CN-121981093-A - Automatic collection and presentation generation method for monitoring data of centralized control station based on RPA

CN121981093ACN 121981093 ACN121981093 ACN 121981093ACN-121981093-A

Abstract

The invention discloses an automatic acquisition and presentation generation method of monitoring data of a centralized control station based on RPA (remote procedure alliance), which comprises the following steps of S1, constructing a standardized report template and a data key description word system, S2, deploying an RPA robot and configuring a multisource system access strategy, S3, executing automatic acquisition and preprocessing of data, S4, carrying out intelligent judgment and content generation based on key description words, S5, carrying out personalized screening and content optimization by applying a large language model, automatically completing cross-system data capture, time stamp alignment, duplicate removal and standardized mapping, regenerating a structured text initial manuscript, greatly shortening the overall time consumption of presentation generation, enabling monitors, operation and maintenance managers and other different roles to quickly acquire core information matched with responsibilities of the presentation aiming at presentation requirements of different posts of the centralized control station, avoiding redundant content interference, improving the pertinence and the service efficiency of the presentation, and ensuring the confidentiality of the presentation content by adopting a special encryption key.

Inventors

LV LILI
HU FAN
ZHENG HUIMIN
LIU LIFANG
XU HUAN
MA LING
TU ZHAO
CHEN YALIN
XU GUIFANG
ZHENG FENG

Assignees

国网湖北省电力有限公司黄冈供电公司

Dates

Publication Date: 20260505
Application Date: 20251226

Claims (10)

1. The method for automatically collecting and generating the centralized control station monitoring data and the briefing based on the RPA is characterized by comprising the following steps: S1, constructing a standardized report template and a data key description word system, namely creating a standard report template containing full-area maximum load, heavy overload main transformer, main transformer oil temperature, a large-current switch cabinet, important equipment monitoring and oil pump starting data, and defining data key description words for each type of data; S2, deploying an RPA robot and configuring a multi-source system access strategy, namely deploying an RPA robot flow automation engine under the intranet safety environment of the centralized control station, configuring the automatic login authority of the RPA robot, adopting encryption credential management to ensure the operation compliance, butting a multi-source service system and a database required by the centralized control station data report, and realizing cross-platform data directional capture; S3, executing automatic data acquisition and preprocessing, namely automatically starting an RPA (remote procedure alliance) robot according to a preset period, acquiring original data from a butt-joint multi-source service system and a database, synchronously calling a data preprocessing module, performing preprocessing operations such as time stamp alignment and duplicate removal, equipment type and alarm level standardization mapping and the like on the acquired data, and further performing multidimensional statistical analysis based on the preprocessed standardization data to generate a descriptive index set such as alarm total amount, high-level alarm duty ratio, peak time distribution and the like; S4, intelligent judgment and content generation are carried out based on the key description words, namely the descriptive index set generated in the S3 is matched with the key description words of the core data defined in the S1, abnormal or key events needing to be included in the briefing are automatically identified, and the abnormal or key events are filled into corresponding fields in a standardized briefing template to form a structured text manuscript; And S5, personalized screening and content optimization are carried out by applying a large language model, namely, key contents are automatically extracted by utilizing the large language model according to historical attention points and preferences of different user roles, key information is highlighted, and personalized bulletin is generated.
2. The method for automatically collecting and generating the briefing of the monitoring data of the centralized control station based on the RPA of claim 1, wherein in S1, the creation of the standard report template is used for forming one-to-one correspondence between the template and the description font system, and the method adopts a structural framework design and comprises 6 core data modules: the method comprises the steps of respectively corresponding to full-area maximum load, heavy overload main transformer, main transformer oil temperature, large-current switch cabinet, important equipment monitoring and oil pump starting data; Each module is provided with a data entry field, a state identification field, a threshold reference field and a description supplement field, and the internal structure is that Wherein Is the first A special-purpose module for the class data, , As a result of the data entry field, For the status-indication field, For the threshold reference field, Supplemental fields for illustration; The 6 core data modules based on the template respectively define corresponding data key description words to form a description font system Wherein With the template module One-to-one correspondence is used for defining the judging rule and the acquisition standard of the data in the module.
3. The method for automatically collecting and generating the briefing of the monitoring data of the centralized control station based on the RPA of claim 1, wherein in S1, the specific definition method for defining the data key description words for each type of data is as follows: maximum load module for full area Corresponding key description word Wherein To collect load data of a node at a certain time in a period, To collect the total number of time nodes in a cycle, Is that The corresponding time node is used for the time period, Entry template field , Entry field ; Main transformer module for heavy overload Corresponding key description word Wherein , For the actual load power of the main transformer, For the rated load power of the main transformer, Is that Duration of the super threshold, setting the threshold 、 Threshold entry field When (when) And is also provided with When, field Identifying heavy overload, otherwise, identifying normal; Module for main transformer oil temperature Corresponding key description word , Is the real-time oil temperature value of the main transformer, For an overtemperature threshold, the threshold is entered into a field , Entry field When (when) When, field The mark is an overtemperature alarm; For high-current switchgear modules Corresponding key description word , For the real-time running current of the switch cabinet, For the rated current to be the same, Entry field , Entry field Fields of According to And (3) with Is normal or overloaded; monitoring module for important equipment Corresponding key description word , For the core operating parameters of the device, For a normal threshold range of parameters, a threshold range entry field , Entry field When (when) When, field The identification is that the parameter is abnormal; For oil pump starting module Corresponding key description word , For the number of times the oil pump is started, For counting time window, set up Entry field Starting action and resetting in the same statistical window 1 time, namely , Entry field Fields of The start-up frequency level is identified.
4. The method for automatically collecting and generating the briefing of the monitoring data of the centralized control station based on the RPA of claim 1, wherein in S2, the configuration of the deployment and access strategy of the RPA robot specifically comprises the following steps: ① The RPA robot engine is deployed in the intranet of the centralized control station, and an independent running process and a resource isolation space are configured to avoid process conflict with other systems; ② An automatic login credential management module is configured, and an AES symmetric encryption algorithm is adopted for login credentials The encryption storage is carried out, and the encryption formula is that , Presetting an encryption key for 256 bits, and decrypting by Recovering the certificate, and storing the secret key through a hardware encryption module; ③ Configuration System Access priority sequence Wherein The corresponding D5000 schedule automation system, Corresponding to a new generation of centralized control system, In response to the scheduling of the report management system, Corresponds to the history alarm database and meets the following requirements The RPA robot logs in the system in sequence according to the priority order, if the access of the high priority system fails, a retry mechanism is triggered , Representing the preset maximum retry times, recording logs and jumping to the next priority system if the retry fails; the multi-source business system and the database comprise a D5000 dispatching automation system, a new generation centralized control system, a dispatching report management system and a historical alarm database.
5. The method for automatically collecting and generating the briefing of the monitoring data of the centralized control station based on the RPA of claim 1, wherein in S3, the collecting of the original data is specifically: The RPA robot is arranged according to a preset period Start-up at a fixed time point per day Starting acquisition according to access priority of S2 Sequentially capturing original data from each system to form an original data set Collected data of the four types of systems are respectively corresponding; The operation flow of the data preprocessing module comprises the following steps: ① Time stamp alignment, namely setting the time stamp of the original data as Is the first Class system No Time stamp of stripe data, standard time stamp format is By means of format conversion algorithms The format is unified, so that the consistency of all data time dimensions is ensured; ② Data de-duplication, namely calculating characteristic hash value of each piece of data by adopting a hash check algorithm Wherein For exclusive-or operation, if there is ( ) The data with earlier time stamp is reserved, the repeated data is deleted, and a data set after the duplication removal is obtained ; ③ Standardized mapping, namely establishing a device type mapping dictionary Mapping dictionary with alarm level Wherein For the name of the original device type, For a standard device type name, For the original alert level(s), For standard alarm level, through mapping function 、 Completing the standardized conversion to obtain standardized data ; Multidimensional statistical analysis is based on The development specifically comprises the following steps: Total alarm , The presentation data is alert class data, Representing non-alert data; High level alarm duty cycle , The number of the alarm data is high; peak time distribution, dividing the acquisition period into Calculating the load average value of each interval in 1 hour interval , Is the first A load data set of 1 hour intervals, Is that The number of data pieces in the data string, , Is data of Corresponding load value is taken Maximum of 3 intervals as peak period ; Finally form the descriptive index set And the index sets are disassembled into sub-index sets according to the data types , Key description word of kth class data in S1 One-to-one correspondence.
6. The method for automatically collecting and generating the briefing of the monitoring data of the centralized control station based on the RPA of claim 1, wherein in S4, the matching flow of the descriptive index set and the key description words of the core data is as follows: Establishing a matching operator Wherein As index set Corresponds to the first A sub-set of metrics for the class data, The first defined for S1 Class data key description words, the matching rule is: When (when) In the time-course of which the first and second contact surfaces, Directly entering template fields ; When (when) In the time-course of which the first and second contact surfaces, When the matching result is 1, the identification is overloaded again, and the field is entered ; The rest category data are correspondingly Matching is carried out according to the threshold rule of (2) to obtain a matching result set ; Automatic identification of anomalies or key events is screening Corresponding abnormal or key event data, extracting standardized information thereof Simultaneous extraction of Corresponding normal data basic information to form a data set to be filled , The matching result of the k-th class data; template filling, namely filling information to be filled according to k-th data With the template module Filling data into fields one by one Filling fields according to the matching result Supplementing relevant background information to fields ; Forming a structured text manuscript after filling Wherein Is the first after filling And a class data module.
7. The method for automatically collecting and generating the briefing of the centralized control station monitoring data based on the RPA of claim 1, wherein in S5, the application of the large language model realizes personalized optimization based on user preference modeling, and the specific process is as follows: ① Preference modeling, collection Historical operational data for user-like roles Wherein Including a monitor, an operation and maintenance manager, etc., Is the first Class user pair number Historical access record of class data, and calculation of attention weight through TF-IDF algorithm , Is the first Class user access The frequency of the class data is such that, Is the first Inverse document frequency, weight set of class data Satisfy the following requirements ; ② Important content refinement based on For text manuscript Is ordered by 6 modules, and the weight is reserved before ranking As core emphasis while forcing all reservation Forms a filtered content collection , ; ③ Highlighting key information, namely highlighting core content by adopting a attention mechanism to convert a text manuscript into a word vector matrix , For text word number, calculate attention weight Wherein The query vector is preferred for the user, In the form of a vector dot product, Greater than a preset threshold The vocabulary of (1) is used as key information and is strengthened by thickening, highlighting and other formats; ④ The personalized presentation is generated by integrating the enhanced core content and the reserved abnormal content, and recombining the text structure according to natural language logic to generate the text structure conforming to the first item Personalized presentation of class user requirements , wherein, Ranking weights ahead Is provided with a set of core data module indices, , In order to preset the number of core modules, For the abnormal data module index set, , , For attention weights exceeding a threshold Is a set of key information words of (a), The text recombination function is characterized by comprising a core module, a priority display abnormal module, a high-attention weight vocabulary enhancement module and a personalized presentation text conforming to natural language logic.
8. The method for automatically collecting and generating the centralized control station monitoring data and the briefing based on the RPA of claim 1, wherein in S5, the large language model is a GPT series model or a BERT series model, and the model training process and the data output of the preamble step form linkage optimization, specifically comprising the following steps: The model input data includes a structured text manuscript of S4 User preference weight of S5 The output is personalized brief report ; The model training adopts a supervised learning mode, and a training sample set is formed by historical manuscript data Corresponding user preference data Standard briefing after manual optimization Constructing; Calculation of training error using cross entropy loss function Wherein In order to obtain the number of samples, For the number of text categories to be the number, Is the first The true class labels of the individual samples are, Predicting class probabilities for the model; minimizing the loss function by Adam optimizer, optimizing model parameters Iteratively updating the formula to Wherein For the first order momentum estimation, For the second order momentum estimation, 、 And the momentum attenuation coefficient is used for ensuring that the model output is accurately matched with the user demand.
9. The method for automatically collecting and generating the briefing of the monitoring data of the centralized control station based on the RPA of claim 1, wherein in S2, the data grabbing process of the RPA robot adopts an incremental collecting strategy, specifically: Record last time of acquisition The collection only grabs New data of (2) after acquisition is completed The collection cut-off time is this time; If the record of the first acquisition or the last acquisition is lost, the full acquisition is executed, and an acquisition time stamp log is built after the acquisition is completed Wherein In order to start the time of the acquisition, For the end time of acquisition, the continuity and integrity of data acquisition is ensured.
10. The method for automatically collecting and generating the centralized control station monitoring data and the briefing based on the RPA of claim 9, further comprising S6 of generating the personalized briefing The method comprises the steps of packaging an encrypted text, calling a two-dimension code generation module to encode the encrypted text into a disposable dynamic two-dimension code, wherein the two-dimension code is effective only in a designated time period and does not contain sensitive system information, and a monitor can safely acquire a monitoring briefing on the same day by scanning the two-dimension code through an external network mobile terminal, wherein the specific process is as follows: ① Text encryption, namely, personalized presentation by adopting AES-256 algorithm Encryption is carried out, and the encryption formula is that Wherein Is a global master key dynamically distributed by a key management center and periodically updated in period ; ② Two-dimension code coding, namely calling a QRCode coding module to encrypt text The code is converted into a disposable dynamic two-dimensional code, and the coding process meets the following requirements Wherein version=4 is a two-dimensional code Version, errorLevel =h is a 30% fault tolerance level; ③ Effective period management and control, namely configuring effective time window for two-dimension code Setting up The two-dimensional code automatically fails after exceeding the window, and the two-dimensional code needs to be regenerated after the failure; ④ The security transmission, wherein a monitor scans the two-dimension code through an external network mobile terminal, and a built-in decryption module of the terminal executes And (5) operating, and displaying the complete brief report after decryption.

Description

Automatic collection and presentation generation method for monitoring data of centralized control station based on RPA Technical Field The invention relates to the technical field of operation and maintenance management of power systems, in particular to an automatic collection and presentation generation method of monitoring data of a centralized control station based on RPA. Background In the operation and maintenance management of the power system, the centralized control station bears the core responsibility of monitoring the operation state of the governed transformer substation equipment, and various core monitoring data such as the maximum load of the whole area, the heavy overload main transformer and the like are required to be summarized and reported on time every day, so that data support is provided for equipment operation and maintenance and scheduling decision. The data reporting of the centralized control station in the current industry mainly depends on a multi-source service system, such as a scheduling report management system, a historical alarm data query system and the like, and a plurality of processes of logging in the system, query screening, manual judgment, presentation filling, information conversion and transmission and the like are needed, so that the following problems exist in practical application: Firstly, the existing centralized control station data reporting process is highly dependent on manual operation, a monitor is required to manually log in a plurality of independent service systems to inquire required data, and then scattered data are screened, sorted and judged and then filled in a bulletin template, so that the whole flow link is tedious and time-consuming; in the process of manually copying and screening data, information deviation is easy to generate due to operation fatigue, standard understanding deviation and the like, so that the accuracy of the data reporting is difficult to guarantee, and the requirements on the data reporting efficiency and accuracy under the intelligent development of a centralized control station cannot be met; Secondly, the centralized control station data briefing generated in the prior art is in a unified format, the core demand difference of users at different positions is not considered, a monitor focuses on real-time abnormal event monitoring, an operation and maintenance manager focuses on equipment operation trend data, a technical responsible person needs to master the overall operation profile, but the briefing in the unified format contains a large amount of redundant information, users at different roles need to manually refine core content related to own responsibility from the briefing, so that the additional workload of the users is increased, key data can be omitted incompletely due to information screening, and the practical use value of the briefing is reduced; Third, there is strict security isolation requirement in the internal network and external network of the centralized control station, the data briefing of the existing centralized control station is transferred to the external network after copying the data from the internal network by using the mobile storage medium manually, and then the sending operation is completed; therefore, it is necessary to design an automatic collection and presentation generation method for monitoring data of a centralized control station based on RPA. Disclosure of Invention The invention aims to provide an automatic collection and presentation generation method of monitoring data of a centralized control station based on RPA (remote procedure analysis), which solves the problems that the conventional centralized control station data presentation relies on manual operation to cause complex flow, long time consumption, delay in presentation occurs easily in a large load period and data accuracy is difficult to guarantee in the background art, solves the problems that the conventional presentation format is lack of post personalized adaptation uniformly, redundant information often causes a user to manually refine core content and has low use efficiency, and simultaneously solves the problems that the conventional presentation internal and external network transmission relies on manual mobile storage media to cause long time consumption, virus transmission, sensitive data leakage and other security risks and the data transmission integrity is easy to be influenced. In order to achieve the above purpose, the present invention provides the following technical solutions: In a first aspect, a method for automatically collecting and generating a bulletin of monitoring data of a centralized control station based on RPA is provided, which includes the following steps: S1, constructing a standardized report template and a data key description word system, namely creating a standard report template containing full-area maximum load, heavy overload main transformer, main transformer oil temp