05-08 重庆邮电大学 半导体先进过程控制的参数优化控制方法
CN103092078A (zh) * 2013-01-07 2013-05-08 北京中医药大学 多阶段间歇生产过程的全程优化方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
JIA, LI ET AL.: "A novel iterative learning control for product quality control in batch process", JOURNAL OF THE CHEMICAL INDUSTRY AND ENGINEERING SOCIETY OF CHINA, vol. 60, no. 8, 31 August 2009 (2009-08-31), pages 2017 - 2023 *

Also Published As

Publication number Publication date
CN105373098A (zh) 2016-03-02
US10739758B2 (en) 2020-08-11
CN105373098B (zh) 2017-10-24
US20190324427A1 (en) 2019-10-24
US20190094838A1 (en) 2019-03-28

Similar Documents

Publication Publication Date Title
WO2017088207A1 (zh) 一种基于变量时段分解的间歇过程无模型在线滚动优化方法
WO2017088208A1 (zh) 一种由数据差异驱动的间歇过程自学习动态优化方法
CN110378044B (zh) 基于注意力机制的多时间尺度卷积神经网络软测量方法
CN101158693A (zh) 基于多核独立元分析的批量生产过程故障检测方法
CN104914723A (zh) 基于协同训练偏最小二乘模型的工业过程软测量建模方法
CN110928187B (zh) 一种基于模糊宽度自适应学习模型的污水处理过程故障监测方法
CN105425583B (zh) 基于协同训练lwpls的青霉素生产过程的控制方法
CN104317269A (zh) 一种基于2d理论的综合预测迭代学习控制方法
CN105334831B (zh) 由数据差异驱动的间歇过程双维在线优化方法
CN113617852B (zh) 一种基于数据驱动的楔横轧智能化控制系统
CN113609672A (zh) 一种基于增量模型的数字孪生系统协同进化方法
CN101221437B (zh) 网络信息交换模式下的工业生产全过程优化控制方法
CN108681656B (zh) 一种基于乙烯裂解炉运行数据的过程分析方法
Wang et al. Smart systems engineering contributing to an intelligent carbon-neutral future: opportunities, challenges, and prospects
CN104407642A (zh) 一种基于迭代学习控制的连铸坯感应加热过程温控方法
CN102662324A (zh) 槽式反应器基于在线支持向量机的非线性模型预测控制方法
CN105446292A (zh) 一种基于变量时段分解的间歇过程无模型摄动优化方法
CN115108595A (zh) 基于机器学习的净水剂智能投加方法
CN102609601B (zh) 一种基于类内质网体膜计算的渣油加氢反应动力学模型参数估计方法
CN102486632A (zh) 一种对二甲苯氧化过程中对苯二甲酸结晶粒径在线分析方法
CN103439964A (zh) 一种ocsvm监控模型的在线更新系统及方法
Srinivasan et al. Analysis on tuning strategy of ph control process in paper industries
Xiao et al. Research on the Control Strategy of Battery Roller Press Deflection Device by Introducing Genetic Algorithm to Optimize Integral Separation PID
CN104898416A (zh) 一种基于armax建模的培养基蒸汽喷射灭菌节能控制方法
Cui et al. Prediction of the product formation with the AdaBoost algorithm in bioprocesses

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15909118

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15909118

Country of ref document: EP

Kind code of ref document: A1