CN-121998579-A - Power communication maintenance ticket examination and approval method and system

Abstract

The invention discloses a method and a system for checking and approving an electric power communication maintenance ticket. The method comprises the steps of obtaining electric power communication maintenance ticket data to be examined and approved, carrying out semantic understanding and information extraction on the electric power communication maintenance ticket data to be examined and approved by using a neural network module, outputting structural semantic information, inputting the structural semantic information into a symbol reasoning module, carrying out logic reasoning and compliance verification on the basis of a micro-symbol rule base by the symbol reasoning module, outputting an examination and approval result, converting an electric power communication maintenance ticket examination and approval rule into a symbol logic expression with weight by adopting a differential logic operator, constructing the micro-symbol rule base, constructing a double-layer optimization framework, and jointly optimizing parameters of the neural network module and weights of rules in the micro-symbol rule base on the basis of upper examination and approval task loss and lower symbol rule logic consistency loss. The invention obviously improves the efficiency, accuracy and auditability of approval and is suitable for intelligent processing of power grid communication overhaul service.

Inventors

• AN ZHIYUAN
• Quan Yizhan
• WANG LEI
• FENG HAO
• SHENG LEI
• ZHANG YUJIA
• ZHU PENGYU
• DING RUIQI
• LIU YAN
• JIN JING
• LIU MING
• CAI XINZHONG

Assignees

• 南京南瑞信息通信科技有限公司
• 国网河南省电力公司信息通信分公司

Dates

Publication Date

20260508

Application Date

20251231

Claims (12)

1. 1. The method for checking and approving the power communication maintenance ticket is characterized by comprising the following steps of: acquiring power communication maintenance ticket data to be examined and approved; using a neural network module to perform semantic understanding and information extraction on the power communication maintenance ticket data to be examined and approved, and outputting structural semantic information; inputting the structural semantic information into a symbol reasoning module, and carrying out logic reasoning and compliance verification by the symbol reasoning module based on a micro symbol rule base to output an approval result; The method comprises the steps of converting an electric power communication maintenance ticket approval rule into a symbol logic expression with weight by adopting a differentiable logic operator, and constructing a micro-symbol rule base; and constructing a double-layer optimization framework, and jointly optimizing parameters of the neural network module and weights of rules in the micro-symbol rule base based on upper-layer approval task loss and lower-layer symbol rule logic consistency loss.
2. 2. The power communication service ticket approval method of claim 1, wherein: pre-training the neural network module through a mask language modeling task by using professional texts in the communication field; acquiring multi-source data in an electric power communication operation and maintenance system, and marking the multi-source data set to construct a training set; and fine tuning the neural network module by using the training set, taking the fine-tuned neural network module as an initial strategy of reinforcement learning, and optimizing the neural network module through a compound rewarding function based on the initial strategy.
3. 3. The power communication service ticket approval method of claim 2, wherein: The composite reward function comprises a first reward function, a second reward function and a third reward function, wherein the first reward function is used for evaluating the performance of the neural network module on the approval task, the second reward function is used for evaluating the reasonability of the logic reasoning path of the neural network module, and the third reward function is used for encouraging the neural network module to explore various reasoning paths.
4. 4. The power communication service ticket approval method of claim 2, wherein: The loss function of the pre-training process comprises cross entropy loss and symbol rule consistency loss, and the symbol rule consistency loss function is introduced to strengthen logic output, wherein the symbol rule consistency loss is calculated according to whether the structural semantic information output by the neural network module is consistent with the logic reasoning result of the symbol reasoning module.
5. 5. The power communication service ticket approval method of claim 2, wherein: Labeling the multi-source dataset includes: defining formalized symbol rules based on the communication typical maintenance ticket templates and the safety regulations; marking an approval result for each historical maintenance ticket, marking the satisfaction state of each historical maintenance ticket for each symbol rule, marking the key semantic entity extracted from the historical maintenance ticket text, and establishing the mapping relation between the key semantic entity and the symbol rule.
6. 6. The power communication service ticket approval method of claim 1, wherein: Calculating the underlying symbol rule logical consistency loss includes: Calculating probability distribution about symbolic variables output under the condition of given input text; Calculating the satisfaction degree of the rule corresponding to the symbol variable according to the probability distribution; Taking the complement of the satisfaction degree as the probability distribution of the symbol variable and the distribution distance measurement of the corresponding rule; And carrying out weighted summation on the distribution distance measurement and the weight of the corresponding rule, and then calculating the mathematical expectation of the sum to obtain the logic consistency loss of the lower-layer symbol rule.
7. 7. The power communication service ticket approval method of claim 1, wherein: In the double-layer optimization framework, a dynamic attenuation entropy regularization term is added in an objective function formed by upper-layer approval task loss and lower-layer symbol rule logic consistency loss, wherein the information entropy of probability distribution of a symbol variable output under the condition of given input text is calculated, mathematical expectations are calculated on the information entropy and multiplied by a dynamic attenuation coefficient to obtain the entropy regularization term, and the final total loss function is obtained by subtracting the entropy regularization term from the objective function.
8. 8. The power communication service ticket approval method of claim 1, wherein: The method further comprises the steps of: Establishing a multi-agent cooperative structure, and outputting an approval result through the cooperation of the multi-agents, wherein the multi-agent cooperative structure comprises a master control agent and special agents, the master control agent distributes an approval task to the corresponding special agents, and each special agent respectively outputs an approval result and an inference path containing confidence evaluation to the master control agent after executing the approval task; And constructing intelligent agents based on the optimized neural network module and the symbol reasoning module, wherein each special intelligent agent corresponds to a micro-symbol rule base of different tasks, and loading the corresponding micro-symbol rule base and the prompt word template according to different examination tasks so as to enable each special intelligent agent to execute different examination tasks.
9. 9. The power communication service ticket approval method of claim 7, wherein: The special agent comprises: The basic information checking intelligent body is used for checking basic information of the overhaul ticket, including a title, an overhaul ticket source, overhaul grading and overhaul completion time; The working content examination intelligent body is used for examining the integrity, accuracy and compliance of the working content; the business influence checking agent is used for checking the consistency of business influence conditions and business ledgers.
10. 10. An intelligent power communication ticket approval system for implementing the intelligent power communication ticket approval method of any one of claims 1 to 9, the system comprising: The acquisition module is used for acquiring the to-be-examined and approved power communication maintenance ticket data; the semantic understanding and information extraction module is used for carrying out semantic understanding and information extraction on the power communication maintenance ticket data to be examined and approved by using the neural network module and outputting structural semantic information; The logical reasoning module is used for inputting the structural semantic information into the symbol reasoning module, and the symbol reasoning module carries out logical reasoning and compliance verification based on the micro-symbol rule base and outputs an approval result; the conversion module is used for converting the electric power communication overhaul ticket approval rule into a symbol logic expression with weight by adopting a differentiable logic operator to construct a micro-symbol rule base; And the optimizing module is used for constructing a double-layer optimizing framework, and is used for jointly optimizing parameters of the neural network module and weights of rules in the micro-symbol rule base based on upper-layer approval task loss and lower-layer symbol rule logic consistency loss.
11. 11. An electronic device comprises a processor and a storage medium, and is characterized in that: The storage medium is used for storing instructions; The processor is configured to operate in accordance with the instructions to perform the steps of the power communication service ticket approval method according to any one of claims 1-9.
12. 12. A computer readable storage medium having stored thereon a computer program, characterized in that the program when executed by a processor implements the steps of the power communication service ticket approval method of any one of claims 1-9.

Description

Power communication maintenance ticket examination and approval method and system Technical Field The invention belongs to the technical field of artificial intelligence application of electric power systems, and particularly relates to an electric power communication overhaul ticket examination and approval method and system, which are suitable for intelligent processing of electric power network communication overhaul business. Background The communication maintenance ticket is used as a core business document for guaranteeing the safe and stable operation of the power grid, and an approval link is a key node in the whole maintenance process, so that the communication maintenance ticket has a vital function for the safe compliance development of maintenance operation. At present, the examination and approval work of the ticket is mainly finished manually, and examination and approval staff need to check the ticket face content item by combining complicated rule documents by means of professional knowledge and experience. The mode has obvious limitations that manual approval efficiency is low, high-efficiency operation and maintenance requirements of a modern power grid are difficult to meet, a rule system is complex, multi-dimensional constraint is covered, manual comprehensive grasp is difficult, misjudgment or missed detection is easy to cause, approval decisions are highly dependent on subjective judgment of personnel, and different approval personnel have differences in understanding of ticket face contents and grasping of rule scales, so that objectivity and consistency of approval conclusions are directly affected. Some intelligent approval schemes based on large language models have appeared in the prior art, and attempts are made to improve approval efficiency through semantic understanding capability of a neural network. However, the method has the fundamental defects in practical application that firstly, a model lacks symbol reasoning capability, a typical examination and approval basis such as an examination and approval ticket template, a business ledger and the like cannot be converted into executable logic rules, so that compliance verification accuracy of examination and approval contents is insufficient, secondly, a model decision process does not have interpretability, a traceable path between an examination and approval conclusion and a specific rule is difficult to establish, the requirement of electric power communication security audit cannot be met, thirdly, the model has insufficient processing capability on uncertainty factors in the examination and approval process, excessive confidence or model illusion problems easily occur, and the reliability of the model in a real operation and maintenance environment is influenced. These defects lead to the difficulty in substantial improvement of the examination and approval efficiency of the maintenance ticket, and the operation risk caused by misjudgment and missed detection cannot be effectively controlled, so that the large-scale application in the power communication operation and maintenance system is difficult to realize. Disclosure of Invention In order to solve the defects in the prior art, the application provides a method and a system for checking and approving the power communication maintenance ticket, which realize efficient, accurate and interpretable intelligent checking and approving of the maintenance ticket. The invention adopts the following technical scheme. The first aspect of the invention provides a method for checking and approving an electric power communication service ticket, which comprises the following steps: acquiring power communication maintenance ticket data to be examined and approved; using a neural network module to perform semantic understanding and information extraction on the power communication maintenance ticket data to be examined and approved, and outputting structural semantic information; inputting the structural semantic information into a symbol reasoning module, and carrying out logic reasoning and compliance verification by the symbol reasoning module based on a micro symbol rule base to output an approval result; The method comprises the steps of converting an electric power communication maintenance ticket approval rule into a symbol logic expression with weight by adopting a differentiable logic operator, and constructing a micro-symbol rule base; and constructing a double-layer optimization framework, and jointly optimizing parameters of the neural network module and weights of rules in the micro-symbol rule base based on upper-layer approval task loss and lower-layer symbol rule logic consistency loss. Optionally, pre-training the neural network module by using the professional text in the communication field through a mask language modeling task; acquiring multi-source data in an electric power communication operation and maintenance system, and marking the multi-source data set to