Electronic cooling control method and device based on large language model and electronic equipment

Abstract

The invention provides an electronic cooling control method and device based on a large language model and electronic equipment, and relates to the technical field of electronic cooling control. The method comprises the following steps of: performing semantic analysis on an electronic cooling natural language instruction input by a user by a main control model through a collaborative architecture of the main control model and a sub-control model of each electronic cooling device; the main control model constructs a physical model for controlling the target electronic cooling equipment according to the semantic analysis result, and determines preliminary control parameters for controlling the target electronic cooling equipment according to the physical model and the current actual operation state of the target electronic cooling equipment; and then the target electronic cooling equipment receives the preliminary control parameters, a sub-control model of the target electronic cooling equipment analyzes the preliminary control parameters, whether the preliminary control parameters are adjusted or not is determined according to an analysis result, and intelligent control, fault prediction and natural language interaction of the electronic cooling equipment are achieved.

Description

Technical Field

[0001] This application relates to the technical field of electronic cooling control, and particularly to an electronic cooling control method and device, and an electronic device based on a large language model. Background Art

[0002] In an accelerator, the electronic cooling control system is a core technology for improving the quality of ion beams. Its core function is to reduce the temperature and momentum dispersion of ion beams through the interaction between low-temperature electron beams and high-temperature ion beams, thereby achieving the stability and focusing of the beam.

[0003] Traditional electronic cooling control systems rely on preset rules and manual experience, and have the following problems: 1) Insufficient adaptability: Unable to dynamically respond to complex working condition changes and difficult to optimize beam parameters; 2) Delayed fault diagnosis: Dependent on manual analysis of logs and difficult to predict equipment anomalies in real time; 3) High interaction complexity: Control instructions require professional programming, with high maintenance costs and prone to errors.

[0004] Existing technologies have tried to introduce traditional machine learning methods, but the model generalization ability is limited and natural language interaction cannot be achieved. Therefore, it is urgent to solve this technical problem. Summary of the Invention

[0005] In view of the above problems, this application is proposed to provide an electronic cooling control method and device, an electronic device, a storage medium, and a computer program product based on a large language model that overcome the above problems or at least partially solve the above problems. The technical solutions are as follows: In a first aspect, an electronic cooling control method based on a large language model is provided, including: Deploy a large language model with a first parameter on the master server in advance, obtain the electronic cooling operation log and electronic cooling theoretical literature as the first sample data set, and perform pre-training, instruction fine-tuning, and reinforcement learning on the large language model with the first parameter to obtain a master model that can understand the physical laws of electronic cooling; Deploy a large language model with a second parameter on each electronic cooling device in advance. For the large language model with the second parameter on each electronic cooling device, obtain the operation log and theoretical literature of the electronic cooling device as the second sample data set, and train the large language model with the second parameter on the electronic cooling device to obtain a slave control model of the electronic cooling device, thereby obtaining the slave control models of each electronic cooling device; In response to the user's input of an electronic cooling natural language instruction, the main control model performs semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result that includes controlling the target electronic cooling device; the main control model constructs a physical model for controlling the target electronic cooling device according to the semantic parsing result, and determines preliminary control parameters for controlling the target electronic cooling device based on the physical model and the current actual operating state of the target electronic cooling device. The target electronic cooling device receives the preliminary control parameters, and the sub-control model of the target electronic cooling device analyzes the preliminary control parameters and determines whether to adjust the preliminary control parameters according to the analysis result.

[0006] In a possible implementation manner, the method further includes: An electronic cooling knowledge base including a theory library and an operation library is pre-constructed, where the theory library includes electronic cooling physical theories, electronic cooling technical literature, and historical experimental data; the operation library includes the operation parameters of each electronic cooling device collected in real time.

[0007] In a possible implementation manner, obtaining the electronic cooling operation log and the electronic cooling theory literature as the first sample data set includes: Obtaining the electronic cooling operation log and the electronic cooling theory literature from the pre-constructed electronic cooling knowledge base as the first sample data set; Obtaining the operation log and the theory literature of the electronic cooling device as the second sample data set includes: Obtaining the operation log and the theory literature of the electronic cooling device from the pre-constructed electronic cooling knowledge base as the second sample data set.

[0008] In a possible implementation manner, the main control model performs semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result that includes controlling the target electronic cooling device, including: The main control model combines the electronic cooling knowledge base to perform semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result that includes controlling the target electronic cooling device.

[0009] In a possible implementation manner, pre-training, instruction fine-tuning, and reinforcement learning are performed on the large language model of the first parameter to obtain a main control model that can understand the physical laws of electronic cooling, including: Taking beam stability, energy consumption efficiency, and fault prediction accuracy as optimization objectives, pre-training, instruction fine-tuning, and reinforcement learning are performed on the large language model of the first parameter to obtain a main control model that can understand the physical laws of electronic cooling.

[0010] In a possible implementation manner, determining whether to adjust the preliminary control parameters according to the analysis result includes: If it is determined according to the analysis result that the preliminary control parameters need to be adjusted, a multi-objective optimization solution is generated. If the adjusted control parameters are obtained through the solution, the target electronic cooling device executes according to the adjusted control parameters. If the adjusted control parameters are not obtained through the solution, a reporting message indicating that the adjusted control parameters are not obtained is generated and reported to the main control model.

[0011] In a possible implementation manner, determining whether to adjust the preliminary control parameters according to the analysis result includes: If it is determined according to the analysis result that the preliminary control parameters do not need to be adjusted, the target electronic cooling device executes according to the preliminary control parameters. During the execution of the target electronic cooling device, the sub-control model analyzes and identifies the execution information of the target electronic cooling device. If it is identified that the target electronic cooling device fails, a preset resolution strategy is executed, and an alarm message indicating that the target electronic cooling device fails is generated and reported to the main control model.

[0012] In a second aspect, an electronic cooling control device based on a large language model is provided for implementing the above-mentioned electronic cooling control method based on a large language model. The device includes: A first deployment unit for pre-deploying a large language model with a first parameter on a main control server, obtaining an electronic cooling operation log and electronic cooling theoretical literature as a first sample data set, and performing pre-training, instruction fine-tuning, and reinforcement learning on the large language model with the first parameter to obtain a main control model capable of understanding the physical laws of electronic cooling; A second deployment unit for pre-deploying a large language model with a second parameter on each electronic cooling device. For the large language model with the second parameter on each electronic cooling device, obtaining the operation log and theoretical literature of the electronic cooling device as a second sample data set, and training the large language model with the second parameter on the electronic cooling device to obtain a sub-control model of the electronic cooling device, thereby obtaining sub-control models of each electronic cooling device; A first control unit for responding to an electronic cooling natural language instruction input by a user, performing semantic parsing on the electronic cooling natural language instruction by the main control model to obtain a semantic parsing result including controlling a target electronic cooling device; the main control model constructs a physical model for controlling the target electronic cooling device according to the semantic parsing result, and determines preliminary control parameters for controlling the target electronic cooling device according to the physical model and the current actual operating state of the target electronic cooling device; A second control unit for the target electronic cooling device to receive the preliminary control parameters, and the sub-control model of the target electronic cooling device analyzes the preliminary control parameters and determines whether to adjust the preliminary control parameters according to the analysis result.

[0013] In a third aspect, an electronic device is provided, which includes a processor and a memory. Among them, a computer program is stored in the memory, and the processor is configured to run the computer program to execute the electronic cooling control method based on a large language model described in any one of the above.

[0014] In a fourth aspect, a storage medium is provided. The storage medium stores a computer program, and the computer program is configured to execute the electronic cooling control method based on a large language model described in any one of the above when running.

[0015] In a fifth aspect, a computer program product is provided, including a computer program, and the computer program is configured to execute the electronic cooling control method based on a large language model described in any one of the above when running.

[0016] By means of the above technical solutions, the electronic cooling control method and device, electronic device, storage medium and computer program product based on a large language model provided by the embodiments of the present application. The electronic cooling control method based on a large language model realizes the intelligent control, fault prediction and natural language interaction of electronic cooling devices through the collaborative architecture of the main control model and the sub-control models of each electronic cooling device, and is particularly suitable for optimizing the beam quality and improving the system stability of electronic cooling. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments of the present application will be briefly introduced below.

[0018] Figure 1 Shows a schematic diagram of the electronic cooling principle provided by the embodiments of the present application; Figure 2 Shows a flowchart of the electronic cooling control method based on a large language model provided by the embodiments of the present application; Figure 3 Shows a main control - sub - control collaborative architecture diagram provided by the embodiments of the present application; Figure 4 Shows a main control model control flowchart provided by the embodiments of the present application; Figure 5 Shows a schematic diagram of the design of the reinforcement learning reward mechanism provided by the embodiments of the present application; Figure 6 Shows a structural diagram of the electronic cooling control device based on a large language model provided by the embodiments of the present application; Figure 7 Shows a structural diagram of the electronic cooling control device based on a large language model provided by another embodiment of the present application; Figure 8 Shows a structural diagram of an electronic device provided by the embodiments of the present application. Detailed implementation manners

[0019] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be fully conveyed to those skilled in the art.

[0020] It should be noted that the terms "first", "second", etc. in the description and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such use can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the term "including" and its variants should be interpreted as open-ended terms meaning "including but not limited to".

[0021] As introduced above, in an accelerator, an electron cooling control system reduces the temperature and momentum dispersion of an ion beam through the interaction between a low-temperature electron beam and a high-temperature ion beam, thereby achieving the stability and focusing of the beam. The electron cooling control system may include a high-voltage platform device for generating a high-voltage electric field, an electron gun for emitting an electron beam, a grid for controlling the turn-off of the electron beam, an anode for extracting the electron beam, a filament for generating electrons, a collector for collecting the electron beam, a deflection plate for deflecting the direction of the electron beam, one or more horizontal correction power supplies for correcting the horizontal error of the electron beam, one or more vertical correction power supplies for correcting the vertical error of the electron beam, a resonant power supply for supplying 20 kHz (kilohertz) alternating current to the high-voltage platform device, an accelerating tube, a decelerating tube, a beam position detector, and other electron cooling devices.

[0022] See Figure 1 , by way of illustration, 1 is an electron gun, 2 is an accelerating tube, 3 is a vacuum chamber, 4 is a cooling section, 5 is a collector, and 6 is a high-voltage platform device. The electron gun 1 can generate an electron beam of 0 - 3 A. After the electron beam is accelerated by the accelerating tube 2, it enters the cooling section 4 through the vacuum chamber 3 and deflection of the deflection plate. In the cooling section 4, it interacts with the ion beam. After cooling the ion beam, it enters the collector 5 through the deflection plate. After the electrons are collected by the collector 5, the electrons are emitted again through the electron gun 1; the high-voltage platform device 6 can extract and accelerate the electron beam to a preset energy through the negative high voltage generated by it. Here, the collector 5 is used to improve the utilization efficiency of the electron beam and realize the secondary utilization of electrons.

[0023] The embodiment of the present application provides an electronic cooling control method based on a large language model. Here, the large language model (Large Language Model, LLM), first, as the name implies, is large in scale, with the number of network parameters reaching tens of billions, hundreds of billions or even more; second, general, which means not limited to specific problems or fields; third, emergence, that is, the emergence of unexpected new capabilities.

[0024] As Figure 2 shown, the electronic cooling control method based on the large language model may include the following steps S201 to S204: Step S201, deploy a large language model with the first parameter on the main control server in advance, obtain the electronic cooling operation log and electronic cooling theoretical literature as the first sample data set, and perform pre-training, instruction fine-tuning and reinforcement learning on the large language model with the first parameter to obtain a main control model that can understand the physical laws of electronic cooling.

[0025] In this step, the large language model can be an open-source large language model. As the model base, there can be different parameter versions, such as 1.5 billion, 7 billion, 14 billion, 32 billion, 671 billion and other parameters. The selection of the parameter scale can be combined with computing power resources and performance requirements, etc. This embodiment does not limit this.

[0026] Step S202, deploy a large language model with the second parameter on each electronic cooling device in advance. For the large language model with the second parameter on each electronic cooling device, obtain the operation log and theoretical literature of this electronic cooling device as the second sample data set, and perform training on the large language model with the second parameter on this electronic cooling device to obtain a sub-control model of this electronic cooling device, so as to obtain the sub-control models of each electronic cooling device.

[0027] In this step, when deploying a large language model with the second parameter on each electronic cooling device, considering the hardware resources and actual needs, the second parameter can be smaller than the first parameter.

[0028] In addition, the same large language model with the second parameter can be deployed on different electronic cooling devices. For example, open-source large language models with 1.5 billion parameters are deployed on different electronic cooling devices. Different electronic cooling devices can also deploy different large language models with the second parameter. For example, an open-source large language model with 1.5 billion parameters is deployed on the electron gun, an open-source large language model with 7 billion parameters is deployed on the accelerating tube, and an open-source large language model with 14 billion parameters is deployed on the high-voltage platform device, etc. It should be noted that the examples listed here are only illustrative and do not limit this embodiment.

[0029] See Figure 3, for illustration, an open-source large language model with 671 billion parameters is deployed on the master control server, an open-source large language model with 1.5 billion parameters is deployed on the FPGA (Field Programmable Gate Array) power controller of the high-voltage platform device, an open-source large language model with 1.5 billion parameters is deployed on the FPGA power controller of the electron gun, an open-source large language model with 1.5 billion parameters is deployed on the FPGA power controller of the accelerating tube, an open-source large language model with 1.5 billion parameters is deployed on the FPGA power controller of the collector, and so on. This embodiment does not limit this.

[0030] In addition, it can also be deployed on Figure 3 the first beam position detector shown, an open-source large language model with 1.5 billion parameters, and an open-source large language model with 1.5 billion parameters on the second beam position detector.

[0031] Step S203, in response to the user-entered natural language instruction for electron cooling, the master control model performs semantic parsing on the natural language instruction for electron cooling to obtain a semantic parsing result including controlling the target electron cooling device; the master control model constructs a physical model for controlling the target electron cooling device according to the semantic parsing result, and determines the preliminary control parameters for controlling the target electron cooling device according to the physical model and the current actual operating state of the target electron cooling device.

[0032] Figure 4 This is the master control model control process. First, in response to the user-entered natural language instruction for electron cooling, the master control model performs semantic parsing on the natural language instruction for electron cooling to obtain a semantic parsing result including controlling the target electron cooling device; the master control model constructs a physical model for controlling the target electron cooling device according to the semantic parsing result; the master control model determines the preliminary control parameters for controlling the target electron cooling device according to the physical model and the current actual operating state of the target electron cooling device, that is, control parameter generation.

[0033] For example, the user-entered natural language instruction for electron cooling is "when the electron beam energy is E k , adjust the energy of the deflection plate so that the electron beam can be normally recycled". At this time, the master control model performs semantic parsing on this natural language instruction for electron cooling to obtain a semantic parsing result including controlling the target electron cooling device (here it is the deflection plate); the master control model constructs the following physical model for controlling the deflection plate according to the semantic parsing result:

[0034] Among them, V is the theoretical parameter of the deflection plate energy, m0 is the rest mass of the electron beam, e is the elementary charge, R is the turning radius, c is the speed of light, E 0 is the rest energy of the electron beam, d is the distance between the plates.

[0035] Then, according to the physical model and the current actual operating state of the deflector plate, substitute R = 1 m (meter), d = 0.07 m, E 0 = 511 keV (kilo-electron volts) and other physical and operating parameters of the deflector plate, and we can obtain:

[0036] At the same time, combining with the historical operating data of the deflector plate, the cumulative error of the deflector plate is obtained ΔV , and finally the preliminary control parameters for controlling the deflector plate are obtained V r =V + ΔV .

[0037] It should be noted that the above examples are only illustrative and do not limit this embodiment.

[0038] Step S204, the target electron cooling device receives the preliminary control parameters, and the sub-control model of the target electron cooling device analyzes the preliminary control parameters, and determines whether to adjust the preliminary control parameters according to the analysis results.

[0039] In this embodiment, through the collaborative architecture of the main control model and the sub-control models of each electronic cooling device, it can directly respond to the natural language instructions of electronic cooling input by the user. The main control model performs semantic parsing on the natural language instructions of electronic cooling to obtain a semantic parsing result including controlling the target electronic cooling device. The main control model constructs a physical model for controlling the target electronic cooling device according to the semantic parsing result, and determines the preliminary control parameters for controlling the target electronic cooling device according to the physical model and the current actual operating state of the target electronic cooling device. Furthermore, when the target electronic cooling device receives the preliminary control parameters, the sub-control model of the target electronic cooling device analyzes the preliminary control parameters, and determines whether to adjust the preliminary control parameters according to the analysis result, realizing the intelligent control, fault prediction and natural language interaction of the electronic cooling device, which is especially suitable for optimizing the beam quality and improving the system stability of electronic cooling. The fault prediction here can be that the sub-control model of the target electronic cooling device predicts whether there is a fault in the target electronic cooling device when analyzing the preliminary control parameters, or the sub-control model of the target electronic cooling device analyzes the operation data of the target electronic cooling device to predict whether there is a fault in the target electronic cooling device.

[0040] In a possible implementation provided in the embodiment of the present application, an electronic cooling knowledge base including a theory library and an operation library can also be pre-constructed. Among them, the theory library includes electronic cooling physical theories, electronic cooling technical literature, and historical experimental data; the operation library includes the operation parameters of each electronic cooling device collected in real time. Pre-constructing the electronic cooling knowledge base facilitates providing rich data for training the main control model and each sub-control model of the electronic cooling device, and can also provide data support during the inference stage.

[0041] In a possible implementation provided in the embodiment of the present application, obtaining the electronic cooling operation log and the electronic cooling theory literature as the first sample data set in step S201 above can specifically include obtaining the electronic cooling operation log and the electronic cooling theory literature as the first sample data set from the pre-constructed electronic cooling knowledge base, so as to provide rich data for training the main control model.

[0042] In the electronic cooling operation log, the operation data of all electronic cooling devices at the same moment can be formed into a 1×n array, where n is the number of electronic cooling devices. The data per minute forms a 60×n matrix and is sent to the large language model of the first parameter for training. This training method solves the problems of training data organization and time series for large-scale devices. Here, the 1×n array, such as filament 5.1A, grid 500V, anode 500V, high-voltage platform device voltage 139000V, high-voltage platform leakage current 0.22mA, collector 2.3A, positive deflection plate 13900V, negative deflection plate 13700V, etc. It should be noted that the examples listed here are only illustrative and do not limit this embodiment.

[0043] In an embodiment of the present application, a possible implementation manner is provided. In the above step S202, the operation log and theoretical literature of the electronic cooling device are obtained as the second sample data set. Specifically, it may include obtaining the operation log and theoretical literature of the electronic cooling device from a pre-constructed electronic cooling knowledge base as the second sample data set, so as to provide rich data for the training of the sub-control models of each electronic cooling device.

[0044] In an embodiment of the present application, a possible implementation manner is provided. In the above step S203, the main control model performs semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result including controlling the target electronic cooling device. Specifically, it may be that the main control model combines the electronic cooling knowledge base to perform semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result including controlling the target electronic cooling device. Here, the main control model combines the electronic cooling knowledge base to perform semantic parsing on the electronic cooling natural language instruction, which can improve the accuracy of semantic parsing.

[0045] In an embodiment of the present application, a possible implementation manner is provided. In the above step S201, the large language model of the first parameter is pre-trained, instruction fine-tuned, and reinforcement learned to obtain a main control model that can understand the physical laws of electronic cooling. Specifically, the large language model of the first parameter is pre-trained, instruction fine-tuned, and reinforcement learned with beam current stability, energy consumption efficiency, and fault prediction accuracy as the optimization objectives to obtain a main control model that can understand the physical laws of electronic cooling.

[0046] See Figure 5, the reward function can be designed, including control precision reward, response speed reward, and efficiency reward. Here, the control precision reward aims to optimize beam stability and the accuracy of fault prediction, and the efficiency reward aims to optimize energy consumption efficiency. In addition, the control precision reward adopts PPO (Proximal Policy Optimization), which is a reinforcement learning algorithm designed to improve training stability by restricting the magnitude of policy updates and enhance the stability of the model through online iterative processes.

[0047] For example, ; where V t is t the total reward at time R t is t the immediate reward at time V t+1 is t+1 the total reward at time γ is the discount factor. It determines the extent to which "future rewards" are considered in "current rewards".

[0048] The reward function is designed as R t =R precsion +R response +R effencient , where R precsion is the control precision, R response is the response speed, R effencient is the energy consumption efficiency. It should be noted that the example here is only illustrative and does not limit this embodiment.

[0049] In an embodiment of the present application, a possible implementation manner is provided. In step S204 above, it is determined whether to adjust the preliminary control parameters according to the analysis result. Specifically, it can be the following step A1: Step A1, if it is determined according to the analysis result that the preliminary control parameters need to be adjusted, then a multi-objective optimization solution is generated. If the adjusted control parameters are obtained through the solution, the target electronic cooling device executes according to the adjusted control parameters; if the adjusted control parameters are not obtained through the solution, an upper reporting information indicating that the adjusted control parameters are not obtained through the solution is generated and reported to the main control model.

[0050] In an embodiment of the present application, a possible implementation manner is provided. In step S204 above, it is determined whether to adjust the preliminary control parameters according to the analysis result. Specifically, it can also be the following step A2: Step A2: If it is determined according to the analysis result that the preliminary control parameters do not need to be adjusted, the target electronic cooling device shall execute according to the preliminary control parameters. During the execution of the target electronic cooling device, the sub-control model analyzes and identifies the execution information of the target electronic cooling device. If a fault of the target electronic cooling device is identified, a preset resolution strategy shall be executed, and an alarm message indicating that the target electronic cooling device has a fault shall be generated and reported to the main control model.

[0051] This embodiment can solve the problem of inconsistent instructions between the main control model and the sub-control model through the above Step A1 and Step A2.

[0052] Continuing with the previous example, after obtaining the preliminary control parameters of the deflection plate, the set data is sent to the deflection plate controller. Under normal circumstances, the analysis result of the sub-control model of the deflection plate is that the preliminary control parameters do not need to be adjusted. Then the deflection plate controller operates according to the preliminary control parameters, but there may be two abnormal situations that require conflict resolution: 1) After adding the cumulative error to the theoretical value, the sent parameter exceeds the maximum parameter that the device itself can withstand, and the device rejects the parameter. The sub-control model of the deflection plate generates a multi-objective optimization solution and attempts to adjust the parameters within the device's capabilities. If the task cannot be completed, relevant information is fed back to the upper-layer main control model.

[0053] 2) After the sent parameter is run, the leakage current of the deflection plate is too large, indicating that the parameter calculation is incorrect. The preset resolution strategy is executed, the underlying controller cancels the instruction and attempts a small-range adjustment, and the adjustment result and the alarm message indicating that the device has a fault are fed back to the upper-layer main control model.

[0054] The above has introduced Figure 2 multiple implementation methods for each link of the illustrated embodiment. Next, the electronic cooling control method based on a large language model of the embodiments of the present application will be further described through specific embodiments.

[0055] In this specific embodiment, it may include hardware deployment, model training, and conflict resolution implementation, as follows: 1) Hardware deployment Select an FPGA chip and configure a DDR4 (Double Data Rate Fourth, fourth-generation double data rate)-3200 memory (72-bit width).

[0056] The main control model runs on the main control server, and the sub-control model is deployed on the FPGA embedded device, that is, each electronic cooling device.

[0057] 2) Model training Pre-training stage: Use electronic cooling theory literature and historical operation logs.

[0058] Instruction fine-tuning phase: Annotate natural language control instructions for electronic cooling (such as "Adjust the magnetic field to xx T (Tesla)"), fault codes, and natural language descriptions.

[0059] Reinforcement learning phase: Use control accuracy, response speed, and energy consumption efficiency as the reward function to iteratively optimize the strategy.

[0060] 3) Conflict resolution implementation When the sub-control model detects an execution anomaly (such as power overlimit), it preferentially triggers the local protection mechanism and reports to the main control model at the same time; The main control model regenerates instructions based on the comprehensive global state and resolves conflicts through dynamic weight adjustment.

[0061] In this embodiment, through the main-control sub-control collaborative architecture, FPGA hardware, and automated training system, intelligent control and fault prediction of electronic cooling are realized; the main control model optimizes the global beam parameters, the sub-control model performs high-precision closed-loop control, and combined with the conflict resolution mechanism and natural language interaction, the system stability and maintainability are significantly improved.

[0062] It should be noted that the sequence numbers of the steps in the above embodiments do not indicate the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application. In practical applications, all the above possible implementation manners can be combined arbitrarily to form possible embodiments of the present application, which will not be elaborated herein one by one.

[0063] Based on the electronic cooling control method based on the large language model provided in the above embodiments, based on the same inventive concept, the embodiments of the present application also provide an electronic cooling control device based on the large language model.

[0064] Figure 6 It is the structural diagram of the electronic cooling control device based on the large language model provided by the embodiments of the present application. As Figure 6 shown, the electronic cooling control device based on the large language model may specifically include a first deployment unit 610, a second deployment unit 620, a first control unit 630, and a second control unit 640.

[0065] The first deployment unit 610 is used to pre-deploy a large language model with the first parameters on the main control server, obtain the electronic cooling operation log and electronic cooling theoretical literature as the first sample data set, and perform pre-training, instruction fine-tuning, and reinforcement learning on the large language model with the first parameters to obtain a main control model that can understand the physical laws of electronic cooling; The second deployment unit 620 is used to pre-deploy the large language model of the second parameter on each electronic cooling device. For the large language model of the second parameter on each electronic cooling device, obtain the operation log and theoretical literature of this electronic cooling device as the second sample data set, and train the large language model of the second parameter on this electronic cooling device to obtain the sub-control model of this electronic cooling device, so as to obtain the sub-control models of each electronic cooling device; The first control unit 630 is used to respond to the electronic cooling natural language instruction input by the user. The main control model performs semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result including controlling the target electronic cooling device; the main control model constructs a physical model for controlling the target electronic cooling device according to the semantic parsing result, and determines the preliminary control parameters for controlling the target electronic cooling device according to the physical model and the current actual operating state of the target electronic cooling device; The second control unit 640 is used for the target electronic cooling device to receive the preliminary control parameters. The sub-control model of this target electronic cooling device analyzes the preliminary control parameters, and determines whether to adjust the preliminary control parameters according to the analysis result.

[0066] In an embodiment of the present application, a possible implementation manner is provided, such as Figure 7 shown, the above Figure 6 The device shown may further include a construction unit 710, which is used to pre-construct an electronic cooling knowledge base including a theory library and an operation library, where the theory library includes electronic cooling physical theories, electronic cooling technical literature, and historical experimental data; the operation library includes the operation parameters of each electronic cooling device collected in real time.

[0067] In an embodiment of the present application, a possible implementation manner is provided, and the first deployment unit 610 is further used to: obtain the electronic cooling operation log and electronic cooling theoretical literature from the pre-constructed electronic cooling knowledge base as the first sample data set; The second deployment unit 620 is further used to: obtain the operation log and theoretical literature of this electronic cooling device from the pre-constructed electronic cooling knowledge base as the second sample data set.

[0068] In an embodiment of the present application, a possible implementation manner is provided, and the first control unit 630 is further used to: The main control model combines the electronic cooling knowledge base to perform semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result including controlling the target electronic cooling device.

[0069] In an embodiment of the present application, a possible implementation manner is provided, and the first deployment unit 610 is further used to: Taking beam stability, energy consumption efficiency, and fault prediction accuracy as optimization objectives, pre-train, instruction-tune, and reinforce-learn a large language model for the first parameter to obtain a master model that can understand the physical laws of electron cooling.

[0070] In an exemplary embodiment of the present application, the second control unit 640 is further configured to: If it is determined according to the analysis result that the preliminary control parameters need to be adjusted, generate a multi-objective optimization solution. If the adjusted control parameters are obtained through the solution, the target electron cooling device shall execute according to the adjusted control parameters; if the adjusted control parameters are not obtained through the solution, generate a reporting message indicating that the adjusted control parameters are not obtained through the solution and report it to the master model.

[0071] In an exemplary embodiment of the present application, the second control unit 640 is further configured to: If it is determined according to the analysis result that the preliminary control parameters do not need to be adjusted, the target electron cooling device shall execute according to the preliminary control parameters. During the execution of the target electron cooling device, the sub-control model analyzes and identifies the execution information of the target electron cooling device. If a fault of the target electron cooling device is identified, execute a preset resolution strategy and generate an alarm message indicating that the target electron cooling device has a fault and report it to the master model.

[0072] Based on the same inventive concept, an exemplary embodiment of the present application further provides an electronic device, including a processor and a memory. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the method for controlling electron cooling based on a large language model according to any one of the above embodiments.

[0073] In an exemplary embodiment, an electronic device is provided, such as Figure 8 shown. Figure 8 The electronic device 800 shown includes a processor 801 and a memory 803. Among them, the processor 801 and the memory 803 are connected, such as through a bus 802. Optionally, the electronic device 800 may further include a transceiver 804. It should be noted that in practical applications, the transceiver 804 is not limited to one, and the structure of the electronic device 800 does not constitute a limitation to the embodiments of the present application.

[0074] The processor 801 may be a CPU (Central Processing Unit), GPU (Graphics Processing Unit), DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of this application. The processor 801 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0075] The bus 802 may include a path for transmitting information between the above components. The bus 802 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 802 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 only a thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0076] The memory 803 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or it may also be an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0077] The memory 803 is used to store the computer program code for executing the solution of this application, and is controlled by the processor 801 for execution. The processor 801 is used to execute the computer program code stored in the memory 803 to implement the content shown in the foregoing method embodiments.

[0078] Among them, the electronic device includes but is not limited to: mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The illustrated electronic device is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of this application.

[0079] Based on the same inventive concept, the embodiments of this application also provide a storage medium in which a computer program is stored. The computer program is configured to execute the electronic cooling control method based on a large language model in any of the foregoing embodiments when running.

[0080] Based on the same inventive concept, the embodiments of this application also provide a computer program product, including a computer program, where the computer program is configured to execute the electronic cooling control method based on a large language model in any of the foregoing embodiments when running.

[0081] Those skilled in the art can clearly understand the specific working processes of the above-described systems, devices, and modules, and can refer to the corresponding processes in the foregoing method embodiments. For the sake of brevity, they will not be described in detail here.

[0082] Those of ordinary skill in the art can understand that the technical solution of this application can essentially or all or part of it be embodied in the form of a software product. The computer software product is stored in a storage medium, which includes several program instructions for causing an electronic device (such as a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of this application when running the program instructions. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0083] Alternatively, all or part of the steps of implementing the foregoing method embodiments can be completed by hardware related to program instructions (such as an electronic device such as a personal computer, a server, or a network device). The program instructions can be stored in a computer-readable storage medium. When the program instructions are executed by the processor of the electronic device, the electronic device executes all or part of the steps of the methods described in the embodiments of this application.

[0084] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that within the spirit and principle of the present application, they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the corresponding technical solutions to deviate from the protection scope of the present application.

Claims

1. An electronic cooling control method based on a large language model, characterized in that: include: Deploy a large language model of the first parameter on the main control server in advance, obtain electronic cooling operation logs and electronic cooling theoretical literature as the first sample data set, perform pre-training, instruction fine-tuning and reinforcement learning on the large language model of the first parameter, and obtain a main control model that can understand the physical laws of electronic cooling; Pre-deploy a large language model of the second parameter on each electronic cooling device. For the large language model of the second parameter on each electronic cooling device, obtain the operation log and theoretical literature of the electronic cooling device as a second sample data set, train the large language model of the second parameter on the electronic cooling device, and obtain a sub-control model of the electronic cooling device, thereby obtaining a sub-control model of each electronic cooling device. In response to the electronic cooling natural language instruction input by the user, the main control model performs semantic analysis on the electronic cooling natural language instruction to obtain a semantic analysis result including control of the target electronic cooling device; the main control model constructs a physical model for controlling the target electronic cooling device according to the semantic analysis result, and determines preliminary control parameters for controlling the target electronic cooling device according to the physical model and the current actual operating state of the target electronic cooling device; The target electronic cooling device receives the preliminary control parameters, and the sub-control model of the target electronic cooling device analyzes the preliminary control parameters, and determines whether to adjust the preliminary control parameters according to the analysis results.

2. The electronic cooling control method based on a large language model according to claim 1, characterized in that: Also includes: An electronic cooling knowledge base including a theoretical library and an operational library is pre-built, wherein the theoretical library includes electronic cooling physical theories, electronic cooling technical literature and historical experimental data; the operational library includes the operating parameters of various electronic cooling devices collected in real time.

3. The electronic cooling control method based on a large language model according to claim 2, characterized in that: The electronic cooling operation log and the electronic cooling theory literature are obtained as the first sample data set, including: From the pre-built electronic cooling knowledge base, electronic cooling operation logs and electronic cooling theory literature are obtained as the first sample data set; The operation log and theoretical literature of the electronic cooling equipment are obtained as the second sample data set, including: From the pre-built electronic cooling knowledge base, the operation log and theoretical literature of the electronic cooling device are obtained as the second sample data set.

4. The electronic cooling control method based on a large language model according to claim 2, characterized in that: The main control model performs semantic parsing on the electronic cooling natural language instruction to obtain a semantic parsing result including control of the target electronic cooling device, including: The main control model combines with the electronic cooling knowledge base to perform semantic parsing on the electronic cooling natural language instructions, and obtains the semantic parsing results including the control of the target electronic cooling equipment.

5. The electronic cooling control method based on a large language model according to claim 1, characterized in that: The large language model of the first parameter is pre-trained, fine-tuned, and reinforced to learn, and a master control model that can understand the physical laws of electronic cooling is obtained, including: Taking beam stability, energy efficiency and fault prediction accuracy as optimization goals, the large language model of the first parameter is pre-trained, instruction fine-tuned and reinforced learned to obtain a master control model that can understand the physical laws of electron cooling.

6. The electronic cooling control method based on a large language model according to claim 1, characterized in that: Determine whether to adjust the preliminary control parameters based on the analysis results, including: If it is determined based on the analysis results that the preliminary control parameters need to be adjusted, a multi-objective optimization solution is generated. If the adjusted control parameters are obtained, the target electronic cooling device is executed according to the adjusted control parameters. If the adjusted control parameters are not obtained, a report information indicating that the adjusted control parameters are not obtained is generated and reported to the main control model.

7. The electronic cooling control method based on a large language model according to claim 1, characterized in that: Determine whether to adjust the preliminary control parameters based on the analysis results, including: If it is determined based on the analysis results that the preliminary control parameters do not need to be adjusted, the target electronic cooling device will execute according to the preliminary control parameters. During the execution of the target electronic cooling device, the sub-control model analyzes and identifies the execution information of the target electronic cooling device. If it is identified that the target electronic cooling device has a fault, the preset mitigation strategy is executed, and an alarm message indicating that the target electronic cooling device has a fault is generated and reported to the main control model.

8. An electronic cooling control device based on a large language model, characterized in that: For implementing the electronic cooling control method based on a large language model as claimed in claim 1, the device comprises: A first deployment unit is used to deploy a large language model of the first parameter on a main control server in advance, obtain an electronic cooling operation log and electronic cooling theoretical literature as a first sample data set, perform pre-training, instruction fine-tuning and reinforcement learning on the large language model of the first parameter, and obtain a main control model that can understand the physical laws of electronic cooling; A second deployment unit is used to deploy a large language model of the second parameter on each electronic cooling device in advance, and for the large language model of the second parameter on each electronic cooling device, obtain an operation log and theoretical literature of the electronic cooling device as a second sample data set, train the large language model of the second parameter on the electronic cooling device, and obtain a sub-control model of the electronic cooling device, thereby obtaining a sub-control model of each electronic cooling device; A first control unit is used to respond to the electronic cooling natural language instruction input by the user, and the main control model performs semantic analysis on the electronic cooling natural language instruction to obtain a semantic analysis result including control of the target electronic cooling device; the main control model constructs a physical model for controlling the target electronic cooling device according to the semantic analysis result, and determines preliminary control parameters for controlling the target electronic cooling device according to the physical model and the current actual operating state of the target electronic cooling device; The second control unit is used for the target electronic cooling device to receive the preliminary control parameters, and the sub-control model of the target electronic cooling device analyzes the preliminary control parameters, and determines whether to adjust the preliminary control parameters according to the analysis results.

9. An electronic device, characterized in that: The invention comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the electronic cooling control method based on a large language model according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the electronic cooling control method based on a large language model according to any one of claims 1 to 7 when running.

11. A computer program product, comprising a computer program, characterized in that The computer program is configured to execute the electronic cooling control method based on a large language model according to any one of claims 1 to 7 when running.

Images