WO2024016637A1 - Method for constructing parameter setting model and industrial process control method - Google Patents

Abstract

Embodiments of the present application relate to the technical field of industrial automation control, and provide a method for constructing a parameter setting model and an industrial process control method. In the present application, the method comprises: constructing an auxiliary training data set and a validation data set according to loop information of a newly-built device and operation data of an auxiliary device, wherein the newly-built device is a device to which a parameter setting model to be trained is applied, and the auxiliary device is a device completing parameter setting and formally running; constructing a local training data set according to initial operation data of the newly-built device; and training to obtain the parameter setting model according to the auxiliary training data set, the validation data set, and the local training data set.

Description

Parameter tuning model construction method and industrial process control method

cross-citation

This application claims the priority of a Chinese patent application submitted to the China Patent Office on July 22, 2022, with priority number 202210859981. incorporated herein by reference.

Technical field

The present application relates to the field of industrial automation control technology, specifically, to a method for constructing a parameter tuning model and an industrial process control method.

Background technique

PID (proportion, integral, differential) is currently the most widely used control strategy. PID parameter tuning is the core content of its control process. Parameter tuning is to determine the deviation values of its proportional coefficient, integral time and differential time based on the characteristics of the control system's working process.

At present, the PID parameter tuning method is generally based on the internal model. This method establishes a mathematical model for PID parameter tuning, establishes a process mathematical model based on process historical data, and uses the internal model tuning strategy to obtain the PID parameters based on the mathematical model. This method is relatively simple. An effective method for tuning PID parameters.

However, the internal model-based tuning method requires a large amount of valid data to establish a process model. For a process model established on a newly built device, the available data information is less, which is not enough to establish a reliable process model, resulting in The actual effect of PID parameter tuning is poor.

Contents of the invention

The purpose of this application includes, for example, providing a method for constructing a parameter tuning model and an industrial process control method. By constructing an auxiliary training data set for training, it effectively makes up for the shortcomings of less effective data available for newly built devices and improves the efficiency of newly built devices. The accuracy of PID parameter tuning of the parameter tuning model established on the device.

The embodiment of this application can be implemented as follows:

In a first aspect, embodiments of the present application provide a method for constructing a parameter tuning model. The method includes:

An auxiliary training data set and a verification data set are constructed according to the circuit information of the newly-built device and the operating data of the auxiliary device, where the newly-built device is a device to which the parameter tuning model to be trained is applied, and the auxiliary device Set as a device that has completed parameter tuning and is officially in operation;

Construct a local training data set based on the initial operating data of the newly built device;

According to the auxiliary training data set, the verification data set and the local training data set, a parameter tuning model is trained.

In an optional implementation, constructing an auxiliary training data set and a verification data set based on the circuit information of the newly-built device and the operating data of the auxiliary device includes:

Filter out the target data set from the operating data of the auxiliary device according to the circuit type of the new device and the corresponding physical characteristics of the control loop;

The target data set is split to obtain the auxiliary training data set and the verification data set.

In an optional implementation, the target data set is selected from the operating data of the auxiliary device according to the circuit type of the newly-built device and the corresponding physical characteristics of the control loop, including:

According to the circuit type of the newly-built device, screen a plurality of optional operating data matching the circuit type from the operating data of the auxiliary device;

The target data set is selected from the plurality of optional operating data according to the physical characteristics corresponding to the control loop of the newly constructed device.

In an optional implementation, the training to obtain a parameter tuning model based on the auxiliary training data set, the verification data set and the local training data set includes:

Based on the auxiliary training data set and the local training data set, train an intermediate tuning model;

Verify the intermediate tuning model based on the verification data set, and after passing the verification, run the intermediate tuning model according to the preset input parameter values to obtain the output parameter values of the intermediate tuning model;

Obtain the loop PID parameter value after the new device is operated according to the preset input parameter value;

According to the loop PID parameter value and the output parameter value, parameter optimization is performed on the intermediate tuning model to obtain the parameter tuning model.

In an optional implementation, performing parameter optimization on the intermediate tuning model according to the loop PID parameter value and the output parameter value to obtain the parameter tuning model includes:

Determine the parameter error rate according to the loop PID parameter value and the output parameter value;

According to the parameter error rate, the model parameters of the intermediate tuning model are iteratively corrected to obtain the parameter tuning model.

In an optional implementation, the model parameters include: a first intermediate weight vector and an intermediate bias parameter value;

Iteratively correcting the model parameters of the intermediate tuning model according to the parameter error rate to obtain the parameter tuning model includes:

According to the parameter error rate, the intermediate offset parameter value is corrected to obtain a process offset parameter value;

Modify the first intermediate weight vector according to the process offset parameter value to obtain a first process weight vector;

Obtain a new intermediate tuning model according to the process bias parameter value and the first process weight vector, and re-determine the parameter error rate of the intermediate tuning model;

Repeat the above process until the parameter error rate is less than the preset threshold, and use the intermediate tuning model as the parameter tuning model.

In an optional implementation, before training to obtain an intermediate tuning model based on the auxiliary training data set and the local training data set, the method further includes:

Determine the root mean square error and coefficient of determination according to the auxiliary training data set and the local training data set;

According to the root mean square error and the coefficient of determination, determine the number of hidden layer nodes of the initial model for parameter tuning;

Construct the initial parameter tuning model according to the number of hidden layer nodes, the preset number of input layer nodes, and the preset number of output layer nodes;

The initial parameter tuning model is trained according to the auxiliary training data set and the local training data set to obtain the intermediate tuning model.

In a second aspect, embodiments of the present application provide an industrial process control method, which method includes:

Determine the tuning parameter values of the newly-built device to be controlled according to the parameter tuning model, which is obtained based on the construction method of the parameter tuning model described in any one of the first aspects;

The newly created device is controlled to execute a target process according to the setting parameter value.

In a third aspect, embodiments of the present application provide a device for constructing a parameter tuning model, including:

The data set construction module is configured to construct an auxiliary training data set and a verification data set based on the circuit information of the newly created device and the operating data of the auxiliary device, wherein the new device is a device to which the parameter tuning model to be trained is applied, and the Auxiliary devices are devices that have completed parameter setting and are officially in operation;

The data set construction module is also configured to construct a local training data set based on the initial operating data of the newly created device.

A model training module is configured to train and obtain a parameter tuning model based on the auxiliary training data set, the verification data set and the local training data set.

The data set building module is specifically configured to filter out the target data set from the operation data of the auxiliary device according to the circuit type of the new device and the corresponding physical characteristics of the control loop; and disassemble the target data set. Process separately to obtain the auxiliary training data set and the verification data set.

The data set building module is specifically configured to filter multiple optional operating data matching the circuit type from the operating data of the auxiliary device according to the circuit type of the new device; The physical characteristics corresponding to the control loop are used to filter out the target data set from the plurality of optional operating data.

The model training module is specifically configured to: train to obtain an intermediate tuning model based on the auxiliary training data set and the local training data set; verify the intermediate tuning model based on the verification data set, and pass the verification Then, run the intermediate tuning model according to the preset input parameter value to obtain the output parameter value of the intermediate tuning model; obtain the loop PID parameter value of the new device after running according to the preset input parameter value; according to the The loop PID parameter value and the output parameter value are used to optimize the parameters of the intermediate tuning model to obtain the parameter tuning model.

The model training module is specifically configured to determine a parameter error rate based on the loop PID parameter value and the output parameter value; and iteratively correct the model parameters of the intermediate tuning model based on the parameter error rate, to obtain The parameter tuning model.

The model training module is specifically configured such that the model parameters include: a first intermediate weight vector and an intermediate offset parameter value; and the intermediate offset parameter value is corrected according to the parameter error rate to obtain a process offset. Parameter value; correct the first intermediate weight vector according to the process offset parameter value to obtain a first process weight vector; obtain a new intermediate weight vector according to the process offset parameter value and the first process weight vector Tuning the model, re-determine the parameter error rate of the intermediate tuning model; repeat the above process until the parameter error rate is less than the preset threshold, and use the intermediate tuning model as the parameter tuning model.

A model building module 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; and determine a hidden parameter setting initial model based on the root mean square error and the coefficient of determination. The number of layer nodes; according to the number of hidden layer nodes, the preset number of input layer nodes and the preset number of output layer nodes, the initial parameter setting model is constructed; according to the auxiliary training data set and the local training data set , train the initial parameter tuning model to obtain the intermediate tuning model.

In a fourth aspect, embodiments of the present application also provide an industrial process control device, including:

The determination module is configured to determine the tuning parameter values of the newly-built device to be controlled according to the parameter tuning model, which is obtained based on the method for constructing the parameter tuning model described in any one of the first aspects.

A control module configured to control the newly-built device to execute a target process according to the setting parameter value.

In a fifth aspect, embodiments of the present application provide a processing device. The processing device includes: a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the processing When the device is running, the processor and the storage medium communicate through a bus, and the processor executes the machine-readable instructions to perform the method for constructing a parameter tuning model as described in any one of the first aspects. Or the steps of the industrial process control method described in the second aspect.

In a sixth aspect, embodiments of the present application provide a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, any one of the aspects described in the first aspect is implemented. The method for constructing a parameter tuning model or the steps of the industrial process control method described in the second aspect.

The beneficial effects of the embodiments of this application include:

Using the parameter tuning model construction method and industrial process control method provided by this application, an auxiliary training data set can be constructed with the help of the operating data of the auxiliary device, and together with the local training data set constructed using the operating data of the newly built device, training and Establish parameter tuning model. This application makes full use of the operating data of the auxiliary device that has completed parameter setting and is officially in operation to train the parameter setting model, which makes up for the shortcomings of less effective data available for the new device and improves the PID parameter tuning of the parameter setting model established on the new device. Accuracy and efficiency.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application and therefore do not It should be regarded as a limitation of the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without exerting creative efforts.

Figure 1 is a schematic flowchart of the steps of a method for constructing a parameter tuning model provided by an embodiment of the present application;

Figure 2 is a schematic flow chart of the data set construction steps of the parameter tuning model construction method provided by the embodiment of the present application;

Figure 3 is a schematic flow chart of the steps of data set screening in the method for building a parameter tuning model provided by the embodiment of the present application;

Figure 4 is a schematic flow chart of the steps of model training and optimization of the method for building a parameter tuning model provided by the embodiment of the present application;

Figure 5 is a schematic flowchart of the implementation of the method for constructing a parameter tuning model provided by the embodiment of the present application;

Figure 6 is a schematic flowchart of the steps of parameter optimization of the method for building a parameter tuning model provided by the embodiment of the present application;

Figure 7 is a schematic flowchart of another step of parameter optimization in the method for building a parameter tuning model provided by the embodiment of the present application;

Figure 8 is a schematic flowchart of the steps of model construction in the parameter tuning model construction method provided by the embodiment of the present application;

Figure 9 is a schematic flow chart of the steps of the industrial process control method provided by the embodiment of the present application;

Figure 10 is a schematic structural diagram of a device for constructing a parameter tuning model provided by an embodiment of the present application;

Figure 11 is a schematic structural diagram of an industrial process control device provided by an embodiment of the present application;

Figure 12 is a schematic structural diagram of the processing equipment provided by the embodiment of the present application.

Icon: 100-parameter tuning model construction device; 1001-data set construction module; 1002-model training module; 1003-model construction module; 110-industrial process control device; 1101-determination module; 1102-control module; 2001-processing 2002-memory.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments These are part of the embodiments of this application, but not all of them. The components of the embodiments of the present application generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations.

Accordingly, the following detailed description of the embodiments of the application provided in the appended drawings is not intended to limit the scope of the claimed application, but rather to represent selected embodiments of the application. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

It should be noted that similar reference numerals and letters represent similar items in the following figures, therefore, once an item is defined in one figure, it does not need further definition and explanation in subsequent figures.

In addition, if the terms "first", "second", etc. appear, they are only used to differentiate the description and shall not be understood as indicating or implying relative importance.

It should be noted that, as long as there is no conflict, the features in the embodiments of the present application can be combined with each other.

PID parameter tuning is a process of adjusting the three parameters of proportion, integral and differential so that the dynamic and static performance of the system meets the requirements and a certain performance index reaches the optimum. At present, the more common parameter tuning method is based on internal model tuning. This method builds a loop process model by collecting historical input and output relationship parameters on the target device, and obtains the tuning parameter results based on the loop process model based on the internal model tuning strategy. Gained the ability to perform real-time PID parameters on the target device. However, the model obtained in this way needs to be driven by a large amount of valid historical data. For new installations, the available data information is less, which is not enough to train a reliable model, resulting in poor actual results in PID parameter tuning. .

Based on this, the applicant has proposed a method for constructing a parameter tuning model and an industrial process control method after research, which can use the auxiliary device with completed parameter tuning to build an auxiliary training data set, assist in training the local training data set, and establish a parameter tuning model. , avoids the problems of insufficient model training and low accuracy caused by insufficient historical data, and improves the accuracy and efficiency of PID parameter tuning of newly built devices.

Transfer learning simulates the thinking process of the human brain. After solving a problem, people will have better and faster solutions to new related problems. In other words, transfer learning is different from previous machine learning methods. , can use the "knowledge" learned from tasks in the same field as the target field, such as data characteristics, model parameters, etc., to assist the learning process in the new field and obtain a model that can be applied to the target field.

Industrial control systems contain many different types of control loops. Although the scale and data distribution may be different for different industrial control systems, when split to the control loop level, different industrial control systems may The control loops have some similar characteristic information. Based on this, embodiments of the present application provide a method for constructing a parameter tuning model and an industrial process control method. Through transfer learning, the operating data of the auxiliary device whose parameter tuning has been completed is applied to the parameter tuning model of the newly built device. In terms of training, it solves the problem that the existing network model relies heavily on the valid data of the newly built device during the training process, and improves the construction speed and accuracy of the parameter tuning model on the newly built device.

The construction method of a parameter tuning model and the industrial process control method provided by the embodiments of the present application will be explained below with reference to multiple specific application examples.

Figure 1 shows a schematic flowchart of the steps of a method for constructing a parameter tuning model provided by an embodiment of the present application. The execution subject of this method may be a computer device with computing and processing capabilities. As shown in Figure 1, the method includes the following steps:

S101: Construct an auxiliary training data set and a verification data set based on the circuit information of the newly created device and the operating data of the auxiliary device.

Among them, the newly-built device is the device used by the parameter tuning model to be trained, and the auxiliary device is the device whose parameter tuning has been completed and is officially running.

The new device can be a device that has been established and needs to build a parameter tuning model for PID parameter tuning. Exemplarily, the newly-built unit may be multiple units included in the ethylene system, such as an ethylene unit, a pyrolysis gasoline hydrogenation unit, a butadiene extraction unit, an aromatics extraction unit, an MTBE/butene-1 unit, and an ethylene glycol unit. Any one or more of the device and the POX device. In some embodiments, multiple control loops may also be included in each device.

The auxiliary device may be a device that contains multiple control loops, has completed parameter setting, and is in normal operating status. One or more of these control loops can match certain characteristics of the control loop with the newly built device. It can be understood that there may be more than one auxiliary device, and the control loops contained in the multiple auxiliary devices together form a control loop set that matches the multiple control loops in the newly-built device in a one-to-one correspondence.

Based on the operating data generated in the above control loop collection, construct an auxiliary training data set with non-overlapping data. in, When i=1, 2,...,n, validation data set in, When i=1, 2,...,k. are the input parameters and output parameters of the auxiliary training data set _Ta , are the input parameters and output parameters of the verification data set S respectively, X _a is the operating data of the auxiliary device, n is the control loop data included in the auxiliary training data set T _a , and k is the number of control loops included in the verification data set S. .

S102: Construct a local training data set based on the initial operating data of the newly created device.

In some embodiments, some initial input parameters can be input into the new device by manually adjusting the PID parameters on the new device, and the corresponding initial output parameters can be obtained. According to the initial input parameters, the initial output parameters and the two corresponding relationship between them, constructing a local training data set in, When i=1, 2,...,m. in, is the input parameter of the local training data set T _b , is the output parameter of the local training data set T _b , and X _b is the initial operating data of the newly built device.

In addition, since the initial operating data obtained by manual PID parameter tuning is small, the amount of data contained in the auxiliary training data set _Ta and the verification data set S can be much larger than the data amount of the local training data set T _b .

In some embodiments, the input parameters and output parameters of the auxiliary training data set _Ta , the verification data set S, and the local training data set _Tb may be the same. Among them, the input parameters may include: the device to which the control loop belongs, the loop type, the loop's closed-loop steady-state time T _s , the peak time T _p , the maximum overshoot Y _max , the positive and negative effects of the controller, and the corresponding physical characteristics of the control loop. wait. The output parameters may include: loop PID parameters, namely proportional coefficient (proportionality), integration time (minutes), and differential time (minutes).

Among the above input parameters, T _s refers to the response time of the control loop in the transition period from excitation to stability, T _p is the moment when the control loop reaches the maximum value in the transition period, Y _max is the maximum value exceeding the set value in the transition period, The positive and negative effects of the controller refer to the positive or negative feedback effect of the controller on the control loop.

In some embodiments, the operation data X _a of the auxiliary device and the initial operation data X _b of the new device can be stored in a table, and each input parameter and output parameter can be extracted and obtained according to the corresponding data items in the table.

S103. According to the auxiliary training data set, the verification data set and the local training data set, train the parameter tuning model.

The type of parameter tuning model can be a BP neural network model, which includes an input layer, a hidden layer, and an output layer. Each layer contains multiple network parameters, which are used to describe the mapping relationship between input parameters and output parameters.

In order to make the mapping relationship more accurate, the auxiliary training data set T _a , the local training data set T _b , and the verification data set S can be used to jointly train and verify the BP neural network model built based on the newly created model, and the network describing the mapping relationship can be After the model is modified, the parameter tuning model is obtained.

In this embodiment, the auxiliary training data set and the verification data set are used, combined with the local data set, to jointly train and obtain the parameter tuning model. Adding the operating data based on auxiliary devices to the training data of the parameter tuning model makes up for the shortcomings of less effective data available for new devices, and improves the accuracy and efficiency of PID parameter tuning of the parameter tuning model established on new devices.

In some embodiments, as shown in Figure 2, in the above step S101, constructing an auxiliary training data set and a verification data set based on the circuit information of the newly created device and the operating data of the auxiliary device can be implemented by the following steps S201 to S202.

S201: Filter out the target data set from the operating data of the auxiliary device according to the loop type of the newly-built device and the corresponding physical characteristics of the control loop.

According to the above embodiment, the loop type of the newly-built device and the corresponding physical characteristics of the control loop are included in the input parameters of each data set. Among them, the loop type can include: flow type, liquid level type, pressure type, temperature type, etc., and the corresponding physical characteristics of the control loop can include: liquid phase, non-pipeline gas phase, pipeline gas phase, etc., of course, it is not limited to this. .

The loop type and the corresponding physical characteristics of the control loop can form a set of labels to identify the process type of the new device and auxiliary device. It can be understood that when the process type of a certain control loop of the newly built device and the auxiliary device matches, the data similarity between the two is the highest. The operating data of the control loop in the auxiliary device can be used as the target data set. Matching control loops for newly created devices are trained.

S202: Split the target data set to obtain an auxiliary training data set and a verification data set.

Next, the determined target data set based on the operating data of the auxiliary device can be split into an auxiliary training data set and a verification data set with disjoint data. Among them, the data volume of the auxiliary training data set can be larger than the data volume of the verification data set.

The auxiliary training data set can be used to train the parameter tuning model, and the verification data set can be used to verify the accuracy of the trained model after the training is completed.

In this embodiment, after filtering out the target data set, the target data set is split into an auxiliary training data set and a verification data set. The homologous data is used to confirm the training degree of the model and improve the accuracy of the PID parameters output by the parameter tuning model. accuracy.

In some embodiments, as shown in Figure 3, in the above-mentioned step S201, the target data set is filtered out from the operating data of the auxiliary device according to the circuit type of the new device and the corresponding physical characteristics of the control loop. The target data set can be obtained from the following steps S301 to S302 is implemented.

S301. According to the circuit type of the newly-built device, screen multiple optional operating data matching the circuit type from the operating data of the auxiliary device.

You can first determine multiple control loops to be tuned in the new device. For example, you can select 20 control loops in the new device. Each control loop type corresponds to 5 control loops. For a new device including 7 device system, 140 control loops can be identified.

In some embodiments, from the control loops of the auxiliary devices, control loops that match the loop type of the control loop selected in the new device are selected, and the operation data of the control loops of these filtered auxiliary devices are as optional operating data.

S302: Filter out the target data set from the plurality of optional operating data according to the physical characteristics corresponding to the control loop of the newly created device.

Based on the above steps, you can further filter the optional running data. On the basis that the multiple control loops of the auxiliary device match the loop type of the control loop to be tuned in a new device, one or more auxiliary devices that match the physical characteristics corresponding to the control loop to be tuned are selected. control loop, and use the finally filtered operating data of the control loop of the auxiliary device as the target data set.

In this embodiment, based on the matching degree of the loop type of the control loop of the new device and the auxiliary device and the corresponding physical characteristics of the control loop, operating data similar to the control loop to be tuned is selected as the target data set. From the data source This ensures the accuracy of the trained parameter tuning model.

In some embodiments, as shown in Figure 4, in the above step S103, the parameter tuning model is trained according to the auxiliary training data set, the verification data set and the local training data set, which can be implemented by the following steps S401 to S405.

S401. Based on the auxiliary training data set and the local training data set, train the intermediate tuning model.

In some embodiments, the intermediate tuning model may be a BP neural network that has been initially trained. After merging the auxiliary training data set and the local training data set, multiple sets of input parameter values are sequentially input into the initially constructed BP neural network. Since in the process of training the intermediate tuning model, an auxiliary training data set is introduced to increase the amount of training data, the local training data set ensures that the trained intermediate tuning model can better fit the newly built device.

Next, based on the output value of the initially constructed BP neural network, compared with the value of the output parameter corresponding to the input parameter, the comparison difference is determined.

Finally, forward feedback is performed based on the difference in comparison, and the initial first weight vector W, the initial second weight vector P, and the initial bias parameter value β of the initially constructed BP neural network are modified to obtain an intermediate tuning model.

S402: Verify the intermediate tuning model based on the verification data set.

As mentioned in the previous embodiment, the verification data set is a data set that has the same origin as the auxiliary training data set. After inputting the corresponding values of the input parameters of the data set into the intermediate tuning model, the intermediate tuning model outputs the corresponding intermediate tuning model output. Parameter corresponding value.

Then, calculate the difference between the corresponding value of the output parameter of the intermediate tuning model and the output parameter value corresponding to the input parameter value in the verification data set. If the difference is less than the preset difference threshold, then the verification of this set of input parameters is obtained. result. Otherwise, the verification fails. Multiple sets of input parameters in the verification data set are input into the intermediate tuning model in sequence, and multiple sets of verification results are obtained.

Finally, based on the ratio of the number of groups that passed the verification in the multi-group verification results to the total number of groups of data in the verification data set, the passing proportion of the verification data set was determined. In some embodiments, if the pass ratio is greater than the preset pass threshold, the intermediate tuning model may be considered to have been trained. Otherwise, if the pass ratio is less than or equal to the preset pass threshold, the above steps can be repeated to continue training to obtain a new intermediate tuning model until the pass ratio is greater than the preset pass threshold.

S403: After passing the verification, run the intermediate tuning model according to the preset input parameter value to obtain the output parameter value of the intermediate tuning model.

If the intermediate tuning model is verified in the above steps, the intermediate tuning model can be run on the new device to further improve the accuracy of the output parameters of the intermediate tuning model on the new device.

In some embodiments, a set of preset input parameter values can be input to the intermediate tuning model, and the intermediate tuning model outputs corresponding output parameter values. Among them, the preset input parameter values can be multiple sets of test input values set by humans in advance, and the parameters corresponding to the preset input parameter values and output parameter values can correspond to the input parameter values and output parameter values in the above-mentioned auxiliary training data set. The parameters are the same.

S404: Obtain the loop PID parameter value of the newly created device after operating according to the preset input parameter value.

You can also input the above preset input parameter values into the control loop corresponding to the new device that is not included in the local training data set. After the new device is actually run, the loop PID parameters output by the new device are obtained, where the loop PID parameters include: Proportional coefficient (proportionality), integration time (minutes), derivative time (minutes).

S405, perform parameter optimization on the intermediate tuning model according to the loop PID parameter value and output parameter value, and obtain to parameter tuning model.

Compare the loop PID parameter values and output parameter values corresponding to the same preset input parameter, and obtain the comparison difference. Based on the comparison difference, the parameters of the intermediate tuning model are optimized to obtain the parameter tuning model.

It should be noted that the preset input parameter values can be input into the control loops in the newly built device that have not participated in the construction of the local training data set, and then the intermediate tuning model is adjusted based on the comparison results between the loop PID parameter values output by these control loops and the output parameter values. Optimize to improve the coverage of the parameter tuning model for each new device in the control system to be tuned.

In summary, combined with the above embodiments, the process of training and optimizing the parameter tuning model is shown in Figure 5.

First, the operating data of the auxiliary device was extracted to obtain the target data set, and then the target data set was further divided to generate the auxiliary training data set and verification data set.

You can also select a part of the control loop in the newly built device, perform PID parameter tuning, obtain initial operating data, and extract these initial operating data to obtain a local training data set.

Then, the auxiliary training data set and the local training data set were combined as the training data set to train the constructed BP neural network model, and the intermediate tuning model was obtained.

On this basis, the intermediate tuning model can be verified through the verification data set constructed in the above steps until the passing ratio is greater than the preset passing threshold. Otherwise, the above training process is repeated.

Finally, the parameters of the intermediate tuning model are further optimized in each new model of the system to be tuned, which does not participate in the construction of the local training data set, and the parameter tuning model is obtained.

In this embodiment, the intermediate tuning model is verified and further parameter optimized to obtain a parameter tuning model, which further improves the accuracy of the output parameter value of the parameter tuning model.

In some embodiments, as shown in Figure 6, in the above step S405, the parameters of the intermediate tuning model are optimized according to the loop PID parameter value and the output parameter value to obtain the parameter tuning model, which can be implemented by the following steps S501 to S502.

S501: Determine the parameter error rate based on the loop PID parameter value and the output parameter value.

The parameter error rate ε _t can be calculated by the following formula:

Among them, 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. is the first weight vector of the intermediate tuning model of ε _t in the i-th control loop. h _t ( _xi ) is the intermediate integer The output parameter value output by the fixed value model, c( _xi ) is the loop PID parameter value output by the newly created device.

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 preset input parameter value.

S502: Iteratively correct the model parameters of the intermediate tuning model according to the parameter error rate to obtain the parameter tuning model.

On this basis, the maximum mean difference method can be used to determine the model data distribution composed of the preset input parameter values and output parameter values of the intermediate tuning model, and the data distribution composed of the preset input parameter values and loop PID parameter values. The difference can be represented by the above parameter error rate ε _t .

Then, based on the parameter error rate ε _t , the mapping between the input value and the output value describing the intermediate tuning parameters is modified to make it closer to the mapping relationship with the new model, and the parameter tuning model is obtained.

In this embodiment, the parameter error rate is determined based on the loop PID parameter value and output parameter value running on the newly created model, and the parameter setting is further corrected accordingly, thereby improving the mapping accuracy of the parameter setting model.

In some embodiments, the model parameters include: a first weight vector and a bias parameter value.

The first weight vector may represent the weight of each layer of the multi-layer BP neural network that constitutes the intermediate tuning model, and the bias parameter value is the control parameter value for the activation state of the neurons in the BP neural network. In some embodiments, a second weight vector P can also be set in the BP neural network, which together with the first weight vector determines the mapping relationship from the input value to the output value of the intermediate tuning model.

As shown in Figure 7, in the above step S502, the model parameters of the intermediate tuning model are iteratively corrected according to the parameter error rate to obtain the parameter tuning model, which can be implemented by the following steps S601 to S604.

S601: Correct the intermediate offset parameter value according to the parameter error rate to obtain the process offset parameter value.

According to the parameter error rate ε _t determined in the above steps, the above-mentioned intermediate offset parameter value is corrected, which can be expressed by the following formula:

β _t =ε _t /(1-ε _t ) ^b , when i=n+1,...,n+m

Then, the original intermediate offset parameter value is replaced with the corrected intermediate offset parameter value, that is, the process offset parameter value.

S602: Modify the first intermediate weight vector according to the process bias parameter value to obtain the first process weight vector.

On this basis, continue to modify the first intermediate weight vector with the process offset parameters determined in the above steps according to the following formula:

Replacing the first intermediate weight vector with the corrected first process weight vector completes a correction process of the first intermediate weight vector.

S603: Obtain a new intermediate tuning model based on the process bias parameter value and the first process weight vector, and re-determine the parameter error rate of the intermediate tuning model.

In this way, a new intermediate tuning model is formed by the process bias parameter value, the first process weight vector, and the second weight vector. The preset input parameters are input into the new device and the new intermediate tuning model respectively. According to the above formula and the new The loop PID parameter values of the device and the output parameter values of the new intermediate tuning model were recalculated to determine the new parameter error rate.

S604: Determine whether the parameter error rate is less than a preset threshold.

In some embodiments, a preset threshold can be set, and when the redetermined parameter error rate is greater than or equal to the preset threshold, the above steps can be repeatedly performed to continue correcting the intermediate tuning model.

S605, repeat the above process until the parameter error rate is less than the preset threshold, and use the intermediate tuning model as the parameter tuning model.

When the redetermined parameter error rate is less than the preset threshold, the intermediate tuning model can be considered to have been corrected, and the corrected intermediate tuning model will be used as the final parameter tuning model.

In this embodiment, the intermediate tuning model is iteratively corrected based on the parameter error rate, which improves the degree of fit between the intermediate tuning model and the new model, and improves the accuracy of the parameter tuning model's output of parameter values on the new model.

In some embodiments, as shown in Figure 8, in the above step S401, based on the auxiliary training data set and the local training data set, before training to obtain the intermediate tuning model, the following steps may also be included:

S701: Determine the root mean square error and coefficient of determination based on the auxiliary training data set and the local training data set.

In some embodiments, the root mean square error RMSE may be determined by:

The coefficient of determination R ² can be determined by the following formula:

Among them, m represents the sample size of the training set composed of the auxiliary training data set and the local training data set, yi represents the true value of the i-th sample of the training set on the new device and the auxiliary device, and y′ _i represents the i-th sample The predicted value of the output of the sample on the built model, is the average of the true values of the sample.

S702: Determine the number of hidden layer nodes of the initial model for parameter tuning based on the root mean square error and the coefficient of determination.

Then, based on the values of the root mean square error and coefficient of determination determined above, select the number of hidden layer nodes of the initial model for parameter tuning. For example, the root mean square error and the coefficient of determination can be divided into multiple ranges, which correspond to the number of hidden layer nodes. When the root mean square error and the coefficient of determination fall into this range, the corresponding hidden layer node can be determined. quantity.

In some embodiments, in the embodiment of the present application, the number of hidden layer nodes may be 5.

S703: Build an initial parameter tuning model based on the number of hidden layer nodes, the preset number of input layer nodes, and the preset number of output layer nodes.

The preset number of input layer nodes can be determined according to the number of input parameters, which can be, for example, 7. The preset number of output layer nodes can be determined according to the number of output parameters, which can be, for example, 3.

In this way, based on the number of hidden layer nodes, the preset number of input layer nodes, and the preset number of output layer nodes, the network structure of the BP neural network can be initially built.

In some embodiments, on this basis, when constructing the parameter tuning initial model, the initial first weight vector W, the initial second weight vector P, and the initial bias value β of the parameter tuning initial model may also be initialized. In some embodiments, the initial weight vector is in,

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.

The initial bias parameter value β can be set as N is the value of n+m.

The initial second weight vector can be set to Corresponds to the initial weight vector.

After constructing the network structure and network parameters of the BP neural network, the initial parameter tuning model is obtained.

S704: Train the initial parameter tuning model based on the auxiliary training data set and the local training data set to obtain an intermediate tuning model.

Finally, according to the above embodiment, the constructed auxiliary training data set and the local training data set are trained to construct the initial parameter tuning model, and the parameters of the initial parameter tuning model are modified to obtain the intermediate tuning model.

In this embodiment, the number of hidden layer nodes is determined based on the root mean square error and coefficient of determination, and an initial parameter tuning model is further constructed, so that the constructed model is more consistent with the data volume of the training set and has stronger fitting ability. , the training speed is faster.

As shown in Figure 9, the embodiment of the present application also provides an industrial process control method, which can be applied to processing equipment that can perform PID parameter tuning for newly built devices. Referring to Figure 9, the method can include the following steps:

S801. Determine the setting parameter values of the newly-built device to be controlled according to the parameter setting model. The parameter setting model is obtained based on the construction method of the parameter setting model in any of the foregoing embodiments.

In some embodiments, a parameter tuning model can be determined according to the above embodiments, and the model can be applied to a new device. According to the input parameter values in the new device, the corresponding PID parameter value is output, that is, the new device to be controlled. Tuning parameter values.

It can be understood that since in the process of training the parameter tuning model, multiple new devices in the industrial control system to be tuned are introduced, the parameter tuning model is suitable for the multiple new devices in the industrial control system to be tuned. Able to output accurate tuning parameter values.

S802: Control the newly created device to execute the target process according to the setting parameter value.

The target process may be a process in which the new device is run according to the setting parameter value. It can be understood that the setting parameter value is the value output by the parameter setting model and needs to be adjusted for the new device. Therefore, tuning parameter values can be input into the newly built device to control its operation and execute the target process.

In this embodiment, the operation of the newly-built device is controlled based on the tuning parameter value output by the parameter-tuning model. The preset control effect can be achieved with higher efficiency on the newly-built device, and the degree of automation of parameter tuning is improved.

Referring to Figure 10, an embodiment of the present application also provides a device 100 for constructing a parameter tuning model, which includes:

The data set construction module 1001 is configured to construct an auxiliary training data set and a verification data set based on the circuit information of the newly created device and the operating data of the auxiliary device, where the new device is the device to which the parameter tuning model to be trained is applied, and the auxiliary device is The device whose parameter setting has been completed and is officially in operation;

The data set construction module 1001 is also configured to construct a local training data set based on the initial operating data of the newly created device.

The model training module 1002 is configured to train and obtain a parameter tuning model based on the auxiliary training data set, the verification data set and the local training data set.

The data set construction module 1001 is also specifically configured to filter out the target data set from the operating data of the auxiliary device according to the circuit type of the new device and the corresponding physical characteristics of the control loop; split the target data set to obtain auxiliary training data set and validation data set.

The data set construction module 1001 is also specifically configured to filter multiple optional operating data matching the loop type from the operating data of the auxiliary device according to the loop type of the newly-built device; and select multiple optional operating data matching the loop type according to the physical characteristics corresponding to the control loop of the newly-built device. Filter out the target data set from optional run data.

The model training module 1002 is specifically configured to train to obtain an intermediate tuning model based on the auxiliary training data set and the local training data set; verify the intermediate tuning model based on the verification data set; and after passing the verification, run according to the preset input parameter values Intermediate tuning model, obtain the output parameter value of the intermediate tuning model; obtain the loop PID parameter value after the new device operates according to the preset input parameter value; perform parameter optimization on the intermediate tuning model according to the loop PID parameter value and output parameter value, and obtain the parameters Tuning the model.

The model training module 1002 is specifically configured to determine the parameter error rate based on the loop PID parameter value and the output parameter value; and iteratively correct the model parameters of the intermediate tuning model based on the parameter error rate to obtain the parameter tuning model.

The model training module 1002 is specifically configured as follows: the model parameters include: a first intermediate weight vector and an intermediate offset parameter value; according to the parameter error rate, the intermediate offset parameter value is corrected to obtain a process offset parameter value; according to the process offset Modify the first intermediate weight vector with the parameter value to obtain the first process weight vector; obtain a new intermediate tuning model based on the process offset parameter value and the first process weight vector, and re-determine the parameter error rate of the intermediate tuning model; repeat the above process until the parameter error rate is less than the preset threshold, and the intermediate tuning model is used as the parameter tuning model.

The model building module 1003 is configured to determine the root mean square error and the coefficient of determination based on the auxiliary training data set and the local training data set; determine the number of hidden layer nodes of the initial model for parameter tuning based on the root mean square error and the coefficient of determination; The number of nodes, the preset number of input layer nodes and the preset number of output layer nodes are used to construct an initial parameter tuning model; based on the auxiliary training data set and the local training data set, the initial parameter tuning model is trained to obtain an intermediate tuning model.

Referring to Figure 11, an embodiment of the present application also provides an industrial process control device 110, including:

The determination module 1101 is configured to determine the tuning parameter values of the new device to be controlled based on the parameter tuning model. The parameter tuning model is obtained based on the construction method of the parameter tuning model in any of the previous embodiments.

The control module 1102 is configured to control the newly created device to execute the target process according to the setting parameter value.

Please refer to Figure 12. This embodiment also provides a processing device. The processing device includes: a processor 2001, a memory 2002 and a bus. The memory 2002 stores machine-readable instructions executable by the processor 2001. When the processing device When ready to run, the above machine-readable instructions are executed, and the processor 2001 communicates with the memory 2002 through a bus. The processor 2001 is configured to execute the steps of the parameter setting model construction method or the industrial process control method in the above embodiments.

The memory 2002, the processor 2001, and the components of the bus are directly or indirectly electrically connected to each other to realize data transmission or interaction. For example, these components may be electrically connected to each other through one or more communication buses or signal lines. The parameter setting model building system or the data processing device of the industrial process control system includes at least one software function that can be stored in the memory 2002 in the form of software or firmware or solidified in the operating system (OS) of the processing device. module. The processor 2001 is configured to execute executable modules stored in the memory 2002, such as software function modules and computer programs included in a parameter setting model building system or a data processing device of an industrial process control system.

Among them, the memory 2002 can be, but is not limited to, random access memory (Random Access Memory, RAM), read-only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), and can Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc.

In some embodiments, the present application also provides a storage medium. A computer program is stored on the storage medium. When the computer program is run by a processor, the steps of the above method embodiments are executed. The specific implementation methods and technical effects are similar and will not be described again here.

Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be described again in this application. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components may be combined or can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the coupling or direct coupling or communication connection between each other shown or discussed may be through some communication interfaces, indirect coupling or communication connection of devices or modules, and may be in electrical, mechanical or other forms.

In addition, each functional unit in various embodiments of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in various embodiments of this application. step. The aforementioned storage media include: U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application. All are covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Industrial applicability

The technical solution provided by the embodiments of the present application can be applied to the field of industrial automation control technology. By using the parameter setting model construction method and the industrial process control method provided by the present application, an auxiliary training data set can be constructed with the help of the operating data of the auxiliary device, and the auxiliary training data set can be constructed. Together with a local training data set constructed from operational data from a newly built installation, a parameter tuning model is trained and built. This application makes full use of the operating data of the auxiliary device that has completed parameter setting and is officially in operation to train the parameter setting model, which makes up for the shortcomings of less effective data available for the new device and improves the PID parameter tuning of the parameter setting model established on the new device. Accuracy and efficiency.

Claims (10)

1. A method for constructing a parameter tuning model, the method includes:

   An auxiliary training data set and a verification data set are constructed based on the circuit information of the newly-built device and the operating data of the auxiliary device, where the new device is a device to which the parameter tuning model to be trained is applied, and the auxiliary device is a device whose parameter tuning has been completed and Officially operating device;

   Construct a local training data set based on the initial operating data of the newly built device;

   According to the auxiliary training data set, the verification data set and the local training data set, a parameter tuning model is trained.
2. The method of constructing a parameter tuning model according to claim 1, wherein said constructing an auxiliary training data set and a verification data set based on the circuit information of the newly-built device and the operating data of the auxiliary device includes:

   Filter out the target data set from the operating data of the auxiliary device according to the circuit type of the new device and the corresponding physical characteristics of the control loop;

   The target data set is split to obtain the auxiliary training data set and the verification data set.
3. The method of constructing a parameter tuning model according to claim 2, wherein the target data set is selected from the operating data of the auxiliary device according to the loop type of the newly-built device and the corresponding physical characteristics of the control loop, including :

   According to the circuit type of the newly-built device, screen a plurality of optional operating data matching the circuit type from the operating data of the auxiliary device;

   The target data set is selected from the plurality of optional operating data according to the physical characteristics corresponding to the control loop of the newly constructed device.
4. The method for constructing a parameter tuning model according to any one of claims 1 to 3, wherein the parameter tuning model is obtained by training according to the auxiliary training data set, the verification data set and the local training data set, include:

   Based on the auxiliary training data set and the local training data set, train an intermediate tuning model;

   Verify the intermediate tuning model based on the verification data set, and after passing the verification, run the intermediate tuning model according to the preset input parameter values to obtain the output parameter values of the intermediate tuning model;

   Obtain the loop PID parameter value after the new device is operated according to the preset input parameter value;

   According to the loop PID parameter value and the output parameter value, parameter optimization is performed on the intermediate tuning model to obtain the parameter tuning model.
5. The method of constructing a parameter tuning model according to claim 4, wherein the parameter optimization of the intermediate tuning model is performed according to the loop PID parameter value and the output parameter value to obtain the parameter tuning model, including :

   Determine the parameter error rate according to the loop PID parameter value and the output parameter value;

   According to the parameter error rate, the model parameters of the intermediate tuning model are iteratively corrected to obtain the parameter tuning model.
6. The method of constructing a parameter tuning model according to claim 5, wherein the model parameters include: a first intermediate weight vector and an intermediate bias parameter value;

   Iteratively correcting the model parameters of the intermediate tuning model according to the parameter error rate to obtain the parameter tuning model includes:

   According to the parameter error rate, the intermediate offset parameter value is corrected to obtain a process offset parameter value;

   Modify the first intermediate weight vector according to the process offset parameter value to obtain a first process weight vector;

   Obtain a new intermediate tuning model according to the process bias parameter value and the first process weight vector, and re-determine the parameter error rate of the intermediate tuning model;

   Repeat the above process until the parameter error rate is less than the preset threshold, and use the intermediate tuning model as the parameter tuning model.
7. The method for constructing a parameter tuning model according to claim 4, wherein before training to obtain an intermediate tuning model based on the auxiliary training data set and the local training data set, the method further includes:

   Determine the root mean square error and coefficient of determination according to the auxiliary training data set and the local training data set;

   According to the root mean square error and the coefficient of determination, determine the number of hidden layer nodes of the initial model for parameter tuning;

   Construct the initial parameter tuning model according to the number of hidden layer nodes, the preset number of input layer nodes, and the preset number of output layer nodes;

   The initial parameter tuning model is trained according to the auxiliary training data set and the local training data set to obtain the intermediate tuning model.
8. An industrial process control method, the method includes:

   Determine the tuning parameter values of the newly-built device to be controlled according to the parameter tuning model, which is obtained based on the construction method of the parameter tuning model described in any one of claims 1-7;

   The newly created device is controlled to execute a target process according to the setting parameter value.
9. A processing device, the processing device includes: a processor, a storage medium and a bus, the storage medium stores machine readable instructions executable by the processor, when the processing device is running, the processor and The storage media communicate through a bus, and the processor executes the machine-readable instructions to execute the method for constructing a parameter tuning model as claimed in any one of claims 1-7 or the industrial method as claimed in claim 8. Steps of process control methods.
10. A computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the method for constructing a parameter setting model as described in any one of claims 1-7 is implemented. The steps of the industrial process control method according to claim 8.