Modular intelligent control method and device based on hydraulic cylinder system and medium

Abstract

The application provides a modular intelligent control method and device based on a hydraulic cylinder system, equipment and medium, collects hydraulic cylinder system operation data, extracts current working condition characteristics through analysis, identifies the working scene of the hydraulic cylinder system, collects historical experience data and system design parameters, decouples the control task of the hydraulic cylinder system into multiple independent algorithm units, each unit is built-in with a basic control law for a specific scene, forms a configurable modular algorithm library, dynamically selects and combines corresponding algorithm units from the modular algorithm library according to the working scene of the hydraulic cylinder system, generates the final control instruction adapted to the scene through a neural network, and realizes scene adaptive control; the application realizes adaptive matching of complex working conditions through online scene identification and neural network dynamic weight distribution, and achieves tracking accuracy improvement, overshoot reduction and performance stable operation in the whole life cycle in combination with a parameter self-tuning mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control, and particularly to a modular intelligent control method, device and medium based on a hydraulic cylinder system. Background Art

[0002] In recent years, the development speed of China's hydraulic industry has been relatively fast and the industrial scale is relatively large. However, compared with foreign enterprises, the progress of key supporting parts is relatively slow. In the cylinder industry, the product performance generally fails to meet the requirements of the main engine's operating performance, and there is a lack of technical data in the product trial production and test stages. There is a lack of experimental research related to reliability and lifespan; especially for cylinders with high pressure, high temperature, large pressure shocks and corrosion resistance, the domestic products have a greater gap, such as the cylinders supporting mining machinery such as hydraulic supports and coal mining machines in fully mechanized coal mining faces. Cylinder products have problems such as short lifespan, poor sealing effect, and poor low-speed stability; the product performance is single and the expandability is less.

[0003] At present, in order to meet the precise control requirements of mining machinery, the requirements for the design optimization and processing accuracy of cylinders are getting higher and higher. To sum up, it is very important to carry out the design research work of high-performance integrated cylinders. Foreign scholars comprehensively use genetic algorithms, discrete sliding mode control, particle swarm algorithms, nonlinear dynamic effects, etc. to optimize and adjust the PID control parameters in electro-hydraulic control systems. Although domestic research and development of electro-hydraulic control systems with multiple actuators have started, there is still a certain gap in control accuracy and automation level compared with foreign countries. In addition, although domestic and foreign scholars have gradually increased the research on the cooperative control of multi-actuator systems, traditional disturbance observers often take single-actuator systems as the research object and are difficult to be directly applied to multi-actuator systems with mutual coupling relationships. In addition, due to cost and installation space limitations, the acquisition of system states has always been a difficult problem to solve. Generally speaking, relevant technologies have made great progress in design methods, processing technologies, controllers, multi-actuator cooperation, etc. in recent years. However, existing research and applications are difficult to meet the precise control requirements of cylinders for mining machinery in complex environments, especially in coal mine environments. Summary of the Invention

[0004] The present invention provides a modular intelligent control method, device and medium based on a hydraulic cylinder system, which realizes the adaptive matching of complex working conditions through online scenario recognition and neural network dynamic weight allocation, and combines a parameter self-tuning mechanism to achieve improved tracking accuracy, reduced overshoot and stable performance operation throughout the life cycle, so as to solve the problems in the background art.

[0005] To achieve the above object, the technical solution of the present invention is as follows: A modular intelligent control method based on a hydraulic cylinder system, which is executed by a computer device through the following steps: S1: Data acquisition and scene recognition. Collect pressure and displacement operation data of the hydraulic cylinder system, and extract the characteristics of the current working scene by analysis to identify the specific scene in which the hydraulic cylinder system is located. S2: Modular algorithm library construction, collecting historical experience data and system design parameters, decoupling the control task of the hydraulic cylinder system into multiple independent algorithm units, each unit has built-in basic control laws for specific scenarios, forming a configurable modular algorithm library; S3: Scenario-based decision arbitration. Based on the specific working scenario of the hydraulic cylinder system, the corresponding control modules are dynamically selected and combined from the modular algorithm library. The final control command adapted to the scenario is generated through the neural network to achieve scenario-adaptive control. S4: Parameter self-tuning optimization. Based on the error feedback and system state changes after the final control command is executed, the internal parameters of each algorithm module are automatically adjusted, and the optimized parameters are updated back to the algorithm library.

[0006] Preferably, the specific process steps of S1 are as follows: S11, multi-source synchronous acquisition, synchronously acquires cylinder displacement and dual-chamber pressure data, as well as historical output data of hydraulic cylinder system control, including the servo valve core displacement command of the previous cycle in the hydraulic cylinder system, and completes multi-sensor timestamp alignment and hardware self-test. S12, Data preprocessing: Kalman filtering is used to suppress noise and remove outliers; the position tracking error is calculated from the cylinder displacement acquired by the given target displacement command, and the error rate of change is obtained by differentiating the position tracking error with respect to time; the real-time piston speed is obtained by differentiating the acquired cylinder displacement. S13, Scene Recognition: Based on a sample library of typical hydraulic working conditions, a lightweight LightGBM classification model is trained offline and used for online inference to identify the current working scene of the hydraulic cylinder system. The posterior probability of the current working scene corresponding to each standard scene is calculated as the confidence level for identifying the current working scene, with the value range limited to 0-1.

[0007] Preferably, the specific process steps of S2 are as follows: S21, Control Task Decoupling: Based on the control requirements of hydraulic cylinders, the hydraulic cylinder system is decoupled into multiple standardized algorithm units with unified input / output interfaces. S22 is designed and built with exclusive basic control laws for each type of scenario. The basic control laws include PID, feedforward compensation, sliding mode control, fuzzy control and impedance control. Each algorithm unit adopts a standardized input and output interface, with unified input target value and feedback value; and unified output servo valve core displacement control quantity, forming a set of customizable and configurable modular algorithm library. S23 establishes a mapping rule base from working scenarios to algorithm units, stores the historical optimal parameters, interface specifications, and physical constraint boundaries of each standardized algorithm unit, and supports online configuration and invocation of the controller.

[0008] Preferably, the process of generating the final control command adapted to the scenario through a neural network in S3 is as follows; Suppose k algorithm units are selected, and each algorithm unit outputs an independent control quantity. ,u represents the control command, which corresponds to the servo valve core displacement command in the hydraulic cylinder system, and t represents time; Solving for the optimal weight vector matrix ,satisfy 'i' represents the i-th module, which enables the final control command. :

[0009] in, This indicates the final control command, specifically the final servo valve spool displacement command. represents the weight vector matrix, * denotes the transpose of the matrix, and w represents the weight vector.

[0010] Preferably, the weight vector matrix The specific calculation process is as follows: The weight vector matrix is ​​calculated using an RBF neural network with specific constraints introduced for hydraulic scenarios. The architecture is as follows: Input layer, 5-dimensional input vector:

[0011] in: For position tracking error, The rate of change of error, For real-time load pressure, For the real-time speed of the piston, This represents the rate of change of the control quantity in the previous cycle. Hidden layer: Assume m RBF neurons, with Gaussian radial basis functions.

[0012] in: Let j be the center vector of the j-th neuron. Let be the base width of the j-th neuron. For the Euclidean norm, This represents the output vector of the j-th neuron. Output layer: Employs the Softmax activation function, satisfying the constraint that the sum of the weights is 1.

[0013] in: Let be the output layer weight vector of the hidden layer corresponding to the i-th module. This is the output vector of the hidden layer.

[0014] Preferably, after S3 generates the final control command, it also includes introducing a scenario transition coefficient for transitional operating conditions. Smooth out control commands:

[0015] Among them, the scene transition coefficient It is dynamically determined by a linear limiting rule that takes the ratio of confidence level to a preset reliability threshold as input, and minimizes it. This is used to balance the control response speed and shock resistance stability during hydraulic system condition switching. T represents the fixed control cycle length, t... T represents the previous control time. This indicates the final control command that will be output at the current time t. It is the final control command that was output to the servo valve in the previous moment.

[0016] Preferably, the specific process steps of S4 are as follows: S41, based on the feedback data after the execution of control commands, constructs a comprehensive performance evaluation function with tracking error, control energy consumption, and system stability as the core, and quantifies the control effect; S41 adopts an adaptive particle swarm optimization algorithm, with the performance evaluation function as the optimization target, and iteratively adjusts the core control parameters of each algorithm module online to adapt to the time-varying working conditions and characteristic drift of the hydraulic system. S41 synchronizes the optimized parameters to the algorithm library, and continuously iterates and optimizes the scene recognition and RBF neural network model offline based on long-term running data, thereby enhancing control capabilities.

[0017] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0018] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0019] As can be seen from the above technical solution compared with the prior art, the present invention has the following beneficial effects: This invention achieves adaptive matching of control strategies to complex operating conditions of hydraulic cylinder systems through online scene recognition and dynamic weight allocation of neural networks, resulting in optimized control effects of improved tracking accuracy and reduced overshoot. By constructing a modular algorithm library and introducing a parameter self-tuning mechanism, it realizes online optimization of control parameters and iterative accumulation of knowledge, thereby achieving performance maintenance and robust operation of the system throughout its entire lifecycle. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the method steps of the present invention; Figure 2 This is a schematic diagram of the specific calculation process of S3 in this invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0022] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention, but should not be used to limit the scope of the present invention.

[0023] This invention provides a modular intelligent control method based on a hydraulic cylinder system, such as... Figure 1 As shown, perform the following steps using a computer device: S1: Data acquisition and scene recognition. Collect pressure and displacement operation data of the hydraulic cylinder system, and extract the characteristics of the current working scene by analysis to identify the specific scene in which the hydraulic cylinder system is located. S2: Modular algorithm library construction, collecting historical experience data and system design parameters, decoupling the control task of the hydraulic cylinder system into multiple independent algorithm units, each unit has built-in basic control laws for specific scenarios, forming a configurable modular algorithm library; S3: Scenario-based decision arbitration. Based on the specific working scenario of the hydraulic cylinder system, the corresponding control modules are dynamically selected and combined from the modular algorithm library. The final control command adapted to the scenario is generated through the neural network to achieve scenario-adaptive control. S4: Parameter self-tuning optimization. Based on the error feedback and system state changes after the final control command is executed, the internal parameters of each algorithm module are automatically adjusted, and the optimized parameters are updated back to the algorithm library.

[0024] Example: like Figure 2As shown in this embodiment, the hydraulic machinery used in a large-scale project is used as the application object. The working environment of this equipment is harsh, the load changes drastically, and the requirements for control precision and response speed are extremely high.

[0025] Traditional control methods face challenges. Fixed PID parameters are ill-suited to the aforementioned variable load scenarios. During heavy-load excavation, slow speeds or large impacts are common, while during no-load reset, excessively rapid response can cause swaying. Sudden disturbances can even lead to jitter or boom drop, severely impacting work efficiency and operator comfort. Therefore, this application introduces a modular intelligent control method based on a hydraulic cylinder system.

[0026] The specific implementation is as follows: Real-time displacement and dual-chamber pressure data of the hydraulic cylinder are synchronously acquired at a frequency of 1kHz. The servo valve core displacement command output in the previous cycle is also synchronously acquired to complete the multi-sensor timestamp alignment and hardware self-test. Pressure pulsation noise suppression and outlier removal are completed through Kalman filtering. The position tracking error, error change rate, piston running speed, real-time load pressure, and change rate of control quantity in the previous cycle of each cylinder are calculated in real time to extract the working condition feature vector.

[0027] Based on the pre-trained offline LightGBM scene classification model (with a labeled sample library covering 6 typical working conditions), the controller completes an online scene inference every 10ms, outputting the scene type and confidence value (0~1) corresponding to the current working condition, providing core input for subsequent decision-making.

[0028] To address the control requirements of this lifting system, control tasks were decoupled beforehand, and a standardized, modular algorithm library was constructed. All modules share a unified input / output interface: inputs are target values ​​and real-time feedback values, and outputs are servo valve spool displacement control values. Specific modules are divided into three categories: Basic control law modules: Adaptive PID module, sliding mode robust control module, cross-coupled synchronous control module, fuzzy PID voltage holding module, and feedforward compensation module; Hydraulic characteristic compensation modules: servo valve dead zone compensation module, friction compensation module, cylinder leakage compensation module; Safety constraint modules include a stroke / pressure limiting module and an emergency stop safety module. A scenario-module mapping rule base is also established to define the module combination corresponding to each type of operating condition: for example, the synchronous lifting condition matches "sliding mode robust control module + cross-coupled synchronous control module + dead zone compensation module," and the steady-state pressure holding condition matches "fuzzy PID pressure holding module + leakage compensation module." The historical optimal parameters and physical constraint boundaries of each module are stored, supporting online configuration and invocation by the controller.

[0029] Then, based on the specific working scenario of the hydraulic cylinder system, corresponding control modules are dynamically selected and combined from the modular algorithm library. A neural network is then used to generate final control commands adapted to the scenario, achieving scenario-adaptive control, such as... Figure 2 As shown; Suppose k algorithm units are selected, and each algorithm unit outputs an independent control quantity. ,u represents the control command, which corresponds to the servo valve core displacement command in the hydraulic cylinder system, and t represents time; Solving for the optimal weight vector matrix ,satisfy 'i' represents the i-th module, which enables the final control command. :

[0030] in, This indicates the final control command, specifically the final servo valve spool displacement command. represents the weight vector matrix, * denotes the transpose of the matrix, and w represents the weight vector.

[0031] Weight vector matrix The specific calculation process is as follows: The weight vector matrix is ​​calculated using an RBF neural network with specific constraints introduced for hydraulic scenarios. The architecture is as follows: Input layer, 5-dimensional input vector:

[0032] in: For position tracking error, The rate of change of error, For real-time load pressure, For the real-time speed of the piston, This represents the rate of change of the control quantity in the previous cycle. Hidden layer: Assume m RBF neurons, with Gaussian radial basis functions.

[0033] in: Let j be the center vector of the j-th neuron. Let be the base width of the j-th neuron. For the Euclidean norm, This represents the output vector of the j-th neuron. Output layer: Employs the Softmax activation function, satisfying the constraint that the sum of the weights is 1.

[0034] in: Let be the output layer weight vector of the hidden layer corresponding to the i-th module. This is the output vector of the hidden layer.

[0035] After generating the final control command, the process also includes introducing a scenario transition coefficient for transitional operating conditions. Smooth out control commands:

[0036] Among them, the scene transition coefficient It is dynamically determined by a linear limiting rule that takes the ratio of confidence level to a preset reliability threshold as input, and minimizes it. This is used to balance the control response speed and shock resistance stability during hydraulic system condition switching. T represents the fixed control cycle length, t... T represents the previous control time. This indicates the final control command that will be output at the current time t. It is the final control command that was output to the servo valve in the previous moment.

[0037] Finally, throughout the entire operation process, the controller collects feedback data in real time after the control commands are executed, and constructs a comprehensive performance evaluation function with synchronous tracking error, control quantity fluctuation, and system pressure stability as the core. The adaptive particle swarm optimization algorithm is adopted, with the evaluation function as the optimization target, and the core parameters of each algorithm module are adjusted online iteratively, such as the proportional / integral / derivative coefficients of PID, the switching gain of sliding mode control, and the weights of RBF neural network. The optimized parameters are synchronously updated to the modular algorithm library. After the operation is completed, the scene recognition model and RBF neural network model are iteratively optimized offline based on the full process operation data to achieve continuous iteration of control capabilities.

[0038] This invention enables closed-loop intelligent control of hydraulic presses under extreme and variable working conditions, achieving perception, decision-making, execution, and optimization. It realizes adaptive matching of control strategies to complex working conditions of hydraulic cylinder systems, achieving optimized control effects of improved tracking accuracy and reduced overshoot.

[0039] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0040] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0041] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the modular intelligent control methods based on a hydraulic cylinder system in the above embodiments.

[0042] It is understood that the systems, devices, and storage media provided in the embodiments of the present invention correspond to the methods provided in the embodiments of the present invention, and the explanations, examples, and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0043] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0044] For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media.

[0045] The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid state disks (SSDs)).

[0046] It should be noted that in this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0047] Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0048] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0049] The embodiments of the present invention are given for the purposes of illustration and description. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A modular intelligent control method based on hydraulic cylinder system, characterized in that, Perform the following steps using a computer device: S1 collects pressure and displacement operation data of the hydraulic cylinder system, and analyzes and extracts the current working condition characteristics to identify the specific working scenario of the hydraulic cylinder system. S2 collects historical experience data and system design parameters, decouples the control task of the hydraulic cylinder system into multiple independent algorithm units, each of which has a built-in basic control law for a specific scenario, forming a configurable modular algorithm library. S3, based on the specific working scenario of the hydraulic cylinder system, dynamically selects and combines corresponding algorithm units from the modular algorithm library, and generates the final control command adapted to the scenario through the neural network to achieve scenario adaptive control; S4 adjusts the internal parameters of each algorithm module based on the error feedback and system state changes after the final control command is executed, and updates the optimized parameters back to the algorithm library.

2. The modular intelligent control method based on hydraulic cylinder system as claimed in claim 1, wherein: The specific process steps of S1 are as follows: S11, multi-source synchronous acquisition, synchronously acquires cylinder displacement and dual-chamber pressure data, as well as historical output data of hydraulic cylinder system control, including the servo valve core displacement command of the previous cycle in the hydraulic cylinder system, and completes multi-sensor timestamp alignment and hardware self-test. S12, Data preprocessing, using the Kalman filter criterion to complete noise suppression and outlier removal; The position tracking error is calculated by acquiring the cylinder displacement based on the given target displacement command, and the error rate of change is obtained by differentiating the position tracking error with respect to time. The real-time piston speed is obtained by differentiating the acquired cylinder displacement. S13, Scene Recognition: Based on a sample library of typical hydraulic working conditions, a lightweight LightGBM classification model is trained offline and used for online inference to identify the current working scene of the hydraulic cylinder system. The posterior probability of the current working scene corresponding to each standard scene is calculated as the confidence level for identifying the current working scene, with the value range limited to 0-1.

3. The modular intelligent control method based on a hydraulic cylinder system as described in claim 1, characterized in that: The specific process steps of S2 are as follows: S21, Control Task Decoupling: Based on the control requirements of hydraulic cylinders, the hydraulic cylinder system is decoupled into multiple standardized algorithm units with unified input / output interfaces. S22 is designed and built with exclusive basic control laws for each type of scenario. The basic control laws include PID, feedforward compensation, sliding mode control, fuzzy control and impedance control. Each algorithm unit adopts a standardized input and output interface, with unified input target value and feedback value; and unified output servo valve core displacement control quantity, forming a set of customizable and configurable modular algorithm library. S23 establishes a mapping rule base from working scenarios to algorithm units, stores the historical optimal parameters, interface specifications, and physical constraint boundaries of each standardized algorithm unit, and supports online configuration and invocation of the controller.

4. The modular intelligent control method based on a hydraulic cylinder system as described in claim 1, characterized in that: The process of generating the final control command adapted to the scenario through the neural network in S3 is as follows; Suppose k algorithm units are selected, and each algorithm unit outputs an independent control quantity. ,u represents the control command, which corresponds to the servo valve core displacement command in the hydraulic cylinder system, and t represents time; Solving for the optimal weight vector matrix ,satisfy , where i represents the i-th module, which enables the final control command. : in, This indicates the final control command, specifically the final servo valve spool displacement command. represents the weight vector matrix, * denotes the transpose of the matrix, and w represents the weight vector.

5. The modular intelligent control method based on a hydraulic cylinder system as described in claim 4, characterized in that: The weight vector matrix The specific calculation process is as follows: The weight vector matrix is ​​calculated using an RBF neural network with specific constraints introduced for hydraulic scenarios. The architecture is as follows: Input layer, 5-dimensional input vector: in: For position tracking error, The rate of change of error, For real-time load pressure, For the real-time speed of the piston, This represents the rate of change of the control quantity in the previous cycle. Hidden layer: Assume m RBF neurons, with Gaussian radial basis functions. in: Let j be the center vector of the j-th neuron. Let be the base width of the j-th neuron. For the Euclidean norm, This represents the output vector of the j-th neuron; Output layer: Employs the Softmax activation function, satisfying the constraint that the sum of the weights is 1. in: Let be the output layer weight vector of the hidden layer corresponding to the i-th module. This is the output vector of the hidden layer.

6. The modular intelligent control method based on a hydraulic cylinder system as described in claim 5, characterized in that: After S3 generates the final control command, it also includes introducing a scenario transition coefficient for transitional operating conditions. Smooth out control commands: Among them, the scene transition coefficient It is dynamically determined by a linear limiting rule that takes the ratio of confidence level to a preset reliability threshold as input, and minimizes it. This is used to balance the control response speed and shock resistance stability during hydraulic system condition switching. T represents the fixed control cycle length, t... T represents the previous control time. This indicates the final control command that will be output at the current time t. It is the final control command that was output to the servo valve in the previous moment.

7. The modular intelligent control method based on a hydraulic cylinder system as described in claim 1, characterized in that: The specific process steps of S4 are as follows: S41, based on the feedback data after the execution of control commands, constructs a comprehensive performance evaluation function with tracking error, control energy consumption, and system stability as the core, and quantifies the control effect; S41 adopts an adaptive particle swarm optimization algorithm, with the performance evaluation function as the optimization target, and iteratively adjusts the core control parameters of each algorithm module online to adapt to the time-varying working conditions and characteristic drift of the hydraulic system. S41 synchronizes the optimized parameters to the algorithm library, and continuously iterates and optimizes the scene recognition and RBF neural network model offline based on long-term running data, thereby enhancing control capabilities.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the computer program is executed by the processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

Images