CN-120894006-B - Intelligent auxiliary decision-making method and system for large language model driven treatment system

CN120894006BCN 120894006 BCN120894006 BCN 120894006BCN-120894006-B

Abstract

The invention relates to a data learning technology, in particular to an intelligent auxiliary decision-making method and system of a large language model driving system, which comprises a data acquisition module, a data processing module, a large language model module and a decision output module, wherein when the periodic abnormal reasons are judged, the abnormal root cause positioning time is shortened through the accurate comparison of a business process period and an equipment maintenance period, the abnormal processing response speed is improved, when the non-periodic abnormal reasons are attributed, the differential response is realized through the dynamic analysis of a frequency change value, the intelligent decision-making of maintenance opportunity is realized through the comprehensive calculation of an abnormal grade, the immediate maintenance is triggered when the judgment is urgent, the loss caused by excessive equipment wear is avoided, the non-urgent situation is advanced, the service life of key equipment is prolonged, the maintenance cost is reduced, the environment data in a closed loop pipe is filtered in a linkage way from the abnormal identification, the attribution and decision-making to the verification to form the full process risk closed loop management, and the system can still keep stable operation under extreme working conditions.

Inventors

CAI CHUANJUN
WANG SHAOFEI
WANG BO
Song Yingshi

Assignees

山东中绿环智能科技有限公司

Dates

Publication Date: 20260508
Application Date: 20250721

Claims (7)

1. The intelligent auxiliary decision-making method of the large language model driven treatment system is characterized in that the intelligent auxiliary decision-making method is implemented through an intelligent auxiliary decision-making system, the intelligent auxiliary decision-making system comprises a data acquisition module, a data processing module, a large language model module and a decision output module, and the intelligent auxiliary decision-making method of the treatment system comprises the following steps: S1, carrying out abnormal value analysis on acquired data by a data processing module, if the number of abnormal values is larger than a preset number, expanding the fluctuation range of the abnormal values, and if the number of the abnormal values is still larger than the preset number, marking the moment as abnormal moment; S2, historical data are called and classified according to the use time, abnormal time frequency in each classification is calculated, and if the frequencies are the same, the abnormality is judged to have periodicity; further comparing the abnormal period with the business process period or the equipment maintenance period, and determining whether the abnormality is caused by business operation or maintenance operation according to whether the time period coincidence ratio or the time period is consistent; S3, if the frequencies are different, calculating a frequency change value, if the frequency change value is continuously increased, checking whether manual operation exists, judging that the system spontaneously recovers and records conditions, if the operation does not exist, confirming that the manual decision is effective and maintaining parameter monitoring, if the frequency change value is continuously reduced, judging that the risk is increased, generating an early warning signal when the threshold value is reached, if the frequency change value is continuously fluctuated, checking that the manual operation exists, re-detecting after finishing, and if the operation does not exist, generating an alarm signal, and triggering an audible and visual alarm or information reminding according to the frequency change amplitude; And S4, after the decision output module receives the signal, if the signal is the maintenance signal in advance, calculating the fluctuation ratio of the abnormal data mean value at the abnormal moment to the mean value at the adjacent moment to judge the abnormal grade, calculating the comprehensive abnormal grade by combining the ratio of the abnormal period length to the preset period length, determining the maintenance time according to the comparison result of the comprehensive abnormal grade and the preset threshold, if the signal is the early warning signal or the warning signal, notifying a worker to overhaul, recording the change of data during overhaul, comparing the recorded data with the preset fluctuation range, calculating the difference value, if the difference value is larger than 1, judging that the operation is effective and recording, otherwise, neglecting, and filtering accidental factors for the effective data to optimize the large language model.
2. The intelligent auxiliary decision-making method of the large language model driven purifying system according to claim 1, wherein the step of determining abnormal time by the data processing module is as follows: k1, sorting the collected data according to time, and averaging multiple data of the corresponding item collected at the same time And standard deviation B, and taking the mean value Detecting the fluctuation range of the data corresponding to the standard deviation B B is a positive integer; k2 marking the detection data of the corresponding item not in the fluctuation range as abnormal value, and marking the abnormal value quantity of the corresponding item Counting the total number of data detected by the corresponding items at the same time If (if) If the detected data is abnormal, the value b is counted and incremented by one to narrow the judging range of abnormal value, Is a preset proportion coefficient, and the ratio coefficient is a preset proportion coefficient, A sequence number representing the corresponding item data; K3, if the value b is counted and added by one, still meeting the requirement Marking the detection time as abnormal time, if Removing abnormal values, and averaging the rest corresponding item detection data To calculate the mean value As the corresponding item data detected at this time.
3. The intelligent auxiliary decision-making method of the large language model driven purifying system according to claim 2, wherein the data processing module performs the analysis step of periodically at abnormal time as follows: M1, calling historical data of corresponding item data, classifying the corresponding item historical data according to the time length of each use, calculating the frequency of abnormal time in the corresponding item data in each classification, and judging that the occurrence of the abnormality has periodicity if the frequency of the abnormal time in each classification is the same; M2, comparing the period of the abnormal moment with the service flow period of the clean system and the maintenance period of the equipment, if the period of the abnormal moment is the same as the service flow period of the clean system, primarily judging that the period caused by the service operation is abnormal, then comparing the period corresponding to the period of the abnormal moment with the period corresponding to the service flow period of the clean system, and if the period corresponding to the period of the abnormal moment occupies more than 80% of the period corresponding to the service flow period of the clean system, determining that the period caused by the service operation is abnormal; M3, if the period of the abnormal moment is the same as the maintenance period of the equipment, primarily judging that the period caused by the maintenance operation is abnormal, then comparing the period of the abnormal moment with the period of the maintenance period of the equipment, if the period of the two periods is the same, determining that the period caused by the maintenance operation is abnormal, and if the period of the abnormal moment is found by comparison to be not caused by the service flow period of the clean system and the maintenance period of the equipment, generating an advanced maintenance signal, and transmitting the advanced maintenance signal to the decision output module.
4. A large language model driven intelligent auxiliary decision-making method of a purifying system according to claim 3, wherein the data processing module performs the aperiodic analysis of abnormal time as follows: N1, calculating a frequency change value of an abnormal moment, if the frequency change value of the abnormal moment is continuously increased, acquiring manual operation data of a corresponding item in a set time period before a current time point, if the manual operation data of the corresponding item is not available, judging that the system is spontaneously recovered, recording recovery conditions, and providing reference for subsequent similar abnormal adjustment; N2, if the frequency change value at the abnormal moment is continuously reduced, judging that the risk of system failure or water quality deterioration is gradually increased, generating an early warning signal when the frequency change value at the abnormal moment reaches a preset frequency change threshold value, and transmitting the early warning signal to a decision output module; And N3, if the frequency change value of the abnormal moment continuously fluctuates, acquiring manual operation data of a corresponding item in a set time period before the current time point, if the manual operation data of the corresponding item exists, detecting the frequency change value of the abnormal moment again after the manual operation is finished, and if the manual operation data of the corresponding item does not exist, generating a warning signal and transmitting the warning signal to the decision output module.
5. The intelligent auxiliary decision-making method of the large language model driven purifying system according to claim 2, wherein the data processing module performs the following judgment of the effective operation data: The method comprises the steps of G1, detecting and recording changes of various items of data in the overhaul process, recording values of the corresponding items of data after debugging, and comparing the recorded values of the corresponding items of data after the changes with a preset fluctuation range of the corresponding items of data, wherein the preset fluctuation range is between the minimum value and the maximum value of the corresponding items of data; If the value after the change of the corresponding item data is not in the preset fluctuation range of the corresponding item data, calculating the absolute value of the difference between the value after the change of the corresponding item data and the preset fluctuation range of the corresponding item data, taking the minimum value of the absolute value of the calculated difference as the difference between the value after the change of the corresponding item data and the preset fluctuation range of the corresponding item data, and calculating the difference value between the value after the change of the corresponding item data and the preset fluctuation range of the corresponding item data, wherein the difference value is equal to the difference divided by the difference value of the adjacent values in the fluctuation range of the corresponding item data; And G3, if the difference value is larger than 1, judging that the operation adjustment is effective, and recording the data change caused by the operation adjustment, otherwise, judging that the operation adjustment is ineffective, and not recording the data change caused by the operation adjustment.
6. The intelligent auxiliary decision-making method of the large language model driven purifying system according to claim 2, wherein the filtering step of the accidental factors by the data processing module is as follows: p1, acquiring operation data recorded by corresponding operation in historical data, calling the implementation times of the corresponding operation data, and filtering the operation data with the implementation times less than the preset times to finish one-time data filtering and screening operation; P2, marking operation data obtained after the primary data filtering and screening operation is completed, marking environment data corresponding to the marked operation data again, acquiring environment data corresponding to the operation data identical to the marked operation data in the historical data, and comparing the marked environment data with corresponding items in the acquired environment data; P3, if the fluctuation of the environmental data corresponding to the two operation data is within the preset fluctuation range of the corresponding item, reserving the two operation data, otherwise, screening out the two groups of operation data to finish the secondary data filtering and screening operation; and P4, recording operation data after the secondary data filtering and screening operation is completed, substituting the operation data into the large language model, and optimizing the trained large language model.
7. The intelligent auxiliary decision-making method of large language model driven purifying system as claimed in claim 4, wherein the decision-making output module performs the following steps for the early maintenance signal: q1 mean value of abnormal data at abnormal time Calculating, acquiring corresponding item data of adjacent times of abnormal time, and calculating average value Will be the average And Dividing the difference value of (2) by a preset fluctuation value by a fluctuation ratio c, if The abnormality level is determined to be the d level, Is a preset comparison value; q2 cycle length at abnormal time And a preset period length Comparing and judging the abnormal grade as A stage; q3 calculation Sum of (2) And sum the sum value Comparing with a preset comparison value Comparing if Judging to maintain the original maintenance period if Then determine advance A maintenance operation is performed such that, Is a preset proportion coefficient, and the ratio coefficient is a preset proportion coefficient, For the duration between the current time point and the next maintenance period, if It is determined that the maintenance operation is performed immediately.

Description

Intelligent auxiliary decision-making method and system for large language model driven treatment system Technical Field The invention relates to a data learning technology, in particular to an intelligent auxiliary decision-making method and system for a large language model driving net treatment system. Background When the traditional purifying and treating system performs exception management, depending on a set fixed threshold, the fixed threshold is easy to cause a large number of false alarms or missed alarms when the water inflow load of a sewage treatment plant is suddenly changed, so that the accuracy of marking the exception time is lower, the time for manually checking the exception source is longer, and the response speed of the system is seriously influenced; the application provides a solution to the technical problem. Disclosure of Invention The invention aims to solve the problems in the background technology and provides an intelligent auxiliary decision-making method and system for a large language model driven clean-up system. The aim of the invention can be achieved by the following technical scheme: the intelligent auxiliary decision making method of the large language model driven treatment system is implemented through the intelligent auxiliary decision making system, the intelligent auxiliary decision making system comprises a data acquisition module, a data processing module, a large language model module and a decision output module, and the intelligent auxiliary decision making method of the treatment system comprises the following steps: s1, analyzing abnormal values of the data transmitted by the data acquisition module by the data processing module, expanding a fluctuation range for judging the abnormal values when the number of the abnormal values is larger than a preset number, and marking the moment as abnormal moment if the number of the abnormal values is still larger than the preset number; s2, historical data are called, abnormal time frequency in each category is calculated according to the use time category, if the frequency is the same, the abnormality is judged to have periodicity, if the abnormal period is the same as the business process period, the time period overlap ratio is further compared, if the abnormal period occupies more than 80% of the business process period, the operation is determined to be caused by business operation, if the abnormal period is the same as the equipment maintenance period, the time period of the abnormal period is compared with the equipment maintenance period, the operation is determined to be caused by maintenance operation, if the abnormal period is not related to the business process and the maintenance period, an early maintenance signal is generated and transmitted to a decision output module; s3, if the frequencies are different, calculating a frequency change value, if the frequency change value is continuously increased, checking whether manual operation exists, judging that the system spontaneously recovers and records conditions if no operation exists, confirming that the manual decision is effective if the operation exists, maintaining parameters and monitoring, if the frequency change value is continuously reduced, judging that the risk is increased, generating an early warning signal when the threshold value is reached, and if the frequency change value is continuously fluctuated, checking that the manual operation exists, detecting again after the operation is finished, otherwise, generating a warning signal, and triggering an audible and visual alarm or information reminding according to the frequency change amplitude; And S4, the decision output module receives the signal and informs a worker to carry out maintenance operation, in the maintenance operation process, the data after maintenance is recorded, the data is compared with a preset fluctuation range, a gap value is calculated, if the gap value is greater than 1, the operation is judged to be effective and recorded, otherwise, the operation is ignored, the effective data is filtered, accidental factors are removed, and the reserved effective data is used for optimizing a large language model after the filtration is completed. As a preferred embodiment of the present invention, the data processing module performs the following steps of: K1, sequencing collected data according to time, calculating a mean value A 1 and a standard deviation B of a plurality of data of corresponding items collected at the same time, setting a fluctuation range [ A 1-bB,A1 +bB ] of corresponding item detection data according to the mean value A 1 and the standard deviation B, wherein B is a positive integer; K2, marking the detection data of the corresponding item which is not in the fluctuation range as abnormal values, counting the abnormal value quantity Y a of the corresponding item, wherein the total number of the data detected by the corresponding item at the same time is Z a,