Parameter setting model construction method and industrial process control method

Abstract

The embodiments of the present application provide a method for constructing a parameter setting model and a method for controlling an industrial process, which relates to the technical field of industrial automation control. The present application constructs an auxiliary training data set and a validation data set based on loop information of a new device and operation data of an auxiliary device, the new device being a device to which the parameter setting model to be trained is applied, and the auxiliary device being a device that has completed parameter setting and is officially operated, and a local training data set is constructed based on the initial operation data of the new device, and a parameter setting model is obtained by training based on the auxiliary training data set, the validation data set and the local training data set.

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority from Chinese Patent Application No. 202210859981.X, entitled "Method for constructing a parameter setting model and a method for controlling an industrial process," filed with the State Intellectual Property Office of the People's Republic of China on July 22, 2022, the entire contents of which are incorporated herein by reference.

[0002] The present application relates to the technical field of industrial automation control, and more particularly to a method for constructing a parameter setting model and a method for controlling an industrial process. [Background technology]

[0003] PID (proportion, integral, differential) is the most widely applied control strategy at present. PID parameter setting is the main content of its control work, which determines the deviation values ​​of its proportional coefficient, integral time and differential time based on the operating characteristics of the control system.

[0004] Currently, the PID parameter setting method is generally based on an internal model. In this method, a mathematical model for PID parameter setting is constructed based on process history data, and the PID parameters are obtained using an internal model setting policy based on the mathematical model. This is an effective PID parameter setting method.

[0005] However, the internal model-based tuning method requires a large amount of valid data to build a process model. For a process model built on a new device, the available data information is small and insufficient to build a reliable process model, so the actual effectiveness of PID parameter tuning is low. Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present application is to provide a parameter setting model construction method and an industrial process control method that effectively compensates for the drawback of limited valid data available for new equipment, for example, by constructing an auxiliary training data set and performing training, and improves the accuracy of PID parameter setting in the parameter setting model constructed on the new equipment. [Means for solving the problem]

[0007] The embodiment of the present application is realized as follows.

[0008] According to a first aspect, an embodiment of the present application comprises: constructing an auxiliary training data set and a validation data set based on loop information of a new device and operation data of an auxiliary device, where the new device is a device to which a parameter setting model to be trained is applied, and the auxiliary device is a device for which parameter setting has been completed and which has been officially operated; building a local training data set based on initial operating data of the new device; training based on the auxiliary training data set, the validation data set and the local training data set to obtain a parameter setting model.

[0009] In a preferred embodiment, the step of constructing the auxiliary training data set and the validation data set based on the loop information of the new device and the operating data of the auxiliary device comprises: selecting a target data set from the auxiliary device's operating data based on the new device's loop type and physical characteristics corresponding to the control loop; performing a splitting process on the target dataset to obtain the supplemental training dataset and the validation dataset.

[0010] In a preferred embodiment, the step of selecting a target data set from the operational data of the auxiliary device based on a loop type and physical characteristics corresponding to a control loop of the new device comprises: selecting, based on the loop type of the new device, a plurality of optional operation data items matching the loop type from the operation data of the auxiliary device; and selecting the target data set from the plurality of optional operating data based on physical characteristics corresponding to a control loop of the new device.

[0011] In a preferred embodiment, the step of training to obtain a parameter-setting model based on the supplemental training data set, the validation data set, and the local training data set comprises: training a mid-set model based on the auxiliary training data set and the local training data set; validating the mid-settlement model based on the validation data set, and after passing the validation, executing the mid-settlement model according to predetermined input parameter values ​​to obtain output parameter values ​​of the mid-settlement model; obtaining loop PID parameter values ​​after the new device has operated according to the predetermined input parameter values; and performing parameter optimization on the intermediate settling model based on the loop PID parameter values ​​and the output parameter values ​​to obtain the parameter settling model.

[0012] In a preferred embodiment, the step of performing parameter optimization on the intermediate settling model based on the loop PID parameter values ​​and the output parameter values ​​to obtain the parameter settling model includes: determining a parameter error rate based on the loop PID parameter values ​​and the output parameter values; and iteratively modifying model parameters of the intermediate settling model based on the parameter error rate to obtain the parameter settling model.

[0013] In a preferred embodiment, the model parameters include a first intermediate weight vector and an intermediate offset parameter value; Iteratively modifying model parameters of the intermediate settling model based on the parameter error rate to obtain the parameter settling model, modifying the intermediate offset parameter value based on the parameter error rate to obtain a process offset parameter value; modifying the first intermediate weight vector based on the process offset parameter value to obtain a first process weight vector; deriving a new mid-settlement model based on the process offset parameter value and the first process weight vector, and redetermining a parameter error rate of the mid-settlement model; repeating the steps above until the parameter error rate becomes smaller than a predetermined threshold, and setting the intermediate settling model as the parameter settling model.

[0014] In a preferred embodiment, before the step of training to obtain an intermediate settling model based on the auxiliary training data set and the local training data set, the method further comprises: determining a root mean square error and a coefficient of determination based on the supplemental training data set and the local training data set; determining the number of hidden layer nodes of the parameter-setting initial model based on the root mean square error and the coefficient of determination; constructing the parameter-setting initial model based on the number of hidden layer nodes, a predetermined number of input layer nodes, and a predetermined number of output layer nodes; The method further includes training the parameter-set initial model based on the auxiliary training data set and the local training data set to obtain the intermediate parameter-set model.

[0015] According to a second aspect, an embodiment of the present application comprises: determining setting parameter values ​​of a new device to be controlled based on the parameter setting model, the parameter setting model being obtained by the parameter setting model construction method according to any one of the first aspect; and controlling the new device to perform a target process based on the settling parameter values.

[0016] According to a third aspect, an embodiment of the present application provides an apparatus for constructing a parameter setting model, including a dataset construction module, a model training module, and a model construction module.

[0017] the dataset construction module is configured to construct an auxiliary training dataset and a validation dataset based on loop information of a new device and operation data of an auxiliary device, the new device being a device to which a parameter setting model to be trained is applied, and the auxiliary device being a device whose parameter setting has been completed and which has been officially operated; The dataset construction module is further configured to construct a local training dataset based on initial operating data of the new device.

[0018] A model training module is configured to train to obtain a parameter-setting model based on the supplemental training data set, the validation data set, and the local training data set.

[0019] The dataset construction module is specifically configured to select a target dataset from the operation data of the auxiliary device based on the loop type and physical characteristics corresponding to the control loop of the new device, and then perform a division process on the target dataset to obtain the auxiliary training dataset and the validation dataset.

[0020] The dataset construction module is specifically configured to select, based on the loop type of the new device, a plurality of optional driving data matching the loop type from the driving data of the auxiliary device, and to select the target dataset from the plurality of optional driving data based on physical characteristics corresponding to the control loop of the new device.

[0021] The model training module is specifically configured to train based on the auxiliary training data set and the local training data set to obtain a mid-settlement model, verify the mid-settlement model based on the verification data set, execute the mid-settlement model according to predetermined input parameter values ​​after the verification is passed, obtain output parameter values ​​of the mid-settlement model, obtain loop PID parameter values ​​after the new device operates according to the predetermined input parameter values, and perform parameter optimization on the mid-settlement model based on the loop PID parameter values ​​and the output parameter values ​​to obtain the parameter setting model.

[0022] The model training module is specifically configured to determine a parameter error rate based on the loop PID parameter values ​​and the output parameter values, and iteratively modify model parameters of the intermediate settling model based on the parameter error rate to obtain the parameter settling model.

[0023] The model training module is specifically configured to: the model parameters include a first intermediate weight vector and an intermediate offset parameter value; modify the intermediate offset parameter value based on the parameter error rate to obtain a process offset parameter value; modify the first intermediate weight vector based on the process offset parameter value to obtain a first process weight vector; obtain a new intermediate settling model based on the process offset parameter value and the first process weight vector; re-determine a parameter error rate of the intermediate settling model; and repeat the above steps until the parameter error rate is smaller than a predetermined threshold, and use the intermediate settling model as the parameter settling model.

[0024] The model construction module is configured to determine a root mean square error and a coefficient of determination based on the auxiliary training data set and the local training data set, determine the number of hidden layer nodes of a parameter-set initial model based on the root mean square error and the coefficient of determination, construct the parameter-set initial model based on the number of hidden layer nodes, a predetermined number of input layer nodes, and a predetermined number of output layer nodes, and train the parameter-set initial model based on the auxiliary training data set and the local training data set to obtain the intermediate setting model.

[0025] According to a fourth aspect, an embodiment of the present application comprises: a determination module that determines setting parameter values ​​of a new device to be controlled based on the parameter setting model, the parameter setting model being obtained by the parameter setting model construction method according to any one of the first aspect; and a control module configured to control the new device to perform a target process based on the settling parameter values.

[0026] According to a fifth aspect, an embodiment of the present application provides a processing device including a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the processing device is operating, the processor and the storage medium communicate via the bus, and the processor executes the machine-readable instructions to perform steps of the parameter setting model construction method described in any one of the first aspect or the industrial process control method described in the second aspect.

[0027] According to a sixth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, the computer program implementing, when executed by a processor, steps of the parameter setting model construction method according to any one of the first aspect or the industrial process control method according to the second aspect. [Effects of the Invention]

[0028] The effects of the embodiment of the present application are as follows.

[0029] According to the parameter setting model construction method and industrial process control method of the present application, an auxiliary training data set is constructed using the operating data of an auxiliary device, and a parameter setting model is trained and constructed together with a local training data set constructed using the operating data of a new device. In this application, the operating data of the auxiliary device that has already completed parameter setting and has been officially operated is fully utilized to train the parameter setting model, thereby compensating for the shortcoming of the limited valid data available for the new device and improving the accuracy and efficiency of PID parameter setting of the parameter setting model constructed on the new device. [Brief explanation of the drawings]

[0030] In order to more clearly explain the technical solutions in the embodiments of the present application, the following briefly describes the drawings that need to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without any creative effort.

[0031] [Figure 1] 1 is a flowchart of steps in a method for constructing a parameter setting model according to an embodiment of the present application. [Figure 2] 1 is a flowchart of a data set construction step of a parameter setting model construction method according to an embodiment of the present application. [Figure 3] 1 is a flowchart of a data set selection step of a parameter setting model construction method according to an embodiment of the present application. [Figure 4] 1 is a flowchart of a model training optimization step of a method for building a parameter setting model according to an embodiment of the present application. [Figure 5] 1 is a flowchart illustrating an implementation of a parameter setting model construction method according to an embodiment of the present application. [Figure 6] 1 is a flowchart of a parameter optimization step of a parameter setting model construction method according to an embodiment of the present application. [Figure 7] 10 is a flowchart of another parameter optimization step of the parameter setting model construction method according to an embodiment of the present application. [Figure 8] 1 is a flowchart of a model construction step of a parameter setting model construction method according to an embodiment of the present application. [Figure 9] 1 is a flowchart of steps of an industrial process control method according to an embodiment of the present application. [Figure 10] 1 is a schematic configuration diagram of a parameter setting model construction device according to an embodiment of the present application. [Figure 11] 1 is a schematic configuration diagram of an industrial process control device according to an embodiment of the present application. [Figure 12]1 is a schematic configuration diagram of a processing device according to an embodiment of the present application; [Explanation of symbols]

[0032] 100 Parameter setting model construction device 1001 Dataset Construction Module 1002 Model Training Module 1003 Model Building Module 110 Industrial process control equipment 1101 Decision Module 1102 Control Module 2001 processor 2002 Memory DETAILED DESCRIPTION OF THE INVENTION

[0033] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application, and of course, the described embodiments are only some embodiments of the present application, but not all embodiments. Generally, the components of the embodiments of the present application described and illustrated in the drawings herein can be arranged and designed in various different configurations.

[0034] Therefore, the detailed description of the embodiments of the present application provided in the accompanying drawings below is not intended to limit the scope of the present application as claimed, but to illustrate only selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments that can be obtained by a person skilled in the art without any creative work fall within the scope of protection of the present application.

[0035] Note that in the following figures, like numbers or letters represent like items, so once an item is defined in one figure, it does not need to be further defined or explained in subsequent figures.

[0036] Additionally, terms such as "first" and / or "second" are used solely to distinguish between descriptions and should not be understood as indicating or implying relative importance.

[0037] It should be noted that the features of the embodiments of the present application can be combined with each other without conflict.

[0038] PID parameter tuning is a process of adjusting three parameters (proportional, integral, and differential) to achieve system dynamic and static performance requirements and optimize certain performance indicators. Currently, the most common parameter tuning method is based on internal model tuning. This method involves collecting historical input-output relationship parameters for a target device to build a loop process model, and then using an internal model tuning policy to obtain parameter tuning results based on the loop process model, thereby obtaining real-time PID parameters for the target device. However, the model obtained in this way requires a large amount of valid historical data as its driving force. For new devices, the available data information is limited, making it insufficient to train a reliable model, resulting in low PID parameter tuning effectiveness.

[0039] In view of this, the applicant has conducted research and proposed a method for constructing a parameter setting model and a method for controlling an industrial process, which uses an auxiliary device for which parameter setting has been completed to construct an auxiliary training data set, and assists training with a local training data set to construct a parameter setting model, thereby avoiding the problem of insufficient model training and low accuracy due to a lack of historical data, and improving the accuracy and efficiency of PID parameter setting for new devices.

[0040] Transfer learning mimics the thought process of the human brain, allowing humans to solve one problem and then have a better and faster solution to a new related problem. In other words, transfer learning differs from traditional machine learning methods in that it uses "knowledge" learned from tasks in the same field as the target field, such as data features and model parameters, to assist the learning process in the new field, thereby obtaining a model that can be applied to the target field.

[0041] An industrial control system includes multiple different types of control loops, and different industrial control systems may have different scales and data distributions, but at the control loop level, there may be some similar characteristic information between the control loops of different industrial control systems. Based on this, in an embodiment of the present application, a parameter setting model construction method and an industrial process control method are provided, which use transfer learning to apply operating data of auxiliary devices whose parameters have been set to training of a parameter setting model for a new device, thereby solving the problem that the existing network model is highly dependent on the valid data of the new device during training, and improving the speed and accuracy of building a parameter setting model for the new device.

[0042] Hereinafter, a method for constructing a parameter setting model and a method for controlling an industrial process according to an embodiment of the present application will be described with reference to several specific application examples.

[0043] 1 is a flowchart of steps of a method for constructing a parameter setting model according to an embodiment of the present application, which may be performed by a computer device having calculation and processing capabilities. As shown in FIG. 1, the method includes the following steps S101 to S103.

[0044] In S101, an auxiliary training data set and a validation data set are constructed based on the loop information of the new device and the operation data of the auxiliary device.

[0045] The new device is a device to which the parameter setting model to be trained is applied, and the auxiliary device is a device for which parameter setting has been completed and which has been officially operated.

[0046] The new unit may be a unit whose construction has been completed and for which a parameter setting model for PID parameter setting needs to be constructed. Exemplarily, the new unit may be any one or more of a plurality of units included in an ethylene system, such as an ethylene unit, a cracked-gasoline hydrogenation unit, a butadiene extraction unit, an aromatic hydrocarbon extraction unit, an MTBE / butene-1 unit, an ethylene glycol unit, and a POX unit. In some embodiments, each unit may further include a plurality of control loops.

[0047] The auxiliary device may be a device that includes multiple control loops, has completed parameter setting, and is in a normal operating state, and one or more of these control loops can match some characteristics of the control loops of the new device. Note that the auxiliary device may be one or more, and the control loops included in the multiple auxiliary devices jointly constitute a control loop set that matches the multiple control loops of the new device in a one-to-one correspondence.

[0048] a supplementary training data set based on the operating data generated in the control loop set, the training data set having no overlapping data; [Number 1] TIFF0007760744000001.tif19170 (where, TIFF0007760744000002.tif719, where i=1, 2, ..., n), and a validation dataset [Number 2] TIFF0007760744000003.tif15170 (where, TIFF0007760744000004.tif719, where i=1, 2, ..., k). TIFF0007760744000005.tif76, TIFF0007760744000006.tif1117 is the auxiliary training dataset T a are the input and output parameters of TIFF0007760744000007.tif76, TIFF0007760744000008.tif1117 are the input and output parameters of the validation dataset S, respectively, and X a is the driving data of the auxiliary device, and n is the auxiliary training data set T a is the control loop data included in the validation data set S, and k is the number of control loops included in the validation data set S.

[0049] In S102, a local training data set is constructed based on the initial operating data of the new device.

[0050] In some embodiments, a new device is manually configured for PID parameters by inputting some initial input parameters into the new device, and obtaining corresponding initial output parameters. A local training dataset is then generated based on the initial input parameters, the initial output parameters, and the corresponding relationship between them. [Number 3] TIFF0007760744000009.tif16170 can be constructed where, TIFF0007760744000010.tif819, i=1, 2, ..., m, where: TIFF0007760744000011.tif87 is the local training dataset T b are the input parameters of TIFF0007760744000012.tif1117 is the local training dataset T b is the output parameter of X b is the initial operating data for the new equipment.

[0051] In addition, because there is little initial operating data obtained by manually setting the PID parameters, the auxiliary training dataset T a, the amount of data contained in the validation dataset S is b The amount of data may be much larger than the

[0052] In some embodiments, the auxiliary training data set T a , validation dataset S, local training dataset T b The input parameters and output parameters may be the same. The input parameters include the device to which the control loop belongs, the loop type, and the closed-loop steady-state time T s , peak time T p , maximum overshoot amount Y max , the positive and negative effects of the controller, the physical characteristics corresponding to the control loop, etc. The output parameters may include the loop PID parameters, which are proportionality coefficient (proportional band), integral time (minutes), and derivative time (minutes).

[0053] Among the above input parameters, T s refers to the response time of the control loop from excitation to stability, and T p is the time of maximum value in the transient period of the control loop, and Y max is the maximum value that exceeds the setpoint during the transient period, and the positive or negative action of the controller refers to the positive or negative feedback action of the controller on the control loop.

[0054] In some embodiments, the auxiliary device operating data X a , initial operation data of new equipment X b may be stored as a table, and each input parameter and output parameter may be obtained by extraction based on the corresponding data item in the table.

[0055] In S103, a parameter setting model is obtained by training based on the auxiliary training data set, the validation data set and the local training data set.

[0056] The type of parameter setting model may be a BP neural network model including an input layer, a hidden layer, and an output layer, and each layer includes multiple network parameters to describe the mapping relationship between the input parameters and the output parameters.

[0057] To make the mapping relationship more accurate, we use an auxiliary training dataset T a , local training dataset T b ,The validation dataset S can be used to train and validate the BP neural network model ,constructed based on the new model, and the network model that ,explains the mapping relationship can be modified to obtain a ,parameter setting model.

[0058] In this embodiment, the parameter setting model is obtained by training it with the local data set using the auxiliary training data set and the validation data set. By adding the operating data based on the auxiliary device to the training data of the parameter setting model, the drawback of the limited valid data available for the new device is compensated for, and the accuracy and efficiency of PID parameter setting of the parameter setting model built on the new device are improved.

[0059] In some embodiments, as shown in FIG. 2, in the above step S101, constructing an auxiliary training data set and a validation data set based on the loop information of the new device and the operation data of the auxiliary device can be realized by the following steps S201 to S202.

[0060] In S201, a target data set is selected from the operational data of the auxiliary device based on the loop type and physical characteristics corresponding to the control loop of the new device.

[0061] As described in the above embodiment, the loop type of the new device and the physical properties corresponding to the control loop are both included in the input parameters of each data set. The loop type may include a flow rate type, a liquid level type, a pressure type, a temperature type, etc., and the physical properties corresponding to the control loop may include, but are not limited to, a liquid phase, a non-pipeline gas phase, a pipeline gas phase, etc.

[0062] The loop type and the physical characteristics corresponding to the control loop can form a set of labels for identifying the process types of the new device and the auxiliary device. Note that when the process types of a control loop of the new device and the auxiliary device match, the data similarity between them is highest, and the operation data of the control loop in the auxiliary device can be used as the target data set to train the matching control loop of the new device.

[0063] In S202, a splitting process is performed on the target dataset to obtain an auxiliary training dataset and a validation dataset.

[0064] The target data set determined based on the auxiliary device operation data can then be divided into auxiliary training and validation data sets with non-overlapping data, where the amount of data in the auxiliary training data set may be greater than the amount of data in the validation data set.

[0065] The auxiliary training data set is used in the process of obtaining a parameter-setting model by training, and the validation data set can be used to verify the accuracy of the trained model after training is completed.

[0066] In this embodiment, after selecting the target dataset, the target dataset is divided into an auxiliary training dataset and a validation dataset, and data from the same source is used to confirm the training level of the model and improve the accuracy of the PID parameters output from the parameter setting model.

[0067] In some embodiments, as shown in FIG. 3, in the above step S201, selecting a target data set from the operation data of the auxiliary device based on the loop type and physical characteristics corresponding to the control loop of the new device can be realized by the following steps S301 to S302.

[0068] In S301, based on the loop type of the new device, a plurality of optional operation data that match the loop type is selected from the operation data of the auxiliary device.

[0069] First, a number of control loops to be settling can be determined in the new device. Illustratively, 20 control loops can be selected in the new device, with each control loop type corresponding to five control loops, thereby determining 140 control loops for a system including seven new devices.

[0070] In some embodiments, control loops of the auxiliary devices that match the loop type of the control loop selected in the new device are selected, and the operating data of the selected control loops of the auxiliary devices is set as optional operating data.

[0071] In S302, the target data set is selected from the plurality of optional operating data based on physical characteristics corresponding to a control loop of the new device.

[0072] Based on the above steps, further selection can be performed on the optional operating data. If the loop types of multiple control loops of an auxiliary device match those of a control loop to be set for a new device, select one or more control loops of the auxiliary device that match the physical characteristics corresponding to the control loop to be set, and finally, the operating data of the selected control loops of the auxiliary device are used as the target data set.

[0073] In this embodiment, based on the loop types of the control loops of the new device and the auxiliary device and the matching degree of the physical characteristics corresponding to the control loops, the operating data similar to the control loop to be adjusted is selected as the target data set, and the accuracy of the trained parameter adjustment model is ensured from the data source.

[0074] In some embodiments, as shown in FIG. 4, in the above step S103, training to obtain a parameter setting model based on the auxiliary training data set, the validation data set, and the local training data set can be realized by the following steps S401 to S405.

[0075] In S401, an intermediate settling model is obtained by training based on the auxiliary training data set and the local training data set.

[0076] In some embodiments, the mid-set model may be an initially trained BP neural network, and after integrating the auxiliary training dataset and the local training dataset, multiple sets of input parameter values ​​are sequentially input into the initially constructed BP neural network. During the training of the mid-set model, the auxiliary training dataset is introduced to increase the amount of training data, and the local training dataset ensures that the trained mid-set model is more compatible with new devices.

[0077] Next, the output values ​​of the initially constructed BP neural network are compared with the values ​​of the output parameters corresponding to the input parameters to determine a comparison difference value.

[0078] Finally, forward feedback is performed based on the comparison difference value to modify the initial first weight vector W, the initial second weight vector P and the initial offset parameter value β of the initially constructed BP neural network to obtain an intermediate settling model.

[0079] In S402, the mid-set model is validated based on the validation data set.

[0080] As described in the above embodiment, the validation dataset is a dataset from the same source as the auxiliary training dataset, and the input parameter corresponding values ​​of the dataset are input to the mid-set model, and then the corresponding output parameter corresponding values ​​of the mid-set model are output from the mid-set model.

[0081] Then, a difference between the corresponding output parameter values ​​of the mid-settlement model and the output parameter values ​​corresponding to the input parameter values ​​in the validation data set is calculated, and if the difference is smaller than a predetermined difference threshold, the result of passing the validation of this set of input parameters is obtained. Otherwise, the validation is not passed. By sequentially inputting multiple sets of input parameters in the validation data set into the mid-settlement model, multiple sets of validation results are obtained.

[0082] Finally, a pass ratio of the validation dataset is determined based on the ratio of the number of pairs of validation results that pass validation to the total number of pairs of data in the validation dataset. In some embodiments, if the pass ratio is greater than a predetermined pass threshold, the training of the mid-set model is considered to be complete. Otherwise, if the pass ratio is less than or equal to the predetermined pass threshold, the above steps can be repeated to obtain a new mid-set model by continuing training.

[0083] In S403, after passing the verification, the mid-settling model is executed according to predetermined input parameter values ​​to obtain output parameter values ​​of the mid-settling model.

[0084] If the mid-set model passes the verification in the above steps, the mid-set model can be run on the new device, which further improves the accuracy of the parameters output by the mid-set model on the new device.

[0085] In some embodiments, a set of predetermined input parameter values ​​may be input to the mid-settlement model, and corresponding output parameter values ​​may be output from the mid-settlement model. The predetermined input parameter values ​​may be a plurality of sets of artificially set test input values, and the parameters corresponding to the predetermined input parameter values ​​and output parameter values ​​may be the same as the parameters corresponding to the input parameter values ​​and output parameter values ​​in the supplemental training data set.

[0086] In S404, the loop PID parameter values ​​after the new device operates according to the predetermined input parameter values ​​are obtained.

[0087] Furthermore, the above-mentioned predetermined input parameter values ​​may be input into the corresponding control loops of the new device that are not included in the local training data set, and after the new device is actually operated, the loop PID parameters output from the new device are obtained, where the loop PID parameters include a proportional coefficient (proportional band), an integral time (minutes), and a derivative time (minutes).

[0088] In S405, parameter optimization is performed on the intermediate settling model based on the loop PID parameter values ​​and the output parameter values ​​to obtain a parameter settling model.

[0089] The loop PID parameter values ​​and the output parameter values ​​corresponding to the same predetermined input parameters are compared to obtain a comparison difference value, and the parameters of the intermediate settling model are optimized based on the comparison difference value to obtain a parameter settling model.

[0090] In addition, predetermined input parameter values ​​may be input to control loops that are not involved in building a local training data set in the new device, and based on the comparison results between the loop PID parameter values ​​output from these control loops and the output parameter values, intermediate setting models are optimized to improve the coverage of the parameter setting models for each new device in the control system to be set.

[0091] To sum up, based on the above embodiment, the process of training and optimizing to obtain a parameter setting model is as shown in FIG.

[0092] First, the operation data of the auxiliary device is extracted to obtain a target dataset, and then the target dataset is divided to generate an auxiliary training dataset and a validation dataset.

[0093] Alternatively, some control loops may be selected from the new device, PID parameter settings may be performed, initial operating data may be obtained, and these initial operating data may be extracted to obtain a local training data set.

[0094] Then, the auxiliary training dataset and the local training dataset are integrated as the training dataset, and the constructed BP neural network model is trained to obtain an intermediate settling model.

[0095] Based on this, the mid-set model can be validated using the validation data set constructed in the above step, and the training process can be repeated until the pass rate is greater than a predetermined pass threshold.

[0096] Finally, the intermediate settling model is subjected to further parameter optimization in control loops that do not involve building a local training data set for each new model of the system being settling to obtain a parameter settling model.

[0097] In this embodiment, verification and further parameter optimization are performed on the intermediate settling model to obtain a parameter settling model, and the accuracy of the output parameter values ​​of the parameter settling model is further improved.

[0098] In some embodiments, as shown in FIG. 6, in the above step S405, parameter optimization is performed on the intermediate settling model based on the loop PID parameter values ​​and the output parameter values ​​to obtain a parameter settling model, which can be realized by the following steps S501 to S502.

[0099] In S501, a parameter error rate is determined based on the loop PID parameter values ​​and the output parameter values.

[0100] Parameter error rate ε t can be calculated by the following formula: [Number 4] TIFF0007760744000013.tif23170

[0101] where n is the number of control loops in the auxiliary training data set and m is the number of control loops in the local training data set. TIFF0007760744000014.tif77 is the ε value of the intermediate settling model in the ith control loop. t is the first weight vector of h t (x i ) is the output parameter value output from the intermediate settling model, and c(x i ) are the loop PID parameter values ​​output from the new device.

[0102] In this way, the parameter error rate is calculated based on the difference between the output parameter value and the loop PID parameter value corresponding to the same predetermined input parameter value.

[0103] In S502, the model parameters of the intermediate settling model are iteratively corrected based on the parameter error rate to obtain a parameter settling model.

[0104] Based on this, the maximum average difference method can be used to determine the difference between the model data distribution situation consisting of the predetermined input parameter values ​​and output parameter values ​​of the mid-set model and the data distribution situation consisting of the predetermined input parameter values ​​and loop PID parameter values, and the difference can be calculated as the parameter error rate ε t It can be shown by

[0105] And further, the parameter error rate ε tBased on this, the mapping from input values ​​to output values ​​of the intermediate settling parameters is modified to be closer to the mapping relationship of the new model, and a parameter settling model is obtained.

[0106] In this embodiment, a parameter error rate is determined based on the loop PID parameter values ​​and output parameter values ​​executed on the new model, and the parameter setting is then modified based on this to improve the mapping accuracy of the parameter setting model.

[0107] In some embodiments, the model parameters include a first weight vector and an offset parameter value.

[0108] The first weight vector can indicate the weights of each layer of a multi-layer BP neural network that constitutes the mid-settlement model, and the offset parameter value is a control parameter value for the activation state of neurons in the BP neural network. In some embodiments, a second weight vector P may be set in the BP neural network, and together with the first weight vector, it may specify a mapping relationship from input values ​​to output values ​​of the mid-settlement model.

[0109] As shown in FIG. 7, in the above step S502, iteratively correcting the model parameters of the intermediate settling model based on the parameter error rate to obtain the parameter settling model can be realized by the following steps S601 to S604.

[0110] In S601, the intermediate offset parameter value is modified based on the parameter error rate to obtain the process offset parameter value.

[0111] The parameter error rate ε determined in the above step t Modifying the intermediate offset parameter value based on the following equation can be shown: [Number 5] TIFF0007760744000015.tif16170

[0112] Then, the original intermediate offset parameter value is replaced with the modified intermediate offset parameter value, that is, the process offset parameter value.

[0113] In S602, the first intermediate weight vector is modified based on the process offset parameter value to obtain a first process weight vector.

[0114] Based on this, the first intermediate weight vector continues to be modified with the process offset parameters determined in the above step according to the following formula: [Number 6] TIFF0007760744000016.tif20170

[0115] When the first intermediate weight vector is replaced with the corrected first process weight vector, the correction process of the first intermediate weight vector is completed once.

[0116] In S603, a new mid-settling model is obtained based on the process offset parameter value and the first process weight vector, and the parameter error rate of the mid-settling model is determined again.

[0117] In this way, a new mid-settling model is constructed using the process offset parameter value, the first process weight vector, and the second weight vector. Predetermined input parameters are input to the new device and the new mid-settling model, respectively. A new parameter error rate is then recalculated and determined based on the above equation, the loop PID parameter values ​​of the new device, and the output parameter values ​​of the new mid-settling model.

[0118] In S604, it is determined whether the parameter error rate is smaller than a predetermined threshold.

[0119] In some embodiments, a predetermined threshold may be set, and if the redetermined parameter error rate is equal to or greater than the predetermined threshold, the above steps may continue to be performed iteratively to continue refining the mid-settlement model.

[0120] In S605, the above steps are repeated until the parameter error rate becomes smaller than a predetermined threshold, and the intermediate settling model is set as the parameter settling model.

[0121] If the re-determined parameter error rate is smaller than the predetermined threshold, it is considered that the correction of the intermediate settling model is complete, and the corrected intermediate settling model is set as the final parameter settling model.

[0122] In this embodiment, the intermediate settling model is iteratively modified based on the parameter error rate to improve the fitting degree between the intermediate settling model and the new model, and improve the accuracy of the output parameter values ​​of the parameter settling model in the new model.

[0123] In some embodiments, as shown in FIG. 8, in the above step S401, before training based on the auxiliary training data set and the local training data set to obtain an intermediate settling model, the following steps S701 to S704 may be further included.

[0124] In S701, the root mean square error and the coefficient of determination are determined based on the auxiliary training data set and the local training data set.

[0125] In some embodiments, the root mean square error RMSE may be determined as follows: [Number 7] TIFF0007760744000017.tif23170

[0126] Coefficient of determination R 2 can be determined by the following formula: [Number 8] TIFF0007760744000018.tif27170

[0127] where m is the number of samples in the training set consisting of the auxiliary training dataset and the local training dataset, and y iindicates the sample true value of the i-th sample in the training set for the new device and auxiliary device, TIFF0007760744000019.tif86 shows the output predicted value of the constructed model for the i-th sample. TIFF0007760744000020.tif65 is the average value of the sample true values.

[0128] In S702, the number of hidden layer nodes of the parameter setting initial model is determined based on the root mean square error and the coefficient of determination.

[0129] Then, the number of hidden layer nodes of the parameter-setting initial model is selected based on the determined values ​​of the root mean square error and the coefficient of determination. For example, the root mean square error and the coefficient of determination are divided into multiple ranges, and are assigned a one-to-one correspondence with the number of hidden layer nodes. If the root mean square error and the coefficient of determination are within the range, the corresponding number of hidden layer nodes can be determined.

[0130] In some embodiments, for example in the present application, the number of hidden layer nodes may be five.

[0131] In S703, a parameter setting initial model is constructed based on the number of hidden layer nodes, a predetermined number of input layer nodes, and a predetermined number of output layer nodes.

[0132] The predetermined number of input layer nodes may be determined according to the number of input parameters, and may be, for example, 7. The predetermined number of output layer nodes may be determined according to the number of output parameters, and may be, for example, 3.

[0133] In this way, the network structure of the BP neural network can be constructed based on the number of hidden layer nodes, a predetermined number of input layer nodes, and a predetermined number of output layer nodes.

[0134] In some embodiments, when constructing a parameter setting initial model based on this, an initial first weight vector W, an initial second weight vector P, and an initial offset value β of the parameter setting initial model may be initialized. In some embodiments, the initial weight vector is [Number 9] TIFF0007760744000021.tif19170, [Number 10] TIFF0007760744000022.tif21170

[0135] n is the number of control loops in the auxiliary training data set and m is the number of control loops in the local training data set.

[0136] The initial offset parameter value β is [Number 11] TIFF0007760744000023.tif18170, where N is the value of n+m.

[0137] The initial second weight vector is [Number 12] TIFF0007760744000024.tif24170, which corresponds to the initial weight vector.

[0138] After constructing the network structure and network parameters of the BP neural network, the initial model for parameter setting is obtained.

[0139] In S704, the parameter-set initial model is trained based on the auxiliary training data set and the local training data set to obtain an intermediate parameter-set model.

[0140] Finally, the constructed parameter-set initial model is trained based on the auxiliary training data set and the local training data set constructed in the above embodiment, and the parameters of the parameter-set initial model are modified to obtain an intermediate setting model.

[0141] In this embodiment, the number of hidden layer nodes is determined based on the root mean square error and the coefficient of determination, and an initial parameter-setting model is constructed. This allows the constructed model to better fit the amount of data in the training set, with higher fitting ability and faster training speed.

[0142] As shown in FIG. 9, an embodiment of the present application further provides an industrial process control method, which may be applied to a processing device capable of performing PID parameter setting for a new device. Referring to FIG. 9, the method includes the following steps S801 to S802.

[0143] In S801, setting parameter values ​​of a new device to be controlled are determined based on a parameter setting model, which is obtained by the parameter setting model construction method described in any of the above embodiments.

[0144] In some embodiments, a parameter settling model can be determined based on the above embodiments, and the model can be applied to a new device to output corresponding PID parameter values, i.e., settling parameter values, of the new device to be controlled based on input parameter values ​​in the new device.

[0145] During the training of the parameter setting model, multiple new devices are introduced into the industrial control system to be set, and the parameter setting model can output accurate setting parameter values ​​for the multiple new devices in the industrial control system to be set.

[0146] In S802, the new device is controlled to perform the target process based on the settling parameter values.

[0147] It is understood that the target process may be a process that the new equipment operates based on settling parameter values, which are output from the parameter settling model and are values ​​to be adjusted for the new equipment, and thus the settling parameter values ​​can be input to the new equipment to control its operation to perform the target process.

[0148] In this embodiment, the operation of the new device is controlled based on the setting parameter values ​​output from the parameter setting model, thereby achieving a predetermined control effect with high efficiency in the new device and improving the degree of automation of parameter setting.

[0149] Referring to FIG. 10, the embodiment of the present application further provides a parameter setting model construction apparatus 100, which includes a dataset construction module 1001, a model training module 1002 and a model construction module 1003.

[0150] The dataset construction module 1001 is configured to construct an auxiliary training dataset and a validation dataset based on the loop information of the new device and the operation data of the auxiliary device, where the new device is a device to which the parameter setting model to be trained is applied, and the auxiliary device is a device whose parameter setting has been completed and which has been officially operated; The dataset construction module 1001 is further configured to construct a local training dataset based on the initial operating data of the new equipment.

[0151] The model training module 1002 is configured to train and obtain a parameter-setting model based on the auxiliary training data set, the validation data set, and the local training data set.

[0152] Specifically, the dataset construction module 1001 is configured to select a target dataset from the operation data of the auxiliary device based on the loop type and physical characteristics corresponding to the control loop of the new device, and perform a division process on the target dataset to obtain an auxiliary training dataset and a validation dataset.

[0153] Specifically, the dataset construction module 1001 is configured to select, based on the loop type of the new device, a plurality of optional operating data matching the loop type from the operating data of the auxiliary device, and to select a target dataset from the plurality of optional operating data based on the physical characteristics corresponding to the control loop of the new device.

[0154] The model training module 1002 is specifically configured to: train based on the auxiliary training dataset and the local training dataset to obtain an intermediate settling model; verify the intermediate settling model based on the verification dataset; after passing the verification, execute the intermediate settling model according to predetermined input parameter values ​​to obtain output parameter values ​​of the intermediate settling model; obtain loop PID parameter values ​​after the new device operates according to the predetermined input parameter values; and perform parameter optimization on the intermediate settling model based on the loop PID parameter values ​​and the output parameter values ​​to obtain a parameter settling model.

[0155] Specifically, the model training module 1002 is configured to determine a parameter error rate based on the loop PID parameter values ​​and the output parameter values, and iteratively modify the model parameters of the intermediate settling model based on the parameter error rate to obtain a parameter settling model.

[0156] Specifically, the model training module 1002 is configured to: the model parameters include a first intermediate weight vector and an intermediate offset parameter value; modify the intermediate offset parameter value based on a parameter error rate to obtain a process offset parameter value; modify the first intermediate weight vector based on the process offset parameter value to obtain a first process weight vector; obtain a new intermediate settling model based on the process offset parameter value and the first process weight vector; re-determine a parameter error rate of the intermediate settling model; repeat the above steps until the parameter error rate is smaller than a predetermined threshold; and take the intermediate settling model as the parameter settling model.

[0157] The model construction module 1003 is configured to determine a root mean square error and a coefficient of determination based on the auxiliary training data set and the local training data set, determine the number of hidden layer nodes of a parameter-set initial model based on the root mean square error and the coefficient of determination, construct a parameter-set initial model based on the number of hidden layer nodes, a predetermined number of input layer nodes and a predetermined number of output layer nodes, and train the parameter-set initial model based on the auxiliary training data set and the local training data set to obtain an intermediate parameter-set model.

[0158] Referring to FIG. 11, an embodiment of the present application includes: A determination module 1101 configured to determine setting parameter values ​​of a new device to be controlled based on a parameter setting model, where the parameter setting model is obtained by the method for constructing a parameter setting model described in any of the above embodiments; Further provided is an industrial process control device comprising: a control module configured to control the new device to perform the target process based on the settling parameter values.

[0159] Referring to FIG. 12 , this embodiment further provides a processing device including a processor 2001, a memory 2002, and a bus, wherein the memory 2002 stores machine-readable instructions executable by the processor 2001, and the machine-readable instructions are executed when the processing device operates, the processor 2001 and the memory 2002 communicate via the bus, and the processor 2001 is configured to execute steps of the parameter setting model construction method or industrial process control method in the above embodiment.

[0160] The memory 2002, the processor 2001, and the bus electrically connect each element to each other directly or indirectly, enabling data transmission or exchange. For example, these elements may be electrically connected to each other by one or more communication buses or signal lines. The data processing device of the parameter setting model construction system or the industrial process control system includes at least one software function module stored in the memory 2002 in the form of software or firmware or cured in the operating system (OS) of the processing device. The processor 2001 is configured to execute executable modules stored in the memory 2002, such as software function modules and computer programs included in the data processing device of the parameter setting model construction system or the industrial process control system.

[0161] The memory 2002 may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc.

[0162] In some embodiments, the present application further provides a storage medium storing a computer program, which, when executed by a processor, performs the steps of the above method embodiments. The specific implementation manners and technical effects are similar, and therefore will not be described again in this application.

[0163] Those skilled in the art will clearly understand that for the sake of convenience and conciseness of description, the specific work processes of the above-mentioned systems and devices can be referenced to the corresponding processes in implementing the method, and descriptions thereof will be omitted in this application. It should be understood that in some embodiments of the present application, the disclosed systems, devices, and methods can be realized in other forms. The above-described device embodiments are merely exemplary. For example, the division of the above modules is merely a division of logical functions. In actual implementation, other division forms may be used. For example, multiple modules or components may be combined or integrated into another system, or some features may be omitted or not implemented. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be indirect couplings or communication connections through several communication interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0164] Furthermore, in each embodiment of the present application, each functional unit may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above functions may be realized in the form of software functional units and stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or a portion of the technical solution, may essentially be embodied in the form of a software product, and the computer software product is stored in a storage medium and includes several commands that cause a computer device (which may be a personal computer, a server, a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0165] Although specific embodiments of the present application have been described above, the scope of protection of the present application is not limited thereto, and all modifications or substitutions that a person skilled in the art can easily conceive within the technical scope disclosed by the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application is based on the scope of protection of the claims. [Industrial Applicability]

[0166] The technical solutions according to the embodiments of the present application are applicable to the technical field of industrial automation control, and the parameter setting model construction method and industrial process control method according to the present application can construct an auxiliary training dataset using operating data of an auxiliary device, and train and construct a parameter setting model together with a local training dataset constructed using operating data of a new device. The present application makes full use of operating data of an auxiliary device that has already completed parameter setting and been officially operated to train the parameter setting model, thereby compensating for the shortcoming of the limited valid data available for new devices, and improving the accuracy and efficiency of PID parameter setting of the parameter setting model constructed on the new device.

Claims

1. constructing an auxiliary training data set and a validation data set based on loop information of a new device and operation data of an auxiliary device, wherein the new device is a device to which a parameter setting model to be trained is applied, and the auxiliary device is a device for which parameter setting has been completed and which has been officially operated, the loop information includes a loop type and physical properties corresponding to a control loop, the loop type is a control object of the new device, and the loop information and operation data of the auxiliary device are included in the auxiliary training data set and the validation data set; building a local training data set based on initial operating data of the new device; training a parameter setting model based on the auxiliary training data set, the validation data set, and the local training data set, wherein the validation data set is for validating the parameter setting model; The step of training to obtain a parameter setting model based on the auxiliary training data set, the validation data set, and the local training data set includes: training a mid-set model based on the auxiliary training data set and the local training data set; Validating the mid-settlement model based on the validation data set, and after passing the validation, executing the mid-settlement model according to predetermined input parameter values ​​to obtain output parameter values ​​of the mid-settlement model; obtaining loop PID parameter values ​​after the new device has operated according to the predetermined input parameter values; performing parameter optimization on the intermediate settling model based on the loop PID parameter values ​​and the output parameter values ​​to obtain the parameter settling model.

2. said step of constructing an auxiliary training data set and a validation data set based on the loop information of the new device and the operation data of the auxiliary device, selecting a target data set from the auxiliary device's operating data based on the new device's loop type and physical characteristics corresponding to the control loop; 2. The method of claim 1, further comprising: performing a splitting process on the target data set to obtain the supplemental training data set and the validation data set.

3. selecting a target data set from the auxiliary device's operating data based on the new device's loop type and physical characteristics corresponding to the control loop, selecting, based on the loop type of the new device, a plurality of optional operation data items matching the loop type from the operation data of the auxiliary device; and selecting the target data set from the plurality of optional operating data sets based on physical characteristics corresponding to a control loop of the new device.

4. The step of performing parameter optimization on the intermediate settling model based on the loop PID parameter values ​​and the output parameter values ​​to obtain the parameter settling model includes: determining a parameter error rate based on the loop PID parameter values ​​and the output parameter values; and iteratively modifying model parameters of the intermediate settling model based on the parameter error rate to obtain the parameter settling model.

5. the model parameters include a first intermediate weight vector and an intermediate offset parameter value; the step of iteratively modifying model parameters of the intermediate settling model based on the parameter error rate to obtain the parameter settling model includes: modifying the intermediate offset parameter value based on the parameter error rate to obtain a process offset parameter value; modifying the first intermediate weight vector based on the process offset parameter value to obtain a first process weight vector; deriving a new mid-settlement model based on the process offset parameter value and the first process weight vector, and redetermining a parameter error rate of the mid-settlement model; 5. The method for constructing a parameter setting model according to claim 4, further comprising the step of repeating the steps above until the parameter error rate becomes smaller than a predetermined threshold, and designating the intermediate setting model as the parameter setting model.

6. Before the step of training to obtain an intermediate settling model based on the auxiliary training data set and the local training data set, the method includes: determining a root mean square error and a coefficient of determination based on the supplemental training data set and the local training data set; determining the number of hidden layer nodes of the parameter-setting initial model based on the root mean square error and the coefficient of determination; constructing the parameter-setting initial model based on the number of hidden layer nodes, a predetermined number of input layer nodes, and a predetermined number of output layer nodes; The method of claim 1 , further comprising: training the parameter-setting initial model based on the auxiliary training data set and the local training data set to obtain the intermediate parameter-setting model.

7. a step of determining setting parameter values ​​of a new device to be controlled based on a parameter setting model, the parameter setting model being obtained by the parameter setting model construction method according to any one of claims 1 to 6; and controlling the new device to perform a target process based on the settling parameter values.

8. A processing device including a processor, a storage medium, and a bus, The storage medium stores machine-readable instructions executable by the processor; When the processing device is in operation, the processor and the storage medium communicate via a bus, and the processor executes the machine-readable instructions to perform the steps of the parameter setting model construction method described in any one of claims 1 to 6.

9. A processing device including a processor, a storage medium, and a bus, The storage medium stores machine-readable instructions executable by the processor; 8. A processing device, wherein when the processing device operates, the processor and the storage medium communicate via a bus, and the processor executes the machine-readable instructions to perform the steps of the industrial process control method of claim 7.

10. A computer program configured to cause a computer to execute the method for constructing a parameter setting model according to any one of claims 1 to 6.

11. A computer program configured to cause a computer to carry out the industrial process control method of claim 7.

Images