System control method and apparatus, and computer device
The hybrid adaptive PID control framework addresses the challenges of manual parameter adjustment in PID controllers and MPC limitations by integrating linear and nonlinear models, ensuring adaptability and robustness in industrial applications.
Patent Information
- Application Number
- PCT/CN2024/096146
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-05-29
- Publication Date
- 2025-12-04
AI Technical Summary
Existing PID controllers require significant manual adjustment of parameters, which is time-consuming and risky, and model predictive control (MPC) is underutilized due to data limitations and compatibility issues, limiting their widespread adoption in industrial applications.
A system control method utilizing a hybrid adaptive PID control framework that incorporates linear and nonlinear models, enabling flexible switching between traditional PID control, linear MPC, and machine learning-based MPC, with multiple policies for optimizing control parameters.
The method reduces the need for on-site manual adjustments, enhances adaptability and robustness, and achieves faster convergence to target states, making it suitable for various industrial scenarios.
Smart Images

Figure CN2024096146_04122025_PF_FP_ABST
Abstract
Description
SYSTEM CONTROL METHOD AND APPARATUS, AND COMPUTER DEVICETECHNICAL FIELD
[0001] This application relates to the control field, and in particular, to a system control method and apparatus, a computer device, and a storage medium.BACKGROUND
[0002] Proportional-integral-derivative (PID) controllers are recognized as most important controllers in industrial application due to significance and wide application. Despite of existing various configuration rules, finding a proper parameter value, particularly a parameter value of a nonlinear or time-varying system, is always a challenge. In an actual application, engineers usually need to spend a plurality of weeks in adjusting parameters of an industrial PID controller on site. In this case, a lot of manpower is required, and a potential risk exists in a production process. It should be noted that, in most cases, the PID parameters are manually adjusted, and only a small number (10%) of factories use a model predictive control (MPC) technology in a control system. Widespread market research shows that there is a large demand for the MPC. Nevertheless, due to limitations such as insufficient historical data, model inaccuracy, or a problem about compatibility with existing software, many factories cannot directly adopt MPC.SUMMARY
[0003] The summary part of the present invention is provided to introduce selected concepts in a simplified form. The concepts are further described in the following detailed description part. The summary part of the present invention is neither intended to identify any key feature or necessary feature of the theme that claims protection, nor intended to help determine the scope of the theme that claims protection.
[0004] Based on the above, this application discloses a system control method, including: initializing and training a prediction model, where the prediction model includes a linear model and a nonlinear model; receiving current system state information, and predicting a future state of a system based on the current system state information and the prediction model; receiving an objective function, and selecting, based on the objective function, at least one of a plurality of policies to calculate an optimization control parameter of the system, where the plurality of policies include a first policy and a second policy, the first policy is based on the linear model, and the second policy is based on the nonlinear model; and applying the optimization control parameter to the system.
[0005] In the above manner, the technical solutions of this application have an adaptive capability, so that on-site manual parameter adjustment is not required, thereby greatly saving manpower, and have a plurality of different control policies incorporated therein, so that high flexibility is achieved, thereby ensuring adaptability and robustness in various control scenarios.
[0006] Further, the selecting at least one of a plurality of policies to calculate an optimization control parameter of the system includes:
[0007] selecting the first policy, where the optimization control parameter of the system is calculated for the linear model through a linear solver; or
[0008] selecting the second policy, where the optimization control parameter of the system is calculated for the nonlinear model through a heuristic algorithm.
[0009] In the above manner, a plurality of different control policies are incorporated, so that high flexibility is achieved, thereby ensuring adaptability and robustness in various control scenarios.
[0010] Further, the plurality of policies further include a third policy, and the selecting at least one of a plurality of policies to calculate an optimization control parameter of the system includes:
[0011] selecting the third policy, where a manually inputted control parameter is received as the optimization control parameter of the system.
[0012] In the above manner, the corresponding optimization control parameter can be quickly obtained through a manual means before the model is fully trained, thereby providing a user with increased use flexibility.
[0013] Further, the initializing and training a prediction model, where the prediction model includes a linear model and a nonlinear model includes:
[0014] receiving an initial control parameter, a training time step, and a control period of the system; and
[0015] training the linear model and the nonlinear model based on the initial control parameter, the training time step, and the control period.
[0016] In the above manner, a model that can optimize control parameters can be trained, to provide a basis for subsequent parameter optimization.
[0017] Further, the receiving current system state information includes: receiving a sensor measurement parameter value and a control component output value.
[0018] In the above manner, corresponding specific system state parameters can be obtained. The system state parameters provide guidance and reference for subsequent system state prediction and control parameter optimization.
[0019] Further, after the initializing and training a prediction model, where the prediction model includes a linear model and a nonlinear model, the method further includes: saving a trained prediction model.
[0020] In the above manner, the trained prediction model can be continuously saved, and the prediction model can be continuously upgraded to improve performance, thereby providing a more accurate model for a solving process of the optimization control parameter.
[0021] In addition, this application discloses a system control apparatus, including:
[0022] an initialization module, configured to initialize and train a prediction model, where the prediction model includes a linear model and a nonlinear model;
[0023] a state module, configured to receive current system state information, and predict a future state of a system based on the current system state information and the prediction model;
[0024] an optimization module, configured to receive an objective function, and select, based on the objective function, at least one of a plurality of policies to calculate an optimization control parameter of the system, where the plurality of policies include a first policy and a second policy, the first policy is based on the linear model, and the second policy is based on the nonlinear model; and
[0025] an application module, configured to apply the optimization control parameter to the system.
[0026] This application further provides a computer device, including a memory and a processor. The memory stores a computer program. The processor, when executing the computer program, implements the above method.
[0027] This application further provides a computer-readable storage medium, storing a computer program. The computer program, when executed by a processor, implements the above method.
[0028] This application further provides a computer program product. The computer program product is tangibly stored in a computer-readable medium and includes computer-executable instructions. The computer-executable instructions, when executed, cause at least one processor to perform the above method.BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Implementations of the present disclosure are described in drawings in a form of an example rather than a limitation, and similar reference numerals in the drawings indicate same or similar components.
[0030] FIG. 1 is a schematic flowchart of a system control method according to an implementation of this application.
[0031] FIG. 2 is a schematic diagram of a system control apparatus according to an implementation of this application.
[0032] FIG. 3 is a schematic diagram of a system-controlled computer device according to an implementation of this application.
[0033] FIG. 4 is a schematic diagram of a system control framework according to an implementation of this application.
[0034] Reference numerals are as follows:
[0035] S101-S104: Step
[0036] 200: Apparatus
[0037] 201: Module
[0038] 202: Module
[0039] 203: Module
[0040] 204: Module
[0041] 300: Computer device
[0042] 302: Processor
[0043] 304: MemoryDETAILED DESCRIPTION
[0044] In the following specification, a large number of specific details are described for a purpose of explanation. However, it may be understood that, implementations of the present invention may be implemented without the specific details. In other examples, well-known circuits, structures, and technologies are not shown in detail to avoid affecting the understanding of the specification.
[0045] Although hybrid adaptive control policies are increasingly popular and are developed to some extent, some challenges still need to be resolved in an actual application, and a gap still exists between a theoretical research and an actual application. One of the key challenges is to ensure robustness. The hybrid adaptive control policies need to cope with uncertainties, disturbances, and changes in system dynamics. The robustness is crucial for maintaining control performance and adaptability in an actual industrial environment. Researchers are exploring methods for improving the robustness of the policies. For example, adaptive learning and malfunction detection are performed by using a robust control technology, developing an adaptive algorithm, or integrating a machine learning method.
[0046] To overcome the challenge, academia and industry need to collaborate to verify the policies in actual industrial applications and improve the policies based on an actual situation. As research and development go increasingly deeper, the hybrid adaptive control policies may completely change control systems in all industries.
[0047] Therefore, in an implementation scenario of this application, a robust hybrid adaptive proportional-integral-derivative (PID) control framework is provided, which can flexibly switch among traditional PID control, model predictive control (MPC) using a linear physical model, and MPC using a machine learning model for a complex physical scenario. The framework not only inherits advantages of the MPC, but also can be seamlessly integrated with PID systems commonly used in many factories.
[0048] Specifically, this application discloses a system control method, including the following steps:
[0049] S101: Initialize and train a prediction model, where the prediction model includes a linear model and a nonlinear model.
[0050] The prediction model is configured to predict a future state of a system. The system may be a target system or a control scenario in industrial manufacturing, for example, a furnace temperature control system of a monocrystalline silicon manufacturing furnace. The linear model may also be referred to as a theoretical physical model or an experimental model. The name of linear model is due to a general composition of linear formulas with physical significance. The nonlinear model may also be referred to as a neural network model or a black box model, including a hidden layer. The name of nonlinear model is due to general nonlinearity.
[0051] For initialization and training of the prediction model, historical data and a hyperparameter may be used as an input for training the prediction model. The historical data may include parameters such as a pressure, a temperature, an argon gas flow, and a power.
[0052] The above parameters and hyperparameters after self-adjustment are inputted into the prediction model, to respectively train the linear model and the nonlinear model, and the trained prediction model is stored after the model is verified to be valid.
[0053] Specifically, the initializing and training a prediction model, where the prediction model includes a linear model and a nonlinear model includes:
[0054] receiving an initial control parameter, a training time step, and a control period of the system; and
[0055] training the linear model and the nonlinear model based on the initial control parameter, the training time step, and the control period.
[0056] In the above specific manner, a model that can optimize control parameters can be trained, to provide a basis for subsequent parameter optimization.
[0057] S102: Receive current system state information, and predict a future state of a system based on the current system state information and the prediction model.
[0058] The receiving current system state information includes: receiving a sensor measurement parameter value and a control component output value.
[0059] Specifically, the system may include a sensor and a control component. The control component may be a programmable logic controller (PLC) . For example, the furnace temperature control system of the above monocrystalline silicon manufacturing furnace includes a sensor and a control component. In some embodiments, the current system state information is received, and current detection data and output data of the sensor and the control component in a manufacturing factory are received. Then, a future state of the furnace temperature control system of monocrystalline silicon manufacturing furnace is predicted based on the current system state information and the above trained prediction model. In a temperature example, a predicted temperature of the manufacturing factory can be obtained through the prediction model.
[0060] S103: Receive an objective function, and select, based on the objective function, at least one of a plurality of policies to calculate an optimization control parameter of the system. The plurality of policies include a first policy and a second policy, the first policy is based on the linear model, and the second policy is based on the nonlinear model.
[0061] The objective function may be a minimized system prediction error. In other words, an error between the future state of the system predicted by the prediction model and a desired target state of the system is expected to be minimized. The control parameter may be understood as a parameter of a control system, for example, a PID parameter. Because some systems have PID controllers embedded therein, control parameters of the systems are actually the PID parameters. The PID parameters can be used to control a heating power of the monocrystalline silicon manufacturing furnace, thereby indirectly controlling a furnace temperature. In a temperature example, the objective function is intended to find an optimal PID parameter, to obtain a most proper power output manner of the heating furnace. In this way, a predicted temperature in a future period of time provided by the prediction model is stable and has a minimized deviation from a target temperature. The selecting at least one of a plurality of policies to calculate an optimization control parameter of the system means selecting a proper prediction model and solver based on different conditions or scenarios, to calculate the corresponding optimization control parameter. For example, when an amount of training data is relatively small, the prediction model is in an early stage of the training, or the optimization control parameter needs to be quickly provided, the linear model and a corresponding linear solver may be used for solving. The training of the linear model only requires a small amount of data, and the linear solver requires only a short calculation time. In addition, the linear model may be customized based on a user need. If the training data is relatively abundant, the prediction model is in a later stage of the training, or a relatively high precision of the optimization control parameter is required, the nonlinear model and a heuristic algorithm may be used for solving. The nonlinear model is more in line with an actual system, and the heuristic algorithm obtains a more precise optimization control parameter. When the training data is insufficient or the prediction model is unavailable, a manually inputted control parameter may be directly received as a subsequent control parameter of the system.
[0062] Specifically, the selecting at least one of a plurality of policies to calculate an optimization control parameter of the system includes:
[0063] selecting the first policy, where the optimization control parameter of the system is calculated for the linear model through a linear solver; or
[0064] selecting the second policy, where the optimization control parameter of the system is calculated for the nonlinear model through the heuristic algorithm.
[0065] In the above specific manner, a plurality of different control policies are incorporated, so that high flexibility is achieved, thereby ensuring adaptability and robustness in various control scenarios.
[0066] Further, the plurality of policies further include a third policy, and the selecting at least one of a plurality of policies to calculate an optimization control parameter of the system includes:
[0067] selecting the third policy, where a manually inputted control parameter is received as the optimization control parameter of the system.
[0068] In the above manner, the corresponding optimization control parameter can be quickly obtained through a manual means before the model is fully trained, thereby providing a user with increased use flexibility.
[0069] As shown in FIG. 4, in some embodiments, in a temperature example, a target temperature, a constraint, the prediction model, and a previous experimental or empirical PID parameter of the system are loaded, solving is performed for the linear model by using the linear solver, and solving is performed for the nonlinear solver by using the heuristic algorithm, to respectively find a PID parameter set. The PID parameter set can minimize the objective function. For example, the PID parameter set can minimize an error or a variance between a target temperature of the system and a predicted temperature of the system in the objective function.
[0070] In some embodiments, three different PID parameter sets may be obtained through the above three policies, and one of the PID parameter sets is selected as a final determined optimization control parameter. In some embodiments, one of the policies may be used for solve a PID parameter set, and the obtained unique PID parameter set may be directly used as the optimization control parameter.
[0071] During the above solving of the PID parameter, a plurality of iterations and model adjustments are required. For example, during the solving, an intermediate PID parameter set is obtained, and then the intermediate PID parameter set may be applied to the prediction model to calculate a temperature-related variable, for example, a power. For example, a loaded prediction model is selected, and a system state, for example, a current temperature, a predicted temperature, a power, and an error between the target temperature and the current temperature is inputted into the prediction model for retraining, to improve performance of the prediction model. After the retraining, parameters of the prediction model are adjusted, and then the trained model is saved. In this way, the prediction model is continuously trained and improved, so that a more accurate PID parameter set can be obtained through the solver, thereby forming a virtuous cycle.
[0072] In some embodiments, during the iteration, an intermediate system state obtained by using the intermediate PID parameter set is also saved, to facilitate accumulation of historical data and training of other prediction models. For example, a system temperature, a predicted temperature, a power, and an error between the target temperature and an intermediate temperature obtained by using the intermediate PID parameter set are saved.
[0073] S104: Apply the optimization control parameter to the system.
[0074] Finally, the optimization control parameter is applied to the system, to enable the system to achieve a target system state.
[0075] Application scenario:
[0076] The system framework is applicable to various control scenarios. To demonstrate effectiveness of the system framework, this application implements the system and the method in a furnace temperature control usage case of a furnace temperature control system of a monocrystalline silicon manufacturing furnace. The system uses a key technology including three different control policies, to cope with various situations.
[0077] The first policy is MPC using an experimental or theoretical physical model. The policy is applicable to a situation that can be accurately described by a physical model. In addition, the system allows a user to customize a function instead of relying on a default function.
[0078] The second policy is to use a time-series neural network model as the MPC prediction model. The method is applicable to a situation in which sufficient historical data is collected from a controlled object, and has a strong control capability.
[0079] The third policy allows the user to manually input a PID parameter. The method is particularly effective in an initial operating stage of a device that does not have sufficient data for model training.
[0080] Demonstration results show that a convergence curve of temperature control is smoother, indicating more stable control and a stronger adaptive capacity to a complex situation, and that the second policy has faster temperature convergence, indicating higher control efficiency. The system framework can cope with a situation with high uncertainty and a large amount of manual intervention.
[0081] It is to be further noted that, for the above system and method, inputs may include a system model, a control target, a control period, an optimization algorithm, and historical data. Outputs include controller parameters.
[0082] In some embodiments, specific steps may be as follows:
[0083] 1. Initialization: an initial controller parameter is set, a time step and a control period are defined, and a physical model and a machine learning model are trained.
[0084] 2. The following steps are repeated:
[0085] a. A current system state is read, including a sensor measurement value and a controller output.
[0086] b. A control model is read.
[0087] c. State prediction is performed based on the control model.
[0088] System state changes within a future time range are predicted by using the current system state and a PID parameter.
[0089] d. The PID parameter is optimized:
[0090] An objective function is defined. For example, a prediction error is minimized.
[0091] An optimization algorithm is applied. For example, an optimal PID parameter is calculated for the physical model by using a linear solver and a machine learning prediction model through heuristic search.
[0092] e. A control action is performed. The control action is calculated based on a calculated optimal controller parameter, and is applied to a controlled system.
[0093] f. A system state and time are updated.
[0094] The process waits for a next control period.
[0095] A new system state is read, and current time is updated.
[0096] 3. The algorithm is ended.
[0097] An embodiment of this application can provide all of the above three policies, to enable a user to select a policy based on a need or a current target condition. The selection from a plurality of policies can reflect characteristics of wide applicability and desirable robustness of the embodiments of this application.
[0098] Compared with the prior art, this application has the following advantages:
[0099] 1. Different from a commonly used PID controller, the present invention has an adaptive capability without a need of on-site manual parameter adjustment. The characteristic significantly saves manpower.
[0100] 2. The present invention is a hybrid system, and has three different control policies incorporated therein, so that high flexibility is achieved, thereby ensuring adaptability and robustness in various control scenarios. Different from a pure MPC system, the present invention omits data collection before use, thereby avoiding any delay.
[0101] 3. Compared with a conventional PID controller, the MPC control mode of the present invention has a significantly improved control effect.
[0102] 4. In the present invention, the PID parameter is optimized, so that the PID parameter and the control variable can be simultaneously outputted. Therefore, the present invention is compatible with a main PID control system of most manufacturers. The characteristic is acceptable for customer engineers.
[0103] Experimental verification:
[0104] A control result in a test environment further shows the following: A control effect generated by a PID parameter manually determined by a control engineer stage by stage is that a temperature response stabilizes at a temperature close to the target temperature after approximately 600 time steps. A control effect generated by an adaptive PID algorithm of this application is that the MPC achieves a stable temperature close to the target temperature after approximately 450 time steps, and that the neural network-based prediction control achieves a stable temperature close to the target temperature after approximately 120 time steps. A temperature change curve is smooth, and required time is only one-fifth of a baseline, indicating a higher control level. The test environment is developed based on real thermal field data, including random interference. In addition, the adaptive PID algorithm adopts different models to ensure fairness during the test.
[0105] In addition, this application has the following advantages:
[0106] 1. The system of this application has a unique YAML ain't markup language (yml) configuration file structure, which is easy to identify. Configuration data used by a plagiarist is analyzed. The structure is particularly effective for detecting a plagiarism behavior.
[0107] 2. The system of this application has a unique interface specially designed for inputting physical model parameters. If others provide a similar interface for a user to adjust parameters, potential plagiarism suspicion is caused.
[0108] The quotations throughout the specification to "an implementation" , "implementation" , "an exemplary implementation" , "some implementations" , "various implementations" , and the like indicate that the described implementations of the present invention may include specific features, structures, or characteristics. However, it does not mean that every implementation needs to include the specific features, structures, or characteristics. In addition, some implementations may have some or all of the features described for other implementations, or do not have the features described for other implementations.
[0109] It should be understood that, although the steps in the flowchart of FIG. 1 are displayed in sequence based on indication of arrows, the steps are not necessarily performed successively in the order indicated by the arrows. Unless otherwise explicitly specified in this application, execution of the steps is not strictly limited, and the steps may be performed in another order. Moreover, at least some of the steps in FIG. 1 may include a plurality of steps or a plurality of stages. The steps or stages are not necessarily performed at the same moment, and may be performed at different moments. The steps or stages are not necessarily performed successively, and may be performed alternately with other steps or at least some of steps or stages of other steps.
[0110] FIG. 2 provides a system control apparatus 200. The apparatus 200 includes:
[0111] an initialization module 201, configured to initialize and train a prediction model, where the prediction model includes a linear model and a nonlinear model;
[0112] a state module 202, configured to receive current system state information, and predict a future state of a system based on the current system state information and the prediction model;
[0113] an optimization module 203, configured to receive an objective function, and select, based on the objective function, at least one of a plurality of policies to calculate an optimization control parameter of the system, where the plurality of policies include a first policy and a second policy, the first policy is based on the linear model, and the second policy is based on the nonlinear model; and
[0114] an application module 204, configured to apply the optimization control parameter to the system.
[0115] It should be noted that the apparatus may include more or fewer modules to implement the described functions.
[0116] For example, at least one module in FIG. 2 may be further divided into a plurality of different submodules. Each of the submodules is configured to perform at least part of operations described herein in combination with the corresponding module. In addition, in some examples, the apparatus 200 may further include an additional module configured to perform other operations already described in the specification. In addition, it may be understood by a person skilled in the art that, an exemplary apparatus 200 may be implemented by software, hardware, firmware, or any combination thereof.
[0117] FIG. 3 provides a computer device. According to an implementation, a computer device 300 may include a processor 302. The processor 302 executes a computer program stored in a memory 304. The computer program, when executed by the processor, implements the above method.
[0118] It should be understood by a person skilled in the art that, the structure shown in FIG. 3 is merely a block diagram of a partial structure related to the solution of this application, and does not constitute a limitation on the computer device to which this application is applied. The computer device may specifically include more or fewer components than those shown in the figure, or have some merged components, or may have different component arrangements.
[0119] A person of ordinary skill in the art may understand that, all or some processes of the method in the above implementation may be implemented by instructing relevant hardware through a computer program. The computer program may be stored in a non-volatile computer-readable storage medium. When the computer program is executed, the processes of the above implementations of the method may be implemented. Any reference to the memory, the storage medium, the database, or other media used in the implementations provided in this application may include at least one of a non-volatile memory and a volatile memory. The non-volatile memory may include a read-only memory (ROM) , a magnetic tape, a floppy disk, a flash memory, an optical memory, or the like. The volatile memory may include a random access memory (RAM) or an external cache. As a description rather than a limitation, the RAM may have various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM) .
[0120] This application further provides a computer-readable storage medium, storing a computer program. The computer program, when executed by a processor, implements the above steps.
[0121] This application further provides a computer program product. The computer program product is tangibly stored in a computer-readable medium and includes computer-executable instructions. The computer-executable instructions, when executed, cause at least one processor to perform the above method.
[0122] Further, the computer program may be stored and run in a cloud to perform the method. Further, constitutions of the program may be laid out on a plurality of devices and clouds. For example, the corresponding steps may be laid out and run on native or local computers, or run on different cloud devices to send a signal through communication connection, or may be laid out and run on native or local computers. In this application, the manner or the method is not limited, corresponding technologies can be flexibly laid out and deployed, and the method is performed and completed by fully using devices and technologies such as a cloud, big data, and a supercomputing capability.
[0123] Some implementations of the present disclosure may include a product. The product may include a storage medium configured to store logic. Examples of the storage medium may include one or more types of computer-readable storage media that can store electronic data, including a volatile or non-volatile memory, a removable or non-removable memory, an erasable or non-erasable memory, a writable or rewritable memory, and the like. Examples of the logic may include various software units, such as a software component, a program, an application, a computer program, an application, a system program, a machine program, operating system software, middleware, firmware, a software module, a routine, a subroutine, a function, a method, a process, a software interface, an application program interface (API) , an instruction set, computing code, computer code, a code segment, a computer code segment, a field, a value, a symbol, or any combination thereof. In some implementations, for example, the product may store executable computer program instructions that, when executed by the processor, cause the processor to perform the method and / or operations described herein. The executable computer program instructions may include any appropriate type of code, for example, source code, compiled code, interpretive code, executable code, static code, and dynamic code. The executable computer program instructions may be implemented based on a predefined computer language, manner, or syntax configured to command a computer to execute a specific function. The instructions may be implemented by using any appropriate high-level, low-level, object-oriented, visualized, compiled, and / or interpretive programming language.
[0124] The examples including the disclosed architecture have been described above. It is certainly not possible to describe every conceivable combination of components and / or methods, but a person skilled in the art may understand that, many other combinations and arrangements are also feasible. Therefore, the novel architecture is intended to cover all such substitutions, modifications, and variations that fall within the spirit and scope of the appended claims.
Claims
1.A system control method (100) , comprising:(101) initializing and training a prediction model, wherein the prediction model comprises a linear model and a nonlinear model;(102) receiving current system state information, and predicting a future state of a system based on the current system state information and the prediction model;(103) receiving an objective function, and selecting, based on the objective function, at least one of a plurality of policies to calculate an optimization control parameter of the system, wherein the plurality of policies comprise a first policy and a second policy, the first policy is based on the linear model, and the second policy is based on the nonlinear model; and(104) applying the optimization control parameter to the system.2.The method (100) according to claim 1, wherein the selecting at least one of a plurality of policies to calculate an optimization control parameter of the system comprises:selecting the first policy, wherein the optimization control parameter of the system is calculated for the linear model through a linear solver; orselecting the second policy, wherein the optimization control parameter of the system is calculated for the nonlinear model through a heuristic algorithm.3.The method (100) according to claim 1, wherein the plurality of policies further comprise a third policy, and the selecting at least one of a plurality of policies to calculate an optimization control parameter of the system comprises:selecting the third policy, wherein a manually inputted control parameter is received as the optimization control parameter of the system.4.The method (100) according to claim 1, wherein the initializing and training a prediction model, wherein the prediction model comprises a linear model and a nonlinear model comprises:receiving an initial control parameter, a training time step, and a control period of the system; andtraining the linear model and the nonlinear model based on the initial control parameter, the training time step, and the control period.5.The method (100) according to claim 1, wherein the receiving current system state information comprises:receiving a sensor measurement parameter value and a control component output value.6.The method (100) according to claim 1, wherein after the initializing and training a prediction model, wherein the prediction model comprises a linear model and a nonlinear model, the method further comprises:saving a trained prediction model.7.A system control apparatus (200) , comprising:an initialization module (201) , configured to initialize and train a prediction model, wherein the prediction model comprises a linear model and a nonlinear model;a state module (202) , configured to receive current system state information, and predict a future state of a system based on the current system state information and the prediction model;an optimization module (203) , configured to receive an objective function, and select, based on the objective function, at least one of a plurality of policies to calculate an optimization control parameter of the system, wherein the plurality of policies comprise a first policy and a second policy, the first policy is based on the linear model, and the second policy is based on the nonlinear model; andan application module (204) , configured to apply the optimization control parameter to the system.8.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program, implements the steps of the method according to any of claims 1 to 4.9.A computer-readable storage medium, storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the method according to any of claims 1 to 4.10.A computer program product, tangibly stored in a computer-readable medium and comprising computer-executable instructions, wherein the computer-executable instructions, when executed, cause at least one processor to perform the method according to any of claims 1 to 4.
Citation Information
Patent Citations
Industrial process control platform for realizing MVC
CN117389237A
Temperature control method and system based on multi-path pid temperature controller
CN117519365A
System for controling greenhouse environment and parameter of controller of greenhouse
KR1020170028721A
Method and system for process control
US20090043546A1