Device fault diagnosis method and system based on Bayesian network

Abstract

The invention relates to the technical field of fault diagnosis, in particular to a device fault diagnosis method and system based on a Bayesian network, and the method comprises the steps: obtaining the historical experience data of an automation and safety device of a power system; constructing a fault diagnosis Bayesian network based on the historical experience data; and diagnosing a device fault based on the fault diagnosis Bayesian network. According to the method, the accurate fault diagnosis Bayesian network is constructed by integrating historical empirical data of the power system automation and safety device, the device fault can be efficiently and accurately diagnosed and processing suggestions can be provided by utilizing the inference capability of the Bayesian network, the network structure and parameters are continuously optimized through actual feedback, and the fault diagnosis efficiency is improved. The accuracy and reliability of fault diagnosis are remarkably improved, and powerful technical support is provided for stable operation of a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a device fault diagnosis method and system based on a Bayesian network. Background Art

[0002] With the continuous large-scale development of power systems, the number of power system automation and safety devices in field operation has increased dramatically year by year. In some power companies, the total number of these devices has reached hundreds of thousands. The massive growth has not only improved the operating efficiency of the power system, but also brought unprecedented challenges. In particular, the on-site fault maintenance and defect location analysis of these devices have become a major manpower and cost burden faced by power companies.

[0003] At present, the fault location and analysis of on-site operating devices mainly rely on the rich experience of technical experts. This process is not only time-consuming, but also requires a lot of valuable human resources. In view of the current situation, how to quickly and accurately locate the cause of the fault and conduct effective defect analysis and processing has become the key to reducing costs, increasing efficiency and saving manpower for power companies.

[0004] Therefore, developing a technology and method that can quickly locate the cause of a fault to replace the traditional manual analysis method has extremely important economic significance and practical needs for improving the operational efficiency of power companies and reducing manpower and time costs. Summary of the invention

[0005] The purpose of the present invention is to provide a device fault diagnosis method and system based on a Bayesian network in order to solve at least one of the above technical problems.

[0006] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0007] A device fault diagnosis method based on Bayesian network, comprising:

[0008] Obtain historical experience data of power system automation and safety devices;

[0009] Based on the historical experience data, construct a fault diagnosis Bayesian network;

[0010] Device faults are diagnosed based on the fault diagnosis Bayesian network.

[0011] Furthermore, a fault diagnosis Bayesian network is constructed, including:

[0012] Sort out and classify all possible source factors that trigger device failure as the first layer of nodes;

[0013] Sort out all the abnormal phenomena recorded in the internal logs of the classification device as the second-layer node;

[0014] Sort out all possible external display anomalies of the classification device as the third-layer nodes;

[0015] The abnormal phenomenon of the warning light of the combing device is used as the fourth layer node;

[0016] According to the direct causal relationship between nodes in each layer, a directed acyclic Bayesian network topological connection relationship between nodes is established;

[0017] Based on the historical experience data and the Bayesian network topology connection relationship, a conditional probability distribution table of each node under one or more of its parent nodes is established, namely, the fault diagnosis Bayesian network.

[0018] Further, diagnosing device faults based on the fault diagnosis Bayesian network includes:

[0019] Obtaining target abnormal phenomena triggered by unknown faults on site, and calculating and sorting the conditional probabilities of all fault factor nodes based on the target abnormal phenomena and the fault diagnosis Bayesian network;

[0020] The fault factor node whose conditional probability meets the preset conditions is taken as the device fault inference result;

[0021] Device failure diagnosis is performed based on the device failure estimation result.

[0022] Furthermore, the conditional probabilities of all fault factor nodes are calculated and sorted, including:

[0023] If the target abnormal phenomenon is triggered, the node value corresponding to the target abnormal phenomenon in the fault diagnosis Bayesian network is set to true; otherwise, the corresponding node value in the fault diagnosis Bayesian network is set to false;

[0024] Obtain a node value instance set E of all target abnormal phenomena corresponding to the unknown fault on site;

[0025] Based on the fault diagnosis Bayesian network and the node value instance set E, the conditional probability of each fault factor node is calculated and sorted from high to low according to the conditional probability.

[0026] Furthermore, the conditional probability of each fault factor node is calculated, including:

[0027] Based on the fault diagnosis Bayesian network, obtaining the target conditional probability of the target fault factor node under the parent node in the Bayesian network topology connection relationship;

[0028] Based on the target conditional probability, calculate the complete joint distribution probability instance P(((C i =true)^E),C j≠i ),

[0029]

[0030] Among them, C i , C j Respectively represent the i-th element and j-th element in the first layer node, j≠i, C i =true means C i The abnormal phenomenon corresponding to the node is triggered; P(x m |parents(X m )) is the node x m At the parent node X m The target conditional probability under is obtained by querying the fault diagnosis Bayesian network;

[0031] Based on the complete joint distribution probability instance, the probability distribution P(C i |E),

[0032] P(C i |E)=ɑ×P(C i ,E)

[0033] Among them, P(C i ,E) represents the first probability distribution (P((C i =true)^E),P((C i =false)^E)); ɑ represents the normalization constant;

[0034]

[0035] Based on the probability distribution P(C i |E), calculate the conditional probability P(C i =true|E);

[0036] P(C i |E)=(P(C i =true|E),P(C i =false|E))

[0037] Among them, P(C i =true|E)+P(C i =false|E)=1.

[0038] Furthermore, the preset conditions include: the conditional probability is higher than a set threshold, or the ranking of the conditional probability is within a preset range.

[0039] Further, the first layer nodes are fault source factor nodes, including: CPU board hardware abnormality, HMI board hardware abnormality, sampling circuit abnormality, IO board hardware abnormality, IO bus signal abnormality, HSB inter-board high-speed bus signal abnormality, CPU board and HMI board internal communication signal abnormality, FPGA hardware interface abnormality, memory hardware abnormality, file system bottom driver abnormality, software load overload, external GOOSE storm overload, HMI and background remote communication overload, background remote communication configuration error, HSB high-speed communication configuration error, external GOOSE communication configuration error, device program version mismatch error, program pointer runaway, program memory read and write out of bounds, program memory continuous leakage, program file descriptor not released correctly, program network port resource not released correctly, program floating point task flag not opened, interface operation error, CPU board and HMI internal communication overload; and / or,

[0040] The second layer nodes are internal abnormal phenomenon nodes, including: frequent hard reset of CPU board, frequent hard reset of HMI board, frequent soft reset of CPU board, frequent soft reset of HMI board, frequent reset of IO board, communication check error of IO board, communication interruption error of IO board, sampling point loss, sampling interruption, invalid sampling data, fixed value setting failure, communication interruption between device HMI and background, external GOOSE communication interruption of device, communication interruption between device HSB boards, internal communication interruption between device HMI and CPU, device program file corruption error, file format error; and / or,

[0041] The third layer node is an external abnormal phenomenon node, including: the device screen is black and flickering, the device reports a set value setting error event, the device reports a sampling abnormality error event, the device reports an input and output error event, the device reports a software version verification error event, the device reports an FPGA interface error event, the device reports an HMI and background communication interruption event, the device reports an HMI and CPU internal communication error event, the device reports an HSB inter-board high-speed communication error event, and the device reports an external GOOSE communication error event; and / or,

[0042] The fourth layer node is an abnormal phenomenon node of the warning light, including: the device warning light is on.

[0043] A device fault diagnosis system based on Bayesian network, comprising:

[0044] An acquisition module is used to obtain historical experience data of power system automation and safety devices;

[0045] A construction module, used for constructing a fault diagnosis Bayesian network based on the historical experience data;

[0046] A diagnosis module is used to diagnose device faults based on the fault diagnosis Bayesian network.

[0047] An electronic device comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the device fault diagnosis method based on a Bayesian network as described above is implemented.

[0048] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the device fault diagnosis method based on a Bayesian network as described above is implemented.

[0049] The beneficial effects of the present invention are:

[0050] The Bayesian network-based fault diagnosis method and system of the present invention can efficiently integrate and analyze the historical fault data of power system automation and safety devices, use the fault trigger source, internal record abnormalities, external display abnormalities, and warning light flashing as network nodes, and combine the fault causal logic to build a network topology and conditional probability table. By applying the present invention, when a field fault occurs, the conditional probability of each fault source can be quickly inferred, providing accurate fault location reference for operation and maintenance personnel, greatly reducing dependence on technical experts and operation and maintenance costs.

[0051] The present invention realizes automatic inference of fault causes and has universality, improves the fault operation and maintenance efficiency of power company's power system automation and safety devices, provides strong support for rapid location and processing of on-site faults, and greatly reduces the manpower demand for technical experts and on-site operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A simplified structural block diagram of a power system automation and safety device of a certain power company according to one embodiment of the present invention;

[0053] Figure 2 A flow chart of a device fault diagnosis method based on a Bayesian network according to an embodiment of the present invention;

[0054] Figure 3 A flow chart of a device fault diagnosis method based on a Bayesian network according to another embodiment of the present invention;

[0055] Figure 4 A Bayesian network topological connection relationship between the first layer nodes and the second layer nodes in one implementation mode of the present invention;

[0056] Figure 5 A Bayesian network topological connection relationship between the second layer nodes and the third layer nodes in one implementation mode of the present invention;

[0057] Figure 6A Bayesian network topological connection relationship between the third layer nodes and the fourth layer nodes in one implementation mode of the present invention;

[0058] Figure 7 This is a structural diagram of a device fault diagnosis system based on a Bayesian network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The content of the present invention will now be discussed with reference to exemplary embodiments. It should be understood that the embodiments discussed are only to enable those skilled in the art to better understand and thus implement the content of the present invention, rather than implying any limitation on the scope of the present invention.

[0060] As used herein, the term “including” and variations thereof are to be interpreted as open-ended terms meaning “including, but not limited to.” The term “based on” is to be interpreted as “based, at least in part, on.” The terms “one embodiment” and “an embodiment” are to be interpreted as “at least one embodiment.”

[0061] Embodiment 1

[0062] Figure 1 A simplified structural block diagram of a power system automation and safety device of a certain power company according to one embodiment of the present invention; Figure 2 FIG. 1 is a flow chart of a device fault diagnosis method based on a Bayesian network according to an embodiment of the present invention. Figure 1-2 As shown, according to one embodiment of the present invention, a device fault diagnosis method based on a Bayesian network includes the following steps:

[0063] Step S10, obtaining historical experience data of power system automation and safety devices;

[0064] Step S20, constructing a fault diagnosis Bayesian network based on historical experience data;

[0065] Step S30, diagnose the device fault based on the fault diagnosis Bayesian network.

[0066] In this embodiment, a device fault diagnosis method based on a Bayesian network is proposed for fault diagnosis of power system automation and safety devices. First, in step S10, historical experience data of power system automation and safety devices are obtained, and a fault history database is formed based on the historical experience data. The historical experience data includes, but is not limited to, fault records, device information, operation data, and expert experience. Among them, the fault record includes the time and location of the fault, the type of fault (such as short circuit, overload, open circuit, etc.), the fault manifestation (such as abnormal indicator light, alarm information, abnormal current and voltage, etc.), and the fault handling measures and results. The device information includes the device model, manufacturer, production date, installation location, operating environment (such as temperature, humidity, electromagnetic interference, etc.) and maintenance records. The operation data includes real-time monitoring data of the device during operation, such as current, voltage, power factor, frequency, etc., as well as the change trend and historical records of these data. Expert experience includes expert judgment of the fault, handling suggestions, and summary analysis after the fault is solved. Next, in step S20, a fault diagnosis Bayesian network is constructed based on the massive historical experience data of device fault phenomena in the fault history database. Through learning and training, the structure and parameters of the Bayesian network are continuously optimized to improve the accuracy and reliability of diagnosis. In step S30, the device fault is diagnosed based on the fault diagnosis Bayesian network. Specifically, it includes: inputting the relevant data of the device to be diagnosed (such as fault manifestation, device information, operation data, etc.) into the Bayesian network; performing inference calculations based on the input data and the probability model of the Bayesian network to obtain possible fault types and probability distributions; determining the most likely fault type based on the results of the inference calculations, and giving corresponding processing suggestions; comparing the diagnosis results with the actual situation, and providing feedback and corrections to the Bayesian network to improve the accuracy of subsequent diagnosis.

[0067] The present invention integrates historical empirical data of power system automation and safety devices to construct an accurate fault diagnosis Bayesian network. By utilizing the reasoning ability of the Bayesian network, it can efficiently and accurately diagnose device faults, provide processing suggestions, and continuously optimize the network structure and parameters through actual feedback, thereby significantly improving the accuracy and reliability of fault diagnosis and providing strong technical support for the stable operation of the power system.

[0068] Figure 4 A Bayesian network topological connection relationship between the first layer nodes and the second layer nodes in one implementation mode of the present invention; Figure 5 A Bayesian network topological connection relationship between the second layer nodes and the third layer nodes in one implementation mode of the present invention; Figure 6 The Bayesian network topological connection relationship between the third layer nodes and the fourth layer nodes in one embodiment of the present invention. Figure 4-6 As shown, according to one embodiment of the present invention, step S20 includes:

[0069] Step S21, sorting out and classifying all possible source factors that trigger device failure as first-layer nodes;

[0070] Preferably, the set of first-layer nodes is called C, and the first-layer nodes are fault source factor nodes, including: CPU board hardware abnormality (c101), HMI board hardware abnormality (c102), sampling circuit abnormality (c103), IO board hardware abnormality (c104), IO bus signal abnormality (c105), HSB inter-board high-speed bus signal abnormality (c106), internal communication signal abnormality between CPU board and HMI board (c107), FPGA hardware interface abnormality (c108), memory hardware abnormality (c109), file system bottom driver abnormality (c110), software load overload (c111), external GOOSE storm overload (c112) , HMI and background remote communication overload (c113), background remote communication configuration error (c114), HSB high-speed communication configuration error (c115), external GOOSE communication configuration error (c116), device program version mismatch error (c117), program pointer runaway (c118), program memory read and write out of bounds (c119), program memory continuous leakage (c120), program file descriptor not released correctly (c121), program network port resources not released correctly (c122), program floating-point task flag not turned on (c123), interface operation error (c124), CPU board and HMI internal communication overload (c125), etc.

[0071] Step S22, sorting out all abnormal phenomena recorded in the internal logs of the classification device as the second layer nodes;

[0072] Preferably, the set of second-layer nodes is called E2, and the second-layer nodes are internal abnormal phenomenon nodes, including: frequent hard reset of CPU board (e201), frequent hard reset of HMI board (e202), frequent soft reset of CPU board (e203), frequent soft reset of HMI board (e204), frequent reset of IO board (e205), IO board communication verification error (e206), IO board communication interruption error (e207), sampling point loss (e208), sampling interruption (e209), sampling data invalid (e210), fixed value setting failure (e211), device HMI and background communication interruption (e212), device external GOOSE communication interruption (e213), device HSB inter-board communication interruption (e214), device HMI and CPU internal communication interruption (e215), device program file corruption error (e216), file format error (e217), etc.;

[0073] Step S23, sorting out all possible external display abnormalities of the classification device as the third layer nodes;

[0074] Preferably, the set of third-layer nodes is called E3, and the third-layer nodes are external abnormal phenomenon nodes, including: the device screen is black and flashing (e301), the device reports a set value setting error event (e302), the device reports a sampling abnormality error event (e303), the device reports an input and output error event (e304), the device reports a software version verification error event (e305), the device reports an FPGA interface error event (e306), the device reports an HMI and background communication interruption event (e307), the device reports an HMI and CPU internal communication error event (e308), the device reports an HSB inter-board high-speed communication error event (e309), the device reports an external GOOSE communication error event (e310), etc.

[0075] Step S24, combing the abnormal phenomenon of the warning light of the device as the fourth layer node;

[0076] Preferably, the set of fourth-layer nodes is called E4, and the fourth-layer nodes are warning light abnormal phenomenon nodes, including: the device warning light is on (e401).

[0077] Step S25, establishing a directed acyclic Bayesian network topological connection relationship between nodes according to the direct causal relationship between nodes in each layer;

[0078] Step S26, based on historical experience data and the Bayesian network topology connection relationship, establish a conditional probability distribution table for each node under one or more of its parent nodes, that is, a fault diagnosis Bayesian network.

[0079] In this embodiment, the content of constructing the fault diagnosis Bayesian network in step S20 is described in detail. First, step S21 sorts out and classifies all possible source factors that trigger the device failure. These factors are used as the first layer nodes of the Bayesian network. They cover factors ranging from hardware anomalies to software errors, configuration errors and human operation errors. These factors constitute the root cause of the device failure and provide key clues for subsequent diagnosis. Next, steps S22 to S24 sort out the internal log record abnormalities, external display abnormalities and warning light abnormalities of the device respectively, and use them as the second layer nodes, third layer nodes and fourth layer nodes of the Bayesian network respectively. These nodes represent various abnormal phenomena exhibited by the device after the failure occurs, which are important observation and judgment basis in the diagnosis process. By classifying and sorting these abnormal phenomena, the present invention can more accurately grasp the manifestation and scope of the fault. In step S25, according to the direct causal relationship between the nodes of each layer, a directed acyclic Bayesian network topological connection relationship between the nodes is established. This relationship reveals the intrinsic connection between the fault source factor and various abnormal phenomena, and provides a logical basis for subsequent reasoning and calculation. By constructing such a network structure, the process of fault propagation and evolution can be more clearly presented. Finally, in step S26, based on the fault history database formed by historical experience data and the Bayesian network topology connection relationship, a conditional probability distribution table (i.e., a fault diagnosis Bayesian network) is established for each node under one or more of its parent nodes. These conditional probability distribution tables reflect the transition probability and conditional probability between the states of each node, and serve as the basis for fault diagnosis of power system automation and safety devices. By using the fault diagnosis Bayesian network, intelligent diagnosis of device faults can be achieved, and the accuracy and reliability of diagnosis can be improved.

[0080] The present invention constructs a fault diagnosis model based on Bayesian network, namely, fault diagnosis Bayesian network, by deeply analyzing and utilizing historical experience data of power system automation and safety devices. This model can accurately capture the intrinsic connection between fault source factors and various abnormal phenomena, realize intelligent diagnosis and precise positioning of faults, and provide strong technical guarantee for the stable operation of the power system and power supply safety.

[0081] According to one embodiment of the present invention, step S30 includes:

[0082] Step S31, obtaining the target abnormal phenomenon triggered by the unknown fault on site, and calculating the conditional probabilities of all fault factor nodes and sorting them based on the target abnormal phenomenon and the fault diagnosis Bayesian network;

[0083] Step S32, taking the fault factor node whose conditional probability meets the preset condition as the device fault inference result;

[0084] Step S33: diagnose the device failure based on the device failure estimation result.

[0085] Preferably, step S31 includes:

[0086] If the target abnormal phenomenon is triggered, the node value corresponding to the target abnormal phenomenon in the fault diagnosis Bayesian network is set to true; otherwise, the corresponding node value in the fault diagnosis Bayesian network is set to false;

[0087] Obtain a node value instance set E of all target abnormal phenomena corresponding to the unknown faults on site;

[0088] Based on the fault diagnosis Bayesian network and the node value instance set E, the conditional probability of each fault factor node is calculated and sorted from high to low according to the conditional probability.

[0089] Preferably, in step S32, the preset conditions include: the conditional probability is higher than a set threshold, or the ranking of the conditional probability is within a preset range.

[0090] In this embodiment, the content of step S30 is described in detail. Specifically, when a field device fails and triggers a series of target abnormal phenomena, we first map these abnormal phenomena to the corresponding nodes in the fault diagnosis Bayesian network. If the abnormal phenomenon is triggered, the corresponding node value is set to true, otherwise it is set to false, thereby obtaining the node value instance set E of all target abnormal phenomena corresponding to the current fault. Subsequently, based on the node value instance set E and the fault diagnosis Bayesian network, we traverse and calculate the conditional probability of all fault factor nodes (i.e., possible fault sources) under given evidence, i.e., P(C i =true|E), this process realizes the reverse reasoning from the observed abnormal phenomenon to the potential cause of the fault. After the calculation is completed, we sort all the fault factor nodes according to the conditional probability, and take the fault factor nodes with conditional probability higher than the set threshold (such as the fault probability exceeding 25%) or within the preset sorting range (such as the top 3 from high to low) as the inferred results of the device failure. Since the fault factor with a higher probability indicates that the current fault phenomenon is more likely to be triggered, these inferred results are displayed to the fault maintenance analysts in order of probability, providing them with an important basis and reference for troubleshooting, helping to quickly locate and solve device failures, thereby effectively improving the efficiency and accuracy of fault handling.

[0091] The present invention constructs a fault diagnosis Bayesian network and utilizes the target abnormal phenomenon triggered by the unknown faults on site to realize the conditional probability calculation and sorting of the fault factor nodes, so as to quickly and accurately infer the possible sources of the device failure. The present invention not only improves the efficiency of troubleshooting, but also provides an important reference basis for fault maintenance analysts, effectively improving the accuracy and timeliness of fault handling.

[0092] According to an embodiment of the present invention, the conditional probability of each fault factor node is calculated, including:

[0093] Based on the fault diagnosis Bayesian network, the target conditional probability of the target fault factor node under the parent node in the Bayesian network topology connection relationship is obtained;

[0094] Based on the target conditional probability, calculate the complete joint distribution probability instance P(((C i =true)^E),C j≠i ),

[0095]

[0096] Among them, C i , C j Respectively represent the i-th element and j-th element in the first layer node, j≠i, C i =true means C i The abnormal phenomenon corresponding to the node is triggered; P(x m |parents(X m )) is the node x m At the parent node X m The target conditional probability under is obtained by querying the fault diagnosis Bayesian network; m and n represent the element numbers in the corresponding layer respectively;

[0097] Based on the complete joint distribution probability instance, calculate the probability distribution P(C i |E),

[0098] P(C i |E)=ɑ×P(C i ,E)

[0099] Among them, P(C i ,E) represents the first probability distribution (P((C i =true)^E),P((C i =false)^E)); ɑ represents the normalization constant;

[0100]

[0101] Based on the probability distribution P(C i|E), calculate the conditional probability P(C i =true|E);

[0102] P(C i |E)=(P(C i =true|E),P(C i =false|E))

[0103] Among them, P(C i =true|E)+P(C i =false|E)=1.

[0104] In this embodiment, a method for calculating the conditional probability of each fault factor node is described in detail. Specifically, the method first obtains the target conditional probability of each target fault factor node under its parent node according to the topological structure of the Bayesian network. Then, based on these target conditional probabilities, a complete joint distribution probability instance P(((C i =true)^E),C j≠i ). Next, the probability distribution P(C i ,E), and obtain the probability distribution of the target fault factor node under the given abnormal phenomenon set E. Finally, the conditional probability that the target fault factor node is true is extracted from the probability distribution, that is, P(C i =true|E). This process realizes the reverse reasoning from the observed abnormal phenomenon set to the potential fault factor node, providing the fault maintenance analysts with accurate fault source inference results, which helps to quickly locate and solve the device fault.

[0105] The present invention constructs a fault diagnosis Bayesian network to accurately calculate the conditional probability of each fault factor node under a given set of abnormal phenomena, thus realizing reverse reasoning from observed abnormal phenomena to potential fault sources. This method provides fault maintenance analysts with accurate fault source speculation, greatly improving the efficiency and accuracy of fault location and resolution.

[0106] Embodiment 2

[0107] Figure 3 FIG. 1 is a flow chart of a device fault diagnosis method based on a Bayesian network according to another embodiment of the present invention. Figure 3 As shown, according to one embodiment of the present invention, a device fault diagnosis method based on a Bayesian network includes the following steps:

[0108] Step S100, constructing a Bayesian network for fault diagnosis of power system automation and safety devices; specifically comprising:

[0109] Step S101, for the massive historical experience data of power system automation and safety devices, summarize and classify all possible trigger source factors of device failures as the first layer nodes of the Bayesian network, and the factor (cause) node set is called C.

[0110] Among them, the fault source factor nodes include:

[0111] CPU board hardware abnormality (c101), HMI board hardware abnormality (c102), sampling circuit abnormality (c103), IO board hardware abnormality (c104), IO bus signal abnormality (c105), HSB inter-board high-speed bus signal abnormality (c106), internal communication signal abnormality between CPU board and HMI board (c107), FPGA hardware interface abnormality (c108), memory hardware abnormality (c109), file system underlying driver abnormality (c110), software load overload (c111), external GOOSE storm overload (c112), HMI and background remote communication overload (c113), 13), background remote communication configuration error (c114), HSB high-speed communication configuration error (c115), external GOOSE communication configuration error (c116), device program version mismatch error (c117), program pointer runaway (c118), program memory read and write out of bounds (c119), program memory continuous leakage (c120), program file descriptor not released correctly (c121), program network port resources not released correctly (c122), program floating-point task flag not turned on (c123), interface operation error (c124), CPU board and HMI internal communication overload (c125), etc.

[0112] Step S102, analyze all abnormal phenomenon logs recorded internally after the device fails, summarize and classify all possible abnormal phenomena (evidence) recorded internally in the device, and use them as the second layer nodes of the Bayesian network to form an internal abnormal phenomenon node set called E2.

[0113] Among them, the internal abnormal phenomenon nodes include:

[0114] CPU board frequently hard resets (e201), HMI board frequently hard resets (e202), CPU board frequently soft resets (e203), HMI board frequently soft resets (e204), IO board frequent resets (e205), IO board communication verification error (e206), IO board communication interruption error (e207), sampling point loss (e208), sampling interruption (e209), sampling data invalid (e210), constant value setting failure (e211), device HMI and background communication interruption (e212), device external GOOSE communication interruption (e213), device HSB board communication interruption (e214), device HMI and CPU internal communication interruption (e215), device program file corruption error (e216), file format error (e217), etc.

[0115] Step S103, analyze all abnormal phenomena displayed on the external human-machine interaction interface and the monitoring background after the device fails, summarize and classify all possible external abnormal phenomena (evidence) of the classified device, and use the third-layer nodes of the Bayesian network to form a set of external abnormal phenomenon nodes called E3.

[0116] Among them, the external abnormal phenomenon nodes include:

[0117] The device screen goes black and flashes (e301); the device reports a set value setting error event (e302); the device reports a sampling abnormality error event (e303); the device reports an input and output error event (e304); the device reports a software version verification error event (e305); the device reports an FPGA interface error event (e306); the device reports an HMI and background communication interruption event (e307); the device reports an HMI and CPU internal communication error event (e308); the device reports an HSB inter-board high-speed communication error event (e309); the device reports an external GOOSE communication error event (e310), etc.

[0118] Step S104, the abnormal phenomenon (evidence) of the device warning light lighting up triggered by the device failure is used as the fourth layer node of the Bayesian network, and the set of abnormal warning light node is called E4.

[0119] Among them, the abnormal phenomenon node of the warning light includes: the device warning light is on (e401).

[0120] Step S105, based on the device fault triggering source factors and the direct cause-effect relationship between various internal and external fault phenomena of the device, a directed acyclic Bayesian network topology connection relationship between the nodes is established.

[0121] Step S106: Based on the massive device fault history database (the database is formed by historical experience data), a conditional probability distribution table of each node under one or more parent nodes is obtained, which serves as the basis for subsequent fault diagnosis.

[0122] Step S200, according to the phenomenon of a certain fault feedback at the device site, based on the principle of Bayesian network, the cause of the fault is diagnosed.

[0123] Step S201, if a certain fault phenomenon is triggered, the corresponding phenomenon node value is true, otherwise the phenomenon node value is false.

[0124] For example, if the device alarm light is on, the corresponding e401 abnormal phenomenon node is true, otherwise it is false.

[0125] For example, if the fixed value setting fails, the corresponding e211 abnormal phenomenon node is true, otherwise it is false.

[0126] Step S202: Based on the abnormal phenomenon instance set E (E2+E3+E4) of the current fault, traverse and infer each node C in the node set C of the fault trigger source factor of the query device according to the Bayesian network. i = true, that is, P(C i =true|E);

[0127] Step S203, P(C i =true|E) is calculated as:

[0128] To infer (C i =true|E), we can first get the distribution of conditional probability P(C i |E), P(C i |E)=(P(C i =true|E), P(C i =false|E)).

[0129] This can be achieved by the following method.

[0130] a)P(C i |E)=ɑ×P(C i ,E),

[0131] Where ɑ is a normalization constant, ensuring that P(C i =true|E)+P(C i =false|E)=1,

[0132] And P(C i ,E) is the first probability distribution P(P((C i=true)^E),P((C i =false)^E)).

[0133] b) According to the marginalization rule of probability:

[0134]

[0135] Among them, C i and C j Represent the i-th factor and j-th factor in set C respectively.

[0136] c) For each complete joint distribution probability instance P(((C i =true)^E),C j≠i ), can be calculated using the chain rule of Bayesian networks:

[0137]

[0138] Among them, P(x m |parents(X m )) represents x m The conditional probability under the parent node in the Bayesian network, and this conditional probability is known.

[0139] d) Calculate the probability P((C i =false)^E)).

[0140] e) According to the normalization rule, the conditional probability P(C i =true|E).

[0141] Step S204: for all fault factor nodes C i , according to the calculated probability P(C i =true|E) Sort by high and low, and put C with higher conditional probability i The factors are displayed to the fault maintenance analysts to provide a basis and reference for locating the cause of the fault.

[0142] The present invention significantly optimizes the fault location process of power system automation and safety devices. By accurately constructing a Bayesian network containing multi-dimensional fault factors and abnormal phenomenon nodes, it can quickly and accurately infer the potential source of the fault based on the fault phenomenon fed back from the site. The present invention not only provides a scientific and reliable basis for fault location for on-site maintenance analysts, but also greatly shortens the troubleshooting time and reduces the dependence on technical expert resources, thereby achieving effective savings in labor costs.

[0143] Compared with the traditional manual fault location method, the present invention not only improves the efficiency and accuracy of fault diagnosis, but also promotes the stability and safety of power system operation. Through intelligent diagnostic means, this method can help enterprises achieve the goal of reducing costs and increasing efficiency, and further improve the operation and maintenance management level of power equipment.

[0144] Embodiment 3

[0145] This embodiment is an application example of a device fault diagnosis method based on a Bayesian network according to an implementation mode of the present invention, and is used to illustrate the features, purposes and advantages of the present invention.

[0146] This is a case study of fault location based on a Bayesian network according to a specific fault phenomenon. It includes the following steps:

[0147] Step S301, establishing the fault location Bayesian network diagram for the power system automation and safety device of a certain power company according to the fault triggering causal logic of the device and the massive historical fault database in advance.

[0148] Step S302, confirm the status of the device warning light.

[0149] The warning light is on, indicating that the operating device has a fault, and the value of the "device warning light is on" node e401 in the Bayesian network is true.

[0150] Step S303, collecting the abnormal node states in the device external abnormal phenomenon node set E3.

[0151] It has been confirmed that if the device interface reports an abnormal input or output event, the value of the "Device reports input or output error" node e304 is true.

[0152] Step S304, collecting abnormal node states in the abnormal phenomenon node set E2 inside the device.

[0153] Confirmed:

[0154] (1) If the device abnormality log records frequent soft resets of the IO board, the value of the "Frequent soft resets of the IO board" node e205 is true.

[0155] (2) If the device abnormality log records an IO board communication interruption record, the value of the "IO board communication interruption" node e207 is true.

[0156] The fault anomaly node instance set corresponding to this fault

[0157] E=(e207=true,e205=true,e304=true,e other =false)

[0158] Step S305: for any fault source factor C i = true, for example, IO bus signal abnormality factor c105,

[0159] First, the chain rule of the Bayesian network can be used to calculate each complete joint distribution probability instance for subsequent calculations.

[0160] P(((c105=true)^(e207=true,e205=true,e304=true,e other =false)),C j≠i ).

[0161] Step S306: for the fault source factor C i = true, for example, the IO bus signal abnormality factor c105, according to the marginalization rule of probability, calculate the probability

[0162]

[0163] Step S307, according to the marginalization rule of probability, similarly calculate the probability

[0164] P(c105=false^E)

[0165] =P(((c105=false)^(e207=true,e205=true,e304=true,e other =false)))

[0166] Step S308, according to the normalization principle

[0167] Probability distribution

[0168] P(c105|E)

[0169] =(P(c105=true|E),P(c105=false|E))

[0170] =ɑ×P(c105,E)

[0171] =ɑ×(P(c105=true,E),P(c105=false),E).

[0172] Among them, P(c105=true^E) and P(c105=false^E) are known from the previous calculation, ɑ is the normalization constant, ensuring that P(C i =true|E)+P(C i =false|E)=1.

[0173] Step S309: obtain the conditional probability of the IO bus signal abnormality factor as above

[0174] P(c105=true|E)

[0175] =P(((c105=true)|(e207=true,e205=true,e304=true,e other =false)))

[0176] Step S310, similar to the above steps, obtain the conditional probability that the fault source factor of all other fault trigger source factors c101-c125 is true under the current fault phenomenon condition,

[0177] P(C i =ture|E)=P(((C i =true)|(e207=true,e205=true,e304=true,e other =false)))

[0178] Step S311: for all conditional probabilities P(C i =ture|E), sort them in order from largest to smallest, and show the top three most likely fault source factors to the on-site fault analysis and maintenance personnel as a basis and reference for locating the fault cause.

[0179] The method of the present invention has been applied in the on-site operation and maintenance of the power system automation and safety device of a certain power company. The power company has tens of thousands of products in service, and the workload of product operation and maintenance is huge. The technical expert manpower of the R&D center invested in analyzing the on-site problems of the devices every year is very considerable. Since the use of the device fault location analysis software applying the method of the invention, the initial fault location accuracy has reached 85%, and the technical expert manpower and time occupied are only 1 / 4 of those in previous years, which are mainly used to analyze difficult defects on site.

[0180] Embodiment 4

[0181] Figure 7 FIG. 1 is a structural diagram of a device fault diagnosis system based on a Bayesian network according to an embodiment of the present invention. Figure 7 As shown, according to one embodiment of the present invention, a device fault diagnosis system based on a Bayesian network includes:

[0182] An acquisition module is used to obtain historical experience data of power system automation and safety devices;

[0183] A construction module is used to construct a fault diagnosis Bayesian network based on historical experience data;

[0184] The diagnosis module is used to diagnose device faults based on the fault diagnosis Bayesian network.

[0185] According to one embodiment of the present invention, an electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, any Bayesian network-based device fault diagnosis method of the present invention is implemented.

[0186] According to one embodiment of the present invention, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any Bayesian network-based device fault diagnosis method of the present invention is implemented.

[0187] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices and media can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0188] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the present application.

[0189] It should be understood that the size of the serial numbers of each step in the content of the invention and the embodiments of the present invention does not absolutely mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

Claims

1. A device fault diagnosis method based on Bayesian network, characterized in that: include: Obtain historical experience data of power system automation and safety devices; Based on the historical experience data, construct a fault diagnosis Bayesian network; Device faults are diagnosed based on the fault diagnosis Bayesian network.

2. The device fault diagnosis method based on Bayesian network according to claim 1 is characterized in that: Construct a Bayesian network for fault diagnosis, including: Sort out and classify all possible source factors that trigger device failure as the first layer of nodes; Combing and sorting out all the abnormal phenomena recorded in the internal logs of the classification device as the second-layer node; Sort out all possible external display anomalies of the classification device as the third-layer nodes; The abnormal phenomenon of the warning light of the combing device is used as the fourth layer node; According to the direct causal relationship between nodes in each layer, a directed acyclic Bayesian network topological connection relationship between nodes is established; Based on the historical experience data and the Bayesian network topology connection relationship, a conditional probability distribution table of each node under one or more of its parent nodes is established, namely, the fault diagnosis Bayesian network.

3. The device fault diagnosis method based on Bayesian network according to claim 2 is characterized in that: Diagnosing device faults based on the fault diagnosis Bayesian network includes: Obtaining target abnormal phenomena triggered by unknown faults on site, and calculating and sorting the conditional probabilities of all fault factor nodes based on the target abnormal phenomena and the fault diagnosis Bayesian network; The fault factor node whose conditional probability meets the preset conditions is taken as the device fault inference result; Device failure diagnosis is performed based on the device failure estimation result.

4. The device fault diagnosis method based on Bayesian network according to claim 3 is characterized in that: Calculate the conditional probabilities of all fault factor nodes and sort them, including: If the target abnormal phenomenon is triggered, the node value corresponding to the target abnormal phenomenon in the fault diagnosis Bayesian network is set to true; otherwise, the corresponding node value in the fault diagnosis Bayesian network is set to false; Obtain a node value instance set E of all target abnormal phenomena corresponding to the unknown fault on site; Based on the fault diagnosis Bayesian network and the node value instance set E, the conditional probability of each fault factor node is calculated and sorted from high to low according to the conditional probability.

5. The device fault diagnosis method based on Bayesian network according to claim 4 is characterized in that: The conditional probability of each fault factor node is calculated, including: Based on the fault diagnosis Bayesian network, obtaining the target conditional probability of the target fault factor node under the parent node in the Bayesian network topology connection relationship; Based on the target conditional probability, calculate the complete joint distribution probability instance P(((C i =true)^E),C j≠i ), Among them, C i , C j Respectively represent the i-th element and j-th element in the first layer node, j≠i, C i =true means C i The abnormal phenomenon corresponding to the node is triggered; P(x m |parents(X m )) represents the node x m At the parent node X m The target conditional probability under is obtained by querying the fault diagnosis Bayesian network; Based on the complete joint distribution probability instance, the probability distribution P(C i |E), P(C i |E)=ɑ×P(C i ,AND) Among them, P(C i ,E) represents the first probability distribution (P((C i =true)^E),P((C i =false)^E)); ɑ represents the normalization constant; Based on the probability distribution P(C i |E), calculate the conditional probability P(C i =true|E); PRAÇA i |E)=(P(C i =true|E),P(C i =false|E)) Among them, P(C i =true|E)+P(C i =false|E)=1.

6. The device fault diagnosis method based on Bayesian network according to claim 3 is characterized in that: The preset conditions include: the conditional probability is higher than a set threshold, or the ranking of the conditional probability is within a preset range.

7. The device fault diagnosis method based on Bayesian network according to claim 2 is characterized in that: The first layer nodes are fault source factor nodes, including: CPU board hardware abnormality, HMI board hardware abnormality, sampling circuit abnormality, IO board hardware abnormality, IO bus signal abnormality, HSB inter-board high-speed bus signal abnormality, CPU board and HMI board internal communication signal abnormality, FPGA hardware interface abnormality, memory hardware abnormality, file system bottom driver abnormality, software load overload, external GOOSE storm overload, HMI and background remote communication overload, background remote communication configuration error, HSB high-speed communication configuration error, external GOOSE communication configuration error, device program version mismatch error, program pointer runaway, program memory read and write out of bounds, program memory continuous leakage, program file descriptor not released correctly, program network port resource not released correctly, program floating point task flag not opened, interface operation error, CPU board and HMI internal communication overload; and / or, The second layer nodes are internal abnormal phenomenon nodes, including: frequent hard reset of CPU board, frequent hard reset of HMI board, frequent soft reset of CPU board, frequent soft reset of HMI board, frequent reset of IO board, communication check error of IO board, communication interruption error of IO board, sampling point loss, sampling interruption, invalid sampling data, fixed value setting failure, communication interruption between device HMI and background, external GOOSE communication interruption of device, communication interruption between device HSB boards, internal communication interruption between device HMI and CPU, device program file corruption error, file format error; and / or, The third layer node is an external abnormal phenomenon node, including: the device screen is black and flickering, the device reports a set value setting error event, the device reports a sampling abnormality error event, the device reports an input and output error event, the device reports a software version verification error event, the device reports an FPGA interface error event, the device reports an HMI and background communication interruption event, the device reports an HMI and CPU internal communication error event, the device reports an HSB inter-board high-speed communication error event, and the device reports an external GOOSE communication error event; and / or, The fourth layer node is an abnormal phenomenon node of the warning light, including: the device warning light is on.

8. A device fault diagnosis system based on Bayesian network, characterized in that: include: An acquisition module is used to obtain historical experience data of power system automation and safety devices; A construction module, used to construct a fault diagnosis Bayesian network based on the historical experience data; A diagnosis module is used to diagnose device faults based on the fault diagnosis Bayesian network.

9. An electronic device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the method for diagnosing device faults based on a Bayesian network as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the device fault diagnosis method based on a Bayesian network as described in any one of claims 1 to 7 is implemented.

Images