Control method, device, system and equipment of production system and storage medium

Abstract

The embodiment of the present specification provides a kind of control method, device, system, equipment and storage medium of production system, the method is applied to predictive control system, the predictive control system includes linear model and nonlinear model, the method comprises: at each sampling time, the following steps are iteratively executed until the target function value that meets the preset condition is solved: determine input sequence;The first output sequence that the linear model is predicted to the input sequence determined this time and the second output sequence that the nonlinear model is predicted to the input sequence determined this time are fused, and then substituted into the preset target function to obtain the target function value by calculation;Using the input sequence corresponding to the target function value that meets the preset condition solved, control instruction is generated to control the production system.

Description

Technical Field

[0001] This disclosure relates to the field of industrial control technology, and in particular to control methods, devices, systems, equipment and storage media for production systems. Background Technology

[0002] Advanced control technology is the core technology for realizing the upgrade of intelligent manufacturing. By establishing models, it can effectively handle systems with large time delays and process constraints, and realize fully automatic control of the production process. It has been successfully applied in industrial production industries such as cement, steel or oil refining.

[0003] Traditional model predictive control schemes use a single model for trend prediction and control action calculation. Specifically, production equipment has input and output parameters. Traditional schemes construct a model representing the linear relationship between these parameters and train it using historical production data. Online control action calculation is essentially solving an optimization problem. By continuously adjusting different future input sequences, the corresponding future output sequence is predicted based on the trained linear relationship until the optimal target input sequence is found.

[0004] In the field of industrial production technology, equipment control needs to meet real-time requirements. Therefore, traditional solutions use linear models for rapid optimization. However, actual production processes are very complex and may be subject to external disturbances. The relationship between input and output parameters is not a simple linear one; there is often a nonlinear dynamic relationship between them. Although traditional linear models are simple to implement and can quickly perform optimization, they are insufficient to accurately describe the relationship between input and output parameters. Therefore, the model cannot accurately predict the output parameters, resulting in poor actual control performance. Summary of the Invention

[0005] To overcome the problems existing in the related technologies, this disclosure provides a control method, apparatus, computer equipment and storage medium for a production system.

[0006] According to a first aspect of the embodiments of this specification, a control method for a production system is provided, the method being applied to a predictive control system, the predictive control system including a linear model and a nonlinear model;

[0007] The linear model and the nonlinear model are respectively used to predict the output sequence corresponding to the input sequence;

[0008] The method includes:

[0009] At each sampling time, the following steps are iteratively executed until the objective function value that satisfies the preset conditions is obtained: Determine the input sequence; merge the first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence, and substitute them into the preset objective function to calculate the objective function value;

[0010] Using the input sequence corresponding to the objective function value that satisfies the preset conditions, control commands are generated to control the production system.

[0011] According to a second aspect of the embodiments of this specification, a control device for a production system is provided, the device being applied to a predictive control system, the predictive control system including a linear model and a nonlinear model;

[0012] The linear model and the nonlinear model are respectively used to predict the output sequence corresponding to the input sequence;

[0013] The device includes:

[0014] The execution module is used to iteratively execute the following steps at each sampling time until the objective function value that satisfies the preset conditions is obtained: determine the input sequence; fuse the first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence, and substitute them into the preset objective function to calculate the objective function value;

[0015] The control module is used to generate control commands to control the production system using the input sequence corresponding to the objective function values ​​that satisfy preset conditions.

[0016] According to a third aspect of the embodiments of this specification, an industrial control system is provided, the industrial control system including a predictive control system and a production system, the predictive control system being connected to the production system, the predictive control system including:

[0017] A modeling subsystem is used to train linear and nonlinear models using historical production data from the production system.

[0018] The control subsystem is used to acquire real-time operating data of the production system. At each sampling time, iteratively executes the following steps until a target function value that satisfies preset conditions is obtained: determining the current input sequence; fusing the first output sequence predicted by the linear model for the current input sequence and the second output sequence predicted by the nonlinear model for the current input sequence, and substituting them into the preset target function to calculate the target function value; using the current input sequence corresponding to the target function value that satisfies the preset conditions, generating control commands to control the production system.

[0019] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method embodiments described in the first aspect above.

[0020] According to a fifth aspect of the embodiments of this specification, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method embodiments described in the first aspect above.

[0021] The technical solutions provided in the embodiments of this specification may include the following beneficial effects:

[0022] In the embodiments of this specification, the predictive control system includes a linear model and a nonlinear model. The nonlinear model is not directly used for optimization; its nonlinear characteristics enable accurate prediction of the output sequence. The structure of the nonlinear model is unrestricted, and compared to conventional nonlinear model predictive control algorithms, it has strong versatility. The linear model is used for optimization. During the solution process, the output sequence incorporates the predictions of the nonlinear model, thus improving the prediction accuracy of the output sequence. Furthermore, due to the linear characteristics of the linear model, the model parameters are interpretable and easy to tune, thus effectively avoiding possible control action anomalies or suboptimal results based on black-box nonlinear model optimization. It exhibits strong robustness across the entire operating range. At the same time, the algorithm has high computational efficiency and can meet the high-frequency computational requirements of industrial production systems.

[0023] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this specification and, together with the description, serve to explain the principles of this specification.

[0025] Figure 1 This specification is a control diagram illustrating a related art based on an exemplary embodiment.

[0026] Figure 2A and Figure 2B These are flowcharts illustrating a control method for a production system according to an exemplary embodiment of this specification.

[0027] Figure 2C This is a schematic diagram illustrating the prediction of a model according to an exemplary embodiment of this specification.

[0028] Figure 2D This is a schematic diagram illustrating an optimization solution according to an exemplary embodiment of this specification.

[0029] Figure 2E This is a schematic diagram illustrating a predictive control system according to an exemplary embodiment of this specification.

[0030] Figure 2F This is a flowchart illustrating another control method for a production system according to an exemplary embodiment of this specification.

[0031] Figure 3 This is a block diagram of a computer device containing a control device for a production system, as illustrated in this specification according to an exemplary embodiment.

[0032] Figure 4 This is a block diagram illustrating a control device for a production system according to an exemplary embodiment. Detailed Implementation

[0033] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this specification as detailed in the appended claims.

[0034] The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of this specification. The singular forms “a,” “the,” and “the” as used in this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0035] It should be understood that although the terms first, second, third, etc., may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this specification, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0036] Complex industrial manufacturing processes (such as cement, steel, or petrochemicals) often exhibit strong nonlinearity, meaning that the response characteristics of the production process change nonlinearly under different operating conditions. For such complex systems, traditional controllers have poor adaptability, and the actual control effect fluctuates greatly, failing to meet actual production requirements. They often require operators to constantly monitor key variables and make manual adjustments, resulting in high workload for front-line production personnel and the inability of the system to maintain optimal production conditions, leading to loss of efficiency.

[0037] like Figure 1 As shown, this is a predictive control scheme in related technologies. Traditional model predictive control algorithms use the same model for trend prediction and control action calculation. Moreover, the model is a linear predictive model, which cannot accurately predict nonlinear process variables. Therefore, the control effect obtained cannot meet the actual production requirements.

[0038] In traditional approaches, a control model is trained using historical production data. This control model is then connected to the production system, and it needs to perform the following prediction and optimization process:

[0039] After acquiring production data from the production system, a future input sequence is adjusted based on constraints. The corresponding future output sequence is predicted, and the corresponding target value (objective function value) is calculated. This adjustment process is repeated until the optimal target future input sequence is found. Using the found optimal input sequence, control commands are generated and sent to the production system.

[0040] Because the dynamic relationship between input and output often exists nonlinearly in actual production processes, linear models cannot accurately predict it. Therefore, the improvement direction of related technologies is nonlinear model predictive control (MMC). Its main improvement is to establish a nonlinear model that accurately describes the system characteristics and then optimize based on this model to obtain the optimal input sequence that satisfies the control objective. However, the core of this type of algorithm still relies on the same model for trend prediction and control action calculation. This approach has the following problems in practical applications:

[0041] 1. AI (Artificial Intelligence) based nonlinear models can only guarantee the prediction accuracy of the model output variables, but the model's task also includes the optimal calculation of control actions. The derivative information in the nonlinear model often differs greatly from the actual process. Therefore, during the optimization process, due to the deviation of derivative information and the uncertainty in causal relationships, the controller solution is very likely to enter the wrong optimization direction, resulting in unsuitable control actions and a decline in control performance.

[0042] 2. Interpretable nonlinear prediction models are difficult to develop and require a lot of effort to build and train. Furthermore, they need to be modeled and developed in conjunction with process mechanisms, which places high demands on developers and is difficult to maintain. For complex production systems with unclear mechanisms, it is impossible to obtain accurate mechanism models.

[0043] 3. The computational efficiency of optimization control solutions based on traditional nonlinear models is low. Since nonlinear models often involve a large number of variables and complex nonlinear relationships, optimization solutions are time-consuming and have poor robustness, easily getting trapped in local saddle points or failing to solve. Therefore, they cannot meet the real-time (second-level) requirements of industrial production.

[0044] For the reasons mentioned above, nonlinear models have not been widely used in industrial production, and the automatic control of such complex nonlinear systems is a major challenge and pain point in the industry.

[0045] Based on this, embodiments of this specification provide a control method for an industrial production system. In this method, the predictive control system includes a linear model and a nonlinear model. The nonlinear model is not directly used for optimization; its nonlinear characteristics enable accurate prediction of the output sequence. The structure of the nonlinear model is unrestricted, and compared to conventional nonlinear model predictive control algorithms, it has strong versatility. The linear model is used for optimization, incorporating the output sequence predicted by the nonlinear model during the solution process, thus improving the prediction accuracy of the output sequence. Furthermore, due to the linear characteristics of the linear model, the model parameters are interpretable and easy to tune, effectively avoiding abnormal control actions obtained from black-box linear models, and exhibiting strong robustness across the entire operating range. Simultaneously, this algorithm has high computational efficiency, meeting the high-frequency computational requirements of industrial production systems.

[0046] like Figure 2A The diagram shown is a flowchart illustrating a control method for an industrial production system according to an exemplary embodiment of this specification. Figure 2B As shown, this method can be applied to predictive control systems, which include linear and nonlinear models.

[0047] The linear model and the nonlinear model are used to predict the output sequence corresponding to the input sequence, respectively. The method may include the following steps:

[0048] In step 202, at each sampling time, the following steps are performed: Iteratively perform the following steps until the objective function value that satisfies the preset conditions is obtained: Determine the input sequence; After fusing the first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence, substitute them into the preset objective function to calculate the objective function value.

[0049] In step 204, the input sequence corresponding to the objective function value that satisfies the preset conditions is used to generate control commands to control the production system.

[0050] In practical applications, a production system can be a production system in any industrial field, such as a cement production system, a solid waste power generation system, or a steel production system, etc.

[0051] In this embodiment, the production system corresponds to input parameters and output parameters. Input parameters of the production system refer to parameters input to one or more controlled devices within the system; the type of input parameter is greater than or equal to 1. Input parameters may include manipulated variables (MV), which are variables that can be manually adjusted and affect the system output. They are used to control the controlled devices in the production system and are controllable variables. Optionally, input parameters may also include disturbance variables, which are variables that cannot be manually adjusted but still affect the system output. They are not controllable variables, but they influence the output of the production system.

[0052] The output parameters of the production system refer to the parameters output after the aforementioned equipment is controlled; they can also be called controlled variables (CV). Appropriate input and output parameters can be flexibly selected based on the actual production system and process requirements. This embodiment does not impose any limitations on this.

[0053] Taking a cement production system as an example, the controlled equipment in the system is the decomposition furnace, which requires automatic control. The outlet temperature of the decomposition furnace reflects the pre-decomposition status of the material within it. Temperature stability is crucial for furnace operation, and it is influenced by various factors such as raw meal feed rate and composition, as well as numerous unpredictable disturbance variables. For instance, in a cement production system, operational variables among the input parameters may include the coal feed rate to the decomposition furnace, while disturbance variables include uncontrollable secondary air temperature, raw meal feed rate, or fluctuations in coal calorific value. The output parameters of the production system include the decomposition furnace outlet temperature.

[0054] Taking an industrial gas production air separation system as an example, the input data of the production system includes the input air flow rate, the opening degree of the compressor and distillation column valves, that is, the operating variables include the air flow rate entering the compressor, the opening degree of the inlet valve, the venting opening degree, the main column return flow valve and the liquid oxygen generator return flow valve, etc.; the output data of the production system includes the product flow rate, purity and pressure, that is, the output parameters include the purity of the outlet oxygen, argon, nitrogen and other products, the product load, the operating temperature, pressure or liquid level, etc.

[0055] In this embodiment, the input sequence includes input parameters corresponding to multiple sampling times, and the output sequence includes output parameters corresponding to multiple sampling times.

[0056] The control scheme in this embodiment includes two models: a linear model and a nonlinear model. The nonlinear model can be used to accurately predict the output sequence, while the linear model can be used for optimization, thereby ensuring the real-time requirements of online control.

[0057] Linear and nonlinear models predict future trends (i.e., future output sequence CV) based on "historical operation variable (MV) and disturbance variable (DV) sequences, as well as future MV sequences". Linear and nonlinear models can use different MV and DV sequences. In the solution process, only the future MV sequence can change and is also the output of the solution problem.

[0058] Nonlinear models can employ arbitrary structures and be trained using algorithms such as supervised learning, unsupervised learning, or deep learning. They can be flexibly configured according to the specific production system being applied. The task of the nonlinear module is to predict the output sequence from the input sequence of the production system. For example... Figure 2C The figure shown is a prediction diagram of the prediction model according to an exemplary embodiment of this specification. In the figure, t represents time, k is the current time, CV represents the sequence of output variables changing over time, and MV represents the sequence of input variables changing over time.

[0059] Here, u, represented by a dashed line after k, refers to the future input sequence over a future period of time; y, represented by a dashed line after k, refers to the future output sequence over a future period of time.

[0060] The task of the predictive model is to adjust based on historical actions and future actions. The predictive model can obtain the predicted sequence of future CVs, providing a basis for the controller to calculate control actions, effectively improving the controller's response speed and control effect.

[0061] A linear model is an interpretable model, meaning the parameters in the model are interpretable. It describes the control relationship from input parameters to output parameters, such as the control relationship from the manipulated variable to the controlled variable. Optionally, it can also describe the control relationship from the "manipulated variable and disturbance variable" to the controlled variable. In practical applications, the control relationship from the manipulated variable to the controlled variable can be user-defined, or the model can be constructed using system identification algorithms. For example, a transfer function can be used to construct the model. A linear model is an interpretable model; for example, a linear model describes y = f(u), where the function f represents the linear relationship between the input parameter y and the output parameter u, and the parameters in the function f are interpretable.

[0062] In this embodiment, step 202 can be executed at each sampling time, and the interval between each sampling time can be a fixed time. The interval can be flexibly configured as needed, and this embodiment does not limit it.

[0063] In this embodiment, the future output sequence predicted by the linear model is called the first output sequence, and the future output sequence predicted by the nonlinear model is called the second output sequence; during online real-time control, the first output sequence and the second output sequence are fused for optimization.

[0064] like Figure 2D The diagram shown is a schematic representation of the optimization solution according to an exemplary embodiment of this specification. At time t = k, the optimal control trajectory for the next N sampling times can be calculated, and the next control output can be executed. This can be flexibly configured as needed, and this embodiment does not impose any limitations on it.

[0065] In practical applications, the objective function of a linear model can be flexibly configured as needed. In some examples, the objective function of a linear model can be: the deviation of the CV from the set value over a future period, plus the change in the MV. The deviation of the CV from the set value and the change in the MV can each be assigned different weights to represent the relative importance of different objectives.

[0066] In practical applications, solving optimization problems requires adjusting different future input sequences. Based on past and future input sequences, this embodiment uses linear and nonlinear models to obtain future output sequences, which is more accurate than the prediction of a purely linear model. The linear model is used for optimization, and the optimal future output sequence is obtained through iteration.

[0067] In this embodiment, the output sequence predicted by the linear model is called the first output sequence, and the sequence predicted by the nonlinear model is called the second output sequence. The first output sequence and the second output sequence have the same sampling time. In this embodiment, the first output sequence and the second output sequence having the same sampling time means that the linear model and the nonlinear model start predicting at the same time and have the same sampling frequency. The first output sequence and the second output sequence include multiple identical sampling times, and the lengths of the two output sequences can be the same or different. For example, the second output sequence can be used to correct the first output sequence. The corrected output sequence is called the target output sequence, and the correction method can be flexibly determined according to actual needs.

[0068] For example, the step of fusing the first output sequence predicted by the linear model for the currently determined input sequence and the second output sequence predicted by the nonlinear model for the currently determined input sequence, and then substituting them into a preset objective function to calculate the objective function value, includes:

[0069] The first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence are fused to determine the target output sequence; wherein, the output parameter value at each target sampling time in the target output sequence is determined based on the output parameter value at the target sampling time in the first output sequence and the output parameter value at the target sampling time in the second output sequence;

[0070] Substitute each output parameter value in the target output sequence into the preset target function to calculate the target function value.

[0071] For example, the first output sequence, the second output sequence, and the target output sequence all include output parameter values ​​at m future sampling times. The output parameter value at the i-th sampling time in the target output sequence is determined based on the output parameter value at the i-th sampling time in the first output sequence and the output parameter value at the i-th sampling time in the second output sequence.

[0072] Solving optimization problems requires adjusting different future input sequences, i.e., generating multiple different future input sequences. In practical applications, this can be done under unconstrained conditions or with constraints, depending on the specific scenario. The constraints can be flexibly configured as needed. In some examples, determining the current input sequence may include:

[0073] The determination of the input sequence includes:

[0074] If the input sequence is determined for the first time, the actual input parameter value at the previous sampling time is used as the input parameter value at the current sampling time and multiple future sampling times to obtain the input sequence determined this time.

[0075] If the input sequence is not determined for the first time, under the constraints of the preset constraints, the input parameter values ​​at each sampling time in the previously determined input sequence are adjusted to obtain the input sequence determined this time.

[0076] The preset constraints include:

[0077] The values ​​of each input parameter in the input sequence satisfy a first preset numerical range; or,

[0078] The difference between each input parameter value in the input sequence and the input parameter value at the previous sampling time satisfies the preset range of change.

[0079] In this embodiment, the step of determining the input sequence is iteratively executed multiple times at a single sampling time. The initially determined input sequence may use the actual input parameter values ​​from the previous sampling time as the input parameter values ​​for the current sampling time and multiple future sampling times. Subsequent determinations of the input sequence are made by adjusting the input parameter values ​​at each sampling time in the previously determined input sequence, under preset constraints, to obtain the current determined input sequence. The adjustment method can be flexibly configured as needed; for example, each adjustment may target the input parameter values ​​at one or more sampling times in the previously determined input sequence, or it may involve increasing the input parameter values ​​at sampling times in the previously determined input sequence by a set step size, etc.

[0080] The first preset numerical range can be flexibly set according to actual applications. The first preset numerical range can be used to limit the upper and / or lower limits of the input parameter values, so that production systems with requirements for input parameter values ​​can make the input sequence predicted by the model meet production requirements by setting the first preset numerical range.

[0081] Similarly, the second preset numerical range can be flexibly set according to actual applications. This second preset numerical range can be used to limit the upper and / or lower limits of the change in the input parameter value relative to the actual input parameter value at the current time. This allows production systems that require changes in input parameter values ​​to ensure that the input sequence predicted by the model meets production requirements and maintains the stability of the input sequence predicted by the model through the setting of the second preset numerical range.

[0082] The preset constraints may also include other conditions, such as the future trend of CV caused by the change of MV; the future prediction sequence of CV satisfying the upper and lower bound constraints; the future action sequence of MV satisfying the upper and lower bound constraints; the future change sequence of MV satisfying the upper and lower bound constraints of the change amount; the degree of freedom in the optimization problem being the future action sequence of MV, etc. This embodiment does not limit these conditions.

[0083] In practical applications, the objective function can be flexibly set according to the actual application scenario. As an example, calculating the objective function value based on the objective function of the linear model may include:

[0084] The difference between the output parameter value and the set output parameter value in the target output sequence is obtained, and the difference between the input sequence determined this time and the input parameter value at the previous sampling time is obtained. The target function value is determined based on the obtained differences.

[0085] For example, the objective function of a linear model can be: the deviation of CV from the set value and the change in MV over a future period of time; that is, the absolute value of the difference between the output parameter values ​​in the target output sequence and the set value, and the absolute value of the difference between the input sequence determined this time and the input parameter values ​​at the previous sampling time. Optionally, different weights can be assigned to the two differences to represent the relative importance between different objectives, and the magnitude of the weights can be flexibly configured as needed. Through the above embodiments, an accurate output sequence can be obtained during optimization.

[0086] In practical applications, optimization solutions can be achieved in various ways, such as gradient descent, Newton's method, or conjugate gradient method. This embodiment does not limit the specific method used. The preset conditions for finding the objective function value that satisfies the preset conditions are the termination conditions for the optimization solution. In practical applications, these conditions can be flexibly configured and may include reaching a preset number of iterations and / or the solved objective function value satisfying a set threshold range.

[0087] Once the objective function value satisfying the preset conditions is obtained, the current input sequence corresponding to the obtained objective function value can be used to generate control instructions to control the production system. The predicted current input sequence may include output parameter values ​​from multiple sampling times, and the control instructions may include output parameter values ​​from the next sampling time. For example, the control instructions may include the operation variable values ​​corresponding to one or more production devices at the next sampling time. In practical applications, including output parameter values ​​from multiple sampling times in the control instructions is optional.

[0088] In practical applications, due to model mismatch and external disturbances, model predictions often differ from actual measurements. If left unaddressed, these accumulated prediction errors can negatively impact control performance. Therefore, in online real-time control, the two models can be corrected in real time to improve prediction accuracy. For example, the method may further include:

[0089] Obtain the actual output parameter value at the current sampling time;

[0090] The linear model is corrected based on the actual output parameter values ​​and the output parameter values ​​predicted by the linear model at the current sampling time from the previous sampling time; and / or,

[0091] The nonlinear model is corrected based on the actual output parameter values ​​and the output parameter values ​​predicted by the nonlinear model at the current sampling time in the previous sampling time.

[0092] In this embodiment, the current predicted value and future predicted sequence of the model can be corrected based on the actual measurement value at the current moment, which can effectively overcome the impact of system disturbances and model mismatch and improve the robustness of the control system.

[0093] like Figure 2E The diagram shown is a schematic representation of a predictive control system according to an exemplary embodiment of this specification. The system includes an intelligent modeling and evaluation subsystem, a data acquisition and storage subsystem, and an intelligent control subsystem; wherein,

[0094] The cloud-based intelligent modeling and evaluation subsystem includes a data preprocessing module, a machine learning model training module, a system identification module, and a performance evaluation module for predictive and control models. The subsystem's inputs include historical operational data from the production system, and its outputs include predictive and control models that meet usage requirements.

[0095] The cloud-based data acquisition and storage subsystem includes a data cloud storage module and a model cloud storage and publishing module. The inputs to this subsystem include operational data collected from the industrial production system on-site, and predictive and control models obtained from the cloud-based intelligent modeling and evaluation subsystem. The outputs of this subsystem include predictive and control models published to the intelligent control subsystem.

[0096] The intelligent control subsystem includes a machine learning model prediction module, a control model prediction module, a model calibration module, and a rolling optimization module. In the machine learning model prediction module, the prediction model calculates the future predicted sequence of the controlled variable based on real-time data and passes it to the rolling optimization solver. Based on the future predicted sequence, the rolling optimization solver module adjusts different control sequences as inputs to the control model to obtain different closed-loop prediction sequences. Then, according to the user-defined control optimization objective, it calculates the optimal control sequence and issues the control command for the next moment to the actuators of the industrial production system for closed-loop control. For example, the control model can predict the first output sequence at each sampling moment, and the prediction model can predict the second output sequence at each sampling moment. The rolling optimization solver can fuse the first and second output sequences to obtain the target output sequence. There are various fusion methods, which can be determined according to needs in practical applications. For example, the output parameter value at each target sampling moment in the target output sequence can be based on the average of the output parameter values ​​at the target sampling moments in the first and second output sequences, or a custom adjustment based on the average, etc. The rolling optimization solver substitutes the fused target output sequence into the objective function to calculate the objective function value, repeats this process, and finally solves for the optimal control sequence.

[0097] like Figure 2F The diagram shown is a flowchart illustrating another control method for a production system according to an exemplary embodiment of this specification, comprising the following steps:

[0098] Step 21: Data preprocessing.

[0099] This step mainly involves preprocessing historical data from the industrial site before modeling to meet the needs of subsequent modeling.

[0100] Historical data may include:

[0101] 1) Output data of industrial production systems, i.e., controlled variables of the control system;

[0102] 2) Input data of industrial production systems, i.e., the main variables that affect the output data of the system, including operating variables and disturbance variables in the control system.

[0103] Data preprocessing includes:

[0104] 1) Abnormal data processing, including but not limited to outlier removal and missing value imputation;

[0105] 2) Feature extraction, including but not limited to algorithms such as Principal Component Analysis (PCA), Partial Least Squares Regression (PLS), and Kalman filtering, to extract effective features from the original data;

[0106] 3) Variable correlation and causality analysis to identify the main influencing factors affecting the output variables for model construction.

[0107] Step 22: Machine learning model training.

[0108] This step mainly involves using machine learning algorithms to construct a nonlinear AI prediction model based on the preprocessed input data (operational variables and disturbance variables) and output variables (controlled variables) of the production system. This model accurately predicts the future change sequence of the output variables. There are no restrictions on the structure of the prediction model in this step, and common supervised learning, unsupervised learning, or deep learning algorithms can be used for training.

[0109] The training data for the prediction model includes historical production data of the production system, which includes the input data (such as coal feed data) and output data (such as temperature data) of the production system.

[0110] During training, the input to the prediction model is the historical input data (coal feed amount) of the production system. Based on the historical input data (coal feed amount) of the production system, the prediction model obtains the prediction data (i.e., predicts the temperature data). Then, the prediction data is compared with the historical output data (historical temperature data) of the production system to obtain the difference between the two. The model parameters are optimized by minimizing the difference between the two.

[0111] Step 23: Performance evaluation of the prediction model.

[0112] In this step, based on the prediction model trained in Step 2, prediction calculations are performed on the input data in the validation set to obtain the future predicted sequence of the output variable (controlled variable). This predicted sequence is then compared with the actual values ​​of the output variable (controlled variable) in the validation set to obtain the corresponding accuracy index. Based on the evaluation results, it is determined whether the prediction model can be used for the control of the production system. Optionally, the effective prediction length of the prediction model can also be determined, that is, the effective time length in the predicted output sequence, i.e., the number of effective sampling times.

[0113] Step 24: System identification.

[0114] This step mainly involves constructing an interpretable control model based on the preprocessed input data (operated variables) and output data (controlled variables) of the production system using a system identification algorithm. It focuses on accurately describing the control relationship from the operated variables to the controlled variables. The parameters of the control model obtained in this step should be interpretable and conform to the characteristics of the actual production system.

[0115] During training, the input to the control model is the historical output data (temperature data) of the production system, and the output of the control model is the input data (coal feed rate) of the production system.

[0116] Step 25: Performance evaluation of the control model.

[0117] This step, based on the control model trained in Step 4, performs prediction calculations on the input data of the validation set to obtain the future predicted sequence of the output variable (controlled variable), and compares it with the actual value of the output variable (controlled variable) in the validation set to obtain the corresponding accuracy index. At the same time, the parameters of the control model can be evaluated manually or by algorithm to determine whether they conform to the characteristics of the actual production system. Based on the evaluation results, it can be determined whether the control model can be used as an intelligent controller.

[0118] Step 26: Model Release.

[0119] This step involves publishing the qualified machine learning prediction and control models obtained from Steps 1-5 to the intelligent control subsystem for online prediction and real-time control.

[0120] Step 27: Online Prediction by the Predictive Model. This step uses the trained and deployed AI predictive model to make online predictions based on real-time input data (operated variables and disturbance variables) of the on-site production system, generating a predicted sequence of system output variables (controlled variables) for a certain period of time in the future, for use by subsequent intelligent control algorithms.

[0121] Step 28: Model calibration.

[0122] This step, based on real-time field data, compares the deviations between the predicted values ​​of the prediction model and the control model, corrects the parameters of the prediction model and the control model, and uses the corrected prediction model and control model for control calculations at the next time step.

[0123] For example, for the control model, the parameters of the control model are corrected based on the deviation between the real-time field data and the first output sequence predicted by the control model. For the prediction module, the parameters of the prediction model are corrected based on the deviation between the real-time field data and the second output sequence predicted by the prediction model.

[0124] Step 29: Calculate online control commands.

[0125] This step is based on the relationship between the predicted sequence generated by the AI ​​prediction model and the control model. Using an optimization solver, the control variables (i.e., the input of the control model) are adjusted to obtain different closed-loop predicted sequences. Then, according to the control optimization objective set by the user, the optimal control sequence is calculated, and the control command for the next moment is sent to the actuator of the industrial production system for closed-loop control.

[0126] In the next running time, steps 27 to 29 are repeated for rolling optimization to achieve online real-time control.

[0127] This embodiment can be applied to any industrial production scenario. Taking the intelligent control scenario of a cement decomposition furnace as an example, the decomposition furnace is a key piece of equipment in the cement production process. Adding municipal solid waste to the decomposition furnace and reducing coal consumption is an inevitable trend in the industry. Reducing temperature fluctuations in the decomposition furnace is beneficial for increasing waste processing capacity, which is of great significance for energy conservation and emission reduction in the system. However, the deflagration phenomenon caused by the increased waste processing capacity often requires the system to make more precise and timely adjustments, which requires the system to have more accurate predictions of future temperature changes. However, due to the strong nonlinearity and randomness of deflagration, its disturbance to the system cannot be introduced into the model predictive controller in the form of traditional disturbance variables. Therefore, using the solution of this embodiment, a reliable nonlinear prediction model of future temperature change trends can be established using machine learning and integrated into the intelligent controller for online rolling optimization, realizing closed-loop control of the decomposition furnace. This effectively improves the control stability of the decomposition furnace outlet temperature. After the decomposition furnace is controlled stably, more waste can be added to reduce coal consumption, thereby increasing waste processing capacity and effectively reducing system coal consumption and carbon emissions.

[0128] In this embodiment, a nonlinear AI model can be used for feedforward prediction. The optimal control sequence is solved based on the AI ​​prediction sequence and the control relationship of the control model. Under the premise of ensuring stable and reasonable control actions, the nonlinear dynamic characteristics of the production system can be considered more effectively, thereby improving the control effect under complex nonlinear conditions.

[0129] In this embodiment, the proposed intelligent algorithm that integrates machine learning prediction models does not directly use AI models for optimization. Therefore, it has no restrictions on the form of AI models and can select appropriate machine learning algorithms for modeling different production processes. The development threshold is low and it has strong versatility. At the same time, the control action solution does not directly optimize complex AI models, so it can effectively avoid the abnormal control actions obtained based on black-box AI prediction models. It has strong robustness across the entire operating range, and the optimization process has high computational efficiency, which can meet the high-frequency computing requirements (second-level) of industrial manufacturing systems.

[0130] In response to the highly nonlinear characteristics of complex production processes, this embodiment proposes a solution that, compared to traditional model predictive control algorithms that use the same model for both trend prediction and control action calculation, decouples the controller's prediction model from the control model and integrates an AI model for feedforward prediction. Based on the AI ​​prediction sequence and the control relationship of the control model, the optimal control sequence is solved. This approach effectively considers the nonlinear characteristics of the system while ensuring stable and reasonable control actions, thereby improving the system's adaptability to complex nonlinear operating conditions.

[0131] The intelligent algorithm proposed in this embodiment, which integrates machine learning prediction models, effectively avoids abnormal control actions obtained based on black-box AI prediction models because the parameters of the control model are interpretable and easy to optimize, and thus has strong robustness across the entire operating range.

[0132] The intelligent controller system proposed in this embodiment integrates machine learning prediction models, combining model training, model quality assessment, model release management, and real-time control subsystems, enabling integrated controller development. Furthermore, this controller system can be based on a cloud-edge collaborative architecture, allowing model training, simulation verification, and parameter tuning on a cloud computing platform, while deploying and running the intelligent controller at the edge for real-time feedback control. This approach combines the stability and real-time performance of the edge platform, ensuring stable and safe production in industrial processes, while leveraging the computing power of the cloud platform to effectively improve controller design and debugging efficiency. The corresponding tuning results can be directly distributed to the edge, enhancing the ease of controller operation.

[0133] Corresponding to the embodiments of the control method for the aforementioned production system, this specification also provides embodiments of the control device for the production system and the computer equipment used therein.

[0134] The embodiments of the control device for the production system described in this specification can be applied to computer equipment, such as servers or terminal devices. The device embodiments can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by a processor that processes the file, loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 3 The diagram shown is a hardware structure diagram of the computer equipment containing the control device of the production system described in this manual. (Except for...) Figure 3 In addition to the processor 310, memory 330, network interface 320, and non-volatile memory 340 shown, the computer device where the control device 331 of the production system in the embodiment is located may also include other hardware depending on the actual function of the computer device, which will not be described in detail here.

[0135] like Figure 4 As shown, Figure 4 This specification is a block diagram illustrating a control device for a production system according to an exemplary embodiment. The device is applied to a predictive control system, which includes a linear model and a nonlinear model.

[0136] The linear model and the nonlinear model are respectively used to predict the output sequence corresponding to the input sequence;

[0137] The device includes:

[0138] The execution module is used to iteratively execute the following steps at each sampling time until the objective function value that satisfies the preset conditions is obtained: determine the input sequence; fuse the first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence, and substitute them into the preset objective function to calculate the objective function value;

[0139] The control module is used to generate control commands to control the production system using the input sequence corresponding to the objective function values ​​that satisfy preset conditions.

[0140] In some examples, the input sequence includes at least one input parameter corresponding to a sampling time, and the output sequence includes at least one output parameter corresponding to a sampling time; the first output sequence and the second output sequence correspond to the same sampling time.

[0141] The execution module performs the following steps: fusing the first output sequence predicted by the linear model for the currently determined input sequence and the second output sequence predicted by the nonlinear model for the currently determined input sequence, and then substituting them into a preset objective function to calculate the objective function value, including:

[0142] The first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence are fused to determine the target output sequence. The output parameter value at each target sampling time in the target output sequence is determined based on the output parameter value at the target sampling time in the first output sequence and the output parameter value at the target sampling time in the second output sequence.

[0143] Substitute each output parameter value in the target output sequence into the preset target function to calculate the target function value.

[0144] In some examples, the output parameter value at each target sampling time in the third output sequence is determined based on the average of the output parameter values ​​at the target sampling time in the first output sequence and the output parameter values ​​at the target sampling time in the second output sequence.

[0145] In some examples, both the linear model and the nonlinear model are pre-trained, and the training data used to train the linear model is partially or entirely the same as the training data used to train the nonlinear model.

[0146] In some examples, the execution module performs the determination of the input sequence, including:

[0147] If the input sequence is determined for the first time, the actual input parameter value at the previous sampling time is used as the input parameter value at the current sampling time and multiple future sampling times to obtain the input sequence determined this time.

[0148] When determining the input sequence for the first time, the input parameter values ​​at each sampling time in the previously determined input sequence are adjusted according to a set step size, and the input parameters at each sampling time after adjustment satisfy the preset constraint conditions to obtain the input sequence determined this time.

[0149] The preset constraints include:

[0150] The values ​​of each input parameter in the input sequence satisfy the first preset numerical range;

[0151] The difference between each input parameter value in the input sequence and the input parameter value at the previous sampling time satisfies the preset range of change.

[0152] In some examples, the execution module performs the substitution into a preset objective function to calculate the objective function value, including:

[0153] The difference between the target output sequence and the set output parameter value is obtained, and the difference between the input parameter value in the input sequence determined this time and the input parameter value at the previous sampling time is obtained. The target function value is determined based on the obtained differences.

[0154] In some examples, the device further includes a calibration module for:

[0155] Obtain the actual output parameter value at the current sampling time;

[0156] The linear model is corrected based on the actual output parameter values ​​and the output parameter values ​​predicted by the linear model at the current sampling time from the previous sampling time; and / or,

[0157] The nonlinear model is corrected based on the actual output parameter values ​​and the output parameter values ​​predicted by the nonlinear model at the current sampling time in the previous sampling time.

[0158] The specific implementation process of the functions and roles of each module in the control device of the above production system can be found in the implementation process of the corresponding steps in the control method of the above production system, and will not be repeated here.

[0159] Accordingly, embodiments of this specification also provide an industrial control system, which includes a predictive control system and a production system, wherein the predictive control system is connected to the production system, and the predictive control system includes:

[0160] A modeling subsystem is used to train linear and nonlinear models using historical production data from the production system.

[0161] The control subsystem is used to acquire real-time operating data of the production system. At each sampling time, iteratively executes the following steps until a target function value that satisfies preset conditions is obtained: determining the current input sequence; fusing the first output sequence predicted by the linear model for the current input sequence and the second output sequence predicted by the nonlinear model for the current input sequence, and substituting them into the preset target function to calculate the target function value; using the current input sequence corresponding to the target function value that satisfies the preset conditions, generating control commands to control the production system.

[0162] Accordingly, this specification also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the aforementioned control method embodiment for the production system.

[0163] Accordingly, embodiments of this specification also provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the control method embodiment for the production system.

[0164] Accordingly, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the control method embodiment for the production system.

[0165] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0166] The above embodiments can be applied to one or more electronic devices. The electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions. The hardware of the electronic device includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0167] The electronic device can be any electronic product that can interact with the user, such as a personal computer, tablet computer, smartphone, personal digital assistant (PDA), game console, interactive network television (IPTV), smart wearable device, etc.

[0168] The electronic device may also include network devices and / or user devices. The network devices include, but are not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of hosts or network servers.

[0169] The networks in which the electronic devices are located include, but are not limited to, the Internet, wide area networks, metropolitan area networks, local area networks, and virtual private networks (VPNs).

[0170] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0171] The steps of the various methods described above are only for clarity. In practice, they can be combined into one step or some steps can be split into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, but without changing the core design of the algorithm and process, are also within the scope of protection of this application.

[0172] The terms "specific example" or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with the embodiments or examples, which are included in at least one embodiment or example of this specification. In this specification, 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.

[0173] Other embodiments of this specification will readily occur to those skilled in the art upon consideration of the specification and practice of the invention claimed herein. This specification is intended to cover any variations, uses, or adaptations that follow the general principles of this specification and include common knowledge or customary techniques in the art not claimed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this specification are indicated by the following claims.

[0174] It should be understood that this specification is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this specification is limited only by the appended claims.

[0175] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

Claims

1. A control method for a production system, the method being applied to a predictive control system, the predictive control system comprising a solver, a linear model, and a nonlinear model; The linear model and the nonlinear model are respectively used to: predict the output sequence containing the controlled variable based on the historical sequence of manipulated variables, the sequence of disturbance variables, and the future sequence of manipulated variables; the manipulated variables include variables used to control the controlled equipment in the production system; The disturbance variables include non-regulation variables that affect the output of the production system; The controlled variable includes the output variable of the controlled device after it is controlled; The method includes: At each sampling time, the following steps are iteratively executed until the objective function value that satisfies the preset conditions is obtained: Determining the input sequence includes: adjusting the future sequence of operational variables by the solver; wherein, if the input sequence is being determined for the first time, the actual input parameter values ​​at the previous sampling time are used as the input parameter values ​​at the current sampling time and multiple future sampling times to obtain the input sequence determined this time; if the input sequence is not being determined for the first time, under the constraints of preset constraints, the input parameter values ​​at each sampling time in the previously determined input sequence are adjusted to obtain the input sequence determined this time. The input sequence, which includes the historical sequence of operational variables, the sequence of perturbation variables, and the future sequence of operational variables adjusted by the solver, is input into the linear model and the nonlinear model, respectively. The first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence are fused together and substituted into a preset objective function to calculate the objective function value; wherein, the solver adjusts multiple different future operation variable sequences in each iteration; Using the input sequence corresponding to the objective function value that satisfies the preset conditions, control commands are generated to control the production system; At the sampling time, the actual output parameter value at the current sampling time is obtained, and the linear model is corrected based on the actual output parameter value and the output parameter value at the current sampling time predicted by the linear model at the previous sampling time.

2. The method according to claim 1, wherein the output sequence includes output parameters corresponding to multiple sampling times; the first output sequence and the second output sequence correspond to the same sampling time; The step of fusing the first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence, and then substituting them into a preset objective function to calculate the objective function value, includes: The first output sequence predicted by the linear model for the determined input sequence and the second output sequence predicted by the nonlinear model for the determined input sequence are fused to determine the target output sequence. The output parameter value at each target sampling time in the target output sequence is determined based on the output parameter value at the target sampling time in the first output sequence and the output parameter value at the target sampling time in the second output sequence. Substitute each output parameter value in the target output sequence into the preset target function to calculate the target function value.

3. The method according to claim 1, wherein both the linear model and the nonlinear model are pre-trained, and the training data used to train the linear model is partially or entirely the same as the training data used to train the nonlinear model.

4. The method according to claim 1, wherein the preset constraint conditions include: The input parameter values ​​in the input sequence satisfy the first preset numerical range; or, The difference between the input parameter values ​​in the input sequence and the input parameter values ​​at the previous sampling time meets the preset range of change.

5. The method according to claim 2, wherein substituting the value into a preset objective function to calculate the objective function value includes: The difference between the output parameter value in the target output sequence and the set output parameter value is obtained, and the difference between the input parameter value in the input sequence determined this time and the input parameter value at the previous sampling time is obtained. The objective function value is determined based on the obtained differences.

6. The method according to claim 1, further comprising: The nonlinear model is corrected based on the actual output parameter values ​​and the output parameter values ​​predicted by the nonlinear model at the current sampling time in the previous sampling time.

7. A control device for a production system, the device being applied to a predictive control system, the predictive control system comprising a solver, a linear model, and a nonlinear model; The linear model and the nonlinear model are respectively used to: predict the output sequence containing the controlled variable based on the historical sequence of manipulated variables, the sequence of disturbance variables, and the future sequence of manipulated variables; the manipulated variables include variables used to control the controlled equipment in the production system; The disturbance variables include non-regulation variables that affect the output of the production system; The controlled variable includes the output variable of the controlled device after it is controlled; The device includes: The execution module is used to iteratively execute the following steps at each sampling time until a target function value satisfying preset conditions is obtained: determining the input sequence, including: adjusting the future sequence of operational variables by the solver; wherein, if the input sequence is determined for the first time, the actual input parameter value at the previous sampling time is used as the input parameter value at the current sampling time and multiple future sampling times to obtain the input sequence determined this time; if the input sequence is not determined for the first time, under the constraints of preset constraints, the input parameter values ​​at each sampling time in the previously determined input sequence are adjusted to obtain the input sequence determined this time; the input sequence containing the historical sequence of operational variables, the perturbation variable sequence, and the future sequence of operational variables adjusted by the solver is input into the linear model and the nonlinear model, respectively; the first output sequence predicted by the linear model for the current determined input sequence and the second output sequence predicted by the nonlinear model for the current determined input sequence are fused and substituted into the preset target function to calculate the target function value; wherein, the solver adjusts multiple different future sequences of operational variables in each iteration; The control module is used to: generate control commands to control the production system using the input sequence corresponding to the objective function value that satisfies the preset conditions; at the sampling time, obtain the actual output parameter value at the current sampling time, and correct the linear model based on the actual output parameter value and the output parameter value at the current sampling time predicted by the linear model at the previous sampling time.

8. An industrial control system, the industrial control system comprising a predictive control system and a production system, the predictive control system being connected to the production system, the predictive control system comprising: A modeling subsystem is used to train linear and nonlinear models using historical production data from the production system. The linear model and the nonlinear model are respectively used to: predict the output sequence containing the controlled variable based on the historical sequence of manipulated variables, the sequence of disturbance variables, and the future sequence of manipulated variables; the manipulated variables include variables used to control the controlled equipment in the production system; the disturbance variables include non-controlled variables that affect the output of the production system; The controlled variable includes the output variable of the controlled device after it is controlled; The control subsystem is used to acquire real-time operating data of the production system. At each sampling moment, iteratively executes the following steps until the objective function value that satisfies the preset conditions is solved: determining the current input sequence, including: adjusting the future sequence of operational variables by the solver; wherein, if the input sequence is determined for the first time, the actual input parameter values ​​at the previous sampling moment are used as the input parameter values ​​at the current sampling moment and multiple future sampling moments to obtain the input sequence determined this time; if the input sequence is not determined for the first time, under the constraints of preset constraints, the input parameter values ​​at each sampling moment in the previously determined input sequence are adjusted to obtain the input sequence determined this time. The input sequence, which includes the historical sequence of operational variables, the sequence of perturbation variables, and the future sequence of operational variables adjusted by the solver, is input into the linear model and the nonlinear model, respectively. The first output sequence predicted by the linear model for the current input sequence and the second output sequence predicted by the nonlinear model for the current input sequence are fused and substituted into a preset objective function to calculate the objective function value. Using the current input sequence corresponding to the objective function value that satisfies the preset conditions, control commands are generated to control the production system. The solver adjusts to generate multiple different future sequences of operational variables during each iteration. At the sampling time, the actual output parameter value at the current sampling time is obtained, and the linear model is corrected based on the actual output parameter value and the output parameter value at the current sampling time predicted by the linear model at the previous sampling time.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method of any one of claims 1 to 6.

Images