Search

CN-121979036-A - Gene library central control method and system based on artificial intelligence

CN121979036ACN 121979036 ACN121979036 ACN 121979036ACN-121979036-A

Abstract

The invention discloses a gene library central control method and system based on artificial intelligence, which are characterized in that multidimensional parameters are monitored in real time through all nodes, risk states are judged, synchronous risks are judged through a set window, after cleaning data are extracted, feature sets are formed through feature screening, cause verification and extraction, and a weighted scoring early warning model is constructed to realize hierarchical early warning and iterative optimization. The scheme can accurately identify the cross-node commonality risk, improve the early warning reliability and ensure the stable operation of the gene refrigeration house cluster.

Inventors

  • GAO KE
  • FAN JUNFANG
  • LIU FUXIANG

Assignees

  • 北京信息科技大学

Dates

Publication Date
20260505
Application Date
20260113

Claims (9)

  1. 1. The gene library central control method based on artificial intelligence is characterized by comprising the following steps: step 1, monitoring all node operation data and risk states in real time, collecting four key parameters of power grid, environment, resource and equipment operation, judging the risk states of all nodes in real time, recording related information, and uploading the information to a distributed fusion center; step 2, judging the multi-node contemporaneous risk, setting a contemporaneous judging window, judging the contemporaneous risk based on the window, and distinguishing the cross-node common risk from the single-node individual risk; Step 3, the contemporaneous risk monitoring data is extracted and cleaned, the contemporaneous risk window and the full-node data of the pre-preset duration are extracted, and an associated data set is generated after cleaning and standardization; And 4, analyzing the cause of the common risk, wherein the method comprises the following steps: 4.1 The feature screening, namely comparing risk with non-risk node data, and screening the features of the risk nodes with general abnormality and the non-risk nodes with normal as potential causes; 4.2 Checking historical contemporaneous risk cases, verifying the relevance of potential features and risk occurrence, and confirming common risk causes; 4.3 Feature extraction, namely converting the confirmed cause into a quantifiable index and definitely giving an early warning threshold value to form a common risk cause feature set; And 5, performing risk early warning application driven by the causal features, constructing an early warning model based on the common risk causal feature set, scoring according to the feature standard condition, classifying the risk grades, and triggering corresponding early warning.
  2. 2. The method according to claim 1, wherein the real-time monitoring of the all-node operation data and risk status in step 1 specifically comprises: 1.1 The monitoring data range covers key operation dimensions of the gene refrigerator cluster, and the standard and frequency are uniformly acquired, wherein the standard and frequency specifically comprise ① power grid parameters, ② environment parameters, ③ resource parameters, ④ equipment operation parameters, wherein the power grid parameters at least comprise voltage, current and power, the environment parameters at least comprise external temperature and humidity and temperature and humidity in the refrigerator, the resource parameters at least comprise liquid nitrogen liquid level and resource occupancy rate, the equipment operation parameters at least comprise compressor load and fan rotation speed, and the acquisition frequency is 200 ms/time; 1.2 Each node determines a local risk state in real time, a risk occurrence mark is 1, a risk-free mark is 0, risk occurrence time and risk level are synchronously recorded, the risk level comprises high, medium and low three levels, and operation data and risk state are uploaded to a distributed fusion center in real time, and SM4 encryption protocol is adopted for transmission.
  3. 3. The method according to claim 1, wherein the multi-node contemporaneous risk determination in step 2 specifically comprises: 2.1 Setting a 30-minute synchronization judging window, and judging that the occurrence time of a plurality of node risks falls in the same window as the synchronization risk; 2.2 Setting a node threshold, judging as a cross-node common risk when the number of the contemporaneous risk nodes is more than or equal to 30% of the total node number, starting subsequent data extraction, and otherwise, performing local recording only for single-node individual risk.
  4. 4. The method of claim 1, wherein the contemporaneous risk monitoring data extraction and cleansing of step 3 specifically comprises: 3.1 The data extraction range is full-node data in a contemporaneous risk window and 60 minutes before the window, and the full-node data comprises risk node data and non-risk node data and is used for subsequent comparative analysis; 3.2 And removing abnormal values by adopting a3 sigma principle, removing time period data with the deletion rate of more than 5%, and generating a contemporaneous risk association data set after carrying out standardized processing on the residual data.
  5. 5. The method of claim 1, wherein the feature screening in step 4.1 is specifically that a feature difference index is built for four major types of monitoring parameters in step 1 based on a contemporaneous risk association data set, a risk node set is set to be R, an un-risk node set is set to be N, a difference degree D calculation formula of a certain feature x is set to be D= | (1/|R|) sigma (R e R|) x_r- (1/|N|) sigma (N e N) x_n|/max (x_max-x_min, epsilon), wherein x_r is a feature value of a risk node, x_max, x_min is the maximum value of the feature total data, epsilon is a minimum value to avoid denominator is 0, and when D is more than or equal to 0.2, the feature is determined to be a significant feature of the risk node and the un-risk node, and the feature is included in a potential cause pool.
  6. 6. The method according to claim 1, wherein the factor verification in step 4.2 specifically comprises the steps of extracting contemporaneous risk historical data of a distributed fusion center for about 12 months to construct a case library, calculating the correlation between occurrence frequency and risk occurrence for each feature in a potential factor pool, setting the total number of historical contemporaneous risk cases as M, setting the number of cases of a certain potential feature f before risk occurrence as K, defining the correlation degree C=K/M, and if C is larger than or equal to 0.7 and the occurrence rate of the feature in a non-risk period is smaller than or equal to 0.3, confirming that the feature is a common risk factor.
  7. 7. The method according to claim 1, wherein the feature extraction in step 4.3 is specifically that the confirmed common risk causes are converted into quantifiable and monitorable standard features, a normal interval [ mu-1.5σ, mu+1.5σ ] of each cause feature is determined through box diagram analysis, wherein mu is a mean value, sigma is a standard deviation, and the abnormality is determined when the mu is a standard deviation and exceeds the interval, and a feature early warning threshold is corrected by combining an industry standard and an equipment threshold requirement, so that a common risk cause feature set comprising a feature name, a data dimension, an abnormality determination threshold and an associated risk level is finally formed.
  8. 8. The method according to claim 1, wherein the causal feature driven risk early warning application in step 5 specifically comprises: 5.1 Based on a cause feature set, a pre-warning model is constructed by adopting a threshold triggering and weighting scoring mechanism, wherein each feature weight in the pre-warning model comprises a power grid fluctuation weight of 0.4, an environment mutation weight of 0.3 and a resource tension weight of 0.3; 5.2 Triggering the score when the characteristic value reaches a threshold value, wherein the comprehensive score is more than or equal to 0.5 and is high in risk, and 0.3-0.5 is medium in risk, and the high risk and the medium in risk trigger early warning; 5.3 And (3) calibrating the characteristic threshold value and the weight every 50 groups of synchronous risk cases, so as to improve the early warning accuracy.
  9. 9. The gene warehouse central control system based on artificial intelligence is characterized in that the method for realizing the gene refrigerator cluster synchronization commonality risk early warning according to any one of claims 1-8 comprises a distributed fusion center and monitoring terminals deployed at nodes of each gene refrigerator; the monitoring terminal is used for performing real-time monitoring on the running data and the risk state of the full node in the step 1, collecting four major key parameters, judging the local risk state, recording related information, and encrypting and uploading the data to the distributed fusion center; the distributed fusion center is used for executing the operations of the steps 2-5, and comprises the steps of judging the synchronous risk of multiple nodes, extracting and cleaning monitoring data, analyzing common risk causes, constructing and operating an early warning model to realize risk early warning and iterative optimization.

Description

Gene library central control method and system based on artificial intelligence Technical Field The invention relates to the technical field of intelligent industry, in particular to a gene library central control method and system based on artificial intelligence. Background The traditional gene library manages the risk of crossing nodes by artificial experience, and intelligent technology is not introduced. The data acquisition focuses on core indexes such as temperature and humidity, liquid nitrogen liquid level and the like in a refrigeration house, ignores key dimensions such as power grid parameters, equipment loads and the like, has no unified standard, mainly comprises manual timing recording or low-frequency automatic acquisition, and has the frequency of minutes to hours. The data are stored in a local mode in a scattered mode, a cross-node shared channel is avoided, the risk judgment needs to manually collect the data afterwards, the relevance is judged by experience, the obvious errors are manually removed in data processing, the cause analysis depends on experience speculation, and the early warning depends on manual inspection and a general emergency scheme is adopted. The method has the obvious problems of incomplete data acquisition dimension, non-uniform standard, poor timeliness, cross-node data fracture, lack of quantitative standard for risk judgment, easy confusion of commonalities and individual risks, low efficiency of data processing, difficult accurate positioning of supporting factors, strong subjectivity of factor analysis, lag of early warning response, no iterative optimization mechanism and high risk repetition probability. Disclosure of Invention The technical scheme aims to solve the problems that in the traditional gene library cross-node risk processing, data acquisition is scattered, standards are different, timeliness is insufficient, risk judgment depends on manual experience, quantification standards are lacked, cause analysis has no data support, early warning response is lagged, an iterative optimization mechanism is not available, and the like, and the accurate identification, cause tracing and active early warning of cross-node common risk are realized. The invention aims to provide a gene library central control method and system based on artificial intelligence, which is characterized by comprising the following steps that step 1, full-node operation data and risk states are monitored in real time, four key parameters of a power grid, environment, resources and equipment operation are collected, the risk states of all nodes are judged in real time, relevant information is recorded, and the information is uploaded to a distributed fusion center; the method comprises the steps of step 2 of multi-node synchronous risk judgment, setting a synchronous judging window, judging synchronous risks based on the window, distinguishing cross-node common risk and single-node individual risk, step 3 of synchronous risk monitoring data extraction and cleaning, extracting the synchronous risk window and front preset time length full-node data, and generating a correlation data set after cleaning and standardization, step 4 of common risk factor analysis, wherein the step 4.1 of characteristic screening comprises the steps of comparing risk with non-risk node data, screening common abnormal characteristics of risk nodes and normal characteristics of non-risk nodes as potential causes, the step 4.2 of factor verification comprises the steps of checking historical synchronous risk cases, verifying the correlation between the potential characteristics and risk occurrence, confirming common risk causes, the step 4.3 of characteristic extraction comprises the steps of converting the confirmed causes into quantifiable indexes, defining early warning thresholds, forming common risk characteristic sets, the early warning application of characteristic driving, constructing an early warning model based on the common risk characteristic sets, grading according to characteristic standard conditions, and triggering corresponding early warning. The invention further provides real-time monitoring of all-node operation data and risk states in the step 1, wherein the real-time monitoring specifically comprises 1.1 of monitoring data range covering key operation dimensions of a gene refrigerator cluster, unified acquisition standards and frequencies, and specifically comprises ① of power grid parameters, ② of environment parameters, wherein the power grid parameters at least comprise voltage, current and power, the environment parameters at least comprise external humiture and humiture in the refrigerator, ③ of resource parameters, the resource parameters at least comprise liquid nitrogen liquid level and resource occupancy rate, ④ of equipment operation parameters at least comprise compressor load and fan rotating speed, the acquisition frequency is 200 ms/time, 1.2 of each node determines the loca