Method for optimizing and constructing soft instrument model based on DCS system

Abstract

The present application relates to the technical field of industrial process control and modeling, in particular to a soft instrument model optimization construction method based on a DCS system, the present application divides production conditions according to fluctuation of monitoring data; analyzes correlation characteristics between relevant monitoring parameters and irrelevant monitoring parameters under the same historical condition slice, and analyzes performance differences of irrelevant monitoring parameters in different historical condition slices on the influence on relevant monitoring parameters, and comprehensively evaluates reference weights of relevant monitoring parameters in the historical condition slice; according to consistency of changes of monitoring parameters in the current condition state window and the historical condition slice, the reference weights of the relevant monitoring parameters under the current condition state window are obtained, and then the soft instrument model of the current condition state window is constructed; the soft instrument model obtained by the method can adapt to different production conditions, effectively process auxiliary variables in the DCS system, and improve from local optimization of a single parameter to comprehensive optimization of all parameters.

Description

Technical Field

[0001] This invention relates to the field of industrial process control and modeling technology, specifically to a method for optimizing and constructing soft instrument models based on DCS systems. Background Technology

[0002] With the improvement of automation levels, DCS (Distributed Control System) systems have gradually replaced conventional instruments, and their superiority has been widely recognized by operators. In modern industrial control processes, maximizing the efficiency of production process control is a basic indicator for measuring the performance of a control system. However, due to limitations such as equipment cost and technology, in complex chemical industrial production processes such as coal gasification furnaces, distillation columns, and polymerization reactors, key quality parameters (such as reactor temperature, purity of top product, melt index, etc.) are often difficult to measure directly due to extreme conditions such as high temperature, high pressure, and strong corrosion, or there is a significant measurement lag.

[0003] The basic idea of ​​soft sensing is to address unmeasurable targets or results that current measuring instruments cannot measure with the required accuracy within a real-world system. By acquiring massive amounts of operational data through a DCS system, and constructing a mathematical model between easily measurable auxiliary variables and difficult-to-measurable dominant variables using molecular models, soft sensing of the difficult-to-measurable dominant variables can be achieved, guiding chemical production and effectively solving this problem. Through soft sensing and optimized control technologies, precise control of the production process can be achieved, improving product quality and production efficiency, and helping to reduce energy consumption and equipment maintenance costs.

[0004] Current technologies typically employ particle swarm optimization (PSO) to select monitoring points from historical databases and construct soft instrumentation models for soft measurement. However, chemical production processes involve multi-level, high-dimensional variable data with severe multicollinearity. Feature selection methods based on local optima do not adequately consider the impact of historical operating conditions on current conditions, resulting in biases in the obtained auxiliary variables for different operating conditions and measurement distortion in the constructed soft instrumentation models. Summary of the Invention

[0005] To address the above technical problems, the present invention aims to provide a method for optimizing and constructing a soft instrument model based on a DCS system.

[0006] The present invention provides a method for optimizing and constructing a soft instrument model based on a DCS system, the specific technical solution of which is as follows:

[0007] Obtain historical monitoring parameters from the factory's DCS system;

[0008] Analyze the data fluctuation characteristics of the historical monitoring parameters in different time periods to obtain historical operating condition slices;

[0009] Obtain relevant and irrelevant monitoring parameters for the historical operating condition slices, and obtain the correlation significance between relevant and irrelevant monitoring parameters under the same historical operating condition slices;

[0010] Based on the correlation significance, the differences in the influence of the unrelated monitoring parameters on the related monitoring parameters in different historical working condition slices are analyzed to obtain the reference factor of the unrelated monitoring parameters on the related monitoring parameters.

[0011] Based on the reference factor and the correlation significance, the reference weights of the relevant monitoring parameters in the historical operating condition slice are obtained.

[0012] Analyze the consistency of changes in monitoring parameters between the current operating condition window and the historical operating condition slices to obtain the matching historical operating condition slices for the current operating condition window. Then, obtain the reference weights of the relevant monitoring parameters under the current operating condition window and construct a soft instrument model under the current operating condition window.

[0013] In some embodiments of the present invention, the data fluctuation characteristics of each historical monitoring parameter over different time periods are analyzed, including:

[0014] By traversing the historical monitoring parameters using a sliding time window, the local maximum value within the time window is obtained, thus identifying the abrupt change point of the historical monitoring parameters.

[0015] Within the time interval between two adjacent mutation points, the fluctuation range of the historical monitoring parameters at continuous times is analyzed, and combined with the average value of the historical monitoring parameters, the data fluctuation parameters of each historical monitoring parameter in different time periods are obtained.

[0016] In some embodiments of the present invention, obtaining historical operating condition slices includes:

[0017] Treating the data fluctuation parameters within each time period as a particle, a coordinate system is established to form a particle diagram.

[0018] The particles were automatically clustered using DBSCA, and each cluster was labeled with a working condition tag to obtain historical working condition slices.

[0019] Each of the historical operating condition slices contains several types of historical monitoring parameters.

[0020] In some embodiments of the present invention, obtaining relevant and irrelevant monitoring parameters of the historical operating condition slice includes:

[0021] Based on the fluctuation range of the historical monitoring parameters at continuous times, and combined with the location information of the monitoring points in the process flow, a topological structure diagram of several historical monitoring parameters in the historical operating condition slice is constructed. The relevant monitoring parameters that directly affect the construction of the soft instrument model of the target position in the historical operating condition slice are extracted, and the relevant and irrelevant monitoring parameters of the historical operating condition slice are obtained.

[0022] In some embodiments of the present invention, obtaining the correlation significance between relevant and irrelevant monitoring parameters under the same historical operating condition slice includes:

[0023] By analyzing the distance relationship between relevant and irrelevant monitoring parameters in the same historical working condition slice, the correlation significance between relevant and irrelevant monitoring parameters in the same historical working condition slice is obtained.

[0024] In some embodiments of the present invention, analyzing the distance relationship between relevant and irrelevant monitoring parameters in the same historical working condition slice includes:

[0025] Calculate the DTW distance between the time series data sequences of relevant monitoring parameters and the time series data sequences of unrelated monitoring parameters in the same historical working condition slice;

[0026] In the topology diagram, the Euclidean distance between relevant and irrelevant monitoring parameters in the same historical working condition slice is obtained;

[0027] By combining the DTW distance and the Euclidean distance, the distance relationship between relevant and irrelevant monitoring parameters in the same historical working condition slice is obtained.

[0028] In some embodiments of the present invention, based on the correlation significance, the differences in the influence of the unrelated monitoring parameters on the related monitoring parameters in different historical operating condition slices are analyzed to obtain the reference factors of the unrelated monitoring parameters on the related monitoring parameters, including:

[0029] The difference in the correlation significance between two adjacent historical working condition slices of the unrelated monitoring parameter in the time series is analyzed, and the influence difference of the unrelated monitoring parameter between adjacent historical working condition slices is obtained by combining the standard deviation of all data of the unrelated monitoring parameter in the historical database.

[0030] By analyzing the difference ratio of the duration of the historical working condition slices and combining it with the influence difference degree, a reference factor for the unrelated monitoring parameter to the related monitoring parameter is obtained.

[0031] In some embodiments of the present invention, analyzing the consistency of changes in monitored parameters between the current operating condition window and the historical operating condition slice to obtain a matching historical operating condition slice for the current operating condition window includes:

[0032] Analyze the differences between the current operating condition status window and each corresponding historical monitoring parameter in the historical operating condition slice to obtain the consistency of changes between the current operating condition status window and the monitoring parameters in each historical operating condition slice;

[0033] The historical working condition slice corresponding to the maximum value of the change consistency is taken as the matching historical working condition slice of the current working condition status window.

[0034] In some embodiments of the present invention, obtaining the reference weights of the relevant monitoring parameters under the current operating condition window and constructing a soft instrument model under the current operating condition window includes:

[0035] The reference weights of the relevant monitoring parameters corresponding to the matched historical working condition slices are used as the reference weights of the relevant monitoring parameters under the current working condition status window.

[0036] Based on the reference weights of the relevant monitoring parameters under the current operating condition window, a soft instrument model under the current operating condition window is constructed.

[0037] In some embodiments of the present invention, after constructing the soft instrument model under the current operating condition window, the method further includes:

[0038] The real-time monitoring parameters in the factory's DCS system are obtained and input into the trained soft instrument model to obtain the soft measurement output value.

[0039] By comparing the simulated dominant variables in the soft measurement output values ​​with the measured dominant variables in the real-time monitoring parameters, the structural parameters of the soft instrument model are improved to obtain an optimized soft instrument model under the current operating condition window.

[0040] Compared with existing technologies, the method for optimizing and constructing soft instrument models based on DCS systems provided by this invention has the following advantages:

[0041] This invention analyzes the fluctuation characteristics of historical monitoring parameters over different time periods to divide production conditions and obtain historical condition slices. Then, it analyzes the significant correlation between relevant and irrelevant monitoring parameters within the same historical condition slice, and further analyzes the differences in the impact of irrelevant monitoring parameters on relevant monitoring parameters in different historical condition slices, comprehensively evaluating the reference weight of relevant monitoring parameters in the historical condition slices. By combining the characteristics of symbiotic or causal relationships of different feature data during the reaction process, this invention enhances the degree of information sharing between particles, effectively handling the complex and redundant feature data in the DCS system. Finally, by analyzing the consistency of changes in monitoring parameters within the current operating condition window and historical condition slices, it obtains matching historical condition slices for the current operating condition window, thereby obtaining the reference weights of relevant monitoring parameters under the current operating condition window and constructing a soft instrument model under the current operating condition window. This soft instrument model is used to extrapolate important parameters that are difficult or impossible to monitor, for controlling actual chemical production. This allows the soft instrument model to adapt to different production conditions, effectively handling the complex and redundant auxiliary variables in the DCS system, and elevating the local optimization of simple parameters in a single loop to the comprehensive analysis and optimization of all quality indicators. Attached Figure Description

[0042] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A basic flowchart illustrating a method for optimizing and constructing a soft instrument model based on a DCS system, as provided in an embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the basic process of another method for optimizing and constructing a soft instrument model based on a DCS system, provided as an embodiment of the present invention. Detailed Implementation

[0045] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the method for optimizing and constructing a soft instrument model based on a DCS system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes that element. Relational terms such as “first” and “second” are used merely to distinguish one entity or operation from another and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0047] The overall logic of this invention is as follows: In chemical production processes, by constructing a soft instrumentation model to realize the changes of difficult-to-measure variables during production, effective prediction and control of the production process can be achieved. This method uses soft instruments to accurately estimate industrial component variables that are difficult to measure directly in industrial processes. Due to changes in environmental indicators, operating instructions, equipment status, and other information, diverse production conditions exist in the production process. The characteristic parameters affecting the reaction process under different conditions have significant differences. Therefore, this invention analyzes relevant monitoring parameters that are correlated with the soft instruments at the target location under different operating conditions, and simultaneously adjusts the weights based on the influence of unrelated monitoring parameters under different operating conditions in the actual production process, enabling the soft instrumentation model to adapt to different production conditions.

[0048] The following description, in conjunction with the accompanying drawings, details the specific scheme of the method for optimizing and constructing a soft instrument model based on a DCS system provided by this invention.

[0049] Please see Figure 1 This illustrates the basic process of a method for optimizing and constructing a soft instrument model based on a DCS system, provided by an embodiment of the present invention.

[0050] like Figure 1 As shown, an embodiment of the present invention provides a method for optimizing and constructing a soft instrument model based on a DCS system, which specifically includes:

[0051] S100: Obtain historical monitoring parameters from the factory's DCS system.

[0052] Historical monitoring parameters are acquired from the plant's DCS (Distributed Control System). Specifically, these parameters include control parameters, equipment status information, and production process data. The DCS system can collect control parameters, equipment status information, and production data from the chemical plant over a continuous six-month period, and perform dynamic monitoring and calculations of the production process data based on these parameters and equipment status information. Control parameters include all sensors deployed throughout the entire production process, with the location information of monitoring points and reactors labeled. This information is used for offline training of soft sensor models to select and construct suitable soft sensor model parameters. Equipment status information includes data on the start-up and shutdown of each reactor during normal operation of the chemical plant within the acquisition period, as well as various states during changes in reactants, reflecting different states under different operating conditions during the production process. Therefore, historical monitoring parameters are acquired from the plant's DCS system.

[0053] The collected historical monitoring parameters are cleaned and standardized to ensure a uniform calculation scale for the parameters generated by different measuring devices. Historical monitoring parameter information for all monitoring points in the production process is compiled (one monitoring parameter is obtained for each monitoring point), and each historical monitoring parameter is arranged in chronological order of collection time to form a historical monitoring parameter data sequence.

[0054] At this point, historical monitoring parameters from the factory's DCS system were obtained.

[0055] Taking the operating data of the methanol-to-olefins (MTO) process as an example, the entire production process includes various historical monitoring parameters such as crude methanol stream, air stream, regenerated catalyst circulating stream, reactor temperature, regenerator temperature, externally supplied steam stream, externally supplied nitrogen stream, and catalyst bed density. Since the concentration of the catalyst stream at the regenerator outlet cannot be directly measured, it is necessary to analyze the data from these various historical monitoring parameters to construct a soft instrumentation model for monitoring the production process. This specifically includes steps S200 to S600.

[0056] S200: Analyze the data fluctuation characteristics of historical monitoring parameters in different time periods to obtain historical operating condition slices.

[0057] In chemical production processes, variations in environmental indicators, operating instructions, and equipment status lead to diverse production conditions. Under different conditions, the characteristic parameters affecting the reaction process exhibit significant differences. While historical monitoring parameters remain constant or change uniformly over time under the same conditions, abrupt changes in historical monitoring parameters typically occur when switching between different conditions. To ensure accurate selection of effective auxiliary variables and enable the constructed instrumentation model to effectively process and analyze data changes under different conditions, it is necessary to divide the collected continuous historical monitoring parameter data into different operating condition segments.

[0058] Based on the above analysis, in the embodiments of the present invention, historical operating condition slices are obtained by analyzing the data fluctuation characteristics of historical monitoring parameters over different time periods. Further, this includes:

[0059] First, a sliding time window is used to traverse historical monitoring parameters, obtaining local maxima within the time window to identify abrupt changes in the historical monitoring parameters. Specifically, taking the first... Taking historical monitoring parameters as an example, firstly, the sliding time window is used to traverse the first... The data sequence of a historical monitoring parameter is used to obtain the local maximum value within each time window. Each local maximum value is taken as the mutation point of the historical monitoring parameter, and all mutation points of the historical monitoring parameter in the current time series are obtained.

[0060] Then, within the time interval between two adjacent mutation points, the fluctuation amplitude of historical monitoring parameters at continuous time points is analyzed. Combined with the average value of historical monitoring parameters, the data fluctuation parameters of each historical monitoring parameter in different time periods are obtained. Specifically, for the first... Historical monitoring parameters, in the first The mutation point and the previous mutation point in the time series (the first mutation point) Within the time period between the mutation points, calculate the absolute value of the difference between two adjacent historical monitoring parameter data, sum the data of all historical monitoring parameters in the time series, and obtain the result. The fluctuation range of historical monitoring parameters over continuous time periods is used to construct the first... Historical monitoring parameters in the first The mutation point and the first The formula for calculating the fluctuation amplitude at consecutive moments between several mutation points is:

[0061]

[0062] In the formula, Indicates the first Historical monitoring parameters in the first The mutation point and the first The fluctuation range at consecutive moments between each mutation point; Indicates the first Historical monitoring parameters in the first The mutation point and the first The consecutive moments between the mutation points One data point; Indicates the first Historical monitoring parameters in the first The mutation point and the first The consecutive moments between the mutation points One data point; Indicates the first Historical monitoring parameters in the first The mutation point and the first The total number of data points between each mutation point; This indicates taking the absolute value.

[0063] For the A type of historical monitoring parameter, calculated by the current time period (the first time period) The mutation point and the first The fluctuation range of historical monitoring parameter data at consecutive moments within a time period (between several mutation points) is used to assess the severity of data fluctuations in the current time period. Combined with the data mean for that current time period, the overall fluctuation of data changes in the current time period is quantified, serving as the first... Historical monitoring parameters in the first The mutation point and the first The data fluctuation parameters between the mutation points over the time interval are used to construct the first mutation point. Historical monitoring parameters in the first The mutation point and the first The formula for calculating the data fluctuation parameter within the time period between several mutation points is:

[0064]

[0065] In the formula, Indicates the first Historical monitoring parameters in the first The mutation point and the first Data fluctuation parameters within the time period between each mutation point; Indicates the first Historical monitoring parameters in the first The mutation point and the first The fluctuation range at consecutive moments between each mutation point; Indicates the first Historical monitoring parameters in the first The mutation point and the first The mean of all data within the time period between each mutation point.

[0066] The ratio of the variation range of historical monitoring parameters to the average value of the data during that period reflects the relative degree of fluctuation of the historical monitoring parameter data. The larger the ratio, the more drastic the fluctuation of the historical monitoring parameter at that time, and the smaller the ratio, the more stable the fluctuation of the historical monitoring parameter at that time.

[0067] Similarly, the data fluctuation parameters of each historical monitoring parameter in the historical database of the DCS system are obtained in different time periods.

[0068] Finally, based on the data fluctuation parameters of each historical monitoring parameter in different time periods, historical operating condition slices are obtained. Specifically, a coordinate system is established with time as the horizontal axis and the data fluctuation parameters of historical monitoring parameters as the vertical axis. The coordinate points of the data fluctuation parameters of all historical monitoring parameters in each time period are taken as particles in the coordinate system, and the historical monitoring parameters of different time periods as a whole form a particle map in the coordinate system. The DBSCA algorithm (Density-Based Spatial Clustering of Applications with Noise) is used to automatically cluster the particles to obtain several clusters. Each cluster is labeled with an operating condition label to obtain historical operating condition slices. Each historical operating condition slice contains several historical monitoring parameters.

[0069] S300: Obtain relevant and irrelevant monitoring parameters for historical operating condition slices, and obtain the correlation significance between relevant and irrelevant monitoring parameters under the same historical operating condition slice.

[0070] The above steps divide all historical monitoring parameters into operating conditions, resulting in historical operating condition slices. In chemical reaction processes, the interactions between various reactants or products, and the overlapping influence of historical monitoring parameter signals between adjacent historical operating condition slices, present challenges. Existing particle swarm optimization algorithms often construct soft measurement models based on single operating conditions, leading to local optima and neglecting the interactions between different reaction parameters or products during production, thus affecting the accuracy of soft measurement results. Therefore, it is necessary to analyze the dynamic information connectivity between different historical monitoring parameters to fully consider the symbiotic or influencing relationships between different parameters in the chemical reaction when determining auxiliary variables for different operating conditions. Furthermore, based on the differences in the manifestation of influence states under different operating conditions, the construction of the soft instrument model can be optimized.

[0071] Based on the above analysis, in the embodiments of the present invention, firstly, relevant and irrelevant monitoring parameters of historical operating condition slices are obtained. Specifically, based on the fluctuation amplitude of historical monitoring parameters at continuous time intervals, combined with the location information of monitoring points in the process flow, a topology diagram of several historical monitoring parameters in the historical operating condition slice is constructed. That is, according to the current industrial production mode, a suitable structure, such as a tree or network structure, is directly selected, and the connection relationship is determined by the correlation between the location information of each monitoring point and the fluctuation amplitude of the historical monitoring parameters of the current monitoring point, thus constructing the topology diagram. Specifically, the method for determining the connection relationship and constructing the topology diagram based on the correlation between the location information of each monitoring point and the fluctuation amplitude of the historical monitoring parameters of the current monitoring point is as follows: calculate the distance between each monitoring point based on the location information of each monitoring point (Euclidean distance, Manhattan distance, or other suitable distance measurement methods can be used), determine the positional correlation between monitoring points based on the distance matrix, and determine the positional correlation between monitoring points based on the historical monitoring parameters. The data fluctuation amplitude is analyzed, and the correlation between the data fluctuation amplitudes of each monitoring point is calculated (methods such as Pearson correlation coefficient and Spearman rank correlation coefficient can be used to measure the correlation of data change amplitudes). The location correlation and data fluctuation amplitude correlation are combined to form a comprehensive correlation matrix (weighted average or other suitable methods can be used to combine these two correlations). Based on the comprehensive correlation matrix, the connection relationships between monitoring points are determined (clustering algorithms can be used to cluster the monitoring points into different adjacent regions. Within each adjacent region, the specific connection relationship between monitoring points is determined based on the comprehensive correlation matrix. For example, a correlation threshold of 0.5 is set; if the comprehensive correlation between two monitoring points in the comprehensive correlation matrix is ​​greater than 0.5, the two monitoring points are connected; otherwise, they are not connected), and a topology diagram is constructed. Based on the topology diagram, relevant monitoring parameters that directly affect the construction of the soft instrument model at the target location are extracted from the historical operating condition slices (historical monitoring parameters that have a direct connection relationship with the target location are relevant monitoring parameters; historical monitoring parameters that affect relevant monitoring parameters but do not have a direct connection relationship with the target location are irrelevant monitoring parameters), thus obtaining the relevant and irrelevant monitoring parameters for the historical operating condition slices.

[0072] Unrelated monitoring parameters affect the target variable's soft measurement results by influencing changes in related monitoring parameters. The impact of these unrelated monitoring parameters varies significantly across different historical operating condition slices. Therefore, it is necessary to obtain the association significance between related and unrelated monitoring parameters within the same historical operating condition slice. Furthermore, by analyzing the distance relationship between related and unrelated monitoring parameters within the same historical operating condition slice, the association significance between related and unrelated monitoring parameters within the same historical operating condition slice can be obtained. Specifically, taking the first... Taking the first historical working condition slice as an example, firstly, calculate the first... Time series data sequences of relevant monitoring parameters in a historical operating condition slice Time series data sequences of unrelated monitoring parameters The DTW (Dynamic Time Warping) distance between them; and, in the topology graph, obtaining the first... The Euclidean distance between relevant and irrelevant monitoring parameters in the first historical operating condition slice is then used. Combining the DTW distance and Euclidean distance, the distance relationship between relevant and irrelevant monitoring parameters in the same historical operating condition slice is obtained, thus yielding the significance of the association between relevant and irrelevant monitoring parameters in the same historical operating condition slice. The first... The formula for calculating the significance of the association between relevant and irrelevant monitoring parameters under a historical working condition slice is as follows:

[0073]

[0074] In the formula, Indicates the first The first historical working condition slice Related monitoring parameters and the first The significance of the association between unrelated monitoring parameters; Indicates the first The first historical working condition slice The location of the relevant monitoring parameters in the topology diagram; Indicates the first The first historical working condition slice for the The first type of relevant monitoring parameter has an impact The location of unrelated monitoring parameters in the topology diagram; Indicates the first Time series data sequences of relevant monitoring parameters in a historical operating condition slice Time series data sequences of unrelated monitoring parameters DTW distance between them; Indicates taking the absolute value; Represented by natural constant An exponential function with base 0.

[0075] Indicates the first The first historical working condition slice Unrelated monitoring parameters and the first The Euclidean distance between the locations of the two related monitoring parameters in the topology diagram is used to determine the correlation significance between them. The smaller the value, the shorter the connection path length between the two historical monitoring parameters in the topology diagram. Indicates the first The first historical working condition slice Data sequences of unrelated monitoring parameters and the first The DTW distance between data sequences of two related monitoring parameters reflects the degree of matching between the two data sequences over time. The smaller the DTW distance, the higher the degree of matching between the data sequences, indicating that the degree of change between the two historical monitoring parameters is more similar, that is, the greater the significance of the association between the two historical monitoring parameters.

[0076] S400: Based on the significance of association, analyze the differences in the influence of unrelated monitoring parameters on related monitoring parameters in different historical working condition slices, and obtain the reference factor of unrelated monitoring parameters on related monitoring parameters.

[0077] Based on the characteristics of the interaction between various reactants or products in the chemical reaction process, the higher the correlation significance, the stronger the possibility of symbiosis between two historical monitoring parameter signals. When there are obvious changes in the symbiotic parameter signals, they need to be given special attention during the model construction process. Therefore, it is necessary to conduct joint analysis of the degree of change of historical monitoring parameter signals with correlation under different operating conditions.

[0078] Based on the above analysis, in the embodiments of the present invention, based on the correlation significance, the differences in the influence of unrelated monitoring parameters on related monitoring parameters in different historical operating condition slices are analyzed to obtain the reference factor of unrelated monitoring parameters on related monitoring parameters. Specifically, the difference in correlation significance between two adjacent historical operating condition slices of unrelated monitoring parameters in the time series is analyzed, and combined with the standard deviation of all data of unrelated monitoring parameters in the historical database, the difference in the influence of unrelated monitoring parameters on related monitoring parameters in adjacent historical operating condition slices is obtained; the difference ratio of the duration of historical operating condition slices is analyzed, and combined with the difference in influence, the reference factor of unrelated monitoring parameters on related monitoring parameters is obtained. Therefore, the first... The type of unrelated monitoring parameter for the first The formula for calculating the reference factor of the relevant monitoring parameters is as follows:

[0079]

[0080] In the formula, Indicates the first The first historical working condition slice The type of unrelated monitoring parameter for the first Reference factors for relevant monitoring parameters; Indicates the first The duration of a historical working condition slice; No. The duration of a historical working condition slice; This indicates taking the maximum value; Indicates the first The first historical working condition slice Related monitoring parameters and the first The significance of the association between unrelated monitoring parameters; Indicates the first The first historical working condition slice Related monitoring parameters and the first The significance of the association between unrelated monitoring parameters; Indicates the first The standard deviation of all data in the historical database for a type of unrelated monitoring parameter; This indicates taking the absolute value.

[0081] The larger the ratio, the greater the difference in duration between two adjacent production conditions, which is used to amplify the production differences between different reaction production conditions; By calculating the first [missing information] between adjacent production conditions Unrelated monitoring parameters and the first The absolute value of the difference in the significance of the association between the relevant monitoring parameters and the current number of parameters. The ratio of the standard deviations of all data in the historical database for a non-correlated monitoring parameter reflects the correlation between adjacent production conditions. The greater the ratio of the information differences among the unrelated monitoring parameters, the more significant the difference. The greater the degree of variation of the first type of unrelated monitoring parameter under different production conditions, the more likely the second type will change. The type of unrelated monitoring parameter for the first The greater the influence of the relevant monitoring parameters, therefore, the first The type of unrelated monitoring parameter for the first The larger the reference factor of a relevant monitoring parameter, the better.

[0082] Similarly, to obtain the first step in the production process All irrelevant monitoring parameters under the first historical operating condition slice for the first Reference factors for various relevant monitoring parameters, and reference factors for all irrelevant monitoring parameters and their corresponding relevant monitoring parameters under each historical working condition slice in the production process.

[0083] S500: Based on the reference factor and the correlation significance, the reference weights of the relevant monitoring parameters in the historical operating condition slices are obtained.

[0084] In some cases, small changes in operating conditions may have a significant impact on relevant monitoring parameters, or changes in operating conditions may cause significant changes in the correlation significance of relevant monitoring parameters. The difference in correlation significance between adjacent historical operating condition slices can reflect the dynamic characteristics of changes between operating conditions, capture the impact of such changes, and help the soft instrument model better adapt to the actual production environment.

[0085] Therefore, based on the reference factor and combined with the correlation significance, the reference weights of relevant monitoring parameters in the historical operating condition slices are obtained. Specifically, this is achieved by calculating the reference weights of adjacent historical operating condition slices (the first slice). A slice of historical working conditions and the first (A slice of historical working conditions) Part 1 Related monitoring parameters and the first The absolute value of the difference in the significance of the association between the two unrelated monitoring parameters, and combined with the first The first historical working condition slice The type of unrelated monitoring parameter for the first Reference factors for relevant monitoring parameters were obtained to obtain the first... The first historical working condition slice The relevant monitoring parameters are used as reference weights when constructing the soft instrument model, then the construction of the first... The first historical working condition slice The formula for calculating the reference weights of relevant monitoring parameters when constructing a soft instrument model is as follows:

[0086]

[0087] In the formula, Indicates the first The first historical working condition slice The relevant monitoring parameters are used as reference weights when constructing the soft instrument model; Indicates the first The first historical working condition slice The type of unrelated monitoring parameter for the first Reference factors for relevant monitoring parameters; Indicates the first The first historical working condition slice Related monitoring parameters and the first The significance of the association between unrelated monitoring parameters; Indicates the first The first historical working condition slice Related monitoring parameters and the first The significance of the association between unrelated monitoring parameters; Indicates taking the absolute value; Indicates the first The first historical working condition slice for the The number of unrelated monitoring parameters that have an impact on the relevant monitoring parameters.

[0088] By traversing the first All irrelevant monitoring parameters in the first historical operating condition slice affect the first The influence of various relevant monitoring parameters is comprehensively determined to identify the first factor that currently affects the target variable. The influence reference weights of relevant monitoring parameters are determined. Then, based on weight analysis and the support vector machine algorithm, the first... A soft instrument model for the target variable under a historical operating condition slice.

[0089] S600: Analyze the consistency of changes in monitoring parameters between the current operating condition window and the historical operating condition slices, obtain the matching historical operating condition slices for the current operating condition window, and then obtain the reference weights of relevant monitoring parameters under the current operating condition window to construct a soft instrument model under the current operating condition window.

[0090] For real-time monitoring parameter data collected during the production process, the attribution information of the real-time monitoring parameter data is first calculated to obtain the production condition status window to which the real-time monitoring parameter data belongs. Then, by analyzing the consistency of changes between the real-time monitoring parameter data in the production condition status window and the historical monitoring parameter data in the historical condition slices (the aforementioned historical condition slices obtained from the historical monitoring parameter data), historical condition slices with high consistency with the current production condition status window are selected. This indicates that the higher the probability that the current production process matches the historical condition slice, the more consistent the soft instrument model in the production process is with the soft instrument model obtained at the historical condition slice.

[0091] Therefore, in embodiments of the present invention, by analyzing the consistency of changes in monitoring parameters (real-time monitoring parameters and historical monitoring parameters) within the current operating condition window and historical operating condition slices, a matching historical operating condition slice for the current operating condition window is obtained, thereby obtaining the reference weights of relevant monitoring parameters under the current operating condition window and constructing a soft instrument model under the current operating condition window. Further, it includes:

[0092] First, the differences between the current operating condition window and each corresponding historical monitoring parameter in the historical operating condition slices are analyzed to obtain the consistency of changes in the monitoring parameters between the current operating condition window and each historical operating condition slice. Specifically, the first... Taking a historical operating condition slice as an example, if the amount of data in the production operating condition status window is greater than the historical slice... The amount of data within a historical operating condition slice is first processed by proportionally shrinking the production operating condition status window to the size of the historical slice. A data segment with an equal amount of data from each historical working condition slice; then calculate the first... Within the historical working condition slice, the first The difference between each data point of a historical monitoring parameter and the previous data point is recorded as the first difference; and the difference between the first and second data points is calculated within the current operating condition window (the production operating condition window to which the real-time monitoring parameter data belongs) for the corresponding category (the first...). The difference between each data point of the real-time monitoring parameter and the previous data point is recorded as the second difference; then the absolute value of the difference between the first difference and the second difference at the corresponding position is calculated; this process is repeated for each data point. The historical operating condition slice and the current operating condition status window contain all data for all types of monitoring parameters. This allows us to obtain the current operating condition status window and the data for the first historical operating condition slice. Consistency of changes in monitored parameters within each historical operating condition slice. Furthermore, constructing the current operating condition status window and the... The formula for calculating the consistency of changes in monitoring parameters within a historical operating condition slice is as follows:

[0093]

[0094] In the formula, The current operating status window is related to the first... Consistency of changes in monitoring parameters within each historical operating condition slice; This indicates the current operating status window. The first type of real-time monitoring parameter The difference between each data point and the previous data point; Indicates the first Within the historical working condition slice, the first The first type of historical monitoring parameter The difference between each data point and the previous data point; Indicates the first The total number of data points in the various monitoring parameters; The current operating status window is related to the first... The total number of the same type of monitoring parameters within a historical working condition slice; Represented by natural constant An exponential function with base 0; This indicates taking the absolute value.

[0095] This indicates the consistency of data changes at corresponding locations; the smaller the difference, the better the consistency between the current operating condition window and the previous one. The data trends of the historical operating condition slices are consistent, and the greater the consistency of data changes, the better the current operating condition status window is compared with the first slice. The greater the likelihood that historical operating condition slices belong to the same operating condition.

[0096] Similarly, the consistency of changes in monitoring parameters between the current operating condition window and all historical operating condition slices is obtained.

[0097] Then, the historical working condition slice corresponding to the maximum value of change consistency is taken as the matching historical working condition slice of the current working condition status window.

[0098] Finally, the reference weights of the relevant monitoring parameters under the current operating condition window are obtained, and a soft instrument model under the current operating condition window is constructed. Specifically, the reference weights of the relevant monitoring parameters corresponding to the historical operating condition slices are used as the reference weights of the relevant monitoring parameters under the current operating condition window, denoted as . Based on the reference weights of the relevant monitoring parameters under the current operating condition window, a soft instrument model under the current operating condition window is constructed.

[0099] Please see Figure 2 This illustrates the basic process of another method for optimizing and constructing a soft instrument model based on a DCS system, provided by an embodiment of the present invention.

[0100] like Figure 2 As shown, in order to enable the soft instrument model to adapt to various working conditions in the actual production process, after constructing the soft instrument model under the current working condition state window, it also includes:

[0101] S700: Acquires real-time monitoring parameters from the factory's DCS system, inputs them into a trained soft instrument model, and obtains soft measurement output values.

[0102] S800: Compares the simulated dominant variables in the soft measurement output values ​​with the measured dominant variables in the real-time monitoring parameters, adjusts the reference weights of relevant monitoring parameters in the soft instrument model, and obtains the optimized soft instrument model under the current operating condition window.

[0103] By comparing the simulated dominant variables in the soft sensor output values ​​with the measured dominant variables in the real-time monitoring parameters, the structural parameters of the soft instrument model are improved to obtain an optimized soft instrument model under the current operating condition window. Specifically, the dominant variable refers to the actual physical variable at the location where the soft instrument model is constructed, which is a variable that cannot be obtained in real time by sensors in the production line, such as flow rate and concentration. The deviation between the simulated and measured dominant variables in the soft sensor output values ​​obtained from the soft instrument model under the current operating condition window is calculated, where the deviation calculation formula is:

[0104]

[0105] In the formula, This indicates the degree of deviation between the simulated dominant variable and the measured dominant variable in the soft measurement output value obtained from the soft instrument model under the current operating condition window; Represents the measured dominant variables at the location where the soft instrument model is constructed; This represents the dominant analog variable in the soft measurement output value obtained from the soft instrument model under the current operating condition window. This represents the minimum value greater than 0. This is to prevent the denominator from being 0, so that it is set ; This represents a linear normalization function that maps numerical values ​​to the range of -1 to 1.

[0106] The construction effect of the soft instrument model under the current operating condition window is determined based on the degree of deviation. A large degree of deviation indicates a poor model fit, requiring adjustment of the reference weights of relevant monitoring parameters in the soft instrument model. The specific adjustment method is as follows: preset a deviation threshold, which can be 0.2; if the deviation degree... If the deviation exceeds the threshold, it indicates a poor model fit. Therefore, the reference weights of the relevant monitoring parameters in the soft instrument model under the current operating condition window are adjusted. The adjusted reference weights of the relevant monitoring parameters are as follows:

[0107]

[0108] In the formula, This indicates the adjusted first [number]th [item] in the soft instrument model under the current operating condition window. Reference weights for relevant monitoring parameters; This indicates the first soft instrument model in the current operating condition window. Reference weights for relevant monitoring parameters; This indicates the degree of deviation between the simulated dominant variable and the measured dominant variable in the soft measurement output value obtained from the soft instrument model under the current operating condition window.

[0109] By optimizing the soft instrument model through steps S700 and S800, the soft instrument model can adapt to various operating conditions in the actual production process. At the same time, real-time control of the production process is carried out based on the degree of deviation between the measurement results and the target state, ensuring precise control of the chemical production process.

[0110] The entire process simulation system was constructed using six months of industrial production data to fit the kinetic parameters in the model. Calibration was performed using data from January to June to ensure that the parameter tuning accuracy and results of the soft instrumentation model met the production requirements of the chemical plant. Then, the production data from July and August were substituted into the soft instrumentation model to calculate the error between the predicted and actual values, thereby verifying the model's broad applicability.

[0111] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0112] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for optimizing and constructing a soft instrument model based on a DCS system, characterized in that, The method includes: Obtain historical monitoring parameters from the factory's DCS system; Analyze the data fluctuation characteristics of the historical monitoring parameters in different time periods to obtain historical operating condition slices; Obtain relevant and irrelevant monitoring parameters for the historical operating condition slices, and obtain the correlation significance between relevant and irrelevant monitoring parameters under the same historical operating condition slices; Based on the correlation significance, the differences in the influence of the unrelated monitoring parameters on the related monitoring parameters in different historical working condition slices are analyzed to obtain the reference factor of the unrelated monitoring parameters on the related monitoring parameters. Based on the reference factor and the correlation significance, the reference weights of the relevant monitoring parameters in the historical operating condition slice are obtained. Analyze the consistency of changes in monitoring parameters between the current operating condition window and the historical operating condition slices to obtain the matching historical operating condition slices for the current operating condition window. Then, obtain the reference weights of the relevant monitoring parameters under the current operating condition window and construct a soft instrument model under the current operating condition window.

2. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 1, characterized in that, Analyze the data fluctuation characteristics of each historical monitoring parameter over different time periods, including: By traversing the historical monitoring parameters using a sliding time window, the local maximum value within the time window is obtained, thus identifying the abrupt change point of the historical monitoring parameters. Within the time interval between two adjacent mutation points, the fluctuation range of the historical monitoring parameters at continuous times is analyzed, and combined with the average value of the historical monitoring parameters, the data fluctuation parameters of each historical monitoring parameter in different time periods are obtained.

3. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 2, characterized in that, Obtain historical operating condition slices, including: Treating the data fluctuation parameters within each time period as a particle, a coordinate system is established to form a particle diagram. The particles were automatically clustered using DBSCA, and each cluster was labeled with a working condition tag to obtain historical working condition slices. Each of the historical operating condition slices contains several types of historical monitoring parameters.

4. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 3, characterized in that, The relevant and irrelevant monitoring parameters of the historical operating condition slices are obtained, including: Based on the fluctuation range of the historical monitoring parameters at continuous times, and combined with the location information of the monitoring points in the process flow, a topological structure diagram of several historical monitoring parameters in the historical operating condition slice is constructed. The relevant monitoring parameters that directly affect the construction of the soft instrument model of the target position in the historical operating condition slice are extracted, and the relevant and irrelevant monitoring parameters of the historical operating condition slice are obtained.

5. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 4, characterized in that, Obtain the significance of the association between relevant and irrelevant monitoring parameters under the same historical operating condition slice, including: By analyzing the distance relationship between relevant and irrelevant monitoring parameters in the same historical working condition slice, the correlation significance between relevant and irrelevant monitoring parameters in the same historical working condition slice is obtained.

6. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 5, characterized in that, Analyze the distance relationship between relevant and irrelevant monitoring parameters in the same historical working condition slice, including: Calculate the DTW distance between the time series data sequences of relevant monitoring parameters and the time series data sequences of unrelated monitoring parameters in the same historical working condition slice; In the topology diagram, the Euclidean distance between relevant and irrelevant monitoring parameters in the same historical working condition slice is obtained; By combining the DTW distance and the Euclidean distance, the distance relationship between relevant and irrelevant monitoring parameters in the same historical working condition slice is obtained.

7. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 1, characterized in that, Based on the correlation significance, the differences in the influence of the unrelated monitoring parameters on the related monitoring parameters in different historical operating condition slices are analyzed to obtain the reference factors of the unrelated monitoring parameters on the related monitoring parameters, including: The difference in the correlation significance between two adjacent historical working condition slices of the unrelated monitoring parameter in the time series is analyzed, and the influence difference of the unrelated monitoring parameter between adjacent historical working condition slices is obtained by combining the standard deviation of all data of the unrelated monitoring parameter in the historical database. By analyzing the difference ratio of the duration of the historical working condition slices and combining it with the influence difference degree, a reference factor for the unrelated monitoring parameter to the related monitoring parameter is obtained.

8. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 1, characterized in that, Analyze the consistency of changes in monitored parameters between the current operating condition window and the historical operating condition slices to obtain matching historical operating condition slices for the current operating condition window, including: Analyze the differences between the current operating condition status window and each corresponding historical monitoring parameter in the historical operating condition slice to obtain the consistency of changes between the current operating condition status window and the monitoring parameters in each historical operating condition slice; The historical working condition slice corresponding to the maximum value of the change consistency is taken as the matching historical working condition slice of the current working condition status window.

9. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 8, characterized in that, Obtain the reference weights of the relevant monitoring parameters under the current operating condition window, and construct a soft instrument model under the current operating condition window, including: The reference weights of the relevant monitoring parameters corresponding to the matched historical working condition slices are used as the reference weights of the relevant monitoring parameters under the current working condition status window. Based on the reference weights of the relevant monitoring parameters under the current operating condition window, a soft instrument model under the current operating condition window is constructed.

10. The method for optimizing and constructing a soft instrument model based on a DCS system according to claim 1, characterized in that, After constructing the soft instrument model under the current operating condition window, it also includes: The real-time monitoring parameters in the factory's DCS system are obtained and input into the trained soft instrument model to obtain the soft measurement output value. By comparing the simulated dominant variables in the soft measurement output values ​​with the measured dominant variables in the real-time monitoring parameters, the reference weights of the relevant monitoring parameters in the soft instrument model are adjusted to obtain the optimized soft instrument model under the current operating condition window.

Images