Power supply service command decision method and device based on large model, and medium

Abstract

The invention relates to a large-model-based power supply service command decision-making method and device and a medium, and belongs to the technical field of power supply services, and the method comprises the following steps: obtaining a task event, analyzing the resource type of the task event, carrying out word segmentation processing on a task event scheme, calculating the importance score of each segmented word by using methods such as TF-IDF and the like; and taking the segmented words greater than or equal to the score threshold as keywords. And searching similar historical disposal schemes according to the keywords and the resource types, extracting decision variable vectors, objective function coefficient vectors and resource limit vectors in the historical disposal schemes, constructing an objective function with a minimum decision cost as an objective, and solving the objective function to obtain an optimal resource allocation amount. And inputting the task event, the similar historical disposal scheme and the optimal resource allocation into the large model, and outputting a power supply service command decision scheme. According to the method, by means of natural language processing (such as BERT word segmentation) and a similar historical scheme matching mechanism, automatic event analysis and rapid case retrieval are achieved.

Description

TECHNICAL FIELD

[0001] The application relates to a large model-based power supply service command decision method, device and medium, and belongs to the technical field of power supply services. BACKGROUND

[0002] In modern society, as a key infrastructure for economic operation and resident life, the stability and reliability of power supply are of great importance. With the rapid development of social economy, the scale of power supply network continues to expand, and the structure becomes more complex, which puts forward more stringent requirements for power supply service command decision. Power supply service command decision not only needs to accurately allocate resources such as repair personnel, vehicles and materials, but also needs to quickly respond to various fault events to reduce the adverse impact of power failure on social economy.

[0003] However, the traditional power supply service command decision method has exposed defects such as low efficiency and easy to make mistakes when dealing with complex and variable power supply scenes. The existing technology mainly depends on manual scheduling and simple rule matching, and it is difficult to integrate historical data, real-time information and resource constraints, which leads to the difficulty of efficient and accurate emergency disposal and resource allocation. Especially in the face of large-scale fault events, the traditional method has been difficult to meet the needs of modernization of power supply services. SUMMARY

[0004] In order to solve the problems existing in the prior art, the application provides a large model-based power supply service command decision method, device and medium.

[0005] The technical scheme of the application is as follows:

[0006] On the one hand, the application provides a large model-based power supply service command decision method, which comprises the following steps:

[0007] Obtaining a task event, analyzing the resource type of the task event;

[0008] Performing word segmentation processing on the task event scheme, and calculating the importance score of each word segmentation, and taking the word segmentation corresponding to the importance score greater than or equal to the score threshold as a keyword;

[0009] Obtaining a similar historical disposal scheme of the task event based on the keyword and the resource type;

[0010] Extracting the decision variable vector, the target function coefficient vector and the resource limit vector in the similar historical disposal scheme, and constructing a target function with the objective of minimizing the decision cost based on the decision variable vector, the target function coefficient vector and the resource limit vector;

[0011] Solving the target function to obtain an optimal resource allocation;

[0012] The task event, similar historical disposal scheme, and optimal resource allocation are input into a large model, and a power supply service command decision scheme is output.

[0013] Preferably, the task event includes a fault repair task and a power distribution network fault event.

[0014] The fault repair task is obtained through a customer service system.

[0015] The power distribution network fault event is obtained through a power distribution network detection system.

[0016] Preferably, the importance of each word is obtained by using a TF-IDF algorithm.

[0017] Preferably, the similar historical disposal scheme of the task event is obtained based on the keyword and the resource type, and the specific steps are as follows:

[0018] The database corresponding to the resource type is queried to obtain the historical disposal scheme.

[0019] The historical disposal scheme is split into multiple texts, the cosine similarity between the keyword and each text is calculated, and the average value of the cosine similarity between the keyword and each text is taken as the cosine similarity score of the historical disposal scheme.

[0020] The historical disposal scheme with the highest cosine similarity score is taken as the similar historical disposal scheme.

[0021] Preferably, a pre-trained BERT model is used for word segmentation processing of the task event scheme.

[0022] Preferably, the decision variable vector is expressed by a formula as follows:

[0023] X = [x1, x2, x3] T ;

[0024] In the formula, X represents the decision variable vector, x1 represents the number of repair personnel, x2 represents the number of repair vehicles, and x3 represents the number of materials.

[0025] The target function coefficient vector is expressed by a formula as follows:

[0026] C = [c1, c2, c3] T ;

[0027] In the formula, C represents the target function coefficient vector, c1 represents the labor cost per person per hour, c2 represents the operation and maintenance cost per vehicle per hour, and c3 represents the material consumption cost.

[0028] The resource limit vector is expressed by a formula as follows:

[0029] B = [b1, b2, b3, b4] T ;

[0030] In the formula, b1 represents the total number of available personnel, b2 represents the total number of available vehicles, b3 represents the total number of materials, and b4 represents the upper limit of the task time.

[0031] The objective function is expressed in a formula as follows:

[0032] min C T X;

[0033] AX = B;

[0034] In the formula, min represents a minimum value function, C T represents the transpose of the objective function coefficient vector.

[0035] Preferably, the large model is a general large model.

[0036] Preferably, the method further comprises auditing the large model output power supply service command decision scheme by an auditor.

[0037] In another aspect, the present application also provides an electronic device having a computer program stored thereon, wherein the computer program is executed by a processor to implement the large model-based power supply service command decision method according to any one of the embodiments of the present application.

[0038] In another aspect, the present application also provides a computer readable storage medium for storing one or more programs, wherein the one or more programs, when executed by one or more processors, cause the one or more processors to implement the large model-based power supply service command decision method according to any one of the embodiments of the present application.

[0039] The present application has the following beneficial effects:

[0040] 1. The present application realizes automatic analysis of event description and rapid retrieval of similar cases through natural language processing (such as BERT word segmentation and TF-IDF keyword extraction) and similar historical scheme matching mechanism. Further, by constructing a mathematical optimization model (such as a linear programming objective function), an optimal resource allocation scheme is automatically solved, which greatly shortens the decision time. Combined with the reasoning ability of a general large model (such as DeepSeek), the system can generate a command scheme within a few minutes, significantly improving the response speed to sudden fault events, reducing the outage time, and enhancing the real-time performance and reliability of power supply services.

[0041] 2、The application converts the historical treatment scheme into mathematical expression (decision variable vector, resource limit vector, objective function coefficient) through structured modeling, establishes an optimization model with the goal of "minimizing decision cost", and realizes quantitative control of resource allocation. The model comprehensively considers multiple constraint conditions such as manpower, vehicles, materials, etc., avoiding the subjective bias and resource waste commonly seen in traditional methods. At the same time, the large model is used to deeply fuse and analyze the historical scheme and the current situation, output a command decision more in line with actual needs, improve the feasibility and accuracy of the scheme, thereby reducing the operating cost and improving the user satisfaction.

[0042] 3、The application constructs the objective function to clearly define the optimization direction (such as minimizing labor cost, vehicle operation and maintenance cost, and material consumption), and the system can automatically calculate the optimal resource allocation under multiple resource constraints. Not only can it avoid over-investment or insufficient allocation of resources, but also can realize global resource coordination when multiple tasks are concurrent. The introduction of historical data enables the system to have learning and adaptation ability, and can dynamically adjust the resource allocation strategy according to different event types, thereby realizing the optimization of resource use and cost saving in a long-term range, and improving the economic benefit and operation efficiency of the power supply enterprise. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 The flowchart of the method of the application is shown. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0045] It should be understood that the step numbers used herein are only for the convenience of description, and are not limited to the execution sequence of the steps.

[0046] It should be understood that the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in the specification and the appended claims of the application, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0047] The terms "include" and "contain" indicate the presence of the described features, whole, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.

[0048] The term "and / or" means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0049] Embodiment one:

[0050] Referring to Figure 1 The application provides a large model-based power supply service command decision method, comprising the following steps:

[0051] Obtaining a task event, analyzing the resource type of the task event, the resource type including electricity fee consultation, user electricity meter failure and power distribution failure;

[0052] Carrying out word segmentation processing on the task event scheme, and calculating the importance score of each word segmentation, taking the word segmentation corresponding to the importance score greater than or equal to the score threshold as a keyword;

[0053] Obtaining a similar historical disposal scheme of the task event based on the keyword and the resource type;

[0054] Extracting a decision variable vector, a target function coefficient vector and a resource limit vector in the similar historical disposal scheme, and constructing a target function with the minimum decision cost as the target based on the decision variable vector, the target function coefficient vector and the resource limit vector;

[0055] Solving the target function to obtain an optimal resource allocation amount;

[0056] Inputting the task event, the similar historical disposal scheme and the optimal resource allocation amount into a large model, and outputting a power supply service command decision scheme.

[0057] Preferably, the task event includes a failure repair task and a power distribution network failure event;

[0058] The failure repair task is obtained through a customer service system;

[0059] The power distribution network failure event is obtained through a power distribution network detection system.

[0060] Preferably, the importance of each word segmentation is obtained by using a TF-IDF algorithm.

[0061] Suppose the current task event scheme is: "a certain community transformer failure, needs emergency repair".

[0062] The word segmentation result is: ["certain", "community", "transformer", "failure", "needs", "emergency", "repair"];

[0063] Suppose the corpus contains the following historical disposal schemes:

[0064] "transformer overheating causes tripping";

[0065] "short circuit of distribution box in Guangming community, east district";

[0066] "high voltage line fault repair";

[0067] Calculate the TF-IDF of the word "transformer":

[0068] TF (in the current document): appears 1 time, total word count 7, TF = 1 / 7 ≈ 0.143;

[0069] IDF: total number of documents in the corpus 3, number of documents containing "transformer" 1, IDF = log(3 / 2) ≈ 0.176;

[0070] TF-IDF ≈ 0.143 × 0.176 ≈ 0.025;

[0071] If the value exceeds the preset threshold (such as 0.02), "transformer" is selected as the keyword.

[0072] Preferably, based on the keyword and resource type, a similar historical disposal scheme for the task event is obtained, and the specific steps are as follows:

[0073] Query the database corresponding to the resource type to obtain the historical disposal scheme;

[0074] Split the historical disposal scheme into multiple texts, calculate the cosine similarity between the keyword and each text, and take the average of the cosine similarity between the keyword and each text as the cosine similarity score of the historical disposal scheme;

[0075] Take the historical disposal scheme with the highest cosine similarity score as the similar historical disposal scheme.

[0076] Suppose we receive a task event:

[0077] Event description: "Due to rainwater backflow, the transformer in the underground distribution room of No. 3 building in Guangming community, east district, needs emergency repair."

[0078] Resource type: power distribution failure;

[0079] In the previous step, by performing word segmentation and TF-IDF analysis on the event description, the following keywords are extracted:

[0080] East district, distribution room, rainwater backflow, transformer, short circuit, repair;

[0081] S1, query the database corresponding to the resource type to obtain the historical disposal scheme:

[0082] The system will not query the historical scheme library of irrelevant resource types such as "electricity consultation" or "user meter failure", but will directly access the special historical scheme database marked as power distribution failure.

[0083] Suppose we found 3 historical treatment plans from this database:

[0084] HP-101:

[0085] Event description: "10kV cable in the development zone industrial park was cut by construction, causing power outage in the area."

[0086] Resource type: power distribution failure;

[0087] Treatment plan text: "Immediately dispatch 2 repair teams with 8 people, 3 engineering vehicles, 500 meters of spare cable, and 10 cable joint kits to the scene. First isolate the fault point, ensure power supply in the non-fault area, and then replace the cable. Estimated time 6 hours."

[0088] HP-102:

[0089] Event description: "Summer peak electricity consumption, city center commercial street box transformer overload and burn."

[0090] Resource type: power distribution failure;

[0091] Treatment plan text: "Send 1 repair team of 4 people, 1 special transformer transport vehicle, 1 crane, and 1 new transformer. First cut off the power, lift off the old transformer, install the new transformer and debug. Time-consuming 8 hours."

[0092] HP-103:

[0093] Event description: "Heavy rain, water enters the distribution room of the old community in the southern district, causing switch cabinet failure."

[0094] Resource type: power distribution failure;

[0095] Treatment plan text: "Send 1 repair team of 4 people, 1 set of pumping equipment, 2 repair vehicles, and dry equipment and spare switch cabinet components. First perform drainage and drying operations, check equipment damage, and replace damaged components. Time-consuming 5 hours."

[0096] S2, calculate the cosine similarity score of each historical treatment plan:

[0097] The system does not directly compare keywords with the entire plan, but rather splits each historical treatment plan into multiple texts. The simplest way to split is to take the event description and treatment plan text as two independent text segments.

[0098] Next, the system calculates the cosine similarity between our keyword list and each text segment of each historical plan.

[0099] 1. Text vectorization:

[0100] First, all text (the list of keywords and fragments of each historical solution) needs to be converted into vectors. Here, the idea of ​​the Bag-of-Words model is adopted, constructing a vector space based on all keywords and words appearing in the historical text.

[0101] Suppose our constructed vocabulary is as follows: [Eastern District, power distribution room, rainwater backflow, transformer, short circuit, emergency repair, development zone, cable, power outage, industrial park, commercial street, box-type, overload, burnout, rainstorm, Southern District, power distribution room, water ingress, switch cabinet, pumping, drying]

[0102] New task keyword vector:

[0103] [1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]; (because the keyword only contains the first 6 words)

[0104] HP-103 event description vector:

[0105] [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,0,0]; (Corresponding words: rainstorm, Chengnan District, power distribution room, flooding, switch cabinet)

[0106] HP-103 Disposal Plan Vector:

[0107] [0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1](Corresponding words: emergency repair, switch cabinet, pumping, drying);

[0108] 2. Calculate cosine similarity:

[0109] Cosine similarity is determined by measuring the cosine of the angle θ between two vectors, cos(θ). The formula is:

[0110] cos(θ)=(A·B) / (||A||*||B||);

[0111] Where A and B are vectors, "·" represents the dot product, and "||||" represents the magnitude of the vector.

[0112] Calculate the similarity to the description of the HP-103 incident:

[0113] Task vector A = [1,1,1,1,1,1,...];

[0114] HP-103 description vector B1 = [0,0,0,0,0,0,...];

[0115] A·B1=(1*0+1*0+...)=0;

[0116] cos(θ)_description = 0 / (|A|*|B1|) = 0;

[0117] Calculate the similarity to the HP-103 treatment plan:

[0118] HP-103 scheme vector B2 = [0,0,0,0,0,1,...];

[0119] A·B2=(1*0+1*0+1*0+1*0+1*0+**1*1**)=1

[0120] |A|=sqrt(6)≈2.45;

[0121] |B2|=sqrt(4)=2 (assuming the four words "emergency repair, switch cabinet, pumping, drying");

[0122] cos(θ)_scheme = 1 / (2.45*2)≈0.204;

[0123] 3. Calculate the average similarity score of HP-103:

[0124] HP-103 score = (cos(θ)_description + cos(θ)_scheme) / 2 = (0 + 0.204) / 2 = 0.102;

[0125] The system will repeat this process, calculating the average similarity score between HP-101 and HP-102. Assume the results are as follows:

[0126] HP-101 score: 0.05 (because the keywords are not very relevant to "cable" and "industrial park");

[0127] HP-102 score: 0.08 (because they share the common word "transformer," but the causes of "overload burnout" and "rain short circuit" are different);

[0128] HP-103 score: 0.102;

[0129] S3. Select the historical handling plan with the highest score:

[0130] Comparing the scores of the three scenarios, HP-103 (flooding of the power distribution room due to heavy rain) had the highest cosine similarity score (0.102). Therefore, the system will determine HP-103 as the "most similar historical handling scenario" to the current task event.

[0131] Preferably, a pre-trained BERT model is used to perform word segmentation on the task event scheme.

[0132] Preferably, the decision variable vector is expressed in a formula as follows:

[0133] X = [x1, x2, x3] T ;

[0134] In the formula, X represents the decision variable vector, x1 represents the number of repair personnel, x2 represents the number of repair vehicles, and x3 represents the number of materials;

[0135] The objective function coefficient vector is expressed in a formula as follows:

[0136] C = [c1, c2, c3] T ;

[0137] In the formula, C represents the objective function coefficient vector, c1 represents the labor cost per person per hour, c2 represents the operation and maintenance cost per vehicle per hour, and c3 represents the material consumption cost;

[0138] The resource limit vector is expressed in a formula as follows:

[0139] B = [b1, b2, b3, b4] T ;

[0140] In the formula, b1 represents the total number of available personnel, which is 100 people, b2 represents the total number of available vehicles, which is 20 vehicles, b3 represents the total number of materials, which is 50 tons, and b4 represents the upper limit of the task time, which is 24 hours;

[0141] The objective function is expressed in a formula as follows:

[0142] min C T X;

[0143] AX = B;

[0144] In the formula, min represents the minimum function, C T represents the transpose of the objective function coefficient vector.

[0145] Preferably, the large model is a general large model, including DeepSeek model, Wenxin Yiyang, and kimi model, etc.

[0146] The solution method of the objective function is as follows:

[0147] The objective function is converted into a double-objective form, which is expressed in a formula as follows:

[0148] min z = α·f1(X) + (1-α)·f2(X);

[0149] In the formula, α represents the resource consumption weight, f1 represents the resource consumption function, f2 represents the task completion time function, and z represents the decision cost. The result of solving minz is the result of X, that is, the optimal resource allocation.

[0150] The resource consumption function is expressed in a formula as:

[0151] f1(X) = c1x1 + c2x2 + c3x3;

[0152] The task completion time function is expressed in a formula as:

[0153] f2(X) = b1x1 + b2x2 + b3x3;

[0154] Preferably, the method further comprises auditing the large model output power supply service command decision scheme by an auditor. If the content is completely correct, it is directly confirmed; if there is an error in the content, the auditor corrects the error.

[0155] Embodiment two:

[0156] The embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the power supply service command decision method based on a large model according to any embodiment of the present application when executing the program.

[0157] Embodiment three:

[0158] The embodiment provides a computer readable storage medium, which stores a computer program, and the program is executable on a processor to implement the power supply service command decision method based on a large model according to any embodiment of the present application.

[0159] In the embodiments of the present application, "at least one" means one or more, and "multiple" means two or more. The "and / or" describes the association relationship of the associated objects, which means that there can be three kinds of relationships, for example, A and / or B, which can represent the cases of A alone, A and B together, and B alone. Wherein A and B can be singular or plural. The character " / " generally represents an "or" relationship between the front and rear associated objects. "At least one of the following" and the like means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, wherein a, b, and c can be single or multiple.

[0160] Those of ordinary skill in the art can realize that the units and algorithm steps described in the embodiments disclosed herein can be realized by electronic hardware, computer software and a combination of electronic hardware and computer software. Whether the functions are realized in hardware or software mode depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0161] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the system, device and unit described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0162] In several embodiments provided in the present application, any function realized in the form of a software function unit and sold or used as an independent product can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts of the prior art that make contributions or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (Read-Only Memory; hereinafter referred to as ROM), a random access memory (Random Access Memory; hereinafter referred to as RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0163] The above is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent flow transformation, or direct or indirect application in other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A large model-based power supply service command decision method, characterized in that, The method comprises the following steps: obtaining a task event, analyzing a resource type of the task event; performing word segmentation processing on the task event scheme, and calculating importance scores of each word segmentation, taking word segmentation corresponding to an importance score greater than or equal to a score threshold as a keyword; obtaining a similar historical disposal scheme of the task event based on the keyword and the resource type; extracting a decision variable vector, a target function coefficient vector and a resource limit vector in the similar historical disposal scheme, and constructing a target function aiming to minimize a decision cost based on the decision variable vector, the target function coefficient vector and the resource limit vector; solving the target function to obtain an optimal resource allocation amount; inputting the task event, the similar historical disposal scheme and the optimal resource allocation amount into a large model, and outputting a power supply service command decision scheme.

2. The large model-based power supply service command decision method according to claim 1, characterized in that, The task event comprises a fault repair task and a power distribution network fault event. The fault repair task is obtained through a customer service system. The power distribution network fault event is obtained through a power distribution network detection system.

3. The large model-based power supply service command decision method according to claim 1, characterized in that, The importance of each word segmentation is obtained by using a TF-IDF algorithm.

4. The large model-based power supply service command decision method according to claim 1, characterized in that, The specific steps of obtaining the similar historical disposal scheme of the task event based on the keyword and the resource type are as follows: querying a database corresponding to the resource type to obtain historical disposal schemes; splitting the historical disposal schemes into multiple texts, calculating cosine similarity of the keyword and each text, and taking an average value of the cosine similarity of each text as a cosine similarity score of the historical disposal scheme; taking the historical disposal scheme with the highest cosine similarity score as the similar historical disposal scheme.

5. The large model-based power supply service command decision method according to claim 1, characterized in that, The task event scheme is processed by using a pre-trained BERT model.

6. The large model-based power supply service command decision method according to claim 1, characterized in that, The decision variable vector is expressed by a formula as follows: X = [x1, x2, x3] T ; In the formula, X represents the decision variable vector, x1 represents the number of repair personnel, x2 represents the number of repair vehicles, and x3 represents the number of materials. The target function coefficient vector is expressed by a formula as follows: C = [c1, c2, c3] T ; In the formula, C represents the target function coefficient vector, c1 represents the labor cost per person per hour, c2 represents the operation and maintenance cost of each vehicle per hour, and c3 represents the material consumption cost. The resource limit vector is expressed by a formula as follows: B = [b1, b2, b3, b4] T ; In the formula, b1 represents the total number of available personnel, b2 represents the total number of available vehicles, b3 represents the total number of materials, and b4 represents the upper limit of the task time. The target function is expressed by a formula as follows: min C T X; AX = B. where min denotes a minimum function, C T denotes the transpose of the target function coefficient vector.

7. The large model-based power supply service command decision method according to claim 1, characterized in that, The large model is a general large model. 8.The large model-based power supply service command decision method according to claim 1, wherein, The method further comprises auditing the power supply service command decision scheme output by the large model by an auditor.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the method of any one of claims 1 to 8 when executing the program.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program implements the method of any one of claims 1 to 8 when executed by the processor.

Images