Adaptive model prediction control method, electronic equipment and storage medium

Abstract

The invention provides an adaptive model prediction control method, electronic equipment and a storage medium, and relates to the technical field of automatic control. Comprising the steps that when it is detected that a target controlled variable in a target industrial process meets a stable state, a pre-generated composite excitation signal is applied to a target operation variable corresponding to the target controlled variable, and input data of the target operation variable and response data of the target controlled variable during the period of applying the composite excitation signal are recorded; according to the input data of the target operation variable and the response data of the target controlled variable, performing parameter estimation on the initial process model to obtain a target process model; and performing result control of the target controlled variable in the target industrial process based on the target process model. Through intelligent state management, composite excitation testing, process model online identification updating and self-adaptive model prediction control are integrated, the full-automatic, high-precision and high-robustness intelligent control process is achieved, and the control precision and stability of model prediction control are improved.

Description

Technical Field

[0001] This application relates to the field of automation control technology, and more specifically, to an adaptive model predictive control method, electronic device, and storage medium. Background Technology

[0002] As industrial processes become increasingly complex, the demands for control accuracy, stability, and robustness are constantly rising. Model Predictive Control (MPC), due to its ability to explicitly handle constraints, multivariable coupling, and time-delay problems, has become a core technology of Advanced Process Control (APC). However, the performance of MPC is highly dependent on the accuracy of the process model. In practical engineering, establishing accurate mechanistic models is costly, time-consuming, and difficult to cover all operating conditions. Therefore, how to establish accurate process models and achieve adaptive control has become a key bottleneck for the widespread application of MPC.

[0003] Currently, the establishment of process models mainly relies on manual intervention, requiring operators to manually apply test signals and select identification windows. Furthermore, the test signals are relatively simple, resulting in low accuracy of the established process models and affecting the control effect of MPC. Summary of the Invention

[0004] The purpose of this application is to address the shortcomings of the prior art by providing an adaptive model predictive control method, electronic device, and storage medium, so as to improve the modeling efficiency and accuracy of process models in model predictive control, thereby improving the accuracy of MPC optimization control.

[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, embodiments of this application provide an adaptive model predictive control method, including: When the target controlled variable under the target industrial process is detected to meet the steady state, a pre-generated composite excitation signal is applied to the target operating variable corresponding to the target controlled variable, and the input data of the target operating variable and the response data of the target controlled variable are recorded during the application of the composite excitation signal. Based on the input data of the target operational variable and the response data of the target controlled variable, the parameters of the initial process model are estimated to obtain the target process model; Based on the target process model, the results of the target controlled variables under the target industrial process are controlled.

[0006] Optionally, the step of applying a pre-generated composite excitation signal to the target manipulated variable corresponding to the target controlled variable when the target controlled variable in the target industrial process is detected to satisfy a steady state includes: After the target industrial process is started, the value of the target controlled variable is collected in real time. If the standard deviation of the target controlled variable is less than a preset threshold for a preset duration, it is determined that the target controlled variable has reached the stable state. The composite excitation signal is applied to the target operated variable corresponding to the target controlled variable.

[0007] Optionally, the composite excitation signal is obtained by weighted combination of a pseudo-random binary sequence signal and a swept-frequency sinusoidal signal, and the weight coefficients of the pseudo-random binary sequence signal and the swept-frequency sinusoidal signal are adaptively configured according to the response speed of the target industrial process; the composite excitation signal is used to detect the correlation between the target operating variable and the target controlled variable in the target industrial process.

[0008] Optionally, the step of estimating the parameters of the initial process model based on the input data of the target operated variable and the response data of the target controlled variable to obtain the target process model includes: Based on the input data of the target operated variable and the response data of the target controlled variable, the recursive least squares method is used to perform preliminary parameter estimation of the initial process model to obtain the initial gain and the initial time constant. Using the initial gain and initial time constant as initial parameters, the initial process model is optimized by sequential least squares programming algorithm to determine the target gain, target time constant and target time delay parameters corresponding to the initial process model. The target process model is obtained by using the target gain, the target time constant, and the target time delay parameter as model parameters of the initial process model.

[0009] Optionally, the result control of the target controlled variable under the target industrial process based on the target process model includes: Based on the target process model, the incremental information of the target operational variable is periodically predicted, and the current input data of the target operational variable is adjusted according to the prediction results, so as to achieve result control of the target controlled variable.

[0010] Optionally, the step of periodically predicting the incremental information of the target operational variable based on the target process model includes: The target process model performs multi-step open-loop prediction based on the actual measured value of the target controlled variable at the current control moment and the historical increment sequence of the target operated variable, generating a predicted sequence of the target controlled variable for the next P steps corresponding to the current control moment; Based on the actual measured value of the target controlled variable at the current control time and the predicted value of the target controlled variable at the previous control time, the predicted sequence of the target controlled variable is compensated to obtain the compensated predicted sequence of the target controlled variable. Based on the compensated predicted sequence of the target controlled variable, the control increment sequence corresponding to the previous control time, and the constructed objective function, the control increment sequence that minimizes the objective function is solved; the objective function is a function related to the output error weight, control increment weight, setpoint trajectory, and future multi-step control increment sequence. Based on the control increment sequence, determine the increment information of the target operational variable at the current control moment.

[0011] Optionally, the step of solving for the control increment sequence that minimizes the objective function based on the compensated predicted sequence of the target controlled variable, the control increment sequence corresponding to the previous control time, and the constructed objective function includes: Substituting the compensated predicted sequence of the target controlled variable and the control increment sequence corresponding to the previous control time into the objective function, and based on the numerical constraint range of the target operated variable, the numerical constraint range of the target controlled variable, and the input rate constraint range of the target operated variable, the control increment sequence that minimizes the objective function is solved.

[0012] Optionally, determining the increment information of the target manipulated variable at the current control moment based on the control increment sequence includes: The first increment value in the control increment sequence is used as the increment information of the target operating variable at the current control moment.

[0013] Optionally, it also includes: In response to the update trigger event, exit the current control flow and restart the detection of the state of the target controlled variable.

[0014] Secondly, embodiments of this application provide an electronic device, including: a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to implement the adaptive model predictive control method provided in the first aspect.

[0015] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program that is executed by a processor to perform the adaptive model predictive control method provided in the first aspect.

[0016] The beneficial effects of this application are: This application provides an adaptive model predictive control method, electronic device, and storage medium, comprising: when a target controlled variable in a target industrial process is detected to meet a steady state, applying a pre-generated composite excitation signal to the target manipulated variable corresponding to the target controlled variable, and recording the input data of the target manipulated variable and the response data of the target controlled variable during the application of the composite excitation signal; estimating the parameters of an initial process model based on the input data of the target manipulated variable and the response data of the target controlled variable to obtain a target process model; and controlling the result of the target controlled variable in the target industrial process based on the target process model. This solution integrates composite excitation testing, online process model identification and updating, and adaptive model predictive control through intelligent state management, achieving a fully automatic, high-precision, and robust intelligent control process, improving the automation level, control accuracy, and stability of model predictive control. Specifically, by using a composite excitation signal for excitation testing, high and low frequency dynamic characteristics can be simultaneously excited, achieving full-band coverage, shortening the model identification time, and reducing the difficulty of obtaining the process model. Furthermore, acquiring data and constructing the process model through online excitation makes the process model more adaptable to the scenario and changes in operating conditions, improving the model's control accuracy and performance.

[0017] Secondly, through state switching and management mechanisms, automated control and safety logic management of the entire process can be achieved. This ensures that each functional module executes in sequence, avoids accidental triggering of model identification under non-steady-state or abnormal operating conditions, and guarantees the safety and reliability of system operation.

[0018] By verifying the identified target process model and exiting and re-executing the composite stimulus test and identification process when the verification is unsuitable, it can be ensured that the target process model put into use has sufficient accuracy, preventing the control performance degradation caused by low-precision models.

[0019] In the process of model predictive control, by introducing an adaptive feedback correction mechanism, the mismatch of the target process model and the cumulative error caused by external disturbances can be effectively suppressed, thereby improving the closed-loop robustness.

[0020] In the optimization solution of model predictive control, a hot start strategy is adopted, which uses the optimal control increment sequence of the previous control time as the initial guess of the current optimization problem. Combined with shift initialization to handle the overlapping parts of the control time domain, the convergence speed of the nonlinear optimization algorithm can be significantly accelerated, the real-time computing burden can be reduced, and the needs of high sampling rate industrial applications can be met.

[0021] When an update event such as periodic update, significant disturbance, large change in setpoint, or deterioration of control performance is detected, the steady-state detection and identification process is automatically restarted. This ensures that model predictive control always closely reflects the current operating conditions and enhances the system's long-term adaptive capability.

[0022] The provided human-computer interaction interface allows users to configure parameters, manually trigger identification, and view operation logs, greatly improving the system's operability and the convenience of engineering maintenance. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic diagram of the architecture of an adaptive model predictive control system provided in an embodiment of this application; Figure 2 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 1 ; Figure 3 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 2 ; Figure 4 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 3 ; Figure 5 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 4 ; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0026] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0027] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0028] First, let's explain some of the technical terms that may be involved in this plan: MPC (Model Predictive Control): An advanced control method based on process models for multi-step prediction and optimization.

[0029] FOPDT (First-Order Plus Dead Time): A first-order plus time-delay model, often used as a simplified mathematical model to describe the dynamic characteristics of industrial processes.

[0030] RLS (Recursive Least Squares): A method for online parameter estimation used to dynamically update model parameters.

[0031] FSM (Finite State Machine): A finite state machine used to manage the operating state and state transition logic of a control system.

[0032] CV (Controlled Variable): The controlled variable is the actual measured value of the controlled process, such as temperature, pressure, liquid level, etc.

[0033] SP (Set Point): The setpoint, the desired target value of the controlled variable.

[0034] MV (Manipulated Variable): The manipulated variable, the control quantity output by the controller, used to adjust the controlled variable.

[0035] P (Prediction Horizon): Prediction time domain, the number of time steps MPC predicts for future outputs.

[0036] M (Control Horizon): Control time domain, the number of control increment steps that can be optimized in MPC.

[0037] SQP (Sequential Quadratic Programming): A numerical algorithm for solving nonlinear optimization problems.

[0038] Chirp sweep sine signal: A continuous excitation signal whose frequency changes linearly or logarithmically with time.

[0039] PRBS pseudo-random binary sequence: a random signal used for process excitation.

[0040] Figure 1 This is a schematic diagram of the architecture of an adaptive model predictive control system provided in an embodiment of this application, as shown below. Figure 1 As shown, the system adopts a modular layered design, including a physical device layer, a device access layer, a data service layer, a business logic layer, and an application layer. The physical device layer deploys the controlled devices. The device access layer supports multiple data communication protocols, including but not limited to: Modbus protocol, database interface, Open Platform Communications Unified Architecture (OPC UA) protocol, and HyperText Transfer Protocol (HTTP) / Representational State Transfer (REST) ​​protocol. The data service layer includes: a data preprocessing module, a feature engineering module, a storage management module, and a cache service module. The business logic layer deploys an intelligent state management module, a steady-state judgment module, a composite excitation signal module, a model identification module, and a model prediction and control module. The application layer is used to realize human-computer interaction management, including a data visualization module, an alarm management module, and a system configuration module.

[0041] The various layers interact through data flow and control commands, forming a complete closed-loop system from field data acquisition to intelligent decision-making and control, and finally to visual monitoring. The system uses the intelligent state management module as the core scheduler to achieve state control and module coordination throughout the entire lifecycle.

[0042] The data service layer is used to collect the operating variables (MV) and controlled variables (CV) of the industrial process in real time from the Distributed Control System (DCS) / Programmable Logic Controller (PLC) system through the industrial communication protocol in the device access layer, and upload the collected data to the business logic layer after preprocessing.

[0043] The intelligent state management module acts as the system's central scheduler, responsible for state switching and module coordination to ensure a safe and reliable control process. It primarily drives the system to switch safely and orderly between different states based on preset logic and current operating conditions.

[0044] The intelligent state management module is mainly responsible for the switching and transition of the following five states: First: Stop state (State 0): The system is not running, the controlled variable is held at the preset safe value, and it is waiting for the start command; State transition condition: the control loop switch is turned on.

[0045] Second: Steady-state detection state (State 1): The steady-state judgment module is invoked to continuously monitor whether the controlled variables have reached and maintained a steady state for more than 5 minutes, in preparation for applying subsequent excitation signals. State transition condition: All controlled variables have reached steady state after more than 5 minutes.

[0046] Third: Composite Excitation Test State (State 2): Once the process reaches steady state, this state is activated, invoking the composite excitation signal module to apply a composite excitation signal of PRBS and Chirp to the manipulated variable, and simultaneously recording the input data of the manipulated variable and the dynamic response data of the controlled variable. State transition condition: The composite excitation signal has been applied.

[0047] Fourth: Online State Identification (State 3): Using the input data of the manipulated variables and the response data of the controlled variables collected during the application of the composite excitation signal, the model identification module is invoked to perform online parameter estimation of the process model and to verify the identified process model. State transition condition: Process model verification passes.

[0048] Fifth: Adaptive Control State (State 4): Invokes the Model Predictive Control module to implement periodic model predictive control based on the validated process model. In this state, the system continuously monitors, and once the periodic update conditions are met or a significant disturbance is detected, a new round of state transition will be triggered, returning to the "steady-state detection" or "compound excitation test" state, realizing periodic or event-triggered updates of the model.

[0049] The application layer provides users with an intuitive operating interface and a comprehensive system view, offering an integrated human-machine interface that supports remote access and operation. It primarily implements the following functions: Real-time data display: Displays the real-time numerical change curves of the controlled variable, setpoint, and manipulated variable in the form of trend charts. Predictive trajectory display: Displays the predicted trajectory of the controlled variable for the next P steps by model predictive control and the corresponding sequence of control increments for the next M steps. Model parameter display: Presents the model parameters and their confidence intervals of the currently used process model in real time. System status monitoring: Clearly displays the current system status (e.g., "adaptive control status," "composite excitation test status"), key performance indicators, and alarm information. Historical data review: Allows users to review historical operation records, identified data, and performance statistics. Parameter configuration interface: Provides a user-friendly interface for engineers to configure controller parameters (e.g., number of steps P, M, weights, etc.), identified parameters, constraints, etc.

[0050] Through the collaborative work of the above five levels, the system forms a complete closed loop of "perception-decision-modeling-control-monitoring". It solves the core pain point of the difficulty in engineering implementation of traditional model predictive control, and truly realizes intelligent control of complex industrial processes with one-click tuning and continuous optimization.

[0051] Figure 2 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 1 The execution subject of this method is a computer device, and the computer device is equipped with the aforementioned... Figure 1 The system shown; such as Figure 2 As shown, the method includes: S101. When the target controlled variable under the target industrial process is detected to meet the steady state, a pre-generated composite excitation signal is applied to the target operating variable corresponding to the target controlled variable, and the input data of the target operating variable and the response data of the target controlled variable are recorded during the application of the composite excitation signal.

[0052] The target industrial process can refer to any changing process in an industrial control scenario. Specifically, an industrial process refers to a physical or chemical change that occurs in an industrial system, and its state can be described by a series of variables (such as temperature, pressure, flow rate, etc.). The goal of the control system is to regulate this process to make it operate stably in the desired working state.

[0053] Taking the temperature and humidity control scenario in a computer room as an example, the industrial process refers to the dynamic process of heat and humidity exchange between the entire precision air conditioning system and the computer room space. This includes: air flow and heat transfer within the computer room; preheating the air with a hot water valve, cooling the air with a cold water valve, and increasing humidity with an electric humidifier; and the impact of external environmental disturbances (such as personnel entering and exiting, and equipment heating) on ​​temperature and humidity. This is a typical process with large time delays, nonlinearity, and multivariate coupling. For example, after adjusting the water valve opening, the temperature will not change immediately, but will take several minutes to be reflected in the sensor readings (i.e., there is a "time lag"), and temperature and humidity also influence each other.

[0054] The controlled variable is the output variable that needs to be monitored and maintained within the target range in an industrial process; it is usually the actual measured value of the system. The controller adjusts the input operating variables to make the controlled variable as close as possible to the desired setpoint.

[0055] Continuing with the above scenario as an example, in this data center control system, the two key controlled variables are: data center temperature and data center humidity. These two variables are collected in real time by high-precision temperature and humidity sensors installed in the data center and input as feedback signals to the MPC controller.

[0056] Manipulated variables are input variables that a controller can directly adjust to influence the controlled process and thus change the behavior of the controlled variable. Manipulated variables typically correspond to the output commands of actuators (such as valves, motors, and heaters).

[0057] Taking the above scenario as an example, the controller's adjustable operating variables include: preheating water valve opening: used for heating and supplying air in winter or during low-temperature periods; cooling water valve opening: used for cooling air in summer or during high-temperature periods; reheating water valve opening: used for finely adjusting the outlet air temperature to avoid overcooling; electric humidifier power / start / stop control: used to supplement air humidity, etc.

[0058] These manipulated variables are calculated based on the controller and sent to the DCS / PLC system via a communication protocol to drive the corresponding field actuators.

[0059] The setpoint is the target value specified by the operator or the higher-level system for the controlled variable. The task of the controller is to adjust the controlled variable so that it approaches and remains near the target value as quickly and smoothly as possible.

[0060] The MPC controller can predict the temperature and humidity change trends of future P steps (e.g., 30 minutes) based on the currently identified process model, and solve for the optimal operating variable adjustment sequence (i.e., control increment sequence) to ensure that the controlled variable tracks the set value quickly and without overshoot while meeting the valve action rate constraint.

[0061] Before the system starts running, it is in a shutdown state, and the output values ​​of all controlled variables are kept at safe values. When the control loop switch is detected to be turned on, the system enters the steady-state detection state and can continuously detect the target controlled variable. When the target controlled variable is detected to meet the steady state, it enters the composite excitation test state. At this time, a pre-generated composite excitation signal can be applied to the target operated variable, and the input data of the target operated variable and the response data of the target controlled variable are recorded simultaneously.

[0062] In a target industrial process, the target controlled variable is not a single type of controlled variable. For example, the target controlled variable can include both temperature and humidity. Similarly, the target manipulated variable is not a single type of manipulated variable. It corresponds to the controlled variable, and the manipulated variable may be different for each type of controlled variable.

[0063] In this embodiment, the target controlled variable can be determined according to the control requirements of the target industrial process, and the target controlled variable also determines the target operated variable. For example, when the target controlled variable is the machine room temperature, the target operated variable is the opening degree of the water valve; when the target controlled variable is the machine room humidity, the target operated variable is the power of the humidifier, etc.

[0064] Under the combined excitation test state, the system actively and purposefully changes the target manipulated variable by applying a combined excitation signal to stimulate the dynamic characteristics of the system. During normal operation, the target manipulated variable changes smoothly according to the set value. However, in this stage, the target manipulated variable will fluctuate or jump in a specific pattern according to the preset combined excitation signal, thereby breaking the current equilibrium state and forcing the target controlled variable (such as temperature and liquid level) to produce a dynamic response.

[0065] In other words, a specially designed composite excitation signal rich in frequency information is applied as a disturbance signal to the target manipulated variable. This actively changing target manipulated variable, as the system input, will induce a change in the target controlled variable. By synchronously recording the data of this pair of "changing target manipulated variable" and "correspondingly changing target controlled variable," the system can identify the dynamic relationship between the target controlled variable and the target manipulated variable online, thereby constructing a process model.

[0066] S102. Based on the input data of the target operated variable and the response data of the target controlled variable, perform parameter estimation on the initial process model to obtain the target process model.

[0067] Optionally, based on the input data of the target operational variable and the response data of the target controlled variable obtained above, the model parameters of the process model can be estimated online to obtain the target process model.

[0068] It is worth noting that the input data of the target manipulated variable and the response data of the target controlled variable can exist in pairs. During the application of the composite excitation signal, the data of the target manipulated variable may change continuously. Under each input data of the target manipulated variable, the target controlled variable will generate a corresponding response data, thus forming a data pair. Through multiple sets of data pairs, a data set is formed. Based on the data set, the model parameters of the process model can be estimated, and the target process model matching the current target industrial process can be constructed.

[0069] S103. Result control of target controlled variables under target industrial processes based on target process models.

[0070] In this embodiment, the target process model constructed above can be embedded into the model predictive controller, that is, embedded into the MPC controller, to construct a dynamic prediction matrix. In each control cycle, open-loop prediction of the next P steps is performed to predict the output sequence of the target controlled variable in the next P steps. Based on the output sequence of the target controlled variable in the next P steps, the optimal control increment sequence is obtained. The control increment sequence refers to the increment sequence corresponding to the target operated variable. Based on the control increment sequence, the input control of the target operated variable is executed to achieve precise control of the output result of the target controlled variable.

[0071] In summary, the adaptive model predictive control method provided in this embodiment includes: when the target controlled variable under the target industrial process is detected to meet the steady state, applying a pre-generated composite excitation signal to the target manipulated variable corresponding to the target controlled variable, and recording the input data of the target manipulated variable and the response data of the target controlled variable during the application of the composite excitation signal; estimating the parameters of the initial process model based on the input data of the target manipulated variable and the response data of the target controlled variable to obtain the target process model; and controlling the result of the target controlled variable under the target industrial process based on the target process model. This solution integrates composite excitation testing, online process model identification and updating, and adaptive model predictive control through intelligent state management, realizing a fully automatic, high-precision, and robust intelligent control process, improving the automation level, control accuracy, and stability of model predictive control. Specifically, by using composite excitation signals for excitation testing, high- and low-frequency dynamic characteristics can be simultaneously excited, shortening the model identification time and reducing the difficulty of obtaining the process model. Furthermore, acquiring data and constructing the process model through online excitation makes the process model more adaptable to the scenario and changes in operating conditions, improving the model's control accuracy and performance.

[0072] Figure 3 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 2Optionally, in step S101, when the target controlled variable under the target industrial process is detected to satisfy a steady state, a pre-generated composite excitation signal is applied to the target operated variable corresponding to the target controlled variable, including: S201. After the target industrial process starts, the value of the target controlled variable is collected in real time. If the standard deviation of the target controlled variable is less than the preset threshold for a preset duration, it is determined that the target controlled variable has reached a stable state.

[0073] In some embodiments, when entering the steady-state detection state, the value of the target controlled variable can be continuously detected. When the target controlled variable no longer changes significantly for a preset duration (e.g., more than 5 minutes), it is determined that the target controlled variable has reached a stable state, and the system can be considered to be in a dynamic equilibrium state.

[0074] Alternatively, the standard deviation of the target controlled variable can be calculated using the following formula:

[0075] in, It is the standard deviation of the controlled variable CV over a preset time period; N refers to the number of data points within the preset time period; The value of the controlled variable. The average value of the controlled variable.

[0076] If the standard deviation of the target controlled variable is lower than the preset threshold within the preset time period, it can be determined that the target controlled variable has reached a stable state, and the process can be judged to have reached a stable state.

[0077] S202. Apply a composite excitation signal to the target controlled variable corresponding to the target operated variable.

[0078] When the determination process reaches a stable state, a composite excitation signal is applied to the target operation variable.

[0079] Optionally, the composite excitation signal is obtained by weighted combination of pseudo-random binary sequence signal and swept frequency sine signal. The weight coefficients of pseudo-random binary sequence signal and swept frequency sine signal are adaptively configured according to the response speed of the target industrial process. The composite excitation signal is used to detect the correlation between the target operating variable and the target controlled variable in the target industrial process.

[0080] In some embodiments, the pseudo-random binary sequence signal can be generated using the following formula:

[0081] in, Refers to pseudo-random binary sequence signals. The weighting coefficients of the pseudo-random binary sequence signal are s(t), where s(t) represents the state of the m-sequence. At any time t, the output value of the random binary sequence signal comes from the current state bits of a maximum-length sequence (m-sequence) generated by a linear feedback shift register.

[0082] A swept-frequency sinusoidal signal can be generated using the following formula:

[0083] in, A sweep frequency sine wave signal. The weighting coefficients of the sweep frequency sinusoidal signal. The sweep frequency, , Indicates the termination frequency. Indicates the starting frequency. This indicates the duration of application of the composite excitation signal.

[0084] The composite excitation signal is calculated as follows:

[0085] in, Indicates a composite excitation signal. and These are weighting coefficients, which can be dynamically configured based on the process response speed. For fast processes, It can be configured to a higher level, which is beneficial for slow processes. It can be configured to a lower level.

[0086] Pseudo-random binary sequence signals have white noise characteristics, wide bandwidth coverage, and short test time, and are mainly used to excite high-frequency responses. On the other hand, swept sinusoidal signals have smooth frequency changes, uniform energy distribution, and strong anti-interference ability, and are mainly used to excite low-frequency responses. The composite excitation signal formed by the combination of the two can achieve full-band coverage.

[0087] The selection criteria for combining the two signals are: frequency band complementarity: the pseudo-random binary sequence signal covers the 0.1-10Hz high-frequency band, while the swept-frequency sinusoidal signal covers the 0.01-1Hz low-frequency band; engineering practicality: the pseudo-random binary sequence signal is simple to implement, and the swept-frequency sinusoidal signal has strong resistance to process noise; parameter adjustability: different process characteristics can be adapted by adjusting the weighting coefficients of the two signals; testing efficiency: compared to a single signal, the testing time is reduced by 40-60%.

[0088] Figure 4 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 3Optionally, in step S102, based on the input data of the target operated variable and the response data of the target controlled variable, parameter estimation is performed on the initial process model to obtain the target process model, including: S301. Based on the input data of the target operated variable and the response data of the target controlled variable, the recursive least squares method is used to perform preliminary parameter estimation of the initial process model to obtain the initial gain and the initial time constant.

[0089] After the composite excitation signal is applied, the system can enter the online identification state. The model identification module can be called to automatically complete the online identification and verification of the process model based on the input data of the target operated variable and the response data of the target controlled variable. The online identification of the process model can also be understood as the dynamic and automated establishment and updating of the process model.

[0090] In this embodiment, a two-stage identification strategy can be adopted. In the first stage, the input data of the target manipulated variable and the response data of the target controlled variable are used to perform preliminary estimation of the model parameters using the recursive least squares algorithm, and the initial gain and initial time constant are quickly fitted. The purpose of the preliminary estimation in the first stage is to obtain reasonable initial values ​​to accelerate the convergence of subsequent optimization.

[0091] Optionally, the process model used in this embodiment can be a first-order time-delay model, FOPDT. This model has a simple structure, clear physical meaning, is suitable for understanding and debugging in industrial scenarios, and is sufficient to describe most slowly changing processes. It can be converted into step response coefficients for implementing MPC control.

[0092] The continuous-time transfer function of the process model is:

[0093] in, Indicates gain. Represents the time constant. This represents the time delay parameter.

[0094] To estimate the model parameters using the recursive least squares algorithm, it needs to be transformed into a difference equation (discrete-time model). Assume the sampling period is... Ignoring the time delay parameter (or through data alignment), its approximate discrete form can be written as:

[0095] in, , .

[0096] Therefore, as long as we estimate and From this, we can deduce: ,

[0097] Therefore, the recursive least squares algorithm actually first estimates the coefficients in the linear difference model. and Then convert them into parameters of the process model. and .

[0098] The recursive least squares algorithm requires the process model to be written in linear regression form:

[0099] For the above first-order difference model:

[0100] Therefore, the regression vector is: The parameter vector to be estimated .

[0101] Here It is the target operation variable. It is the target controlled variable, so It consists of the input data of the target operational variable and the response data of the target controlled variable.

[0102] The formula for the recursive least squares algorithm is as follows:

[0103]

[0104]

[0105] in, It is the first Step on parameters The estimate, It is the response data of the controlled variable of the target at the current moment. It consists of the input data of the target manipulated variable and the response data of the target controlled variable from the previous moment. It is a forgetting factor that assigns higher weight to new data and is suitable for time-varying or non-stationary systems.

[0106] initialization (such as the zero vector) (A large diagonal matrix, representing high uncertainty), at each sampling step, constructing a new matrix using the input data of the current target variable and the response data of the target controlled variable. Calculate the prediction error Update parameter estimates Repeat this process until the parameters converge. Once the recursive least squares algorithm converges, a stable result can be obtained. and It can be calculated and .

[0107] S302. Using the initial gain and initial time constant as initial parameters, the initial process model is optimized by using the sequential least squares programming algorithm to determine the target gain, target time constant, and target time delay parameters corresponding to the initial process model.

[0108] The second stage, also known as the nonlinear optimization and refinement stage, uses the initial gain and initial time constant estimated in the first stage as initial parameters. It employs a sequential least squares programming algorithm to solve the problem, aiming to minimize the mean square error between the predicted output and the measured output. This optimizes the three parameters of the process model, resulting in the target gain, target time constant, and target time delay parameter of the process model.

[0109] The formula for solving the sequence least squares programming algorithm is as follows:

[0110] in, This represents the mean square error between the predicted output and the measured output. Indicates gain. Represents the time constant. Represents the time delay parameter. This represents the predicted output (i.e., the numerical value of the target controlled variable predicted by the process model). This represents the measured output (i.e., the response data of the target controlled variable collected during the application of the composite excitation signal). This represents the regularization coefficient, used to prevent overfitting or numerical instability.

[0111] S303. The target gain, target time constant, and target time delay parameter are used as model parameters of the initial process model to obtain the target process model.

[0112] By solving the above problems, we can obtain the target gain, target time constant, and target time delay parameters that make the model optimal, thereby constructing the target process model.

[0113] In some embodiments, the constructed target process model also needs to be validated, which can be done by calculating the goodness of fit between the predicted output and the measured output of the process model. ,like If the target process model is deemed valid, it will be reverted to the composite stimulus test stage, and data will be collected again for model parameter estimation.

[0114] Goodness of fit:

[0115] in, Indicates the predicted output. This represents the measured output. This indicates the goodness of fit.

[0116] Once the model is validated, it can enter the adaptive control state, where model predictive control can be executed based on the target process model.

[0117] If the model verification fails, the system can return to the composite stimulus test state, reapply a new composite stimulus signal, re-acquire modeling data, and recreate the process model.

[0118] Optionally, in step S103, the result control of the target controlled variable under the target industrial process based on the target process model includes: periodically predicting the incremental information of the target operating variable based on the target process model, and adjusting the current input data of the target operating variable according to the prediction results, so as to achieve result control of the target controlled variable.

[0119] In some embodiments, multi-step predictive and optimization control can be performed based on the target process model. In each control cycle / control moment, the optimal control increment sequence (incremental information of the target operated variable) is predicted. Based on the prediction results, the input data of the target operated variable at the current control moment can be adjusted to achieve result control of the target controlled variable.

[0120] Among them, based on the optimal control increment sequence, the input data of the target operated variable can be precisely controlled, so that the response result of the target controlled variable can approach the set value as quickly as possible.

[0121] Figure 5 A flowchart illustrating the adaptive model predictive control method provided in the embodiments of this application. Figure 4 Optionally, in the above steps, periodically predicting the incremental information of the target operational variable based on the target process model includes: S401. The target process model performs multi-step open-loop prediction based on the actual measured value of the target controlled variable at the current control time and the historical increment sequence of the target operated variable, generating a predicted sequence of the target controlled variable for the next P steps corresponding to the current control time.

[0122] It should be noted that predicting the open-loop output of the next P steps is the core step in model predictive control. Without applying new control increments, it uses the target process model to predict the natural evolution trend of the target controlled variable over the next P time steps based solely on the current system state and historical inputs.

[0123] In some embodiments, the open-loop output for the next P steps can be predicted using the step response matrix based on the historical incremental sequence and the actual measured value of the target controlled variable at the current control moment, as shown in the following formula:

[0124] in, This represents the predicted output vector (i.e., the predicted sequence of the target controlled variable in the next P steps). The dynamic matrix is ​​represented by step response coefficients, and the identified target process model can directly generate the dynamic matrix. Indicates the control increment sequence. This indicates a free response, meaning that no new adjustment actions will be applied from the current control moment.

[0125] Based on the above formula, substitute the historical increment sequence into... It can solve for the predicted sequence of the target controlled variable for the next P steps corresponding to the current control time. .

[0126] S402. Based on the actual measured value of the target controlled variable at the current control time and the predicted value of the target controlled variable at the previous control time, the predicted sequence of the target controlled variable is compensated to obtain the compensated predicted sequence of the target controlled variable.

[0127] Optionally, the compensated predicted sequence of the target controlled variable and the control increment sequence corresponding to the previous control time can be substituted into the objective function. Based on the numerical constraint range of the target operated variable, the numerical constraint range of the target controlled variable, and the input rate constraint range of the target operated variable, the control increment sequence that minimizes the objective function can be solved.

[0128] This embodiment introduces a feedback correction mechanism, which can use the actual measured value of the target controlled variable at the current control moment and the first predicted value of the predicted sequence of the target controlled variable for the next P steps predicted at the previous control moment to perform constant compensation on the predicted sequence of the target controlled variable obtained at the current control moment. The calculation formula is as follows:

[0129] in, This represents the compensated predicted sequence of the target controlled variable. Indicates the prediction error. ,in, This represents the actual measured value of the target controlled variable at the current control moment. This represents the first predicted value in the prediction sequence of the target controlled variable for the next P steps predicted at the previous control time.

[0130] Based on this, the compensated predicted sequence of the target controlled variable for the next P steps corresponding to the current control time can be calculated.

[0131] Therefore, the compensated predicted sequence of the target controlled variable can be expressed as: ,in, Let be a vector consisting entirely of 1s. = ,but .

[0132] S403. Based on the compensated predicted sequence of the target controlled variable, the control increment sequence corresponding to the previous control time, and the constructed objective function, solve for the control increment sequence that minimizes the objective function; the objective function is a function related to the output error weight, control increment weight, setpoint trajectory, and control increment sequence.

[0133] Next, construct the objective function:

[0134] in, Indicates the output error weight. This indicates the control of incremental weights. This indicates the trajectory of the set value.

[0135] Substitution Then it can be written in vector form: +

[0136] in, , indicating the trajectory of the set value; , represents the output error weight matrix; , representing the control increment weight matrix.

[0137] When expanded, this is a quadratic function of ΔU:

[0138] in, , , This is a constant term.

[0139] Constraint handling: Input constraints (numerical constraint range of the target operation variable): Input rate constraint (range of data change rate constraint for the target operand): Output constraints (numerical constraint range of the target controlled variable): .

[0140] All constraints are expressed as about Inequalities: Input constraints: ,because , , .in, (forward (+1 data points are 1), the whole can be converted to: .

[0141] Input rate constraints: Directly to Component constraints: ,

[0142] Right now , .

[0143] Output constraints: Substitute into the target process model: ,

[0144] Right now: ;

[0145] Start optimization and solution: Provide initial guesses (warm start): Use the optimal solution from the previous control time step. Apply shift initialization strategy:

[0146] That is, discard the first element and pad with zeros at the end (or keep the last value).

[0147] By calling the sequence quadratic programming solver and inputting the objective function, gradient, the aforementioned constraints, and initial values, the optimal control increment sequence can be output. .

[0148] S404. Based on the control increment sequence, determine the increment information of the target manipulated variable at the current control moment.

[0149] Based on the obtained control increment sequence, the incremental information of the target operating variable at the current control moment can be determined from the sequence. Based on the incremental information, the target operating variable can be regulated so that the result of the target controlled variable is triggered to change after the corresponding actuator is controlled to act.

[0150] Optionally, in step S404, determining the increment information of the target operating variable at the current control time based on the control increment sequence includes: using the first increment value in the control increment sequence as the increment information of the target operating variable at the current control time.

[0151] In some embodiments, since the obtained control increment sequence contains the increment values ​​of the target operating variable at M step sizes, only the first element in the sequence is applied as the control quantity when implementing control.

[0152] Assuming the first element is 20, it can be determined that at the current control moment, based on the existing target manipulated variable, the target manipulated variable is increased by 20 and used as input to implement control. Based on the change in the input of the target manipulated variable, the target controlled variable will change synchronously and gradually approach the set value.

[0153] After the current control moment ends, the system waits for the next control moment to arrive. This process is repeated to complete multi-step prediction and optimization control, ultimately enabling the target controlled variable to approach the set value and stabilize near the set value.

[0154] It should be noted that during the execution of optimization control, the collected raw data, such as the collected target controlled variable, can also be preprocessed, including but not limited to: outlier removal, moving average filtering, time-aligned interpolation (linear interpolation), etc., to ensure data quality and improve the accuracy of subsequent optimization control.

[0155] Optionally, this method also includes: in response to an update trigger event, exiting the current control flow and restarting the detection of the state of the target controlled variable.

[0156] In some embodiments, during system operation, it is also possible to continuously monitor whether an update trigger event occurs, including but not limited to: reaching a preset update cycle, detecting a significant external disturbance, a large change in the set value, or a deterioration in the control performance index; if any of the above update trigger events occur, the current control state can be exited, and the steady-state detection and model identification process can be re-executed to achieve periodic or event-driven adaptive updates of the target process model.

[0157] It is worth noting that there are some alternative excitation signals used in this scheme. Pure pseudo-random binary sequence signals, pure frequency-sweeping sine signals, or N-frame go-back (GBN) signals, or multiple sine signals can also be used to achieve excitation under a single signal.

[0158] For parameter estimation of process models, subspace identification methods, neural network identification, or Bayesian methods can also be used. State management mechanisms can also be implemented using Petri nets, behavior trees, or rule-based systems.

[0159] An example of the target industrial process is as follows: The system achieves precise temperature and humidity control in a computer room, reducing temperature fluctuations from 1.2℃ to 0.5℃ and humidity fluctuations from 8% to 3%, with stabilization time decreasing from 15-20 minutes to 8-15 minutes. The core system code is developed using Python.

[0160] Operating variables: preheating water valve, cold water valve, reheating water valve, electric humidifier opening; Controlled variables: computer room temperature, computer room humidity; Control stability: Temperature fluctuation range reduced from ±1.2°C to ±0.5°C; Automation level: Achieve unattended, automated, and optimized operation around the clock.

[0161] In summary, the adaptive model predictive control method provided in this embodiment includes: when the target controlled variable under the target industrial process is detected to meet the steady state, applying a pre-generated composite excitation signal to the target manipulated variable corresponding to the target controlled variable, and recording the input data of the target manipulated variable and the response data of the target controlled variable during the application of the composite excitation signal; estimating the parameters of the initial process model based on the input data of the target manipulated variable and the response data of the target controlled variable to obtain the target process model; and controlling the result of the target controlled variable under the target industrial process based on the target process model. This solution integrates composite excitation testing, online process model identification and updating, and adaptive model predictive control through intelligent state management, realizing a fully automatic, high-precision, and robust intelligent control process, and improving the automation level, control accuracy, and stability of model predictive control. Specifically, by using composite excitation signals for excitation testing, high and low frequency dynamic characteristics can be excited simultaneously to achieve full frequency band coverage, shorten the model identification time, and reduce the difficulty of obtaining the process model. Furthermore, acquiring data and constructing the process model through online excitation makes the process model more adaptable to the scenario and changes in operating conditions, improving the model's control accuracy and performance.

[0162] Secondly, through state switching and management mechanisms, automated control and safety logic management of the entire process can be achieved. This ensures that each functional module executes in sequence, avoids accidental triggering of model identification under non-steady-state or abnormal operating conditions, and guarantees the safety and reliability of system operation.

[0163] By verifying the identified target process model and exiting and re-executing the composite stimulus test and identification process when the verification is unsuitable, it can be ensured that the target process model put into use has sufficient accuracy, preventing the control performance degradation caused by low-precision models.

[0164] In the process of model predictive control, by introducing an adaptive feedback correction mechanism, the mismatch of the target process model and the cumulative error caused by external disturbances can be effectively suppressed, thereby improving the closed-loop robustness.

[0165] In the optimization solution of model predictive control, a hot start strategy is adopted, which uses the optimal control increment sequence of the previous control time as the initial guess of the current optimization problem. Combined with shift initialization to handle the overlapping parts of the control time domain, the convergence speed of the nonlinear optimization algorithm can be significantly accelerated, the real-time computing burden can be reduced, and the needs of high sampling rate industrial applications can be met.

[0166] When an update event such as periodic update, significant disturbance, large change in setpoint, or deterioration of control performance is detected, the steady-state detection and identification process is automatically restarted. This ensures that model predictive control always closely reflects the current operating conditions and enhances the system's long-term adaptive capability.

[0167] The provided human-computer interaction interface allows users to configure parameters, manually trigger identification, and view operation logs, greatly improving the system's operability and the convenience of engineering maintenance.

[0168] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, on which a device can be deployed Figure 1 The system shown.

[0169] The device includes: processor 801 and storage medium 802.

[0170] Storage medium 802 is used to store programs, and processor 801 calls the programs stored in storage medium 802 to execute the above method embodiments. The specific implementation and technical effects are similar, and will not be described in detail here.

[0171] The storage medium 802 stores program code, which, when executed by the processor 801, causes the processor 801 to perform various steps in the adaptive model predictive control method according to various exemplary embodiments of this application as described in the "Exemplary Methods" section above.

[0172] The processor 801 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0173] Storage medium 802, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. The storage medium can include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type storage medium, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage medium, magnetic disk, optical disk, etc. The storage medium is any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, storage medium 802 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0174] Optionally, this application also provides a program product, such as a computer-readable storage medium, including a program that, when executed by a processor, performs the above-described method embodiments.

[0175] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0176] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0177] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0178] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An adaptive model predictive control method, characterized in that, include: When the target controlled variable under the target industrial process is detected to meet the steady state, a pre-generated composite excitation signal is applied to the target operating variable corresponding to the target controlled variable, and the input data of the target operating variable and the response data of the target controlled variable are recorded during the application of the composite excitation signal. Based on the input data of the target operational variable and the response data of the target controlled variable, the parameters of the initial process model are estimated to obtain the target process model; Based on the target process model, the results of the target controlled variables under the target industrial process are controlled.

2. The method according to claim 1, characterized in that, When the target controlled variable in the target industrial process is detected to satisfy a steady state, applying a pre-generated composite excitation signal to the target operated variable corresponding to the target controlled variable includes: After the target industrial process is started, the value of the target controlled variable is collected in real time. If the standard deviation of the target controlled variable is less than a preset threshold for a preset duration, it is determined that the target controlled variable has reached the stable state. The composite excitation signal is applied to the target operated variable corresponding to the target controlled variable.

3. The method according to claim 1, characterized in that, The composite excitation signal is obtained by weighted combination of pseudo-random binary sequence signal and swept frequency sine signal. The weight coefficients of the pseudo-random binary sequence signal and swept frequency sine signal are adaptively configured according to the response speed of the target industrial process. The composite excitation signal is used to detect the correlation between the target operating variable and the target controlled variable in the target industrial process.

4. The method according to claim 1, characterized in that, The step of estimating parameters of the initial process model based on the input data of the target operated variable and the response data of the target controlled variable to obtain the target process model includes: Based on the input data of the target operated variable and the response data of the target controlled variable, the recursive least squares method is used to perform preliminary parameter estimation of the initial process model to obtain the initial gain and the initial time constant. Using the initial gain and initial time constant as initial parameters, the initial process model is optimized by sequential least squares programming algorithm to determine the target gain, target time constant and target time delay parameters corresponding to the initial process model. The target process model is obtained by using the target gain, the target time constant, and the target time delay parameter as model parameters of the initial process model.

5. The method according to claim 1, characterized in that, The result control of the target controlled variable under the target industrial process based on the target process model includes: Based on the target process model, the incremental information of the target operational variable is periodically predicted, and the current input data of the target operational variable is adjusted according to the prediction results, so as to achieve result control of the target controlled variable.

6. The method according to claim 5, characterized in that, The step of periodically predicting incremental information of the target operational variable based on the target process model includes: The target process model performs multi-step open-loop prediction based on the actual measured value of the target controlled variable at the current control moment and the historical increment sequence of the target operated variable, generating a predicted sequence of the target controlled variable for the next P steps corresponding to the current control moment; Based on the actual measured value of the target controlled variable at the current control time and the predicted value of the target controlled variable at the previous control time, the predicted sequence of the target controlled variable is compensated to obtain the compensated predicted sequence of the target controlled variable. Based on the compensated predicted sequence of the target controlled variable, the control increment sequence corresponding to the previous control time, and the constructed objective function, the control increment sequence that minimizes the objective function is solved; the objective function is a function related to the output error weight, control increment weight, setpoint trajectory, and control increment sequence. Based on the control increment sequence, determine the increment information of the target operational variable at the current control moment.

7. The method according to claim 6, characterized in that, The step of solving for the control increment sequence that minimizes the objective function based on the compensated predicted sequence of the target controlled variable, the control increment sequence corresponding to the previous control time, and the constructed objective function includes: Substituting the compensated predicted sequence of the target controlled variable and the control increment sequence corresponding to the previous control time into the objective function, and based on the numerical constraint range of the target operated variable, the numerical constraint range of the target controlled variable, and the input rate constraint range of the target operated variable, the control increment sequence that minimizes the objective function is solved.

8. The method according to claim 6, characterized in that, Determining the increment information of the target manipulated variable at the current control moment based on the control increment sequence includes: The first increment value in the control increment sequence is used as the increment information of the target operating variable at the current control moment.

9. The method according to claim 1, characterized in that, Also includes: In response to the update trigger event, exit the current control flow and restart the detection of the state of the target controlled variable.

10. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores program instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the program instructions to implement the method as described in any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that is executed by a processor to implement the method as described in any one of claims 1 to 9.

Images