CN-121981697-A - Intelligent auxiliary decision-making method and system for power load management operation and maintenance detection

CN121981697ACN 121981697 ACN121981697 ACN 121981697ACN-121981697-A

Abstract

The invention provides an intelligent auxiliary decision-making method and system for power load management operation and maintenance detection, comprising the following steps: collecting real-time detection data of a power load management system terminal and a control loop; generating a detection report and an improvement suggestion through a pre-trained intelligent auxiliary decision-making system according to the real-time detection data; the detection report and the improvement suggestion are used as intelligent auxiliary strategies and are displayed to operation and maintenance detection personnel; the intelligent auxiliary decision-making system can rapidly process and analyze real-time detection data, generate professional detection reports and improvement suggestions, shorten analysis detection time, improve overall detection efficiency, help operation and maintenance personnel to rapidly locate and solve problems, improve overall operation and maintenance quality and efficiency, ensure stability and safety of an electric power load management system, reduce dependence on high-skill professionals, improve detection and analysis efficiency, reduce training and operation and maintenance costs, and save a large amount of manpower and time cost.

Inventors

CHEN KE
CHEN SONGSONG
WANG SHUYANG
GONG TAORONG
GUO QIANG

Assignees

中国电力科学研究院有限公司
国网山东省电力公司
国家电网有限公司

Dates

Publication Date: 20260505
Application Date: 20241031

Claims (14)

1. An intelligent auxiliary decision-making method for detecting operation and maintenance of power load management is characterized by comprising the following steps: Collecting real-time detection data of a power load management system terminal and a control loop; Generating a detection report and an improvement suggestion through a pre-trained intelligent auxiliary decision-making system according to the real-time detection data; The detection report and the improvement suggestion are used as intelligent auxiliary strategies and are displayed to operation and maintenance detection personnel; The intelligent auxiliary decision-making system is based on historical operation and maintenance detection data and historical current detection data of the power load management system and is obtained by training a simplified natural language processing DistilBERT model.
2. The method of claim 1, wherein the training of the intelligent auxiliary decision system comprises: collecting historical operation and maintenance detection data and historical current detection data of a power load management system; based on the historical operation and maintenance detection data and the historical current detection data, generating a historical operation and maintenance detection data word vector and a historical current detection data word vector which can be identified by a simplified natural language processing DistilBERT model by adopting word segmentation processing and a mark embedding method; Training the simplified natural language processing DistilBERT model based on the historical operation and maintenance detection data word vector, the historical current detection data word vector and the training task, and completing training of the simplified natural language processing DistilBERT model when the iteration termination condition is met; The trained simplified natural language processing DistilBERT model is integrated into an intelligent auxiliary decision-making system in a back-end service mode, so that training of the intelligent auxiliary decision-making system is completed; wherein the historical operation and maintenance detection data comprises one or more of historical maintenance record data, historical fault log data and historical operation data.
3. The method of claim 2, wherein training the reduced natural language processing DistilBERT model based on the historical operation and maintenance detection data word vector and the historical current detection data word vector and a training task, when an iteration termination condition is met, completing training of the reduced natural language processing DistilBERT model, comprising: based on the historical operation and maintenance detection data word vector and the historical current detection data word vector, completing a masking language training task and a next sentence prediction training task of a simplified natural language processing DistilBERT model; after each time the masking language training task and the next sentence prediction training task are completed, calculating a loss function of the simplified natural language processing DistilBERT model, and evaluating the performance of the simplified natural language processing DistilBERT model according to the loss function; and when the performance reaches the triggering condition of the early stop system or the training times reach the preset times, the construction of the model of the simplified natural language processing DistilBERT is completed.
4. The method of claim 3, wherein the completing the mask language training task and the next sentence prediction training task of the reduced natural language processing DistilBERT model based on the historical operation and maintenance detection data word vector and the historical current detection data word vector comprises: masking the historical operation and maintenance detection data word vector and the historical current detection data word vector according to a preset masking proportion, and then carrying out masking prediction through the masking language training task to obtain a masking prediction result, thereby completing the masking language training task; And randomly extracting the historical operation and maintenance detection data word vector or the historical current detection data word vector from the historical operation and maintenance detection data word vector and the historical current detection data word vector, predicting the extracted historical operation and maintenance detection data word vector or the next sentence word vector of the historical current detection data word vector through the next sentence prediction training task, and obtaining a judging result according to the predicted next sentence word vector and the actual next sentence word vector to finish the next sentence prediction training task.
5. The method of claim 1, wherein after collecting the real-time detection data of the power load management system terminal and the control loop, before generating the detection report and the improvement suggestion by the intelligent auxiliary decision-making system according to the real-time detection data, further comprising: Performing data cleaning and data format conversion on the real-time detection data to generate real-time detection data after data preprocessing; wherein the data cleaning includes removing noise and removing redundant information.
6. The method of claim 1, wherein exposing the inspection report and improvement advice to an operation and maintenance inspector as intelligent assistance policies comprises: the detection report and the improvement suggestion are used as intelligent auxiliary strategies, and are displayed to operation and maintenance detection personnel through a user interface; the user interface comprises a data input interface, a report display interface and an operation guidance interface.
7. The intelligent auxiliary decision-making system for the power load management operation and maintenance detection is characterized by comprising a data acquisition module, a monitoring report suggestion generation module and a strategy display module; the data acquisition module is used for acquiring real-time detection data of the power load management system terminal and the control loop; the monitoring report suggestion generation module is used for generating a detection report and an improvement suggestion through a pre-trained intelligent auxiliary decision-making system according to the real-time detection data; the strategy display module is used for displaying the detection report and the improvement suggestion to operation and maintenance detection personnel as intelligent auxiliary strategies; The intelligent auxiliary decision-making system is based on historical operation and maintenance detection data and historical current detection data of the power load management system and is obtained by training a simplified natural language processing DistilBERT model.
8. The system of claim 7, wherein the system further comprises an intelligent auxiliary decision system training module, the intelligent auxiliary decision system training module comprising: The training data acquisition sub-module is used for acquiring historical operation and maintenance detection data and historical current detection data of the power load management system; The training data processing sub-module is used for generating a historical operation and maintenance detection data word vector and a historical current detection data word vector which can be identified by a simplified natural language processing DistilBERT model by adopting word segmentation processing and a mark embedding method based on the historical operation and maintenance detection data and the historical current detection data; The model iteration training sub-module is used for training the simplified natural language processing DistilBERT model based on the historical operation and maintenance detection data word vector, the historical current detection data word vector and the training task, and completing training of the simplified natural language processing DistilBERT model when the iteration termination condition is met; The intelligent auxiliary decision system integration sub-module is used for integrating the trained simplified natural language processing DistilBERT model into the intelligent auxiliary decision system in a back-end service mode to complete the training of the intelligent auxiliary decision system; wherein the historical operation and maintenance detection data comprises one or more of historical maintenance record data, historical fault log data and historical operation data.
9. The system of claim 8, wherein the model iterative training sub-module comprises: the task training unit is used for completing a masking language training task and a next sentence prediction training task of the simplified natural language processing DistilBERT model based on the historical operation and maintenance detection data word vector and the historical current detection data word vector; the model performance evaluation unit is used for calculating a loss function of the simplified natural language processing DistilBERT model after each time of completion of a masking language training task and a next sentence prediction training task, and evaluating the performance of the simplified natural language processing DistilBERT model according to the loss function; and the model training termination unit is used for completing the construction of the model for simplifying the natural language processing DistilBERT when the performance reaches the trigger condition of early stop or the training times reach the preset times.
10. The system according to claim 9, wherein the task training unit is specifically configured to: masking the historical operation and maintenance detection data word vector and the historical current detection data word vector according to a preset masking proportion, and then carrying out masking prediction through the masking language training task to obtain a masking prediction result, thereby completing the masking language training task; And randomly extracting the historical operation and maintenance detection data word vector or the historical current detection data word vector from the historical operation and maintenance detection data word vector and the historical current detection data word vector, predicting the extracted historical operation and maintenance detection data word vector or the next sentence word vector of the historical current detection data word vector through the next sentence prediction training task, and obtaining a judging result according to the predicted next sentence word vector and the actual next sentence word vector to finish the next sentence prediction training task.
11. The system of claim 7, wherein the system further comprises a data preprocessing module for: Performing data cleaning and data format conversion on the real-time detection data to generate real-time detection data after data preprocessing; wherein the data cleaning includes removing noise and removing redundant information.
12. The system of claim 7, wherein the policy presentation module is specifically configured to: the detection report and the improvement suggestion are used as intelligent auxiliary strategies, and are displayed to operation and maintenance detection personnel through a user interface; the user interface comprises a data input interface, a report display interface and an operation guidance interface.
13. The electronic equipment is characterized by comprising at least one processor and a memory, wherein the memory and the processor are connected through a bus; the memory is used for storing one or more programs; an intelligent decision-making assist method for power load management operation and maintenance detection according to any one of claims 1 to 6, when said one or more programs are executed by said at least one processor.
14. A readable storage medium having stored thereon an execution program, which when executed, implements an intelligent auxiliary decision method for power load management operation and maintenance detection according to any one of claims 1 to 6.

Description

Intelligent auxiliary decision-making method and system for power load management operation and maintenance detection Technical Field The invention belongs to the technical field of power system load operation and maintenance detection, and particularly relates to an intelligent auxiliary decision-making method and system for power load management operation and maintenance detection. Background In recent years, with continuous optimization of energy structures and intelligent development of power systems, power load management plays an increasingly important role in modern power grids. The load management system monitors and regulates the power load to realize reasonable distribution of power resources and stable operation of the system. However, in the operation and maintenance process, the conventional power load management system often depends on operation and maintenance personnel with rich expertise, which puts high demands on the expertise and experience of operators. The traditional load management system detector needs to have abundant expertise and experience for operators, can accurately analyze detection results, and can formulate corresponding operation and maintenance strategies. This approach not only places high demands on the training and maintenance of professionals, but also in actual operation, data analysis and problem localization often require a lot of time and labor, resulting in lower detection efficiency. In addition, with the continuous expansion of the scale and the increase of the complexity of the power system, the number of professional workers capable of performing detection and analysis work is limited, and the wide operation and maintenance requirements are difficult to meet. These limitations severely impact the efficient operation and maintenance of the electrical load management system. Therefore, the defects of the prior art include (1) that the traditional detector relies on workers with professional knowledge to detect and analyze, so that timeliness and popularity requirements of operation and maintenance detection of a novel power load management system are difficult to meet, (2) that the traditional detection method requires a large amount of time and labor in data analysis and problem positioning, so that detection efficiency is low, and (3) that the number of professional workers is limited along with the continuous expansion of the scale and the increase of complexity of a power system, so that the wide operation and maintenance requirements are difficult to meet. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides an intelligent auxiliary decision-making method for detecting the operation and maintenance of power load management, which comprises the following steps: Collecting real-time detection data of a power load management system terminal and a control loop; Generating a detection report and an improvement suggestion through a pre-trained intelligent auxiliary decision-making system according to the real-time detection data; The detection report and the improvement suggestion are used as intelligent auxiliary strategies and are displayed to operation and maintenance detection personnel; The intelligent auxiliary decision-making system is based on historical operation and maintenance detection data and historical current detection data of the power load management system and is obtained by training a simplified natural language processing DistilBERT model. Preferably, the training of the intelligent auxiliary decision system comprises: collecting historical operation and maintenance detection data and historical current detection data of a power load management system; based on the historical operation and maintenance detection data and the historical current detection data, generating a historical operation and maintenance detection data word vector and a historical current detection data word vector which can be identified by a simplified natural language processing DistilBERT model by adopting word segmentation processing and a mark embedding method; Training the simplified natural language processing DistilBERT model based on the historical operation and maintenance detection data word vector, the historical current detection data word vector and the training task, and completing training of the simplified natural language processing DistilBERT model when the iteration termination condition is met; The trained simplified natural language processing DistilBERT model is integrated into an intelligent auxiliary decision-making system in a back-end service mode, so that training of the intelligent auxiliary decision-making system is completed; wherein the historical operation and maintenance detection data comprises one or more of historical maintenance record data, historical fault log data and historical operation data. Preferably, the training the model of the reduced natural language processing DistilBERT bas