A large model-based power dispatching control method and device

Abstract

The application discloses a power dispatching control method and device based on a large model, acquires a power dispatching request input by a user, inputs the power dispatching request into a preset demand analysis model, and obtains a power dispatching demand corresponding to the power dispatching request; based on the power dispatching demand, confirms a plurality of dispatching subtasks and an execution sequence of each dispatching subtask; inputs each dispatching subtask into an execution analysis model respectively, generates an execution device and control code corresponding to each dispatching subtask, and controls a control device according to the execution sequence of each dispatching subtask based on the execution device and control code. The application analyzes the power dispatching request of the user layer by layer through a large language model, and improves the accuracy of power dispatching.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligence, and in particular to a power dispatching control method and device based on a large model. BACKGROUND

[0002] With the continuous expansion of the scale and the continuous improvement of the complexity of the power system, the traditional manual power dispatching method has been difficult to meet the needs of modern power grids. The conventional dispatching mode of the power system at present is a rule-based power dispatching method, which performs power dispatching through pre-set rules and decision trees. This method automatically generates dispatching instructions according to the real-time power grid state by setting a series of dispatching rules. However, the rule-based power dispatching method lacks flexibility and is difficult to cope with complex and variable power grid situations, and is prone to inaccurate power dispatching.

[0003] Therefore, there is an urgent need for a power dispatching control strategy to solve the problem of inaccurate power dispatching of the rule-based power dispatching method. SUMMARY

[0004] The embodiments of the present application provide a power dispatching control method and device based on a large model to solve the problem of inaccurate power dispatching of the rule-based power dispatching method.

[0005] In order to solve the above problems, an embodiment of the present application provides a power dispatching control method based on a large model, comprising:

[0006] obtaining a power dispatching request input by a user;

[0007] inputting the power dispatching request into a pre-set demand analysis model to obtain a power dispatching demand corresponding to the power dispatching request; wherein the power dispatching demand includes dispatching content and control equipment; training a large language model with historical power dispatching request samples marked with power dispatching demands as input and the demand analysis model as output;

[0008] based on the power dispatching demand, confirming a plurality of dispatching sub-tasks and an execution order of each dispatching sub-task; wherein a dispatching sub-task corresponds to a power dispatching demand;

[0009] inputting each dispatching sub-task into an execution analysis model to generate an execution equipment and control code corresponding to each dispatching sub-task, and based on the execution equipment and control code, controlling the control equipment according to the dispatching content of each dispatching sub-task in the execution order; wherein historical dispatching sub-task samples marked with execution equipment and control code are input into the execution analysis model as output, and a large language model is trained.

[0010] As an improvement of the above-mentioned scheme, the power dispatching requirement is analyzed by a task decomposition algorithm to obtain a plurality of dispatching sub-tasks and a dependency relationship between each of the dispatching sub-tasks.

[0011] The power dispatching requirement is analyzed by a task decomposition algorithm to obtain a plurality of dispatching sub-tasks and a dependency relationship between each of the dispatching sub-tasks.

[0012] Based on the dependency relationship between each of the dispatching sub-tasks, the execution order of each of the dispatching sub-tasks is determined.

[0013] As an improvement of the above-mentioned scheme, the control device is controlled according to the execution order of each of the dispatching sub-tasks.

[0014] It is determined whether the execution order of each of the dispatching sub-tasks exists in parallel or not. The execution order includes parallel order and sequence.

[0015] If it exists, the dispatching sub-tasks in parallel are marked as a target sub-task group, and the control device corresponding to the dispatching sub-tasks is controlled according to the corresponding dispatching content according to the sequence of the target sub-task group and a plurality of non-parallel dispatching sub-tasks. When the target sub-task group is executed, the control device corresponding to the target sub-task group is controlled according to the corresponding dispatching content according to the parallel order.

[0016] If it does not exist, the control device is controlled according to the sequence of each of the dispatching sub-tasks.

[0017] As an improvement of the above-mentioned scheme, the dispatching sub-tasks include the dispatching order of the execution device, the parameters required for the execution device to dispatch, and the processing method of the intermediate results generated by the execution device.

[0018] As an improvement of the above-mentioned scheme, after the control device is controlled according to the execution order of each of the dispatching sub-tasks, the following steps are further included.

[0019] The running data of the control device is collected.

[0020] The type and value of the running data are matched based on a preset fault database.

[0021] When the matching is successful, the fault coping strategy of the control device corresponding to the running data is retrieved from the fault database, and the data collection of the control device is maintained.

[0022] When the matching fails, no operation is performed.

[0023] Correspondingly, the embodiment of the present application also provides a power dispatching control device based on a large model, comprising: a data acquisition module, a demand analysis module, a task splitting module and a control scheduling module.

[0024] The data acquisition module is configured to acquire a power dispatching request input by a user.

[0025] The demand analysis module is configured to input the power dispatching request into a preset demand analysis model to obtain a power dispatching demand corresponding to the power dispatching request; wherein the power dispatching demand comprises dispatching content and a control device; a large language model is trained by taking a historical power dispatching request sample marked with the power dispatching demand as input and taking the demand analysis model as output.

[0026] The task splitting module is configured to confirm a plurality of dispatching subtasks and an execution order of each dispatching subtask based on the power dispatching demand; wherein each dispatching subtask corresponds to a power dispatching demand.

[0027] The control scheduling module is configured to input each dispatching subtask into an execution analysis model respectively to generate an execution device and a control code corresponding to each dispatching subtask, and to control the control device according to the dispatching content in the execution order of each dispatching subtask based on the execution device and the control code; wherein a historical dispatching subtask sample marked with the execution device and the control code is taken as input, the execution analysis model is taken as output, and a large language model is trained.

[0028] As an improvement of the above-mentioned scheme, the confirmation of the plurality of dispatching subtasks and the execution order of each dispatching subtask based on the power dispatching demand comprises:

[0029] The power dispatching demand is analyzed by a task decomposition algorithm to obtain a plurality of dispatching subtasks and a dependency relationship between each dispatching subtask;

[0030] The execution order of each dispatching subtask is determined based on the dependency relationship between one dispatching subtask.

[0031] As an improvement of the above-mentioned scheme, the control of the control device according to the dispatching content in the execution order of each dispatching subtask comprises:

[0032] It is judged whether the execution order of each dispatching subtask has a parallel order; wherein the execution order comprises a parallel order and a sequence order.

[0033] If the parallel scheduling sub-tasks exist, mark the parallel scheduling sub-tasks as target sub-task groups, and control the control devices corresponding to the scheduling sub-tasks according to the scheduling contents in the order of the target sub-task groups and a plurality of non-parallel scheduling sub-tasks; wherein, when the target sub-task groups are executed, the control devices corresponding to the target sub-task groups are controlled according to the scheduling contents in the parallel order.

[0034] If the parallel scheduling sub-tasks do not exist, control the control devices according to the scheduling contents in the order of each scheduling sub-task.

[0035] As an improvement of the above scheme, the scheduling sub-tasks include: scheduling order of the execution device, parameters required for scheduling of the execution device, and processing mode of intermediate results generated by the execution device.

[0036] As an improvement of the above scheme, after the control of the control devices according to the scheduling contents in the order of each scheduling sub-task, the method further includes:

[0037] Collecting running data of the control device;

[0038] Matching the types and values of the running data based on a preset fault database;

[0039] When the matching is successful, the fault coping strategy of the control device corresponding to the running data is retrieved from the fault database, and the data collection of the control device is maintained;

[0040] When the matching fails, no operation is performed.

[0041] As can be seen from the above, the present application has the following beneficial effects:

[0042] The present application provides a power dispatching control method based on a large model, including: obtaining a power dispatching request input by a user; inputting the power dispatching request into a preset demand analysis model to obtain a power dispatching demand corresponding to the power dispatching request; based on the power dispatching demand, confirming a plurality of scheduling sub-tasks and an execution order of each scheduling sub-task; inputting each scheduling sub-task into an execution analysis model respectively to generate an execution device and control code corresponding to each scheduling sub-task, and based on the execution device and the control code, controlling a control device according to the scheduling contents in the order of each scheduling sub-task. The present application determines the power dispatching demand of the power dispatching request input by the user through the large model, determines the scheduling sub-tasks and the corresponding execution order by analyzing the power dispatching demand, finally confirms the execution device and the control code of the scheduling sub-tasks through the large language model, and controls the control device according to the scheduling contents in the order of execution, which realizes the intelligentization of power dispatching. The power dispatching request of the user is analyzed layer by layer through the large language model, which improves the accuracy of power dispatching. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating a power dispatch control method based on a large model provided in an embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the structure of a power dispatch control device based on a large model provided in an embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram of a terminal device structure provided in an embodiment of the present invention. Detailed Implementation

[0046] 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 only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Example 1

[0048] See Figure 1 , Figure 1 This is a flowchart illustrating a power dispatch control method based on a large model, as provided in an embodiment of the present invention. Figure 1 As shown, this embodiment includes steps 101 to 104, and the specific steps are as follows:

[0049] Step 101: Obtain the power dispatch request input by the user.

[0050] Step 102: Input the power dispatch request into the preset demand analysis model to obtain the power dispatch demand corresponding to the power dispatch request; wherein, the power dispatch demand includes: dispatch content and control equipment; use historical power dispatch request samples marked with power dispatch demand as input and the demand analysis model as output to train the large language model.

[0051] In this embodiment, the requirements analysis model can identify key points in user requirements and acquire tool information needed to solve the task. Leveraging the powerful natural language processing capabilities of the large language model, the requirements analysis model can quickly understand complex tasks.

[0052] Step 103: Based on the power dispatching requirements, determine several dispatching subtasks and the execution order of each dispatching subtask; wherein, a dispatching subtask corresponds to a power dispatching requirement.

[0053] In this embodiment, determining several scheduling subtasks and the execution order of each scheduling subtask based on the power dispatching demand includes:

[0054] The power dispatching demand is analyzed by a task decomposition algorithm to obtain several dispatching subtasks and the dependencies between each dispatching subtask.

[0055] Based on the dependencies between the scheduled subtasks, the execution order of each scheduled subtask is determined.

[0056] In a specific embodiment, several scheduled subtasks and the execution order of each scheduled subtask can be represented by a task tree structure. The characteristics of the task tree structure include: nodes: each node represents a subtask; hierarchy: tasks are divided into different levels according to their dependencies; parallelism and sequence: for subtasks that can be processed in parallel, the task tree will arrange for parallel execution; otherwise, they will be executed sequentially.

[0057] In this embodiment, the scheduling subtask includes: the scheduling order of the execution devices, the parameters required for scheduling the execution devices, and the processing method for the intermediate results generated by the execution devices.

[0058] Step 104: Input each of the scheduled subtasks into the execution analysis model to generate the execution device and control code corresponding to each scheduled subtask. Based on the execution device and control code, control the scheduling content of the control device according to the execution order of each scheduled subtask. The large language model is trained by taking historical scheduled subtask samples marked with execution devices and control codes as input and the execution analysis model as output.

[0059] In one specific embodiment, after the scheduling subtask is planned, the corresponding code is written by calling the execution analysis model according to the task requirements. This code is used to execute the device and carry out the scheduling task. To improve the accuracy of code generation, the system introduces the following optimizations:

[0060] 1. Code Template Library: The system maintains a code template library for common power dispatching tasks. These templates can significantly improve the efficiency and accuracy of code generation.

[0061] 2. Syntax Checking and Code Optimization: After code generation, the system automatically performs syntax checks to ensure code correctness. Simultaneously, it optimizes the code to reduce runtime errors.

[0062] It's important to note that the code written for the analysis model will be executed within a secure sandbox to prevent destructive impacts on the system. The sandbox mechanism provides an isolated environment, ensuring the security of tool calls.

[0063] In this embodiment, controlling the scheduling content of the control device according to the execution order of each of the scheduled subtasks includes:

[0064] Determine whether the execution order of each scheduled subtask exists in a parallel order; wherein, the execution order includes: parallel order and sequential order;

[0065] If they exist, the parallel scheduling subtasks are marked as the target subtask group. The control devices corresponding to the scheduling subtasks are controlled according to the order of the target subtask group and several non-parallel scheduling subtasks. When the target subtask group is executed, the control devices corresponding to the scheduling subtasks of the target subtask group are controlled according to the parallel order.

[0066] If not, the control device is controlled according to the order of each of the aforementioned scheduling subtasks.

[0067] In this embodiment, after controlling the scheduling content of the control device according to the execution order of each of the scheduled subtasks, the method further includes:

[0068] Collect operational data from control equipment;

[0069] Based on a pre-set fault database, the types and values ​​of operational data are matched.

[0070] When a match is successful, retrieve the fault response strategy for the control equipment corresponding to the running data from the fault database, and continue to collect data from the control equipment;

[0071] No action is taken when a match fails.

[0072] In one specific embodiment, the execution analysis model and the requirements analysis model have self-diagnostic and debugging capabilities. Specific operations include:

[0073] 1. Error Diagnosis: Analyze error messages and determine the cause of the error;

[0074] 2. Automatic debugging: Based on the diagnostic results, the system performs corresponding debugging operations and attempts to fix errors;

[0075] 3. Limitation identification: The system can also identify problems that it cannot solve, avoiding repeated attempts on the same problem, thereby improving the overall system efficiency.

[0076] In a specific embodiment, information after each task execution by the execution analysis model and the requirements analysis model, including encountered problems and solutions, is stored in an experience pool. The experience pool effectively improves the system's intelligence level, specifically implemented as follows: The experience pool uses an efficient data structure to store information such as problem descriptions, solutions, and success rates. Secondly, a retrieval algorithm: the system uses a semantic similarity algorithm to quickly retrieve similar historical problems from the experience pool, helping the large language model quickly find reference solutions. Finally, an experience update mechanism: based on the latest execution results, the system continuously optimizes and updates the data in the experience pool, ensuring that the experience pool remains efficient and accurate. Regarding the specific implementation of experience pool maintenance, we can introduce the following mathematical formula to describe its core mechanism:

[0077] 1. Data structure design:

[0078] Let the experience pool E be a set, where each element e_i represents an experience record:

[0079] E = {e_1,e_2,...,e_n};

[0080] Each experience record e_i can be represented as a vector:

[0081] e_i = (p_i, s_i, r_i, t_i);

[0082] Where p_i: problem description vector, s_i: solution vector, r_i: success rate (0≤r_i≤1), and t_i: timestamp.

[0083] 2. Retrieval Algorithm:

[0084] Semantic similarity is calculated using cosine similarity. Given a new problem q and problems p_i in the experience pool, the similarity is calculated as follows:

[0085] sim(q,p_i)=cos(q,p_i)=(q·p_i) / (||q||*||p_i||);

[0086] Where represents the vector dot product, and ||x|| represents the Euclidean norm of vector x.

[0087] The retrieval process can be represented as finding the k most similar experiences:

[0088] TopK=argmax_k(sim(q,p_i))for p_iin E.

[0089] 3. Experience update mechanism:

[0090] Define the experience update function U:

[0091] U(e_i,e_new)=(α*p_i+(1-α)*p_new,

[0092] β*s_i+(1-β)*s_new,

[0093] γ*r_i+(1-γ)*r_new,

[0094] t_new)

[0095] Where α, β, and γ are weight parameters (0 ≤ α, β, γ ≤ 1), used to balance new and old experiences. 4. Experience Pool Optimization:

[0096] Define the overall quality Q of the experience pool:

[0097] Q(E) = Σ(w_i*r_i) / |E|;

[0098] Where w_i is the weight of each experience, which can be calculated based on the time decay function: w_i = exp(-λ*(t_current-t_i));

[0099] λ is the time decay coefficient.

[0100] 5. Experience pool maintenance algorithm:

[0101] For the new experience e_new:

[0102] if sim(p_new,p_i)>threshold for any e_iin E:

[0103] e_i = U(e_i, e_new)

[0104] else:

[0105] E=E∪{e_new}

[0106] if |E|>max_size:

[0107] E = E - {argmin_e(w_e*r_e)}

[0108] 6. Perform optimizations regularly:

[0109] while Q(E)<Q_threshold and|E|> min_size:

[0110] E = E - {argmin_e(w_e*r_e)}

[0111] It should be noted that the application scenarios of this embodiment include, but are not limited to:

[0112] - Power flow calculation: Scheduling and invocation tools complete complex power flow analysis;

[0113] - Power dispatching: Helps dispatchers plan and execute tasks efficiently and accurately;

[0114] - Black Start Sequence: Automatically executes the sequential scheduling during the power grid black start process;

[0115] - Incident handling: Quickly diagnose power system incidents and activate emergency response tools;

[0116] -Solution generation: Automatically generate the best execution solution based on different power dispatching task requirements.

[0117] In a specific embodiment, a specific motor control scheduling scheme is given as an example:

[0118] 1. Task Understanding and Planning: The requirements analysis model receives a user's natural language query: "Please develop a 24-hour optimized scheduling plan for 10 high-power motors in an industrial park, considering grid load balancing and cost minimization."

[0119] The requirements analysis model quickly understands the task requirements and formulates a detailed execution plan: a) acquire power grid load forecast data; b) establish a motor control optimization model; c) solve the optimization problem; d) generate specific motor control schemes; e) monitor and adjust in real time.

[0120] This flexible ability to understand and plan tasks is difficult to achieve with traditional rule-based systems.

[0121] 2. Multi-tool collaborative invocation: The execution analysis model automatically selects and invokes multiple professional tools based on task requirements:

[0122] a) Use the load forecasting tool to obtain the 24-hour load curve of the power grid: L(t) = 10000 + 5000 * sin(πt / 12) kW, 0 ≤ t < 24

[0123] b) Use the optimization modeling tool to construct the motor control optimization model: Objective function: minimize max(L(t)) - min(L(t)) Constraints:

[0124] ·ΣP_i(t)≤5000kW, i=1,2,...,10;

[0125] ·0≤P_i(t)≤500kW;

[0126] • Each motor starts and stops ≤ 3 times;

[0127] • Single run time ≥ 2 hours.

[0128] c) Call the optimization solver to obtain the optimal control scheme.

[0129] The seamless integration of these tools demonstrates its powerful tool calling and collaboration capabilities.

[0130] 3. Code generation and execution: Automatically generate Python code to implement optimization solutions: It can automatically generate and execute code according to task requirements, which greatly improves the flexibility and adaptability of the scheduling system.

[0131] 4. Real-time monitoring and dynamic adjustment: During execution, the motor's operating status is continuously monitored.

[0132] When an anomaly is detected, such as when the temperature of motor 1 reaches 90℃ (exceeding the threshold of 85℃) at t=10:00,

[0133] immediately:

[0134] a) Use fault diagnosis tools to analyze the cause. b) Generate a response strategy: reduce P_1(t) from 400kW to 300kW. c) Redistribute 100kW load to other motors. d) Notify maintenance personnel to check the bearings.

[0135] Real-time calculation of adjusted energy consumption and cost: E_new=∫[0,24](P_1_new(t) / η)dt≈

[0136] 5800kWh Cost_new=∫[0,24](P_1_new(t) / η*C(t))dt≈3480 yuan

[0137] This dynamic adjustment capability reflects the intelligent and adaptive characteristics of the execution analysis model and the requirements analysis model.

[0138] 5. Experience learning and knowledge accumulation: After each scheduled task, the execution analysis model and the requirements analysis model store the experience into the experience pool: e_i = (p_i, s_i, r_i, t_i) where p_i is the problem description vector (e.g., "motor overheating, bearing failure"), s_i is the solution vector, r_i is the success rate, and t_i is the timestamp.

[0139] When encountering similar situations in the future, the execution analysis model and the requirements analysis model will use semantic similarity for fast retrieval: sim(q,p_i)=(q·p_i) / (||q||*||p_i||);

[0140] If the similarity is higher than a threshold (e.g., 0.8), the execution analysis model and the requirements analysis model will prioritize using similar solutions to continuously improve scheduling efficiency and reliability.

[0141] This ability to continuously learn and accumulate knowledge is one of the key characteristics that distinguishes execution analysis models and requirements analysis models from traditional systems.

[0142] See Figure 2 , Figure 2 This is a schematic diagram of the structure of a power dispatch and control device based on a large model according to an embodiment of the present invention, including: a data acquisition module 201, a demand analysis module 202, a task splitting module 203, and a control and dispatching module 204;

[0143] The data acquisition module is used to acquire power dispatch requests input by the user;

[0144] The demand analysis module is used to input the power dispatch request into a preset demand analysis model to obtain the power dispatch demand corresponding to the power dispatch request; wherein, the power dispatch demand includes: dispatch content and control equipment; using historical power dispatch request samples marked with power dispatch demand as input and the demand analysis model as output, the large language model is trained;

[0145] The task splitting module is used to determine several scheduling sub-tasks and the execution order of each scheduling sub-task based on the power dispatching requirements; wherein, a scheduling sub-task corresponds to a power dispatching requirement.

[0146] The control and scheduling module is used to input each of the scheduling subtasks into the execution analysis model, generate the execution device and control code corresponding to each scheduling subtask, and control the scheduling content of the control device according to the execution order of each scheduling subtask based on the execution device and control code; wherein, the large language model is trained with historical scheduling subtask samples marked with execution devices and control codes as input and the execution analysis model as output.

[0147] As an improvement to the above scheme, the step of determining several scheduling sub-tasks and the execution order of each scheduling sub-task based on the power dispatching requirements includes:

[0148] The power dispatching demand is analyzed by a task decomposition algorithm to obtain several dispatching subtasks and the dependencies between each dispatching subtask.

[0149] Based on the dependencies between the scheduled subtasks, the execution order of each scheduled subtask is determined.

[0150] As an improvement to the above scheme, the control of the scheduling content of the control device according to the execution order of each of the scheduling subtasks includes:

[0151] Determine whether the execution order of each scheduled subtask exists in a parallel order; wherein, the execution order includes: parallel order and sequential order;

[0152] If they exist, the parallel scheduling subtasks are marked as the target subtask group. The control devices corresponding to the scheduling subtasks are controlled according to the order of the target subtask group and several non-parallel scheduling subtasks. When the target subtask group is executed, the control devices corresponding to the scheduling subtasks of the target subtask group are controlled according to the parallel order.

[0153] If not, the control device is controlled according to the order of each of the aforementioned scheduling subtasks.

[0154] As an improvement to the above scheme, the scheduling subtask includes: the scheduling order of the execution devices, the parameters required for scheduling the execution devices, and the processing method for the intermediate results generated by the execution devices.

[0155] As an improvement to the above scheme, after controlling the scheduling content of the control device according to the execution order of each of the scheduled subtasks, the method further includes:

[0156] Collect operational data from control equipment;

[0157] Based on a pre-set fault database, the types and values ​​of operational data are matched.

[0158] When a match is successful, retrieve the fault response strategy for the control equipment corresponding to the running data from the fault database, and continue to collect data from the control equipment;

[0159] No action is taken when a match fails.

[0160] This embodiment obtains a power dispatch request input by a user; inputs the power dispatch request into a preset demand analysis model to obtain the corresponding power dispatch requirements; based on the power dispatch requirements, confirms several dispatch sub-tasks and the execution order of each dispatch sub-task; inputs each dispatch sub-task into an execution analysis model to generate the execution device and control code corresponding to each dispatch sub-task; and controls the dispatch content of the control device according to the execution order of each dispatch sub-task based on the execution device and control code. This invention determines the power dispatch requirements of the user-input power dispatch request through a large model, determines the dispatch sub-tasks and their corresponding execution order by analyzing the power dispatch requirements, and finally confirms the execution device and control code of the dispatch sub-tasks through a large language model, controlling the dispatch content of the control device according to the execution order, thus realizing intelligent power dispatch. By analyzing the user's power dispatch request layer by layer through a large language model, the accuracy of power dispatch is improved.

[0161] Example 2

[0162] See Figure 3 , Figure 3 This is a schematic diagram of the terminal device structure provided in an embodiment of the present invention.

[0163] One terminal device in this embodiment includes: a processor 301, a memory 302, and a computer program stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program, it implements the steps of the various large-model-based power dispatch control methods described above in this embodiment, for example... Figure 1 The steps of the power dispatch control method based on a large model are shown. Alternatively, when the processor executes the computer program, it implements the functions of each module in the above-described device embodiments, for example: Figure 2 The diagram shows all the modules of a power dispatch and control device based on a large model.

[0164] In addition, embodiments of the present invention also provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the power dispatch control method based on a large model as described in any of the above embodiments.

[0165] Those skilled in the art will understand that the schematic diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown in the diagram, or combine certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc.

[0166] The processor 301 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The processor 301 is the control center of the terminal device, connecting various parts of the terminal device through various interfaces and lines.

[0167] The memory 302 can be used to store the computer programs and / or modules. The processor 301 implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory 302. The memory 302 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0168] Wherein, if the modules / units integrated in the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0169] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0170] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A large model-based power dispatching control method, characterized in that, The method comprises the following steps: acquiring a power dispatch request input by a user; inputting the power dispatch request into a preset demand analysis model to obtain a power dispatch demand corresponding to the power dispatch request; wherein the power dispatch demand comprises dispatch content and a control device; a large language model is trained by taking a historical power dispatch request sample marked with the power dispatch demand as input and taking the demand analysis model as output; based on the power dispatch demand, confirming a plurality of dispatch subtasks and an execution sequence of each dispatch subtask; wherein a dispatch subtask corresponds to a power dispatch demand; inputting each dispatch subtask into an execution analysis model respectively to generate an execution device and control code corresponding to each dispatch subtask, and based on the execution device and control code, controlling the control device according to the execution sequence of each dispatch subtask; wherein a historical dispatch subtask sample marked with the execution device and control code is taken as input, the execution analysis model is taken as output, and a large language model is trained.

2. The large model-based power dispatch control method of claim 1, wherein, The method further comprises the following steps: analyzing the power dispatch demand by a task decomposition algorithm to obtain a plurality of dispatch subtasks and a dependency relationship between each dispatch subtask; based on the dependency relationship between a dispatch subtask, determining the execution sequence of each dispatch subtask. 3.The large model-based power dispatching control method according to claim 1, characterized in that, The method further comprises the following steps: determining whether the execution sequence of each dispatch subtask has a parallel sequence; wherein the execution sequence comprises a parallel sequence and a sequence; if there is, marking the parallel dispatch subtasks as a target subtask group, and controlling the control device corresponding to the dispatch subtask according to the sequence of the target subtask group and a plurality of non-parallel dispatch subtasks; wherein when the target subtask group is executed, the control device corresponding to the dispatch subtask of the target subtask group is controlled according to the parallel sequence; if not, controlling the control device according to the sequence of each dispatch subtask.

4. The large model-based power dispatch control method of claim 1, wherein, The dispatch subtask comprises a dispatch sequence of an execution device, parameters required for dispatch of the execution device, and a processing method for intermediate results generated by the execution device.

5. The large model-based power dispatch control method according to any one of claims 1 to 4, characterized by, After the control device is controlled according to the execution sequence of each dispatch subtask, the method further comprises the following steps: collecting running data of the control device; based on a preset fault database, matching the type and value of the running data; when the matching is successful, calling a fault response strategy of the control device corresponding to the running data from the fault database, and maintaining data collection of the control device; when the matching fails, no operation is performed.

6. A large model-based power dispatch control device, characterized by, The method comprises the following steps: a data acquisition module, a demand analysis module, a task splitting module, and a control scheduling module; the data acquisition module is configured to acquire a power dispatch request input by a user; The demand analysis module is configured to input the power dispatch request into a preset demand analysis model to obtain a power dispatch demand corresponding to the power dispatch request; the power dispatch demand includes dispatch content and a control device; a historical power dispatch request sample marked with the power dispatch demand is input, and the demand analysis model is output to train the large language model; The task splitting module is configured to confirm a plurality of dispatch subtasks and an execution sequence of each dispatch subtask based on the power dispatch demand; each dispatch subtask corresponds to a power dispatch demand; The control dispatch module is configured to input each dispatch subtask into an execution analysis model to generate an execution device and a control code corresponding to each dispatch subtask, and control the control device based on the execution device and the control code according to the execution sequence of each dispatch subtask; a historical dispatch subtask sample marked with the execution device and the control code is input, and the execution analysis model is output to train the large language model.

7. The large model-based power dispatch control device according to claim 6, wherein The confirming of the plurality of dispatch subtasks and the execution sequence of each dispatch subtask based on the power dispatch demand includes: analyzing the power dispatch demand by a task decomposition algorithm to obtain a plurality of dispatch subtasks and a dependency relationship between each dispatch subtask; determining the execution sequence of each dispatch subtask based on the dependency relationship between the dispatch subtasks. 8.The large model-based power dispatching control device according to claim 6, wherein, The controlling of the control device according to the execution sequence of each dispatch subtask includes: determining whether the execution sequence of each dispatch subtask includes a parallel sequence; the execution sequence includes a parallel sequence and a sequence; if yes, marking the parallel dispatch subtasks as a target subtask group, and controlling the control device corresponding to the dispatch subtask according to the sequence of the target subtask group and a plurality of non-parallel dispatch subtasks; when the target subtask group is executed, the control device corresponding to the dispatch subtask of the target subtask group is controlled according to the parallel sequence; if no, controlling the control device according to the sequence of each dispatch subtask. 9.The large model-based power dispatching control device according to claim 6, wherein, The dispatch subtask includes a dispatch sequence of an execution device, parameters required for dispatch of the execution device, and a processing method for intermediate results generated by the execution device.

10. The large model-based power dispatch control device according to any one of claims 6 to 9, characterized by, After the controlling of the control device according to the execution sequence of each dispatch subtask, the method further includes: collecting running data of the control device; matching the type and value of the running data based on a preset fault database; when the matching is successful, retrieving a fault response strategy of the control device corresponding to the running data from the fault database, and continuing to collect the data of the control device; when the matching fails, no operation is performed.

Images