CN117746599A - Real-time early warning method, system and storage medium for park management - Google Patents

Real-time early warning method, system and storage medium for park management Download PDF

Info

Publication number
CN117746599A
CN117746599A CN202311607548.8A CN202311607548A CN117746599A CN 117746599 A CN117746599 A CN 117746599A CN 202311607548 A CN202311607548 A CN 202311607548A CN 117746599 A CN117746599 A CN 117746599A
Authority
CN
China
Prior art keywords
data
early warning
time
real
time sequence
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311607548.8A
Other languages
Chinese (zh)
Inventor
罗顺辉
程树英
詹超
郭志平
刘建芳
吴德铿
陈晓杉
陈彪
叶瀚
李榕桂
王妍
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fujian Wangneng Technology Development Co ltd
State Grid Information and Telecommunication Co Ltd
Original Assignee
Fujian Wangneng Technology Development Co ltd
State Grid Information and Telecommunication Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fujian Wangneng Technology Development Co ltd, State Grid Information and Telecommunication Co Ltd filed Critical Fujian Wangneng Technology Development Co ltd
Priority to CN202311607548.8A priority Critical patent/CN117746599A/en
Publication of CN117746599A publication Critical patent/CN117746599A/en
Pending legal-status Critical Current

Links

Classifications

    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a real-time early warning method, a system and a storage medium for park management, which are used for cleaning, denoising and filling historical time sequence data, clustering the data, dividing the data into a plurality of groups of modal data according to modal characteristics, respectively fitting the plurality of groups of modal data to obtain time sequence graphs, smoothing each time sequence graph, and obtaining the processed one-dimensional time sequence data of each mode. Training an early warning model by adopting one-dimensional time sequence data, and obtaining a current threshold by adopting the trained early warning model; and comparing the real-time monitoring index data of each node with the calculated upper and lower limit constraints of the current threshold, and if the real-time monitoring index data exceeds the range of the upper and lower limit constraints of the current threshold, carrying out early warning. According to the method, the time sequence data are divided into a plurality of groups of modal data, and the final fusion of the data is carried out based on the attention mechanism, so that the accuracy of dynamic threshold prediction is improved. The invention carries out real-time early warning based on the dynamic threshold value, can update the threshold value in real time according to the change of the monitoring index, and has better practicability.

Description

Real-time early warning method, system and storage medium for park management
Technical Field
The invention belongs to the technical field of intelligent early warning of parks, and particularly relates to a real-time early warning method, a real-time early warning system and a storage medium for park management.
Background
With the rapid development of technology and the construction of intelligent parks, data processing and real-time early warning of park management become an important subject. In the operation process of the park, a large amount of data including security monitoring data, environment monitoring data, personnel flow data and the like can be generated, and the data needs to be processed and early-warned efficiently, accurately and timely so as to improve the safety and management efficiency of the park.
In the prior art, a preset threshold monitoring or alarm system is a common way in existing solutions, and these systems typically consist of a monitoring tool and an automation tool for monitoring specific indicators in real time and triggering an alarm when the preset threshold is exceeded. For example, in the IT domain, a network administrator may use monitoring tools such as Zabbix or Grafana to monitor network traffic in real time. When network traffic exceeds a certain preset threshold, the tools trigger an alarm informing the administrator that a problem may occur in the network. The administrator can take corresponding measures based on the alarm, such as adjusting network configuration, upgrading network devices, or investigating network attacks. However, these systems typically require preset thresholds that may not function effectively for dynamically changing environments. At the same time, these systems often fail to alert to unknown or new threats because they can only alert to already preset threats. Secondly, threshold setting is difficult: in a monitoring or alarm system where a threshold is preset, correctly setting the threshold is an important issue. If the threshold is set too tightly, it may result in some important alarms being ignored; if the threshold is set too loosely, it may result in too many false alarms that affect the proper operation of the system. False and missing report of abnormal condition: threshold-based monitoring or alarm systems are typically only capable of detecting certain types of anomalies. If the manifestation of the abnormal situation does not exactly match the set threshold, the system may misreport or miss the report.
In other areas, such as manufacturing, similar monitoring and alarm systems may also be used to monitor the operation of a production line in real time. For example, when a device on a production line fails or performance is degraded, the monitoring system may be pre-warned before the device fails, thereby allowing maintenance personnel sufficient time to prepare for maintenance work to minimize production interruption. However, these existing solutions all have some problems. First, they often only alarm against pre-set threats and cannot pre-warn against unknown or new threats. Second, they typically only monitor specific metrics and do not fully reflect the overall condition of the system. Furthermore, these solutions typically require manual thresholding, which can be complex and time consuming, and the thresholding may not always be accurate.
In summary, the monitoring or alarm system with the preset threshold specifically includes the following drawbacks:
(1) The real-time performance is not enough: the monitoring or alarm system of the preset threshold typically only triggers an alarm when the data reaches the preset threshold, meaning that the system may not be able to detect immediately when an abnormal situation has just occurred.
(2) Requiring periodic maintenance and updating: due to environmental and condition variations, the preset threshold may require periodic maintenance and updating. If the threshold value is not adjusted in time, false alarms and missing alarms of the system can be caused.
(3) The data quality requirement is high: monitoring or alarm systems of preset thresholds typically require high quality, accurate data as input. If there is noise, missing values, or other problems with the data, the performance of the system may be affected.
(4) Lack of adaptivity: the monitoring or alarm system of the preset threshold cannot adaptively adjust the threshold according to the dynamic change of the data. Thus, for some dynamically changing cases, more complex algorithms or models may be required to handle.
Disclosure of Invention
The invention aims to provide a real-time early warning method, a real-time early warning system and a storage medium for park management, and aims to solve the problems. The method and the device perform real-time early warning based on the dynamic threshold value, realize early warning of unknown threats and improve the accuracy and the real-time performance of early warning. We will also enable our system to adaptively adjust the thresholds and processing strategies to accommodate changes in campus environments and threats. In general, our invention will provide a more efficient, intelligent campus management solution that can automatically detect and pre-warn of anomalies by learning and understanding the dynamic changes of data, thereby avoiding problems that may be caused by preset thresholds.
The invention is realized mainly by the following technical scheme:
a real-time early warning method for park management comprises the following steps:
step S1: and (3) data acquisition: acquiring historical time sequence data in a park, wherein the time sequence data comprises electric parameter monitoring data, environment monitoring data and acquisition time;
step S2: data preprocessing: the method comprises the steps of cleaning, denoising and filling collected data, clustering similar data points together through a K-means method, dividing the data into a plurality of groups of modal data according to modal characteristics, fitting the plurality of groups of modal data respectively to obtain time sequence graphs, smoothing each time sequence graph, and obtaining one-dimensional time sequence data of each processed mode;
step S3: building an early warning model, and training the early warning model by adopting the processed one-dimensional time sequence data:
step S31: converting the processed one-dimensional time sequence data of each mode into two-dimensional data;
step S32: feature extraction: extracting the characteristics of two-dimensional data of each mode;
step S33: respectively inputting the features of the two-dimensional data of each mode into a dynamic threshold model, and outputting the dynamic features of influence factors; inputting the dynamic characteristics of each mode into a data fusion module, fusing the dynamic characteristics of a plurality of modes based on an attention mechanism, calculating a threshold value of a corresponding time point, calculating and updating the upper limit constraint range and the lower limit constraint range of the index in real time, and outputting threshold value information;
step S34: testing the early warning model by adopting a test sample;
step S4: obtaining a current threshold value by adopting a trained early warning model; and comparing the real-time monitoring index data of each node with the calculated upper and lower limit constraints of the current threshold, and if the real-time monitoring index data exceeds the range of the upper and lower limit constraints of the current threshold, carrying out early warning.
Preferably, the modulus features include gaussian distribution features, skewness features, and the like.
In order to better implement the present invention, further, in the step S2, when clustering, the euclidean distance is used as a standard for measuring the similarity between the data objects, and the euclidean distance is expressed as follows:
Dist(X i ,X j )=[(X i -X j )*(X i -X j ) T ] 1/2
wherein X represents a row vector having m attributes;
in the clustering process, each iteration, the corresponding cluster center needs to be recalculated: the average value of all the data objects in the cluster is the center of the cluster after updating; defining the Center of the Kth cluster as Center k
Wherein C is k Represents a kth cluster;
∣C k and | represents the number of data objects in the kth class cluster;
summing in the formula refers to cluster C k The sum of all elements in each column of attributes, thus Center k Also a row vector containing m attributes;
reclassifying the classified clusters by iterating continuously and updating cluster centers; the magnitude of the cluster center change is as follows:
ΔJ=Center k new -Center k old
wherein K represents the number of clusters,
and when the difference value delta J of two adjacent iterations is smaller than a set threshold value, ending the iteration, wherein the obtained cluster is the final clustering result.
In order to better implement the present invention, further, in the step S2, a Savitzky-Golay filter is used to implement curve smoothing.
To better implement the present invention, further, in the step S31, the one-dimensional time series data is normalized first, and then a markov transfer matrix is constructed based on the markov transfer field:
dividing the time sequence into Q quantile boxes, wherein the data volume in each quantile box is the same;
changing each data in the time sequence into the serial number of the corresponding bit box;
constructing a transfer matrix:
wherein omega ij Representing the frequency at which bin i transitions to bin j, an
Constructing a Markov transfer matrix:
in order to better implement the present invention, in step S33, the dynamic threshold model is further built and updated in real time based on the peak exceeding critical value theorem and the maximum likelihood estimation fit tail distribution.
In order to better implement the present invention, in step S33, the global average pulling is used to extract the channel characteristics, and the full-connection layer is used to generate weights, where the weight calculation of each channel includes information of all the channel characteristics.
The invention is realized mainly by the following technical scheme:
the real-time early warning system for park management is carried out by adopting the method and comprises a data acquisition module, a data preprocessing module, a model training module and a real-time early warning module;
the data acquisition module is used for acquiring time sequence data, and the data preprocessing module is used for preprocessing the time sequence data and forming one-dimensional time sequence data of each mode; the model training module is used for training an early warning model by adopting one-dimensional time sequence data of each mode, and the real-time early warning module is used for predicting dynamic threshold information based on the trained early warning model and comparing to obtain an early warning result.
In order to better realize the invention, the early warning model further comprises a feature extraction layer, a dynamic threshold model, an attention layer and a data fusion layer, wherein the feature extraction layer is used for extracting features of two-dimensional data, the dynamic threshold model is used for outputting dynamic features, the attention layer is used for realizing information interaction based on an attention mechanism, and the data fusion layer is used for fusing the multi-mode dynamic features and outputting threshold information.
A computer readable storage medium having stored thereon a computer program which when executed by a processor implements the method described above.
The beneficial effects of the invention are as follows:
according to the method, aiming at the different influence trends of each influence factor data in the time sequence data, a plurality of groups of modal data are obtained through division, the final fusion of the data is carried out based on the attention mechanism, the weight proportion of each influence factor is considered based on the attention mechanism, and the accuracy of dynamic threshold prediction is improved. In order to better improve the processing precision, the method converts the one-dimensional time sequence data into the two-dimensional structure data to be input into the neural network model for processing, and performs data preprocessing and noise filtering at the beginning, thereby improving the precision of subsequent data processing. The method and the system perform real-time early warning based on the dynamic threshold, not only can monitor important indexes of the database in real time and send alarm information to abnormal conditions of the park in time, but also can update the threshold in real time according to the change of the monitored indexes, and have good practicability.
Drawings
FIG. 1 is an overall flow chart of a real-time early warning method for campus management according to the present invention;
FIG. 2 is a flow chart of the early warning model training of the present invention.
Detailed Description
Example 1:
a real-time early warning method for park management, as shown in fig. 1 and 2, comprises the following steps:
step S1: and (3) data acquisition: acquiring historical time sequence data in a park, wherein the time sequence data comprises electric parameter monitoring data, environment monitoring data and acquisition time;
step S2: data preprocessing: the method comprises the steps of cleaning, denoising and filling collected data, clustering similar data points together through a K-means method, dividing the data into a plurality of groups of modal data according to modal characteristics, fitting the plurality of groups of modal data respectively to obtain time sequence graphs, smoothing each time sequence graph, and obtaining one-dimensional time sequence data of each processed mode;
step S3: building an early warning model, and training the early warning model by adopting the processed one-dimensional time sequence data:
step S31: converting the processed one-dimensional time sequence data of each mode into two-dimensional data;
step S32: feature extraction: extracting the characteristics of two-dimensional data of each mode;
step S33: respectively inputting the features of the two-dimensional data of each mode into a dynamic threshold model, and outputting the dynamic features of influence factors; inputting the dynamic characteristics of each mode into a data fusion module, fusing the dynamic characteristics of a plurality of modes based on an attention mechanism, calculating a threshold value of a corresponding time point, calculating and updating the upper limit constraint range and the lower limit constraint range of the index in real time, and outputting threshold value information;
step S34: testing the early warning model by adopting a test sample;
step S4: obtaining a current threshold value by adopting a trained early warning model; and comparing the real-time monitoring index data of each node with the calculated upper and lower limit constraints of the current threshold, and if the real-time monitoring index data exceeds the range of the upper and lower limit constraints of the current threshold, carrying out early warning.
Preferably, in the step S2, at the time of clustering, the euclidean distance is used as a standard for measuring the similarity between the data objects, and the euclidean distance is expressed as follows:
Dist(X i ,X j )=[(X i -X j )*(X i -X j ) T ] 1/2
wherein X represents a row vector having m attributes;
in the clustering process, each iteration, the corresponding cluster center needs to be recalculated: the average value of all the data objects in the cluster is the center of the cluster after updating; defining the Center of the Kth cluster as Center k
Wherein C is k Represents a kth cluster;
∣C k and | represents the number of data objects in the kth class cluster;
summing in the formula refers to cluster C k The sum of all elements in each column of attributes, thus Center k Also a row vector containing m attributes;
reclassifying the classified clusters by iterating continuously and updating cluster centers; the magnitude of the cluster center change is as follows:
ΔJ=Center k new -Center k old
wherein K represents the number of clusters,
and when the difference value delta J of two adjacent iterations is smaller than a set threshold value, ending the iteration, wherein the obtained cluster is the final clustering result.
Preferably, in the step S2, a Savitzky-Golay filter is used to implement curve smoothing.
Preferably, in the step S31, the one-dimensional time series data is normalized first, and then a markov transfer matrix is constructed based on the markov transfer field:
dividing the time sequence into Q quantile boxes, wherein the data volume in each quantile box is the same;
changing each data in the time sequence into the serial number of the corresponding bit box;
constructing a transfer matrix:
wherein omega ij Representing the frequency at which bin i transitions to bin j, an
Constructing a Markov transfer matrix:
preferably, in the step S33, the dynamic threshold model is built and updated in real time based on the peak exceeding critical value theorem and the maximum likelihood estimation fit tail distribution.
Preferably, in the step S33, the global average pulling is used to extract the channel features, and the full-connection layer is used to generate weights, where the weight calculation of each channel includes information of all the channel features.
Preferably, as shown in fig. 2, the early warning model includes a feature extraction layer, a dynamic threshold model, an attention layer and a data fusion layer, wherein the feature extraction layer is used for extracting features of two-dimensional data, the dynamic threshold model is used for outputting dynamic features, the attention layer is used for realizing information interaction based on an attention mechanism, and the data fusion layer is used for fusing the multi-mode dynamic features and outputting threshold information.
According to the method, aiming at the different influence trends of each influence factor data in the time sequence data, a plurality of groups of modal data are obtained through division, the final fusion of the data is carried out based on an attention mechanism, the weight proportion of each influence factor is considered based on global features and local features, and the accuracy of dynamic threshold prediction is improved. In order to better improve the processing precision, the method converts the one-dimensional time sequence data into the two-dimensional structure data to be input into the neural network model for processing, and performs data preprocessing and noise filtering at the beginning, thereby improving the precision of subsequent data processing. The method and the system perform real-time early warning based on the dynamic threshold, not only can monitor important indexes of the database in real time and send alarm information to abnormal conditions of the park in time, but also can update the threshold in real time according to the change of the monitored indexes, and have good practicability.
The foregoing description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and any simple modification, equivalent variation, etc. of the above embodiment according to the technical matter of the present invention fall within the scope of the present invention.

Claims (9)

1. The real-time early warning method for park management is characterized by comprising the following steps of:
step S1: and (3) data acquisition: acquiring historical time sequence data in a park, wherein the time sequence data comprises electric parameter monitoring data, environment monitoring data and acquisition time;
step S2: data preprocessing: the method comprises the steps of cleaning, denoising and filling collected data, clustering similar data points together through a K-means method, dividing the data into a plurality of groups of modal data according to modal characteristics, fitting the plurality of groups of modal data respectively to obtain time sequence graphs, smoothing each time sequence graph, and obtaining one-dimensional time sequence data of each processed mode;
step S3: building an early warning model, and training the early warning model by adopting the processed one-dimensional time sequence data:
step S31: converting the processed one-dimensional time sequence data of each mode into two-dimensional data;
step S32: feature extraction: extracting the characteristics of two-dimensional data of each mode;
step S33: respectively inputting the features of the two-dimensional data of each mode into a dynamic threshold model, and outputting the dynamic features of influence factors; inputting the dynamic characteristics of each mode into a data fusion module, fusing the dynamic characteristics of a plurality of modes based on an attention mechanism, calculating a threshold value of a corresponding time point, calculating and updating the upper limit constraint range and the lower limit constraint range of the index in real time, and outputting threshold value information;
step S34: testing the early warning model by adopting a test sample;
step S4: obtaining a current threshold value by adopting a trained early warning model; and comparing the real-time monitoring index data of each node with the calculated upper and lower limit constraints of the current threshold, and if the real-time monitoring index data exceeds the range of the upper and lower limit constraints of the current threshold, carrying out early warning.
2. The real-time early warning method for campus management according to claim 1, wherein in the step S2, the euclidean distance is used as a standard for measuring similarity between data objects during clustering, and the euclidean distance is expressed as follows:
Dist(X i ,X j )=[(X i -X j )*(X i -X j ) T ] 1/2
wherein X represents a row vector having m attributes;
in the clustering process, each iteration, the corresponding cluster center needs to be recalculated: the average value of all the data objects in the cluster is the center of the cluster after updating; defining the Center of the Kth cluster as Center k
Wherein C is k Represents a kth cluster;
∣C k and | represents the number of data objects in the kth class cluster;
summing in the formula refers to cluster C k The sum of all elements in each column of attributes, thus Center k Also a row vector containing m attributes;
reclassifying the classified clusters by iterating continuously and updating cluster centers; the magnitude of the cluster center change is as follows:
ΔJ=Center k new -Center k old
wherein K represents the number of clusters,
and when the difference value delta J of two adjacent iterations is smaller than a set threshold value, ending the iteration, wherein the obtained cluster is the final clustering result.
3. The real-time early warning method for campus management according to claim 2, wherein in step S2, curve smoothing is implemented by using a Savitzky-Golay filter.
4. The real-time early warning method for campus management according to claim 1, wherein in step S31, the one-dimensional time series data is normalized first, and then a markov transfer matrix is constructed based on the markov transfer field: dividing the time sequence into Q quantile boxes, wherein the data volume in each quantile box is the same;
changing each data in the time sequence into the serial number of the corresponding bit box;
constructing a transfer matrix:
wherein omega ij Representing the frequency at which bin i transitions to bin j, anConstructing a Markov transfer matrix:
5. the real-time early warning method for campus management according to claim 1, wherein in step S33, the dynamic threshold model is established and updated in real time based on the peak exceeding threshold theorem and the maximum likelihood estimation fit tail distribution.
6. The method for real-time early warning for campus management according to claim 5, wherein in step S33, the global average pooling is used to extract channel characteristics, the full-connection layer is used to generate weights, and the weight calculation of each channel includes information of all channel characteristics.
7. A real-time early warning system for park management, which is performed by the method of any one of claims 1-6, and is characterized by comprising a data acquisition module, a data preprocessing module, a model training module and a real-time early warning module; the data acquisition module is used for acquiring time sequence data, and the data preprocessing module is used for preprocessing the time sequence data and forming one-dimensional time sequence data of each mode; the model training module is used for training an early warning model by adopting one-dimensional time sequence data of each mode, and the real-time early warning module is used for predicting dynamic threshold information based on the trained early warning model and comparing to obtain an early warning result.
8. The real-time early warning system for campus management according to claim 7, wherein the early warning model comprises a feature extraction layer for extracting features of two-dimensional data, a dynamic threshold model for outputting dynamic features, an attention layer for realizing information interaction based on an attention mechanism, and a data fusion layer for fusing the multi-modal dynamic features and outputting threshold information.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the method of any of claims 1-6.
CN202311607548.8A 2023-11-27 2023-11-27 Real-time early warning method, system and storage medium for park management Pending CN117746599A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311607548.8A CN117746599A (en) 2023-11-27 2023-11-27 Real-time early warning method, system and storage medium for park management

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311607548.8A CN117746599A (en) 2023-11-27 2023-11-27 Real-time early warning method, system and storage medium for park management

Publications (1)

Publication Number Publication Date
CN117746599A true CN117746599A (en) 2024-03-22

Family

ID=90253534

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311607548.8A Pending CN117746599A (en) 2023-11-27 2023-11-27 Real-time early warning method, system and storage medium for park management

Country Status (1)

Country Link
CN (1) CN117746599A (en)

Similar Documents

Publication Publication Date Title
CN115578015B (en) Sewage treatment whole process supervision method, system and storage medium based on Internet of things
JP2004524609A (en) Adaptive modeling of changing states in predictive state monitoring
CN111538311B (en) Flexible multi-state self-adaptive early warning method and device for mechanical equipment based on data mining
CN113570200B (en) Power grid running state monitoring method and system based on multidimensional information
US7716152B2 (en) Use of sequential nearest neighbor clustering for instance selection in machine condition monitoring
CN112911627B (en) Wireless network performance detection method, device and storage medium
CN115858303B (en) Zabbix-based server performance monitoring method and system
CN117171157B (en) Clearing data acquisition and cleaning method based on data analysis
CN116914917A (en) Big data-based monitoring and management system for operation state of power distribution cabinet
CN115524002B (en) Operation state early warning method, system and storage medium of power plant rotating equipment
CN117421994A (en) Edge application health monitoring method and system
CN116665421A (en) Early warning processing method and device for mechanical equipment and computer readable storage medium
CN117041312A (en) Enterprise-level information technology monitoring system based on Internet of things
CN115561546A (en) Abnormity detection and alarm system for power system
CN117560300B (en) Intelligent internet of things flow prediction and optimization system
KR101960755B1 (en) Method and apparatus of generating unacquired power data
CN117368651B (en) Comprehensive analysis system and method for faults of power distribution network
CN117596119A (en) Equipment data acquisition and monitoring method and system based on SNMP (simple network management protocol)
CN117607780A (en) Abnormality analysis system and analysis method for electric energy metering device
CN116030955B (en) Medical equipment state monitoring method and related device based on Internet of things
CN116595353A (en) Remote fault diagnosis and intelligent decision system for camera
CN111934903A (en) Docker container fault intelligent prediction method based on time sequence evolution genes
CN114936614B (en) Operation risk identification method and system based on neural network
CN117746599A (en) Real-time early warning method, system and storage medium for park management
CN116956089A (en) Training method and detection method for temperature anomaly detection model of electrical equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination