Model-based training of industrial control methods and apparatus

Abstract

The application discloses an industrial control method and device based on model training, and belongs to the field of industrial control. The industrial control method based on model training comprises the following steps: using a target algorithm prediction model to perform operation processing on real-time industrial big data in a target industrial scene, so that a target control algorithm and a target independent variable for controlling a target controlled object in the target scene output by the target algorithm prediction model are obtained; and inputting the value of the target independent variable in the real-time industrial big data into a target controller in which the target control algorithm is deployed, so that the value of a target control variable is obtained, to control the target controlled object. The industrial control method based on model training has high precision and accuracy of the obtained target control algorithm, can significantly reduce the labor and time cost, improve the control efficiency and control effect, and has high universality.

Description

Technical Field

[0001] This application belongs to the field of industrial control, and in particular relates to an industrial control method and device based on model training. Background Technology

[0002] With the development of artificial intelligence, industrial big data and AI are increasingly being applied to the field of automatic control technology. In related technologies, the main approach involves manually analyzing industry process mechanisms to obtain corresponding parameter correlations. For example, industry experts can pre-analyze the process mechanisms to obtain correlation parameters related to the controlled parameters, thereby deriving control algorithms for industrial control. However, this method results in control algorithms with low precision and accuracy, requires significant manpower, and demands a high level of expertise from users, thus affecting the efficiency of algorithm acquisition. Furthermore, its applicability is limited, and it cannot be universally applied. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes an industrial control method and device based on model training, which can significantly reduce labor and time costs, improve control efficiency and control effect, and has a high degree of versatility.

[0004] In a first aspect, this application provides an industrial control method based on model training, the method comprising:

[0005] A target algorithm prediction model is used to process real-time industrial big data in a target industrial scenario to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario, output by the target algorithm prediction model. The target algorithm prediction model is obtained by training on historical sample industrial big data. Both the historical sample industrial big data and the real-time industrial big data include values ​​for multiple independent variables and control variables. The historical sample industrial big data represents historical data, and the real-time industrial big data represents current data. The target independent variable is the independent variable corresponding to the target control variable determined from the multiple independent variables in the real-time industrial big data. The target control algorithm is a control algorithm used to represent the relationship between the target independent variable and the target control variable.

[0006] The values ​​of the target independent variables in the real-time industrial big data are input to the target controller equipped with the target control algorithm to obtain the values ​​of the target control variables, thereby controlling the target controlled object.

[0007] According to the model-trained industrial control method of this application, the target algorithm prediction model, which is trained, can directly obtain the target independent variables and the target control algorithm corresponding to the target independent variables based on the input real-time industrial big data. The obtained target control algorithm has high precision and accuracy, and does not require users to manually select independent variables related to the target control variables, which significantly reduces manpower and time costs, lowers the application threshold, and has high universality. Then, industrial control is performed based on the automatically obtained target independent variable values ​​and target control algorithm, which has high control effect.

[0008] According to one embodiment of this application, after processing real-time industrial big data in a target industrial scenario using a target algorithm prediction model to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario output by the target algorithm prediction model, the method further includes:

[0009] The target independent variable and the target control algorithm are input into the simulation system under the target industrial scenario, and the values ​​of the target control variable output by the simulation system are obtained;

[0010] The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system.

[0011] According to one embodiment of this application, the optimization of the target independent variable and the target control algorithm based on the value of the target control variable output by the simulation system includes:

[0012] The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system.

[0013] The optimized target independent variable and the optimized target control algorithm are input into the simulation system, and the value of the target control variable output by the simulation system is obtained again.

[0014] Repeat the steps described above, which optimize the target independent variable and the target control algorithm based on the value of the target control variable output by the simulation system, until the value of the target control variable output by the simulation system meets the target control accuracy.

[0015] According to one embodiment of this application, after inputting the value of the target independent variable in the real-time industrial big data to a target controller deployed with the target control algorithm to obtain the value of the target control variable, the method further includes:

[0016] Based on the value of the target control variable, the target algorithm prediction model is used again to process the updated real-time industrial big data to optimize the target independent variable and the target control algorithm respectively.

[0017] According to one embodiment of this application, the target algorithm prediction model is trained in the following manner:

[0018] Obtain the historical sample industrial big data under the target industrial scenario;

[0019] The historical sample industrial big data is learned by using an initial algorithm prediction model to obtain the correlation between any two elements among multiple independent variables and control variables in the historical sample industrial big data.

[0020] Based on the aforementioned correlation, a target independent variable corresponding to the target control variable is determined from multiple independent variables of the historical sample industrial big data, and a target control algorithm corresponding to the target independent variable corresponding to the historical sample industrial big data is also determined.

[0021] According to one embodiment of this application, obtaining the historical sample industrial big data in the target industrial scenario includes:

[0022] The historical industrial big data acquired in the target industrial scenario is preprocessed to obtain preprocessed historical industrial big data; the historical industrial big data is at least a portion of the historical data in the target industrial scenario.

[0023] The preprocessed historical industrial big data is processed according to the processing methods corresponding to each category of the preprocessed historical industrial big data to obtain the historical sample industrial big data.

[0024] According to one embodiment of this application, the step of processing the preprocessed historical industrial big data based on the processing methods corresponding to the respective categories of the preprocessed historical industrial big data to obtain the historical sample industrial big data includes:

[0025] In the case where the category is the first category, the preprocessed historical industrial big data is identified as the historical sample industrial big data;

[0026] In the case where the category is the second type, based on the database time sequence arrangement method and the time information corresponding to the preprocessed historical industrial big data, the preprocessed historical industrial big data is arranged and labeled to obtain the historical sample industrial big data.

[0027] In the case of the third category, the preprocessed historical industrial big data is split based on the sub-category of the control system under the target industrial scenario, and the classified historical industrial big data corresponding to each sub-category is obtained respectively; based on the time sequence arrangement of the database and the time information corresponding to the classified historical industrial big data, the classified historical industrial big data is arranged and labeled to obtain the historical sample industrial big data.

[0028] According to one embodiment of this application, the historical sample industrial big data includes at least two of the following: operating condition variables, control system output data under the target industrial scenario, process object variables, and control system variables under the target industrial scenario.

[0029] According to one embodiment of this application, the initial algorithm prediction model includes an encoder module and a decoder module connected in sequence. The step of learning the historical sample industrial big data using the initial algorithm prediction model includes:

[0030] The historical sample industrial big data is input into the encoder module to obtain the intermediate representation output by the encoder module. The intermediate representation is obtained by the encoder module after learning the dependency relationship between data at different positions in the data sequence of the historical sample industrial big data and mapping the historical sample industrial big data.

[0031] The intermediate representation is input to the decoder module to obtain the target independent variable and the target control algorithm output by the decoder module that correspond to the historical sample industrial big data.

[0032] According to one embodiment of this application, at least one of the encoder module and the decoder module includes:

[0033] The system comprises a multi-layer self-attention sub-layer and a feedforward neural network layer, wherein the output of the multi-layer self-attention sub-layer is connected to the input of the feedforward neural network layer; wherein...

[0034] The self-attention sublayer is used to process the data at the target location based on the data at other locations in the data sequence besides the target location, and to obtain the correlation between the data at the target location and the data at other locations.

[0035] Secondly, this application provides an industrial control device based on model training, the device comprising:

[0036] The first processing module is used to process real-time industrial big data in a target industrial scenario using a target algorithm prediction model, and to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario, output by the target algorithm prediction model. The target algorithm prediction model is obtained by training on historical sample industrial big data. Both the historical sample industrial big data and the real-time industrial big data include values ​​for multiple independent variables and control variables. The historical sample industrial big data is used to represent historical data, and the real-time industrial big data is used to represent current data. The target independent variable is the independent variable corresponding to the target control variable determined from among the multiple independent variables in the real-time industrial big data. The target control algorithm is a control algorithm used to represent the relationship between the target independent variable and the target control variable.

[0037] The second processing module is used to input the values ​​of the target independent variables in the real-time industrial big data to the target controller deployed with the target control algorithm, so as to obtain the values ​​of the target control variables and control the target controlled object.

[0038] According to the model-trained industrial control device of this application, the target algorithm prediction model, based on real-time industrial big data input, can directly obtain the target independent variables and the target control algorithm corresponding to the target independent variables. The obtained target control algorithm has high precision and accuracy, and does not require users to manually select independent variables related to the target control variables, significantly reducing manpower and time costs, lowering the application threshold, and having high universality. Then, industrial control is performed based on the automatically obtained target independent variable values ​​and the target control algorithm, which has high control effect. The target control algorithm obtained from the automatically obtained target independent variables is used for industrial control, and the obtained target control algorithm has high precision and accuracy.

[0039] Thirdly, this application provides a control system based on the model-training-based industrial control method described in the first aspect, comprising:

[0040] Big data training model system;

[0041] An edge real-time control system, wherein the big data training model system is communicatively connected to the edge real-time control system.

[0042] Fourthly, this application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the model-trained industrial control method as described in the first aspect above.

[0043] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the model-trained industrial control method as described in the first aspect above.

[0044] The above-described one or more technical solutions in the embodiments of this application have at least one of the following technical effects:

[0045] By training a target algorithm prediction model based on real-time industrial big data, the target independent variables and the corresponding target control algorithms can be directly obtained. The obtained target control algorithms have high precision and accuracy, and users do not need to manually select independent variables related to the target control variables, which significantly reduces manpower and time costs, lowers the application threshold, and has high universality. Then, industrial control is carried out based on the automatically obtained target independent variable values ​​and target control algorithms, which has high control effect.

[0046] Furthermore, by inputting historical industrial big data samples into the initial algorithm prediction model, the initial algorithm prediction model is trained to output the independent variables related to the control variables and the control algorithm. This allows the model to obtain the corresponding parameter correlations and control algorithms through big data training without relying on research into the process mechanisms of specific industries. It eliminates the need for users to pre-label sample control variables and their corresponding independent variables, significantly reducing manpower and time costs and improving the model's learning ability and intelligence. It also effectively eliminates the influence of subjective human factors, thereby improving the model's accuracy and precision, making it suitable for various industrial scenarios.

[0047] Furthermore, by automatically acquiring the target independent variables and target control algorithms corresponding to the target control variables based on real-time industrial big data, and then conducting simulation tests based on the target independent variables and target control algorithms, the target independent variables and target control algorithms can be optimized based on the simulation results to achieve the best results. This can further improve the accuracy and precision of the acquired target independent variables and target control algorithms, thereby improving the subsequent control effect.

[0048] Furthermore, based on real-time industrial big data, the system automatically acquires the target independent variables and target control algorithms corresponding to the target control variables. Then, based on the actual values ​​obtained after controlling the target controller using the target control algorithm, it optimizes the target independent variables and target control algorithms, thereby enabling the target independent variables and target control algorithms to reach their optimal state. In addition to effectively eliminating the influence of human subjective factors, it can also automatically adjust the target independent variables and target control algorithms based on the actual control situation, thereby improving control efficiency and control effect.

[0049] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0050] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0051] Figure 1 This is one of the flowcharts illustrating the industrial control method based on model training provided in the embodiments of this application;

[0052] Figure 2 This is the second flowchart of the industrial control method based on model training provided in the embodiments of this application;

[0053] Figure 3 This is the third flowchart of the industrial control method based on model training provided in the embodiments of this application;

[0054] Figure 4 This is a schematic diagram of the structure of the control system provided in the embodiments of this application;

[0055] Figure 5 This is a schematic diagram of the structure of an industrial control device based on model training provided in an embodiment of this application;

[0056] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0057] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0058] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0059] The following description, in conjunction with the accompanying drawings, details the model-training-based industrial control method, model-training-based industrial control device, control system, electronic device, readable storage medium, and computer program product provided in this application, through specific embodiments and application scenarios.

[0060] like Figure 1 As shown, the industrial control method based on model training includes steps 110 and 120.

[0061] Step 110: Use the target algorithm prediction model to process real-time industrial big data in the target industrial scenario to obtain the target control algorithm and target independent variables used to control the target controlled object in the target scenario, as output by the target algorithm prediction model.

[0062] In this step, the target algorithm prediction model is obtained by training on historical sample industrial big data. Both historical sample industrial big data and real-time industrial big data include the values ​​of multiple independent variables and control variables.

[0063] Historical industrial big data is used to represent historical data, while real-time industrial big data is used to represent current data. The target independent variable is the independent variable that corresponds to the target control variable, which is determined from multiple independent variables in the real-time industrial big data. The target control algorithm is the control algorithm used to represent the relationship between the target independent variable and the target control variable.

[0064] Real-time industrial big data refers to at least a portion of the industrial big data generated in a target industrial scenario, which includes the values ​​of multiple independent variables and control variables.

[0065] The categories of real-time industrial big data include, but are not limited to: operating condition variables, control system output data in the target industrial scenario, process object variables, and control system variables in the target industrial scenario.

[0066] It should be noted that not all of the independent variables are related to the control variables.

[0067] In some embodiments, some of the multiple independent variables may be related to the control variable; or multiple independent variables may not be related to the control variable; or multiple independent variables may all be related to the control variable. This application does not limit this.

[0068] The target control variable is the parameter that needs to be controlled, and the target independent variable is the parameter that is associated with the target control variable among multiple independent variables.

[0069] The number of target control variables can be one or more.

[0070] For example, in a temperature control scenario, the parameters affecting the temperature value include pressure, current, and flow rate. Temperature is the target control variable, while pressure, current, and flow rate are the target independent variables. The relationship between temperature and pressure, current, and flow rate forms the target control algorithm.

[0071] In this step, both the target control algorithm and the target independent variable are automatically acquired by the system, rather than being determined manually.

[0072] In actual implementation, the target independent variable can be determined from multiple independent variables based on the correlation between any two elements among the multiple independent variables and control variables.

[0073] Once the target independent variable is obtained, the target control algorithm between the target independent variable and the target control variable can be further determined.

[0074] It is understandable that different control types may have different independent variables and control algorithms.

[0075] In some embodiments, the target algorithm prediction model can be a Transformer model.

[0076] Among them, Transformer is a deep neural network model based on self-attention mechanism, which can efficiently process sequential data in parallel.

[0077] Understandably, each industrial scenario can have a corresponding algorithmic prediction model.

[0078] The algorithm prediction models may differ for different industrial scenarios.

[0079] Step 120: Input the values ​​of the target independent variables in the real-time industrial big data to the target controller with the target control algorithm to obtain the values ​​of the target control variables, so as to control the target controlled object.

[0080] In this embodiment, the target controller is a controller for the target industrial scenario.

[0081] The target industrial scenario can be any industrial scenario, such as: fused magnesia smelting scenario, thermal power generation scenario, air conditioning system control scenario, and semiconductor etching machine reaction chamber temperature control scenario, etc., and this application does not limit it.

[0082] In some embodiments, real-time industrial big data may include industrial big data from any one of the following scenarios: fused magnesia smelting, thermal power generation, air conditioning system control, and semiconductor etching machine reaction chamber temperature control.

[0083] It should be noted that this model-trained industrial control method is applied to a target controller, which is equipped with a target control algorithm. The target control algorithm is used to characterize the control relationship between the target independent variable and the target control variable.

[0084] The target control algorithm is an algorithm used to control the target controller in order to adjust the parameters to obtain the desired value of the target independent variable.

[0085] The target controlled object is the object that ultimately needs to be controlled.

[0086] It is understandable that the target controlled object may be the same as or different from the target control variable. In different situations, the target control variable is used to control the target controlled object.

[0087] In actual execution, once the target independent variables in the target industrial scenario are obtained, the corresponding values ​​of the target independent variables can be extracted from real-time industrial big data and input into the target controller with the target control algorithm to obtain the values ​​of the target control variables in order to control the target controlled object.

[0088] In some embodiments, the target controlled object can also be used as an independent variable that influences the value of the target control variable.

[0089] For example, in PID control, the value of the target controlled object at the previous moment can be used as the independent variable to adjust the value of the target control variable at the next moment.

[0090] During the research and development process, the inventors discovered that in related technologies, the independent variables related to the control variables are mainly obtained through manual analysis or expert research on industry process mechanisms. After obtaining the target independent variable, the control algorithm between the control variable and the independent variable is further predicted based on the manually obtained target independent variable and target control variable. Then, the corresponding industrial control is carried out based on the control algorithm. This method requires prior knowledge of the industrial mechanism of a specific industry. If it involves cross-industry collaboration, it needs to be supported by algorithms from other industries, and it cannot be universally applicable.

[0091] In this application, the target algorithm prediction model, which is trained, can directly obtain the target independent variable and the target control algorithm corresponding to the target independent variable based on the input real-time industrial big data. This eliminates the need for users to manually obtain the target independent variable, significantly reducing manpower and time costs. It has a low operating threshold and is applicable to any industry, thereby improving the efficiency of obtaining the control algorithm and having high universality and versatility.

[0092] In addition, by analyzing the massive amounts of industrial big data input, the calculation error caused by missed detection can be reduced, thereby improving the accuracy and precision of the acquired target control algorithm and thus improving the control effect.

[0093] According to the model-trained industrial control method provided in this application, the target algorithm prediction model, based on real-time industrial big data input, can directly obtain the target independent variables and the target control algorithm corresponding to the target independent variables. The obtained target control algorithm has high precision and accuracy, and does not require users to manually select independent variables related to the target control variables, which significantly reduces manpower and time costs, lowers the application threshold, and has high universality. Then, industrial control is performed based on the automatically obtained target independent variable values ​​and target control algorithm, which has a high control effect.

[0094] The training method for the target algorithm prediction model will be explained below.

[0095] like Figure 2 As shown, in some embodiments, the target algorithm prediction model can be trained in the following manner:

[0096] Acquire historical industrial big data samples from the target industrial scenario;

[0097] By learning from historical industrial big data through an initial algorithm prediction model, the correlation between any two elements among multiple independent and control variables in the historical industrial big data can be obtained.

[0098] Based on the correlation, the target independent variable corresponding to the target control variable is determined from multiple independent variables in historical industrial big data samples, as well as the target control algorithm corresponding to the target independent variable corresponding to the historical industrial big data samples.

[0099] In this embodiment, historical sample industrial big data is sample data used to train the target algorithm prediction model.

[0100] Historical sample industrial big data consists of massive amounts of data.

[0101] In some embodiments, historical sample industrial big data may include at least two of the following: operating condition variables, control system output data under the target industrial scenario, process object variables, and control system variables under the target industrial scenario.

[0102] It should be noted that the sample data used to train the target algorithm prediction model should belong to the same industrial scenario as the application data that is subsequently used to apply the target algorithm prediction model, and should be different data.

[0103] For example, taking the industrial scenario of boiler temperature control in a thermal power plant as an example, the factories that control the boiler temperature in a thermal power plant include Plant A, Plant B, and Plant C. In the actual training process, the target algorithm prediction model can be trained based on the industrial big data of Plant A. The trained target algorithm prediction model can be applied to Plant A, Plant B, and Plant C respectively. In this case, the industrial big data of Plant A used in the training process is the historical sample industrial big data, while in the subsequent application process, the industrial big data of Plant B and Plant C input into the trained target algorithm prediction model is the real-time industrial big data.

[0104] Historical industrial big data samples also include the values ​​of target control variables and multiple independent variables, which may differ from the values ​​of target control variables and multiple independent variables included in real-time industrial big data.

[0105] Taking the industrial scenario of furnace temperature control in a thermal power plant boiler as an example, the target control variable is temperature, and multiple independent variables can include any parameters involved in the thermal power plant boiler.

[0106] The initial algorithm prediction model is the model to be trained.

[0107] In some embodiments, the initial algorithm prediction model can be a Transformer model.

[0108] During the training process, the acquired historical sample industrial big data is input into the initial algorithm prediction model, such as the initial Transformer model, so that the model can output the target independent variable and control algorithm corresponding to the target control variable. The initial algorithm prediction model is trained to obtain the trained target algorithm prediction model.

[0109] For example, historical industrial big data samples can be input into an initial algorithm prediction model, which can then learn the relationships between variables. Based on these relationships, independent variables related to the control variables can be selected from the variables corresponding to the input historical industrial big data samples. The model can then continue training to further learn the control algorithm.

[0110] In this embodiment, historical sample industrial big data is input into the initial algorithm prediction model, which learns the correlation between variables. Based on the correlation, independent variables related to the control variables are selected from the variables corresponding to the input historical sample industrial big data. Then, training continues to learn the control algorithm.

[0111] Understandably, during the training process, any control variable and its corresponding independent variable, as well as the control algorithm, can be obtained.

[0112] The following explanation uses PID control as an example.

[0113] For example, through training, the following control algorithm can be output:

[0114] y(t)=(y(t-1),(t),d())

[0115] Where y(t) represents the value of the controlled object at time t; y(t-1) represents the value of the controlled object at time (t-1); u(t) represents the output of the controller at time t, that is, the value of the target control variable at time t; and d(t) represents the disturbance received by the target control variable at time t.

[0116] The value of the target control variable u(t) at time t can be further expressed as:

[0117] u(t)=(e(t),(i))

[0118] Where p(i) is the value of the target independent variable used in the control algorithm for u(t), and this target independent variable is also obtained by training the target algorithm prediction model.

[0119] e(t) can be expressed as:

[0120] e(t) = (t) - (t)

[0121] Where r(t) is the correction setpoint under PID control, which is further determined based on the target independent variable obtained by training the target algorithm prediction model.

[0122] In the application process, it is only necessary to determine the target control variable from any obtained control variables, and then the target independent variable associated with the target control variable can be accurately obtained from multiple independent variables.

[0123] The trained target algorithm prediction model can be deployed to other control systems in the same industrial scenario. The target control algorithm can be output based on the real-time industrial big data input to control the target control system for industrial control.

[0124] For example, after training a target algorithm prediction model for the industrial scenario of boiler temperature control in a thermal power plant using historical sample industrial big data from Plant A, the target algorithm prediction model can be deployed to Plants B and C, so that Plants B and C can obtain the corresponding target control algorithm based on their respective real-time industrial big data.

[0125] The industrial control method based on model training provided in this application involves inputting historical industrial big data samples into an initial algorithm prediction model. The initial algorithm prediction model is trained to output the independent variables related to the control variables and the control algorithm. This method achieves the acquisition of the corresponding parameter correlations and control algorithms through big data training without relying on research into the process mechanisms of specific industries. It eliminates the need for users to pre-label sample control variables and their corresponding independent variables, significantly reducing manpower and time costs and improving the model's learning ability and intelligence. Furthermore, it effectively eliminates the influence of subjective human factors, thereby improving the model's accuracy and precision, making it suitable for various industrial scenarios.

[0126] In some embodiments, the initial algorithm prediction model may include an encoder module and a decoder module connected in sequence. The initial algorithm prediction model learns from historical industrial big data samples and may include:

[0127] Historical sample industrial big data is input into the encoder module, and the intermediate representation output by the encoder module is obtained. The intermediate representation is obtained by the encoder module after learning the dependency relationship between data at different positions in the data sequence of historical sample industrial big data and mapping the historical sample industrial big data.

[0128] The intermediate representation is input into the decoder module to obtain the target independent variables and target control algorithm corresponding to the historical sample industrial big data output by the decoder module.

[0129] In this embodiment, the initial algorithm prediction model may include an encoder module and a decoder module.

[0130] The encoder maps the input data sequence to a set of intermediate representations, while the decoder converts the intermediate representations into the target sequence for output.

[0131] In this application, the encoder module learns the dependencies between data at different locations in the data sequence of historical sample industrial big data to obtain an intermediate representation; and the decoder module transforms the intermediate representation to output the target independent variable and control algorithm, thereby training the target algorithm prediction model.

[0132] In some embodiments, at least one of the encoder module and the decoder module may include: a multi-layer self-attention sub-layer and a feedforward neural network layer, wherein the output of the multi-layer self-attention sub-layer is connected to the input of the feedforward neural network layer; wherein the self-attention sub-layer is used to process the data at the target position based on the data at other positions in the data sequence besides the target position, and to obtain the correlation between the data at the target position and the data at other positions.

[0133] In this embodiment, the number of layers of the multi-layer self-attention sub-layer can be user-defined, and this application does not impose any limitation.

[0134] The output of the multi-layer self-attention sublayer is connected to the input of the feedforward neural network layer.

[0135] The target location can be any position in the data sequence.

[0136] In some embodiments, both the encoder module and the decoder module may include multiple layers of self-attention sublayers and feedforward neural network layers.

[0137] The self-attention sublayer can learn the dependencies between different positions in the data sequence. That is, when processing information at each position, the model will consider information at all other positions in the sequence.

[0138] Based on multi-layer training, the correlation between variables is obtained, thereby further obtaining a control algorithm for a specific target control variable.

[0139] The industrial control method based on model training provided in the embodiments of this application can effectively predict the correlation between variables based on the input industrial big data by using the Transformer architecture to learn the dependency relationship between different positions in the sequence, thereby obtaining a control algorithm for a certain variable based on the correlation relationship. It has strong learning ability and high model accuracy.

[0140] The following explains how historical industrial big data was obtained.

[0141] Continue to refer to Figure 2 In some embodiments, obtaining historical sample industrial big data in the target industrial scenario may include:

[0142] The historical industrial big data of the target industrial scenario is preprocessed to obtain the preprocessed historical industrial big data; the historical industrial big data is at least part of the historical data of the target industrial scenario.

[0143] Based on the processing methods corresponding to the respective categories of preprocessed historical industrial big data, the preprocessed historical industrial big data is processed to obtain historical sample industrial big data.

[0144] In this embodiment, preprocessing includes data cleaning, integration, transformation, discretization, and reduction.

[0145] Historical industrial big data refers to at least some historical data from the target industrial scenario, and the amount of historical industrial big data is relatively large.

[0146] Historical industrial big data can include: operating condition variables, control system output data under the target industrial scenario, process object variables, and control system variables under the target industrial scenario.

[0147] Among them, the operating condition variables are used to characterize the operating condition data of the factory, such as field sensor, instrument data and equipment operating status data.

[0148] The output data of the control system in the target industrial scenario is used to characterize the information generated by the control system, such as operator operation instructions, alarm information, and intermediate calculation variable results.

[0149] Process object variables are used to characterize the parameters of the process object itself in the plant, such as the VF curve of the frequency converter, the pipe diameter, the pipe material, and the power of the electric heater.

[0150] Control system variables in the target industrial scenario are used to characterize the parameters of the control system itself, such as controller manufacturer, controller CPU type, number of cores, clock speed, controller storage space, controller timer jitter, control task cycle, I / O bus type, number of I / O channels, and I / O channel type.

[0151] Understandably, operating condition variables and control system output data under target industrial scenarios are generally stored in the factory's SCADA system database or the factory's information system database, or can be collected from the control system (PLC / DCS). The above data itself has a timestamp.

[0152] For process object variables, they need to be obtained from the factory design documents. In actual execution, after the factory design phase is completed, the design institute will submit the corresponding design documents to the factory for storage.

[0153] For control system variables in the target industrial scenario, they are generally found in the control system configuration engineering file and the control system technical specification file.

[0154] Based on the differences between operating condition variables, control system output data under the target industrial scenario, process object variables, and control system variables under the target industrial scenario, the preprocessed historical industrial big data can be further divided into multiple categories.

[0155] Then, based on the processing methods corresponding to their respective categories, the preprocessed historical industrial big data is processed to obtain historical sample industrial big data.

[0156] In some embodiments, processing the preprocessed historical industrial big data based on the processing methods corresponding to their respective categories to obtain historical sample industrial big data may include:

[0157] In the case of category 1, the preprocessed historical industrial big data is identified as historical sample industrial big data;

[0158] In the case of category 2, based on the database time sequence arrangement method and the time information corresponding to the preprocessed historical industrial big data, the preprocessed historical industrial big data is arranged and labeled to obtain historical sample industrial big data.

[0159] In the case of category three, the historical industrial big data is preprocessed based on the sub-category of the control system under the target industrial scenario, and the classified historical industrial big data corresponding to each sub-category is obtained respectively; based on the database time sequence arrangement method and the time information corresponding to the classified historical industrial big data, the classified historical industrial big data is arranged and labeled to obtain historical sample industrial big data.

[0160] In this embodiment, the time information is the timestamp corresponding to each data item.

[0161] The first category includes operating condition variables and control system output data under the target industrial scenario.

[0162] The second category includes process object variables.

[0163] The third category includes control system variables in the target industrial scenario.

[0164] For the first type of data, which already has a timestamp, there is no need for secondary labeling.

[0165] For the second type of data, it is necessary to follow the database arrangement and add timestamps for data entry and labeling.

[0166] For the third type of data, the control system configuration engineering file data needs to be classified and split, arranged in the database manner, and timestamps added for data entry and annotation.

[0167] Based on the above processing methods, historical industrial big data can be obtained.

[0168] like Figure 3 As shown, in some embodiments, after processing real-time industrial big data in the target industrial scenario using the target algorithm prediction model to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario output by the target algorithm prediction model, the method may further include:

[0169] The target independent variable and the target control algorithm are input into the simulation system under the target industrial scenario, and the values ​​of the target control variable output by the simulation system are obtained.

[0170] The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system.

[0171] In this embodiment, the target control variable is taken as the simulation value output by the simulation system.

[0172] After predicting the target independent variables and the target control algorithm corresponding to the target independent variables through the target algorithm prediction model, the target control algorithm can be input into the simulation system for simulation testing, so as to optimize the target independent variables and the target control algorithm based on the simulation results.

[0173] The target control variable is set to the simulated value output by the simulation system.

[0174] Understandably, optimizing variables based on simulation values ​​can be repeated multiple times.

[0175] In some embodiments, optimizing the target independent variable and the target control algorithm based on the values ​​of the target control variable output by the simulation system may include:

[0176] The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system.

[0177] The optimized target independent variable and the optimized target control algorithm are input into the simulation system, and the values ​​of the target control variable output by the simulation system are obtained again.

[0178] Repeat the above steps of optimizing the target independent variable and the target control algorithm based on the value of the target control variable output by the simulation system until the value of the target control variable output by the simulation system meets the target control accuracy.

[0179] In this embodiment, the simulation system is the simulation system corresponding to the control system in the target industrial scenario.

[0180] Target control precision can be customized by the user.

[0181] The target independent variable and target control algorithm are optimized based on the value of the target control variable to obtain the optimized target independent variable and target control algorithm. Then, the optimized target independent variable and optimized target control algorithm are input into the simulation system again to obtain new simulation values. Based on the new simulation values, the previously optimized target independent variable and target control algorithm are optimized.

[0182] Repeat the above operation until the value of the target control variable output by the simulation system in the last time meets the target control accuracy, thereby achieving the simulation effect required by the user.

[0183] According to the model training-based industrial control method provided in this application, after automatically acquiring the target independent variables and target control algorithms corresponding to the target control variables based on real-time industrial big data, simulation tests are further conducted based on the target independent variables and target control algorithms to optimize the target independent variables and target control algorithms based on the simulation results, so as to achieve the best results. This can further improve the accuracy and precision of the acquired target independent variables and target control algorithms, thereby improving the subsequent control effect.

[0184] Continue to refer to Figure 2 In some embodiments, after step 120, the method may further include:

[0185] Based on the values ​​of the target control variables, the target algorithm prediction model is used again to process the updated real-time industrial big data to optimize the target independent variables and the target control algorithm respectively.

[0186] In this embodiment, the value of the target control variable is the actual value obtained by performing control calculations through the deployed target control algorithm during the application process.

[0187] By using actual values ​​as new historical data to update historical sample industrial big data, and then updating the target algorithm prediction model based on the updated historical sample industrial big data, the target independent variables and target control algorithm output by the target algorithm prediction model are updated, thereby achieving further optimization of the target independent variables and target control algorithm.

[0188] For example, during operation, the data from the actual algorithm execution is fed back into the target algorithm prediction model for further training to obtain the optimal solution for the control algorithm and the target independent variable, and then the control algorithm package in the control system is updated synchronously.

[0189] According to the model training-based industrial control method provided in this application, the target independent variable and target control algorithm corresponding to the target control variable are automatically obtained based on real-time industrial big data. The target independent variable and target control algorithm are then optimized based on the actual values ​​obtained after controlling the target controller according to the target control algorithm. This optimizes the target independent variable and target control algorithm, thereby achieving the best performance. In addition to effectively eliminating the influence of human subjective factors, the method can also automatically adjust the target independent variable and target control algorithm based on the actual control situation, thereby improving control efficiency and control effect.

[0190] The following explanation uses the process control of fused magnesia smelting as an example.

[0191] The traditional fused magnesia smelting process uses an HMI+PLC control system.

[0192] The main function of the HMI is to input the current setpoint. The PLC mainly uses a PID process to achieve current tracking control, so that the output three-phase electrode current tracks the setpoint input by the HMI, and adjusts the output value u(k) according to the deviation e(t) between the actual current value y(k) and the setpoint r.

[0193] The specific control algorithm is as follows:

[0194]

[0195] Where u(k) is the adjustment value used to adjust the controlled object y(k), that is, the value of the target control variable; T is the target duration, and t is the time.

[0196] The initial algorithm prediction model provided in the embodiments of this application is used to process the collected historical industrial big data to obtain historical sample industrial big data, and to perform training calculations based on the historical sample industrial big data.

[0197] Historical industrial big data can include historical and operational data collected from fused magnesia plants, including but not limited to: target value r of energy consumption per ton. * upper limit value r min The lower limit value r max The current setting value r(k) for each operation, the melting voltage B1, the electrode diameter B2, and the lower limit current value y. min Current upper limit value y max The data include the actual current value y(k), the control cycle T of the control system, and the instantaneous power P(k) of the electric furnace.

[0198] like Figure 3 As shown, after processing or labeling the above-mentioned historical industrial big data to obtain historical sample industrial big data, the historical sample industrial big data is input into the initial algorithm prediction model for training, so that the trained target algorithm prediction model can output other variables related to the current value and the control algorithm between the current value and other variables.

[0199] For example, multiple target independent variables can be obtained that are associated with the target control variable: the target value r of energy consumption per ton. * upper limit value r min The lower limit value r max Melting voltage B1, electrode diameter B2, lower limit current y min Current upper limit value y max And the actual value y(k) of the controlled object, and the control algorithm between them.

[0200] After obtaining the target independent variable and the control algorithm, the simulation system calculates the optimized setpoint y(nk) based on the target independent variable and the control algorithm, and then calculates the optimized setpoint y(nk) based on the current feedback value y.i (+1), current value y(k) and deviation e i The setpoint compensation value required for PID control is automatically calculated, and this setpoint compensation value is added to the optimized setpoint y(nk) to obtain the final setpoint y. sp (), and finally set the final value y sp The optimized control algorithm is deployed to the target controller to obtain the values ​​of the target control variables in order to control the target controlled object.

[0201] The target control algorithm is obtained by training the target algorithm prediction model provided in the embodiments of this application. The target independent variable is optimized based on the target control algorithm. The optimal set value can be automatically obtained according to the on-site smelting environment conditions, replacing the original setting value input based on human experience, eliminating the influence of human subjective factors, and avoiding the problem that the operator cannot identify abnormal working conditions and adjust the current set value in a timely and accurate manner.

[0202] In practical applications, it can improve the stability of current control in fused magnesia furnaces, reduce abnormal operating conditions, and lower power consumption during the fused magnesia smelting process.

[0203] Of course, in other embodiments, the industrial control method based on model training of this application can also be applied to other industrial scenarios, such as for predictive analysis of the economic value of factory operation. Based on the comprehensive analysis of data such as output, energy consumption and raw material consumption, a mathematical model (i.e., control algorithm) between output, energy consumption and raw material consumption is derived, thereby giving the optimal production capacity plan and giving control strategies, such as adjusting production strategies and control strategies by combining peak and off-peak electricity prices and raw material prices in the market.

[0204] The industrial control method based on model training provided in this application can be executed by an industrial control device based on model training. This application uses an industrial control device based on model training executing the industrial control method based on model training as an example to illustrate the industrial control device based on model training provided in this application.

[0205] This application also provides an industrial control device based on model training.

[0206] like Figure 5 As shown, the industrial control device based on model training may include: a first processing module 510 and a second processing module 520.

[0207] The first processing module 510 is used to process real-time industrial big data in the target industrial scenario using the target algorithm prediction model to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario, output by the target algorithm prediction model. The target algorithm prediction model is obtained by training on historical sample industrial big data. Both the historical sample industrial big data and the real-time industrial big data include the values ​​of multiple independent variables and control variables. The historical sample industrial big data is used to represent historical data, and the real-time industrial big data is used to represent current data. The target independent variable is the independent variable corresponding to the target control variable determined from the multiple independent variables in the real-time industrial big data. The target control algorithm is the control algorithm used to represent the relationship between the target independent variable and the target control variable.

[0208] The second processing module 520 is used to input the values ​​of the target independent variables in real-time industrial big data to the target controller with the target control algorithm, so as to obtain the values ​​of the target control variables and control the target controlled object.

[0209] According to the model-trained industrial control device provided in the embodiments of this application, the system automatically determines the target independent variable associated with the target control variable from multiple independent variables based on the correlation between any two elements in the input industrial big data and multiple independent variables and control variables. On this basis, the target control algorithm is further obtained based on the automatically obtained target independent variable and deployed to the control system for corresponding industrial control. Users do not need to manually obtain the target independent variable, which significantly reduces manpower and time costs, has a low operation threshold, and is applicable to any industry, thereby improving the efficiency of obtaining the control algorithm and having high universality and versatility.

[0210] In some embodiments, the model-trained industrial control device may further include: a third processing module, configured to:

[0211] After processing real-time industrial big data in the target industrial scenario using the target algorithm prediction model to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario, the target independent variables and target control algorithm are input into the simulation system in the target industrial scenario to obtain the values ​​of the target control variables output by the simulation system.

[0212] The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system.

[0213] In some embodiments, the third processing module may also be used for:

[0214] The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system.

[0215] The optimized target independent variable and the optimized target control algorithm are input into the simulation system, and the values ​​of the target control variable output by the simulation system are obtained again.

[0216] Repeat the above steps of optimizing the target independent variable and the target control algorithm based on the value of the target control variable output by the simulation system until the value of the target control variable output by the simulation system meets the target control accuracy.

[0217] In some embodiments, the device may further include a fourth processing module, which is used to input the value of the target independent variable in real-time industrial big data to the target controller with the target control algorithm deployed, and after obtaining the value of the target control variable, to use the target algorithm prediction model to process the updated real-time industrial big data again based on the value of the target control variable, so as to optimize the target independent variable and the target control algorithm respectively.

[0218] In some embodiments, the device may further include a fifth processing module for:

[0219] Acquire historical industrial big data samples from the target industrial scenario;

[0220] By learning from historical industrial big data through an initial algorithm prediction model, the correlation between any two elements among multiple independent and control variables in the historical industrial big data can be obtained.

[0221] Based on the correlation, the target independent variable corresponding to the target control variable is determined from multiple independent variables in historical industrial big data samples, as well as the target control algorithm corresponding to the target independent variable corresponding to the historical industrial big data samples.

[0222] In some embodiments, the fifth processing module may also be used for:

[0223] The historical industrial big data of the target industrial scenario is preprocessed to obtain the preprocessed historical industrial big data; the historical industrial big data is at least part of the historical data of the target industrial scenario.

[0224] Based on the processing methods corresponding to the respective categories of preprocessed historical industrial big data, the preprocessed historical industrial big data is processed to obtain historical sample industrial big data.

[0225] In some embodiments, the fifth processing module may also be used for:

[0226] In the case of category 1, the preprocessed historical industrial big data is identified as historical sample industrial big data;

[0227] In the case of category 2, based on the database time sequence arrangement method and the time information corresponding to the preprocessed historical industrial big data, the preprocessed historical industrial big data is arranged and labeled to obtain historical sample industrial big data.

[0228] In the case of category three, the historical industrial big data is preprocessed based on the sub-category of the control system under the target industrial scenario, and the classified historical industrial big data corresponding to each sub-category is obtained respectively; based on the database time sequence arrangement method and the time information corresponding to the classified historical industrial big data, the classified historical industrial big data is arranged and labeled to obtain historical sample industrial big data.

[0229] In some embodiments, the initial algorithm prediction model includes an encoder module and a decoder module connected in sequence, and a fifth processing module, which can also be used for:

[0230] Historical sample industrial big data is input into the encoder module, and the intermediate representation output by the encoder module is obtained. The intermediate representation is obtained by the encoder module after learning the dependency relationship between data at different positions in the data sequence of historical sample industrial big data and mapping the historical sample industrial big data.

[0231] The intermediate representation is input into the decoder module to obtain the target independent variables and target control algorithm corresponding to the historical sample industrial big data output by the decoder module.

[0232] This application also provides a control system.

[0233] like Figure 4 As shown, the control system includes a big data training model system and an edge real-time control system.

[0234] The big data training model system is connected to the edge real-time control system.

[0235] The big data training model system is used to train on historical industrial big data samples to obtain a target algorithm prediction model that can output the target control algorithm and target independent variables.

[0236] The big data training model system and the edge real-time control system are combined to execute the model-trained industrial control method described in any of the above embodiments.

[0237] An edge real-time control system can include a control system development environment and a control system runtime environment.

[0238] For example, a big data training model system can send the generated target algorithm prediction model to the control system development environment. The control system development environment then completes the overall control program coding based on the target control algorithm, configures the input and output of the control task, maps and associates it with physical I / O, and configures the control task running cycle. Finally, it is compiled and downloaded to the control system runtime environment, which is responsible for the signal interface with field sensors and actuators, thereby completing the real-time control of the production site.

[0239] During operation, the data from the actual algorithm execution can be fed back to the target algorithm prediction model for further training, to obtain the optimal solution for the control algorithm and the target independent variable, and then the control algorithm package of the management edge control system development environment can be updated synchronously.

[0240] Through numerous experiments by the inventors, the edge real-time control system of this application has a jitter time level as low as 5µs, which can adapt to high-precision control scenarios such as industrial robot control and CNC machine tool control.

[0241] In some embodiments, the edge real-time control system may also have a cloud-edge collaborative algorithm management interface and a data communication interface. The algorithm management interface is used to obtain the target control algorithm generated by the big data training model system, and the data communication interface is used to synchronize the real-time data of the edge real-time control system to the big data training model system for training.

[0242] like Figure 2 As shown, in some embodiments, a big data training model system may include, in sequence, a data acquisition module, a data processing module, a model training module, an encoding module, and a control simulation module.

[0243] In this embodiment, the control simulation module is electrically connected to both the model training module and the edge real-time control system.

[0244] The data acquisition module is used to acquire historical industrial big data.

[0245] like Figure 4 As shown, the data processing module is used to preprocess and label historical sample industrial big data to obtain historical sample industrial big data.

[0246] The model training module is used to train an initial algorithm prediction model based on historical sample industrial big data, and to obtain the target algorithm prediction model.

[0247] The control simulation module is used to simulate and test the target control algorithm based on the output of the target algorithm prediction model, and to optimize the target control algorithm and target independent variables based on the simulation results.

[0248] The specific execution steps of each module have been described in the above embodiments and will not be repeated here.

[0249] Continue to refer to Figure 2 In some embodiments, the edge real-time control system may include a first module and an execution module.

[0250] In this embodiment, the first module is electrically connected to the control simulation module;

[0251] The first module may include a download module and a deployment module.

[0252] For example, the first module can be the development environment for the aforementioned control system.

[0253] The download module is used to download the target control algorithm and target independent variables obtained by the control simulation module; the deployment module is used to deploy the target control algorithm and target independent variables to the execution module.

[0254] The execution module is electrically connected to both the first module and the data acquisition module.

[0255] The execution module is used to perform control calculations based on the values ​​of target independent variables in the target industrial scenario through the target control algorithm, so as to obtain the values ​​of target control variables and control the target controlled object.

[0256] The execution module can be the aforementioned edge real-time control system.

[0257] In addition, the execution module can also send the actual output value to the data acquisition module to update historical industrial big data, thereby updating the target algorithm prediction model.

[0258] Taking a PID control system as an example, the execution module may include: a current setting control module, a self-optimizing correction module, and a loop control module with output compensation.

[0259] Among them, the current setting control module is used to calculate the optimized setpoint y(nk) based on the target independent variable output by the algorithm prediction model and the target control algorithm.

[0260] The self-optimizing correction module is used to adjust the current feedback value y. i (+1), current value y(k) and deviation e i (), calculate the set compensation value.

[0261] The loop control module includes: a PID controller u i1 (), rate of change compensation of higher-order nonlinear terms u i2 () and the higher-order nonlinear term compensator u from the previous time step i3 (), to jointly achieve current control.

[0262] According to the control system provided in the embodiments of this application, a target algorithm prediction model is obtained by training a big data training model system. Then, an edge real-time control system is deployed based on the target control algorithm output by the target algorithm prediction model for industrial control. This eliminates the need for users to manually obtain target variables, effectively eliminating the influence of human subjective factors and effectively improving the accuracy and precision of the obtained target control algorithm. In addition, the target variables and target control algorithm can be automatically adjusted based on the actual control situation, thereby improving control efficiency and control effect.

[0263] In some embodiments, such as Figure 6 As shown, this application embodiment also provides an electronic device 600, including a processor 601, a memory 602, and a computer program stored in the memory 602 and executable on the processor 601. When the program is executed by the processor 601, it implements the various processes of the above-described industrial control method embodiment based on model training, or industrial control algorithm determination method, or algorithm prediction model training method, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0264] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0265] This application also provides a non-transitory computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described industrial control method embodiment based on model training and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0266] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0267] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described embodiment of the industrial control method based on model training.

[0268] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0269] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described industrial control method embodiment based on model training, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0270] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0271] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0272] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0273] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0274] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0275] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. An industrial control method based on model training, characterized in that, include: A target algorithm prediction model is used to process real-time industrial big data in a target industrial scenario to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario, as output by the target algorithm prediction model. The target algorithm prediction model is obtained by training on historical sample industrial big data. Both the historical sample industrial big data and the real-time industrial big data include values ​​for multiple independent variables and control variables. The historical sample industrial big data represents historical data, and the real-time industrial big data represents current data. The target independent variable is the independent variable corresponding to the target control variable determined from the multiple independent variables in the real-time industrial big data. The target control algorithm is a control algorithm used to represent the relationship between the target independent variable and the target control variable. The values ​​of the target independent variables in the real-time industrial big data are input to the target controller equipped with the target control algorithm to obtain the values ​​of the target control variables, so as to control the target controlled object; The target algorithm prediction model is trained in the following manner: Obtain the historical sample industrial big data under the target industrial scenario; The historical sample industrial big data is learned by using an initial algorithm prediction model to obtain the correlation between any two elements among multiple independent variables and control variables in the historical sample industrial big data. Based on the aforementioned correlation, a target independent variable corresponding to the target control variable is determined from multiple independent variables of the historical sample industrial big data, and a target control algorithm corresponding to the target independent variable corresponding to the historical sample industrial big data is also determined.

2. The industrial control method based on model training according to claim 1, characterized in that, After processing real-time industrial big data in the target industrial scenario using the target algorithm prediction model to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario, as output by the target algorithm prediction model, the method further includes: The target independent variable and the target control algorithm are input into the simulation system under the target industrial scenario, and the values ​​of the target control variable output by the simulation system are obtained; The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system.

3. The industrial control method based on model training according to claim 2, characterized in that, The optimization of the target independent variable and the target control algorithm based on the values ​​of the target control variable output by the simulation system includes: The target independent variable and the target control algorithm are optimized based on the values ​​of the target control variable output by the simulation system. The optimized target independent variable and the optimized target control algorithm are input into the simulation system, and the value of the target control variable output by the simulation system is obtained again. Repeat the steps described above, which optimize the target independent variable and the target control algorithm based on the value of the target control variable output by the simulation system, until the value of the target control variable output by the simulation system meets the target control accuracy.

4. The industrial control method based on model training according to any one of claims 1-3, characterized in that, After inputting the values ​​of the target independent variables from the real-time industrial big data into the target controller equipped with the target control algorithm to obtain the values ​​of the target control variables, the method further includes: Based on the value of the target control variable, the target algorithm prediction model is used again to process the updated real-time industrial big data to optimize the target independent variable and the target control algorithm respectively.

5. The industrial control method based on model training according to claim 1, characterized in that, The acquisition of historical sample industrial big data in the target industrial scenario includes: The historical industrial big data acquired in the target industrial scenario is preprocessed to obtain preprocessed historical industrial big data; the historical industrial big data is at least a portion of the historical data in the target industrial scenario. The preprocessed historical industrial big data is processed according to the processing methods corresponding to each category of the preprocessed historical industrial big data to obtain the historical sample industrial big data.

6. The industrial control method based on model training according to claim 5, characterized in that, The process of processing the preprocessed historical industrial big data according to the respective categories of the preprocessed historical industrial big data to obtain the historical sample industrial big data includes: In the case where the category is the first category, the preprocessed historical industrial big data is identified as the historical sample industrial big data; In the case where the category is the second type, based on the database time sequence arrangement method and the time information corresponding to the preprocessed historical industrial big data, the preprocessed historical industrial big data is arranged and labeled to obtain the historical sample industrial big data. In the case of the third category, the preprocessed historical industrial big data is split based on the sub-category of the control system under the target industrial scenario, and the classified historical industrial big data corresponding to each sub-category is obtained respectively; based on the time sequence arrangement of the database and the time information corresponding to the classified historical industrial big data, the classified historical industrial big data is arranged and labeled to obtain the historical sample industrial big data.

7. The industrial control method based on model training according to claim 1, characterized in that, The historical sample industrial big data includes at least two of the following: operating condition variables, control system output data under the target industrial scenario, process object variables, and control system variables under the target industrial scenario.

8. The industrial control method based on model training according to claim 1, characterized in that, The initial algorithm prediction model includes an encoder module and a decoder module connected in sequence. The learning process using the initial algorithm prediction model on the historical sample industrial big data includes: The historical sample industrial big data is input into the encoder module to obtain the intermediate representation output by the encoder module. The intermediate representation is obtained by the encoder module after learning the dependency relationship between data at different positions in the data sequence of the historical sample industrial big data and mapping the historical sample industrial big data. The intermediate representation is input to the decoder module to obtain the target independent variable and the target control algorithm output by the decoder module that correspond to the historical sample industrial big data.

9. The industrial control method based on model training according to claim 8, characterized in that, At least one of the encoder module and the decoder module includes: The system comprises a multi-layer self-attention sub-layer and a feedforward neural network layer, wherein the output of the multi-layer self-attention sub-layer is connected to the input of the feedforward neural network layer; wherein... The self-attention sublayer is used to process the data at the target location based on the data at other locations in the data sequence besides the target location, and to obtain the correlation between the data at the target location and the data at other locations.

10. An industrial control device based on model training, characterized in that, include: The first processing module is used to process real-time industrial big data in a target industrial scenario using a target algorithm prediction model, and to obtain the target control algorithm and target independent variables for controlling the target controlled object in the target scenario, output by the target algorithm prediction model. The target algorithm prediction model is obtained by training on historical sample industrial big data. Both the historical sample industrial big data and the real-time industrial big data include values ​​for multiple independent variables and control variables. The historical sample industrial big data is used to represent historical data, and the real-time industrial big data is used to represent current data. The target independent variable is the independent variable corresponding to the target control variable determined from among the multiple independent variables in the real-time industrial big data. The target control algorithm is a control algorithm used to represent the relationship between the target independent variable and the target control variable. The second processing module is used to input the value of the target independent variable in the real-time industrial big data to the target controller deployed with the target control algorithm, so as to obtain the value of the target control variable and control the target controlled object. The industrial control device is further used to train the target algorithm prediction model, which is trained by: acquiring historical sample industrial big data under the target industrial scenario; learning the historical sample industrial big data through the initial algorithm prediction model to obtain the correlation between any two elements among multiple independent variables and control variables in the historical sample industrial big data; and, based on the correlation, determining the target independent variable corresponding to the target control variable and the target control algorithm corresponding to the target independent variable in the historical sample industrial big data from the multiple independent variables of the historical sample industrial big data.

11. A control system based on the model-training-based industrial control method as described in any one of claims 1-9, characterized in that, include: Big data training model system; An edge real-time control system, wherein the big data training model system is communicatively connected to the edge real-time control system.

Images