CN112101422B - Typical case self-learning method for power system fault case - Google Patents
Typical case self-learning method for power system fault case Download PDFInfo
- Publication number
- CN112101422B CN112101422B CN202010839126.3A CN202010839126A CN112101422B CN 112101422 B CN112101422 B CN 112101422B CN 202010839126 A CN202010839126 A CN 202010839126A CN 112101422 B CN112101422 B CN 112101422B
- Authority
- CN
- China
- Prior art keywords
- case
- actual
- value
- characteristic value
- data
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 239000011159 matrix material Substances 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 19
- 230000008569 process Effects 0.000 claims description 23
- 238000012544 monitoring process Methods 0.000 claims description 4
- 238000012258 culturing Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 239000000463 material Substances 0.000 abstract description 2
- 238000003745 diagnosis Methods 0.000 description 29
- 238000012423 maintenance Methods 0.000 description 16
- 238000004458 analytical method Methods 0.000 description 10
- 238000004891 communication Methods 0.000 description 9
- 238000013500 data storage Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000001360 synchronised effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
A typical case self-learning method for power system fault cases comprises the following steps: defining a matching characteristic value, wherein the matching characteristic value is a case characteristic value with the characteristic value reliability being greater than or equal to a first threshold value; defining a characteristic value matrix, wherein the characteristic value matrix is a matrix of the matching characteristic values contained in each typical case; after the actual case occurs, acquiring a state value of the actual case; matching the state value with the characteristic value matrix of each typical case, and calculating the case matching degree of the actual case and each typical case to obtain the maximum case matching degree; when the maximum case matching degree is smaller than a second threshold value, performing first processing on the actual case; and when the maximum case matching degree is greater than or equal to the second threshold value, performing second processing on the actual case. The method can automatically generate and divide corresponding typical cases and provide abundant learning materials for power system staff.
Description
Technical Field
The invention relates to the field of power systems, in particular to a typical case self-learning method for a power system fault case.
Background
A power system is a system in which a large number of power stations, substations, distribution stations, subscribers, and the like are connected by transmission and distribution lines. It is usually composed of generator, transformer, bus, power transmission and distribution line and electric equipment. The electrical components, devices and systems are typically in normal operation, but may also fail or be in abnormal operation.
The power system fault refers to a state that electrical components and equipment cannot work according to expected indexes, that is, the electrical components and equipment do not reach functions which the electrical components and equipment should reach, and the fault comprises a generator set fault, a transformer fault, a transmission line fault, a substation fault, a bus fault and the like.
As the scale of the power system becomes larger, the structure becomes more and more complex, and fault generation is unavoidable. The fault processing process of the power system may be that topology change is detected from the running state of the system, fault sign information is detected from the area (unit) associated with the topology change, and analysis processing is performed on the information, so that the specific area and the position (such as a fault range or a fault point) where the fault occurs are judged mainly according to the signal of the protection action. After the fault range or the fault point is determined, the fault area (unit) is firstly ensured to be reliably cut off or reliably isolated, then the power supply recovery of the power loss load is finished, and finally the fault cause investigation and fault defect elimination treatment are carried out.
The special system for power system diagnosis is a corresponding power system fault diagnosis expert system.
In the power system, the types of signal acquisition equipment used for operation and maintenance are more, and system maintenance staff cannot perform skilled application on the signal acquisition equipment, so that the problems of low maintenance efficiency, poor maintenance effect and the like of the power system are caused. For this reason, power enterprises often develop professional case training activities, and issue case training manuals, etc. Although the method can help operation staff to improve business quality to a certain extent, because of more power equipment related to the power system, the requirements of operation and maintenance of the power system cannot be met due to case training activities and case training manuals.
Disclosure of Invention
The invention solves the problem of providing a typical case self-learning method of a power system fault case so as to automatically generate and divide corresponding typical cases and provide abundant learning data for power system staff.
In order to solve the above problems, the present invention provides a typical case self-learning method for a fault case of a power system, including: defining a matching characteristic value, wherein the matching characteristic value is a case characteristic value with the characteristic value reliability being greater than or equal to a first threshold value; defining a characteristic value matrix, wherein the characteristic value matrix is a matrix of the matching characteristic values contained in each typical case; after the actual case occurs, acquiring a state value of the actual case; matching the state value with the characteristic value matrix of each typical case, and calculating the case matching degree of the actual case and each typical case to obtain the maximum case matching degree; when the maximum case matching degree is smaller than a second threshold value, performing first processing on the actual case; and when the maximum case matching degree is greater than or equal to the second threshold value, performing second processing on the actual case.
Optionally, the first processing includes: a new typical case is generated using the actual case.
Optionally, the first processing includes: and generating the state value of the new typical case as a new case feature value, and calculating the feature value credibility of the new case feature value.
Optionally, the second processing includes: defining the typical case corresponding to the maximum case matching degree as the most similar case of the actual case; defining the state value of the actual case, wherein the state value meeting the condition of the matching characteristic value is the actual case state value of the actual case; calculating the characteristic value credibility of the actual case state value as the actual characteristic value credibility; and calculating the case credibility of the actual case as the most similar case according to the actual eigenvalue credibility and the maximum case matching degree.
Optionally, the second process further includes: after the credibility of the actual case as the most similar case is calculated, updating the data of the most similar case, and then, recalculating the credibility of the state value of the actual case; and updating the recalculated actual eigenvalue credibility to the eigenvalue matrix of the most similar case.
Optionally, the state value which is not satisfied as the matching feature value in the state values of the actual case is defined as a sample state value of the actual case, and feature value cultivation is performed on the sample state value.
Optionally, performing eigenvalue culture on the sample state value includes: when the actual case occurs for more than N times, if the ratio of the sample state value in the times is not lower than K%, converting the sample state value into a case characteristic value of the actual case, calculating the characteristic value credibility of the case characteristic value, and judging whether the case characteristic value can be used as a matching characteristic value; wherein N is an integer of 3 or more, and K is a number of 50 to 100.
Optionally, the case feature value and the case state value are generated in panoramic data of the user power grid, the panoramic data including at least one of the following data: SCADA system data of the transformer substation; data of the electric power centralized control SCADA system; power dispatching SCADA system data; protecting information system data; relay protection device data; safety, stability and automatic control device data; intelligent measurement and control device data; fault recorder data; power equipment status monitoring data; production process data.
Optionally, the case feature values have associated attributes.
In one aspect of the technical scheme, a matching characteristic value is defined, wherein the matching characteristic value is a case characteristic value with the characteristic value reliability being greater than or equal to a first threshold value; defining a characteristic value matrix, wherein the characteristic value matrix is a matrix of the matching characteristic values contained in each typical case; after the actual case occurs, acquiring a state value of the actual case; matching the state value with the characteristic value matrix of each typical case, and calculating the case matching degree of the actual case and each typical case to obtain the maximum case matching degree; when the maximum case matching degree is smaller than a second threshold value, performing first processing on the actual case; and when the maximum case matching degree is greater than or equal to the second threshold value, performing second processing on the actual case. The method can automatically generate and divide the corresponding typical cases, namely can realize the operations of generating, matching, warehousing and the like of the typical cases, provides abundant and accurate various typical cases for case learning of staff, namely provides abundant learning materials for the staff of the power system, so that the staff can master better fault processing and analyzing capacity of the power system.
Drawings
FIG. 1 is a schematic diagram of a configuration of a diagnostic system master station and a dispatch center (or a centralized control center) in a power system fault diagnosis expert system according to the first embodiment;
FIG. 2 is a schematic diagram of a diagnostic system master deployment architecture for a power system fault diagnosis expert system according to the first embodiment;
FIG. 3 is a schematic diagram of a diagnostic system master station deployment architecture of a power system fault diagnosis expert system according to the second embodiment;
fig. 4 is a schematic diagram of a scenario step corresponding to a typical case self-learning method of a power system fault case according to the third embodiment of the present invention.
Detailed Description
The present invention will be described in detail with reference to the accompanying drawings for more clear illustration.
Example 1
Referring to fig. 1 and 2 in combination, a power system fault diagnosis expert system provided by the present invention is shown.
The power system fault diagnosis expert system includes a diagnosis system master station, which is directly set by using a network of a dispatch master station (or centralized control master station) (hereinafter, collectively referred to as a dispatch master station).
In fig. 1, the left side of the broken line is the structure of the scheduling master station, and the right side of the broken line is the diagnostic system master station of the fault diagnosis expert system.
As can be seen from fig. 1, the diagnostic system master station of the present embodiment is suspended in the network structure of the scheduling master station.
As shown in fig. 1, the corresponding scheduling master station may include: the system comprises a dispatching master station data storage structure, an engineer station, an operator station, a remote forwarding/dispatching communication unit, a dispatching master station server and the like.
The telemechanical forwarding/scheduling communication unit of the scheduling master station can access the power scheduling network. And the server of the dispatching master station is accessed to SCADA information of each transformer substation in the centralized control area.
The diagnostic system master station can directly access the power system by using the communication equipment of the scheduling master station.
FIG. 2 illustrates a specific deployment configuration of a diagnostic system master station.
As shown in fig. 2, the diagnostic system master station includes: a data storage structure (shown in dashed boxes in fig. 2), an expert knowledge base, a pre-server, an analysis engine, and an operating workstation.
The data storage structure is used for storing data. The expert knowledge base is used for expert knowledge storage. The front-end server is used for collecting the operation parameters of the power system and executing data preprocessing. The analysis engine is used as a real-time reasoning machine, collects observation information required by cache reasoning from the front-end server, searches proper expert knowledge from the expert knowledge base, completes the reasoning, and stores the reasoning process and the reasoning result. The running workstation is used as a user client to display information and the like.
As shown in FIG. 2, in this embodiment, the data storage structure may include a data server and a disk array. The data storage structure of this embodiment includes two data servers. The two data servers may act as historical data servers to store historical cases, historical reports, and stationarity analysis historical data, etc. The redundant configuration of two data servers can ensure the safety of data existence. Disk arrays may be used for individual storage of long-term history data. The number of the magnetic disks can be selected according to the requirement. In other embodiments, other data storage structures may be employed, such as a disk array may be omitted, or only one data server may be used.
As shown in fig. 2, in this embodiment, the expert knowledge base is used to store and update expert knowledge for fault diagnosis of various power systems, and the corresponding expert knowledge may be stored according to a certain rule for calling. The expert knowledge base is adapted for independent configuration.
As shown in fig. 2, in this embodiment, the front-end server may collect the operation parameters of the power system in real time and perform related data preprocessing. The front-end servers are adapted to employ stand-alone deployments.
As shown in fig. 2, in this embodiment, the analysis engine is used as a real-time inference engine, and may collect, from a corresponding front-end server, each observation information required for cache inference, and may search for appropriate expert knowledge from an expert knowledge base, so as to complete the inference, and save the inference process and intermediate conclusions in real time (i.e., the inference result of the analysis engine may be an intermediate conclusion including diagnosis). The analysis engines are preferably deployed independently to make the analytical reasoning process of the diagnostic system more efficient and reliable.
As shown in fig. 2, in the present embodiment, the operation workstation serves as a user client, and the displayed information includes operation information of a user system (client system). The operation workstation can specifically display real-time operation information of a user system, can also be used for displaying expert early warning information and expert diagnosis reports, and can also be used for starting diagnosis tracking, starting case inversion and other functions. And, the running workstation may be used to initiate a remote inquiry cloud expert system function. The operation workstation is arranged in a mode of being deployed separately from the server.
It should be noted that, in conjunction with fig. 1 and 2, the diagnostic system arrangement of the present embodiment is a station-end arrangement (disposed at a station-controlled layer station end). However, in other embodiments, the diagnostic system arrangement may be deployed in other structural locations.
With continued reference to fig. 2, the diagnostic system master station may also include a maintenance workstation. The maintenance workstation is used to effect maintenance of the diagnostic system. The maintenance workstation is specifically configured to allow a user engineer (knowledge engineer) to perform maintenance on the diagnostic system via the workstation. For example, modeling configuration of the power system, knowledge maintenance of an expert database and the like are realized. In this embodiment, the maintenance workstations are deployed independently, which is beneficial to better implementation of the maintenance function thereof. In other embodiments, the maintenance workstation may also be incorporated with an operating workstation of the diagnostic system.
With continued reference to fig. 2, the diagnostic system master station may also include an emergency command center interface server. The emergency command center interface server is used for being in communication connection with the enterprise emergency command center. The emergency command center interface server may be responsible for real-time communication with the enterprise emergency command center. In this embodiment, the emergency command center interface server is deployed independently, and this structure can exert its effect more. In other embodiments, the emergency command center interface server may also be incorporated with an analysis engine or a running workstation.
With continued reference to fig. 2, the diagnostic system master station may also include a WEB server. The WEB server is used for realizing WEB release of information and pushing of short messages (mobile information). The WEB server may issue a report of the failure of the electronic system with WEB, and may timely notify relevant personnel of the corresponding failure information by pushing a short message (mobile information). In other embodiments, the WEB server may be omitted without being employed.
With continued reference to fig. 2, the diagnostic system master station may also include a cloud expert system interface server. The cloud expert system interface server is used for being in communication connection with the cloud expert system. When the cloud expert system interface server is used for realizing communication with the cloud expert system, the fault diagnosis capability of the diagnosis system is expanded, and the realization of fault cloud diagnosis is ensured. In the embodiment, an independent server, namely an independent deployment structure is adopted, so that cloud diagnosis is more efficient, safe, reliable and timely. In other embodiments, the cloud expert system interface server may be combined with the WEB server.
With continued reference to fig. 2, the diagnostic system master station may also include a firewall. The WEB server and cloud expert system interface server are isolated outside the firewall. The firewall is used for safe partition of the system, and the embodiment separates the WEB server and the cloud expert system interface server from other parts of the system, so that other structures are better protected, and the system is more stable.
With continued reference to fig. 2, the diagnostic system master station may also include various network devices. These network devices are used to ensure the communication of the system. As shown in fig. 2, the network device is implemented by a switch, and the diagnostic system master station shown in fig. 2 includes a first switch, a second switch and a third switch. For the first front-end switch of the main station of the diagnosis system, an optical fiber interface can be adopted according to the specific situation of an access system, and a gigabit bandwidth switch is preferably adopted. The second switch and the third switch can also adopt gigabandwidth switches.
With continued reference to fig. 2, the diagnostic system master station may also include an output device. The output device may specifically be a printer, as shown in fig. 2. The printer is used for printing corresponding fault reports, diagnosis reports and the like at any time.
With continued reference to fig. 2, in this embodiment, the system for enabling the diagnostic system master station to access through the front-end server includes a synchronous clock (system), a SCADA system, an IED (system), a security system, a security management platform system, and the like. The synchronous clock is the synchronous clock of the power system and is used for ensuring the clock synchronization of data. The information protection system is a relay protection information processing system and is used for managing relay protection fixed values, fault message information and the like.
As shown in fig. 2, the present embodiment uses a single front-end server, and thus, this structure may be referred to as a single front-end single-network structure. The single front single network structure enables the internal network structure of the main station of the diagnosis system to be of a single network structure, and the structure is simpler, so that the system cost can be reduced.
It should be noted that, in conjunction with the foregoing description of fig. 1 and fig. 2, each node in fig. 2 is a logical function definition node, and when actually deployed, the logical function nodes and the physical nodes may be completely in one-to-one correspondence according to the scheme in the graph, or clipping of the function nodes, merging of the physical nodes, and the like may be performed as needed. As described above, for both logical function nodes of the operation workstation and the maintenance workstation, the physical implementation may be performed by a single workstation computer.
Referring to fig. 1 and fig. 2, in this embodiment, a station end of a power system fault diagnosis expert system is deployed at a station control layer, and a diagnosis system master station may be specifically deployed at a dispatching center station end, a centralized control center station end, or a substation station end. The SCADA system opens a forwarding channel with the expert system station end, and can forward real-time information required by the expert system in each substation of the whole plant to the expert system station end according to IEC 60870-5-104 or IEC61850 standard protocol. The deployment scheme can fully reuse resources and has good practicability for both new projects and existing project transformation.
Example two
Referring to fig. 3, another power system fault diagnosis expert system provided by the present invention is shown.
The fault diagnosis expert system provided in this embodiment is largely identical in structure to the foregoing embodiment, and thus the corresponding contents of the foregoing embodiment can be referred to in combination.
These same include the same parts of the diagnostic system master station, in particular the diagnostic system master station comprises: the system comprises a data storage structure, an expert knowledge base, a front-end server, an analysis engine and an operation workstation; the data storage structure may include a data server and a disk array; in addition, the system also comprises a maintenance workstation, an emergency command center interface server, a WEB server, a cloud expert system interface server, a firewall, network equipment, output equipment (the output equipment can be a printer) and the like; and the diagnosis system master station is connected with the synchronous clock, the SCADA system and the IED system, the security system, the safety management and control platform system and the like through the front-end server. The nature, characteristics and advantages of these structures can be referred to in the correspondence of the previous embodiments.
Unlike the diagnostic system shown in fig. 1 and 2, in the diagnostic system shown in fig. 3, the diagnostic system master station has two front-end servers.
Although the diagnostic system master station has two front-end servers, in the diagnostic system master station of the present embodiment, the network structure within the front-end servers is still a single-network structure, and thus, such a deployment structure may be referred to as a dual-front-end single-network structure.
In the structure, the two front-end servers can not only collect the operation parameters of the power system in real time and execute relevant data preprocessing more quickly and effectively, but also realize load balancing better by adopting redundant deployment of the two servers.
Another configuration that differs from the diagnostic system shown in fig. 1 and 2 is that the diagnostic system shown in fig. 3 further includes a diagnostic system sub-station.
In fig. 3, the diagnostic system substation is arranged to be connected to a front-end server. The arrangement of the substation of the diagnosis system enables the application range of the whole fault diagnosis expert system to be further expanded, and the application area to be further expanded.
In addition, due to the arrangement of the substation, the fault diagnosis expert system of the power system provided by the embodiment can enable the system or the system main body part (diagnosis system main station) to be arranged in a dispatching center, a centralized control center or a substation without being necessarily arranged at a station end of the dispatching main station and the like, so that the applicability is stronger.
Example III
The embodiment of the invention provides a typical case self-learning method for a fault case of a power system, which comprises the following steps:
defining a matching characteristic value, wherein the matching characteristic value is a case characteristic value with the characteristic value reliability being greater than or equal to a first threshold value; in this embodiment, the first threshold may be, for example, 80%, or may be, for example, 50%, and typically the first threshold is 60% or more;
defining a characteristic value matrix, wherein the characteristic value matrix is a matrix of the matching characteristic values contained in each typical case; in this embodiment, each typical case may be located in a typical case library;
after the actual case occurs, acquiring a state value of the actual case;
matching the state value with the characteristic value matrix of each typical case, and calculating the case matching degree of the actual case and each typical case to obtain the maximum case matching degree;
when the maximum case matching degree is smaller than a second threshold value, performing first processing on the actual case; in this embodiment, the second threshold may be, for example, 25%, or 30%;
and when the maximum case matching degree is greater than or equal to the second threshold value, performing second processing on the actual case.
In this embodiment, calculating the case matching degree between the actual case and each typical case may be performed by using various methods. One of the methods may be: when an actual disturbance case occurs to a user power grid (namely, when an actual power grid system fault case occurs), various state data of the actual case are obtained through various data sources in panoramic data, the state value (namely, the state data) of the actual case is matched with matching characteristic values in characteristic value matrixes of various typical cases in a typical case library in advance, so that the matching ratio of the state value (the state data) of the actual case to the matching characteristic values of various types in the characteristic value matrixes of various typical cases is calculated, and then the matching average value of the matching types is calculated and used as the case matching degree of the actual case and the corresponding typical case, wherein the characteristic value matrixes are represented by M, and the case matching degree is represented by MP.
According to the above, the maximum case matching degree may be obtained by: after traversing the eigenvalue matrix M of all the typical cases in the typical case library, determining the maximum case matching degree among all the case matching degrees, namely MP, wherein the maximum case matching degree is denoted as MPmax.
In the above, the state value is generated from panoramic data of the user power grid, where the panoramic data includes at least one of the following data: SCADA system data of the transformer substation; data of the electric power centralized control SCADA system; power dispatching SCADA system data; protecting information system data; relay protection device data; safety, stability and automatic control device data; intelligent measurement and control device data; fault recorder data; power equipment status monitoring data; production process data. It should be noted that, corresponding case feature values are also generated in the panoramic data, so as to ensure consistency of the corresponding values.
The panoramic data can also be used for defining that case characteristic values have ' category ' attributes, and specific categories can be divided into attribute values such as ' topological relation class ', ' relay protection information class ', ' security information class ', ' analog quantity class ', ' state monitoring class ', process information class ' and the like.
The above-described "category" attribute of the present invention is used as an illustrative example only and is not to be construed as limiting the present invention. This corresponds to the calculation of the degree of matching for the previous case.
In this embodiment, the first process may include: a new typical case is generated using the actual case.
Further, the first process may include: and generating the state value of the new typical case as a new case feature value, and calculating the feature value credibility of the new case feature value.
In the above process, for example, the disturbance cases with the maximum case matching degree still lower than 25% can be treated as a new case different from each typical case currently stored in the typical case library. At this time, the system can give a new typical case interface, and the user can decide whether to create the present case as a new typical case or manually perform typical case matching. I.e. the first process may comprise providing a definition interface or an information interaction interface for said new representative new case.
In this embodiment, the second process may include:
defining the typical case corresponding to the maximum case matching degree as the most similar case of the actual case; the most similar case is the case with the highest similarity with the actual case;
defining the state value of the actual case, wherein the state value meeting the condition of the matching characteristic value is the actual case state value of the actual case;
calculating the characteristic value credibility of the actual case state value as the actual characteristic value credibility;
according to the actual feature value credibility and the maximum case matching degree, calculating the case credibility of the actual case as the most similar case:
…………（1）
in the above formula 1, CF is the case confidence level, CFi is the actual eigenvalue confidence level, and MPmax is the maximum case matching level.
If the maximum case matching degree MPmax exceeds, for example, 25%, the similarity between the current fault case (the actual case) and the typical case corresponding to the maximum case matching degree MPmax may be considered to be the highest, i.e., the typical case is confirmed to be the most similar case. The reliability CF of the fault case similar to the typical case can be further calculated according to the matched actual characteristic value reliability CFi and MPmax, and a conclusion is finally given.
In this embodiment, further, the second processing may further include:
after the credibility of the actual case as the most similar case is calculated, updating the data of the most similar case, and then, recalculating the credibility of the state value of the actual case;
the recalculated actual eigenvalue reliability is updated to the eigenvalue matrix M of the most similar case (i.e. the latest calculation result is synchronized to the eigenvalue matrix M of the typical case).
And defining the state value which is not satisfied as the matching characteristic value in the state values of the actual case as a sample state value of the actual case, and culturing the characteristic value by the sample state value.
The culturing the characteristic value of the sample state value comprises the following steps:
when the actual case occurs for more than N times, if the ratio of the sample state value in the times is not lower than K%, converting the sample state value into a case characteristic value of the actual case, calculating the characteristic value credibility of the case characteristic value, and judging whether the case characteristic value can be used as a matching characteristic value; wherein N is an integer of 3 or more, and K is a number of 50 to 100.
In the above process, the sample culture of the case feature value is continued for the state value which is not matched with the typical case corresponding to the maximum case matching degree MPmax of the disturbance case. First, the state value relevant to the disturbance case is selected as a culture sample. For example, when the similar typical case occurs for more than 5 times, the historical big data statistical analysis is performed, when the ratio of the occurrence of a certain culture sample in the historical case is not lower than 80%, the culture sample can be converted into the characteristic value of the typical case, the characteristic value of the characteristic value case is calculated according to the subsequent method of the invention, and whether the characteristic value is stored in the characteristic value matrix M matrix as the matching characteristic value of the typical case is determined according to the conclusion of the method. And if the historical data analysis shows that the culture sample does not meet the case characteristic value conversion condition, continuing to culture. For state values that are not specific to the disturbance case, the state values may be selected for storage in a history repository, and prompt information may be provided to determine whether to further examine the process.
Determining the feature value confidence level includes:
…………（2）
in the above formula 2, cf is the feature value reliability of the case feature value, cf0 is the feature value reliability priori value of the case feature value, kach is the attribute weight of the acquisition mode, N is the occurrence number of cases corresponding to the case feature value, and m is the occurrence number of the case feature value.
Kach is an acquisition mode attribute weight, that is, in this embodiment, the case feature value has an "acquisition mode" attribute in addition to an "association" attribute and a "category" attribute. After the actual case occurs, the case feature values can be reliably acquired in real time from all data sources defined in the panoramic data through network communication, and the attribute of the acquisition mode is defined as automatic. In contrast, if a feature value that can be automatically acquired from the above-described panoramic data cannot be clearly ensured, the "acquisition mode" attribute thereof is defined as "non-automatic". Different acquisition modes, wherein attribute weights Kach of the acquisition modes are different when the credibility of the characteristic values is calculated, and the characteristic values of the automatic acquisition are higher.
As can be seen from the formula 2, in this embodiment, the confidence priori value (which can be determined by the staff according to the assignment mode) of the obtained case feature value is included, the obtained mode attribute is confirmed, the occurrence frequency of the case corresponding to the obtained case feature value, the occurrence frequency of the obtained case feature value, and the like. When the Cf of a certain feature value (case feature value) is not lower than, for example, 50% by calculation of the formula 2, the feature value can be considered as a matching feature value of the present case and stored in the matching feature value matrix M of the present exemplary case.
In this embodiment, the case feature value may be set to have an associated attribute. The association attribute is associated with preprocessing. Therefore, the method provided in this embodiment may further include pre-storing a plurality of typical cases into the case library, and predefining a plurality of case feature values.
In particular, the power system fault expert diagnostic system may have a built-in typical case library. For a new project, several exemplary case libraries may be predefined. Such as "case of internal faults of the transformer body", "case of faults within the line range", and "case of bus voltage loss", etc. For each type of typical case, a corresponding case characteristic value may be defined. The most desirable case feature values are exclusive, i.e., unique to the case, and not common to multiple cases. However, for a systematic fault case, such as a "bus voltage loss case", the characteristic value may be common to multiple cases, for example, the bus voltage loss is caused by the line protection action, and the characteristic value of the "line protection action" is common to the "bus voltage loss case" and the "fault case in the line range". Accordingly, the feature value is assigned to the "relevance" attribute, and the value of the attribute may be set to an attribute value such as "case-specific" or "system-related".
The above-described "relevance" attribute of the present invention is merely for illustrative purposes and is not to be construed as limiting the present invention.
It should be noted that, for the above-mentioned different attributes, the corresponding assignment of the case feature values may be "0/1" (digital type) or may be an analog value type. For the former, a certain state of 0 or 1 is defined as a characteristic value, namely the corresponding case characteristic value can be binarized; for the latter, a value interval is defined as a characteristic value. Thus, when a certain actually collected data meets the corresponding 0/1 state or falls into a defined value interval, the collected data is considered to be matched with the characteristic value corresponding to the typical case, namely, the characteristic value definition of the corresponding case is met.
One specific scenario corresponding to the embodiment of the present invention is as follows, please refer to fig. 4 in combination:
step 100: the actual case occurs; the actual case is the disturbance event and the like described above;
step 10, step 1: acquiring a case state value; as previously described, case state values may be derived from the corresponding disturbance event panorama data;
step 102: performing similarity matching between the (case state value) and M (characteristic value matrix) of each typical case in the (typical) case library; the similarity matching calculation method can be various, and the content can be referred to;
step 103: calculating the matching degree (case state value) between each typical case, namely MP; the calculation method refers to the corresponding content;
step 104: obtaining the highest case matching degree MPmax;
step 105: judging whether MPmax is greater than or equal to 25%, namely whether the MPmax exceeds a corresponding second threshold value;
step 106: when the judgment result of the step 105 is "N" (no), a new case is generated; step 106 corresponds to the first process described above, and the first process selects the actual case as the generation of the new typical case;
step 107: pushing out a typical case definition interface, and interacting with a user to define a new typical case; step 107, which is further content of the first process, allows the user to perform definition and operation of a corresponding more new typical case;
step 108: defining a typical case; step 108, i.e., after step 107, the completed operation may be considered as one way to implement defining a typical case; of course, as mentioned above, it is also possible to perform corresponding definition of the typical case operation when a new item is encountered;
step 109: generating case feature values, wherein the corresponding generation process can be a definition of direct warehouse entry in advance, or refer to the corresponding processes from step 116 to step 120 mentioned later;
step 110: calculating the reliability Cf of the characteristic value; reference may be made to the corresponding content of equation 2 above;
step 111: judging whether Cf is greater than or equal to 50 percent (according to the calculation result of 110);
step 112: when Cf is greater than or equal to 50%, namely the judgment result of 111 is Y (yes), writing the characteristic value into a case characteristic value matching matrix; in contrast, if the determination result in step 111 is "N" (no), the process returns to step 110;
step 113: presetting a case feature value initial value Cf0 (namely a priori value); this step acts on step 110;
step 114: when the judgment result in step 105 is "Y" (yes), calculating the occurrence reliability CF of the M corresponding case from the matched eigenvalues in the M (eigenvalue matrix) corresponding to MPmax (highest case matching degree);
step 115: according to step 114, the present conclusion is drawn;
step 116: performing characteristic value sample culture (namely a further process corresponding to the second processing) on the unmatched state values;
step 117: judging whether the corresponding state value is the characteristic state value or not; if step 117 is "N" (NO), then step 118 is entered;
step 118: the repository is to be checked, namely, the corresponding state value is stored in the corresponding database to be processed later;
conversely, if step 117 is "Y" (Yes), step 119 is entered; step 119: judging whether the number of sample cases (containing corresponding state values) is greater than or equal to 5;
if the result of the determination in step 119 is "no", return to step 116;
if the determination result of step 119 is "Y" (yes), step 120 is entered, i.e., whether the occurrence ratio of the culture status value is greater than or equal to 80% is continued; if the determination result of step 120 is "N" (no), returning to step 116; if the determination result of step 120 is "Y" (yes), step 109 is entered.
According to the scene, the method provided by the embodiment can realize the operations of generating, matching, warehousing and the like of the typical cases, and provides various rich and accurate typical cases for case study of staff, so that the staff can master better power system fault processing analysis capability.
Although the present invention is disclosed above, the present invention is not limited thereto. Various changes and modifications may be made by one skilled in the art without departing from the spirit and scope of the invention, and the scope of the invention should be assessed accordingly to that of the appended claims.
Claims (5)
1. A method for typical case self-learning of power system fault cases, comprising:
defining a matching characteristic value, wherein the matching characteristic value is a case characteristic value with the characteristic value reliability being greater than or equal to a first threshold value;
defining a characteristic value matrix, wherein the characteristic value matrix is a matrix of the matching characteristic values contained in each typical case;
after the actual case occurs, acquiring a state value of the actual case;
matching the state value with the characteristic value matrix of each typical case, and calculating the case matching degree of the actual case and each typical case to obtain the maximum case matching degree;
when the maximum case matching degree is smaller than a second threshold value, performing first processing on the actual case; the first process includes: generating a new typical case by using the actual case, generating the state value of the new typical case as a new case characteristic value, and calculating the characteristic value credibility of the new case characteristic value;
when the maximum case matching degree is greater than or equal to the second threshold value, performing second processing on the actual case; the second process includes: defining the typical case corresponding to the maximum case matching degree as the most similar case of the actual case; defining the state value of the actual case, wherein the state value meeting the condition of the matching characteristic value is the actual case state value of the actual case; calculating the characteristic value credibility of the actual case state value as the actual characteristic value credibility; calculating the case credibility of the actual case as the most similar case according to the actual eigenvalue credibility and the maximum case matching degree;
the case feature values and the case status values are generated in panoramic data of the user power grid, the panoramic data comprising at least one of the following data: SCADA system data of the transformer substation; data of the electric power centralized control SCADA system; power dispatching SCADA system data; protecting information system data; relay protection device data; safety, stability and automatic control device data; intelligent measurement and control device data; fault recorder data.
2. The power system fault case representative case self-learning method of claim 1, wherein the second process further comprises:
after the credibility of the actual case as the most similar case is calculated, updating the data of the most similar case, and then, recalculating the credibility of the state value of the actual case;
and updating the recalculated actual eigenvalue credibility to the eigenvalue matrix of the most similar case.
3. The typical case self-learning method of a power system fault case according to claim 2, wherein the state value which is not satisfied as the matching feature value among the state values of the actual case is defined as a sample state value of the actual case, and the sample state value is subjected to feature value cultivation;
the culturing the characteristic value of the sample state value comprises the following steps:
when the actual case occurs for more than N times, if the ratio of the sample state value in the times is not lower than K%, converting the sample state value into a case characteristic value of the actual case, calculating the characteristic value credibility of the case characteristic value, and judging whether the case characteristic value can be used as a matching characteristic value; wherein N is an integer of 3 or more, and K is a number of 50 to 100.
4. The method of claim 1, wherein the panoramic data further comprises power plant status monitoring data and production process data.
5. The power system fault case typical case self-learning method of claim 1, wherein the case feature values have associated attributes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010839126.3A CN112101422B (en) | 2020-08-19 | 2020-08-19 | Typical case self-learning method for power system fault case |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010839126.3A CN112101422B (en) | 2020-08-19 | 2020-08-19 | Typical case self-learning method for power system fault case |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112101422A CN112101422A (en) | 2020-12-18 |
| CN112101422B true CN112101422B (en) | 2024-02-13 |
Family
ID=73754605
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010839126.3A Active CN112101422B (en) | 2020-08-19 | 2020-08-19 | Typical case self-learning method for power system fault case |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112101422B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113759876B (en) * | 2021-09-14 | 2023-05-19 | 西安交通大学 | Case-reasoning-based wind turbine generator fault diagnosis method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102284517A (en) * | 2011-06-30 | 2011-12-21 | 东北大学 | Post-rolling cooling long-termed self-learning method based on case-based reasoning |
| CN104573062A (en) * | 2015-01-23 | 2015-04-29 | 桂林电子科技大学 | Intelligent learning method based on description logic and case-based reasoning |
| CN108040036A (en) * | 2017-11-22 | 2018-05-15 | 江苏翼企云通信科技有限公司 | A kind of industry cloud Webshell safety protecting methods |
| CN108090567A (en) * | 2018-01-19 | 2018-05-29 | 国家电网公司 | Power communication system method for diagnosing faults and device |
| CN109977146A (en) * | 2019-05-07 | 2019-07-05 | 山东省科学院激光研究所 | Method for diagnosing faults, device and electronic equipment |
| CN111553380A (en) * | 2020-03-22 | 2020-08-18 | 中润普达(十堰)大数据中心有限公司 | Auxiliary diagnosis system for automobile fault symptoms and using method thereof |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1703449A1 (en) * | 2005-03-18 | 2006-09-20 | BRITISH TELECOMMUNICATIONS public limited company | Fault diagnostics |
-
2020
- 2020-08-19 CN CN202010839126.3A patent/CN112101422B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102284517A (en) * | 2011-06-30 | 2011-12-21 | 东北大学 | Post-rolling cooling long-termed self-learning method based on case-based reasoning |
| CN104573062A (en) * | 2015-01-23 | 2015-04-29 | 桂林电子科技大学 | Intelligent learning method based on description logic and case-based reasoning |
| CN108040036A (en) * | 2017-11-22 | 2018-05-15 | 江苏翼企云通信科技有限公司 | A kind of industry cloud Webshell safety protecting methods |
| CN108090567A (en) * | 2018-01-19 | 2018-05-29 | 国家电网公司 | Power communication system method for diagnosing faults and device |
| CN109977146A (en) * | 2019-05-07 | 2019-07-05 | 山东省科学院激光研究所 | Method for diagnosing faults, device and electronic equipment |
| CN111553380A (en) * | 2020-03-22 | 2020-08-18 | 中润普达(十堰)大数据中心有限公司 | Auxiliary diagnosis system for automobile fault symptoms and using method thereof |
Non-Patent Citations (3)
| Title |
|---|
| 国家电网公司武汉高压研究所.电力监测与故障诊断技术全书 发电卷火电篇 4.2005,第1688页. * |
| 基于案例库的电力通信故障诊断算法研究;刘芳;河北电力技术;第38卷(第04期);38-41 * |
| 案例推理的动态学习模型及其在TE 过程;赵辉;中国博士学位论文全文数据库信息科技辑(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112101422A (en) | 2020-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113283704B (en) | Intelligent power grid fault handling system and method based on knowledge graph | |
| CN105790441B (en) | A kind of intelligent substation relay protection cruising inspection system | |
| CN103440367B (en) | Transformer station's panoramic view data modeling method | |
| CN104753178A (en) | Power grid fault handling system | |
| CN104657913A (en) | Intelligent warning system based on all-data model | |
| CN103699967A (en) | Panoramic information platform based on electric network regulation and control integration | |
| CN112072642B (en) | Method for estimating abnormal operation state of power system | |
| CN116500385B (en) | Power transmission network monitoring and checking method, device, equipment and medium | |
| CN111276929B (en) | Information recording method for fault expert diagnosis of power system | |
| CN112463892A (en) | Early warning method and system based on risk situation | |
| CN112101422B (en) | Typical case self-learning method for power system fault case | |
| CN112054510B (en) | Method for estimating abnormal operation state of power system | |
| CN111917183A (en) | Source network load control second-level network remote monitoring analysis platform | |
| CN111327474B (en) | Power system fault diagnosis method based on topology analysis | |
| CN104992307B (en) | A kind of power distribution network risk collaboration managing and control system | |
| CN113740666A (en) | Method for positioning storm source fault of data center power system alarm | |
| CN112054511A (en) | Correlation generation method for accident handling measures and standard tickets of power system | |
| CN116094174A (en) | Knowledge graph-based power grid operation and maintenance monitoring method, system, equipment and medium | |
| CN117118062A (en) | Secondary system data fusion monitoring method and system for power dispatching automation | |
| CN115935201A (en) | Monitoring method and device for key emergency equipment GPS and electric quantity data acquisition | |
| CN112256922B (en) | Method and system for quickly identifying faults and power failure | |
| CN114116252A (en) | System and method for storing operation measurement data of regulation and control system | |
| CN112104073B (en) | Real-time information calibration method for power system | |
| CN113011047A (en) | Reliability analysis method based on secondary system of intelligent substation of 220kV or below | |
| CN112614017A (en) | Power failure information management system, power failure information management method, and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |