Power distribution network defect-to-fault grade real-time judgment method based on Bayesian network analysis

Abstract

The invention discloses a power distribution network defect-to-fault grade real-time judgment method based on Bayesian network analysis, and the method comprises the following steps: 1, building a standard defect library, 2, building an equipment fault library, 3, respectively defining conditional probabilities for defect severity and fault types; calculating each conditional probability by using a maximum likelihood estimation method, and completing parameter learning of the Bayesian network; 5, inputting defect information detected at a certain moment into the Bayesian network constructed according to the above steps to obtain a defect severity probability and a fault type probability at the moment; 6, reasoning the Bayesian network by using a reasoning algorithm to obtain the occurrence probability of each fault type; and 7, calculating the severity of fault occurrence under a certain given fault type by using the conditional probability. According to the invention, real-time identification of power distribution network equipment defect levels and real-time judgment of equipment defect-to-fault possibility are realized.

Description

Technical Field

[0001] This invention relates to the field of distribution network management technology, and in particular to a real-time method for determining the fault level of distribution network defects based on Bayesian network analysis. Background Technology

[0002] Currently, the equipment in power distribution networks is mainly automated, and its inspection and maintenance rely primarily on human experience and subjective judgment, making it impossible to intelligently identify equipment defects. Once a defect occurs, the response time is relatively slow. If equipment defects are not addressed promptly, they can escalate into equipment malfunctions or even shutdowns, affecting the normal operation of the power distribution network.

[0003] To achieve intelligent inspection of defects in power distribution network equipment, existing technologies need to be combined with big data and deep learning technologies to enable rapid and accurate defect detection and to determine the likelihood of equipment defects escalating into equipment failures. Common steps are as follows: First, a standard database of equipment defects in the power distribution network is collected and established. This database can cover various types of equipment defects and equipment defects under different operating conditions, providing sufficient learning samples for the model and ensuring that it can accurately identify various equipment defects in subsequent stages.

[0004] Secondly, features such as faulty device ID, faulty device name, and fault type are extracted from a large number of original fault feature items to model and analyze the faults of distribution network equipment.

[0005] Based on the established dataset, a defect detection model for power distribution network equipment can be constructed, thereby enabling accurate identification of various defects.

[0006] In recent years, many scholars have conducted research on the problem of judging the transformation of defects into faults in power distribution network equipment, mainly based on traditional image processing and machine vision deep learning algorithms for identification and detection.

[0007] Traditional image recognition methods extract features such as edge, shape, scale-invariant features, and gradient histogram of the target based on the characteristics of power distribution network equipment. First, a judgment is made based on the matching of features, and then the Euclidean distance is calculated to determine the location of the defective equipment.

[0008] The above methods have low accuracy and scalability, can identify a limited number of defect types, have weak ability to express equipment defect characteristics, and have a high false detection rate.

[0009] Moreover, due to the complexity and variability of images, this traditional method is no longer effective in handling diverse defect situations. Regarding identification methods incorporating deep learning techniques, researchers have proposed using algorithms such as SIFT, Faster R-CNN, and SSD to identify equipment defects.

[0010] Traditional equipment defect identification technologies suffer from computational difficulties, high requirements for datasets, and a lack of judgment and classification of the degree of equipment defects. Therefore, they not only have problems such as large computational load and large data volume, but also do not classify and judge the level of defects.

[0011] Therefore, how to achieve real-time identification of the probability level of defects in power distribution network equipment turning into faults has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0012] In view of the above-mentioned deficiencies of the prior art, the present invention provides a real-time method for judging the level of defects in distribution networks based on Bayesian network analysis. The purpose is to achieve real-time identification of the defect level of distribution network equipment and real-time judgment of the possibility of equipment defects turning into faults, reduce the amount of computation, and ensure the real-time and fast defect judgment.

[0013] To achieve the above objectives, this invention discloses a real-time method for determining the fault level of a distribution network based on Bayesian network analysis, comprising the following steps: Step 1: Establish a standard defect library: Select features such as defective equipment ID, defective equipment name, and defect occurrence time to model equipment defects, forming a defect model R1 in the standard defect library, as shown in the following formula: R1={Q1(ID_defect,t1),Q2(n1,f1)} The defect model R1 of the standard defect library is divided into Q1 and Q2 parts; In part Q1, ID_defect is the defective device ID, and t1 represents the time when the device defect occurred. In part Q2, n1 represents the name of the device with the defect, and f1 represents the name of the defect type. Step 2: Establish an equipment fault database: Extract faulty equipment ID, faulty equipment name, fault type, and other features from a large original fault feature database to model and analyze equipment faults, forming a fault model R2 for the equipment fault database, as shown in the following formula: R2={Q3(ID_fault,t2),Q4(n2,f2)} The fault model R2 of the equipment fault library is divided into Q3 and Q4 parts; In Q3, ID_fault represents the faulty device ID. The unique device ID can be used to link devices that have had defects and faults, i.e., ID_fault = ID_defect. In part Q4, t2, n2, and f2 represent the time of equipment failure, the name of the faulty equipment, and the name of the fault type, respectively. Step 3: Define conditional probabilities for defect severity and fault type respectively: The conditional probability of defect severity is defined as: P(Defect Severity|Defect Type), which is the probability distribution of defect severity given a defect type. The conditional probability of a fault type is defined as: P(Fault Type|Defect Severity), which is the probability of a fault type occurring given the severity of the defect. Step 4: Using the large amount of historical data collected, calculate the conditional probabilities using the maximum likelihood estimation method to complete the parameter learning of the Bayesian network. Step 5: Input the defect information detected at a certain moment into the Bayesian network constructed according to the above steps to obtain the defect severity probability P (Defect Severity|Defect Type) at that moment, and further obtain the fault type probability P (Fault Type|Defect Severity). Step 6: Use inference algorithms to infer the Bayesian network and obtain the probability of occurrence of each fault type; Step 7: Based on the reasoning results, further determine the severity of the fault. Use conditional probability P(FaultSeverity|FaultType) to estimate the severity of the fault under a given fault type.

[0014] The beneficial effects of this invention are: Compared with traditional equipment defect identification methods, this invention does not involve complex parameter solving and calculation processes, thus having an advantage in terms of computational complexity.

[0015] Compared to traditional identification methods that only identify the existence and type of equipment defects, this invention can automatically classify and determine the severity of equipment defects; and based on the defect type, it uses Bayesian networks to further assess the likelihood of the defect transforming into equipment failure and the potential severity of any subsequent failure. The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0016] Figure 1 A flowchart of an embodiment of the present invention is shown. Detailed Implementation Example

[0017] like Figure 1 As shown, the real-time method for determining the fault level of a distribution network based on Bayesian network analysis includes the following steps: Step 1: Establish a standard defect library: Select features such as defective equipment ID, defective equipment name, and defect occurrence time to model equipment defects, forming a defect model R1 in the standard defect library, as shown in the following formula: R1={Q1(ID_defect,t1),Q2(n1,f1)} The defect model R1 in the standard defect library is divided into Q1 and Q2 parts; In part Q1, ID_defect is the defective device ID, and t1 represents the time when the device defect occurred. In part Q2, n1 represents the name of the device with the defect, and f1 represents the name of the defect type. Step 2: Establish an equipment fault database: Extract faulty equipment ID, faulty equipment name, fault type, and other features from a large original fault feature database to model and analyze equipment faults, forming a fault model R2 for the equipment fault database, as shown in the following formula: R2={Q3(ID_fault,t2),Q4(n2,f2)} The fault model R2 in the equipment fault database is divided into Q3 and Q4 parts; In Q3, ID_fault represents the faulty device ID. The unique device ID can be used to link devices that have had defects and faults, i.e., ID_fault = ID_defect. In part Q4, t2, n2, and f2 represent the time of equipment failure, the name of the faulty equipment, and the name of the fault type, respectively. Step 3: Define conditional probabilities for defect severity and fault type respectively: The conditional probability of defect severity is defined as: P(Defect Severity|Defect Type), which is the probability distribution of defect severity given a defect type. The conditional probability of a fault type is defined as: P(Fault Type|Defect Severity), which is the probability of a fault type occurring given the severity of the defect. Step 4: Using the large amount of historical data collected, calculate the conditional probabilities using the maximum likelihood estimation method to complete the parameter learning of the Bayesian network. Step 5: Input the defect information detected at a certain moment into the Bayesian network constructed according to the above steps to obtain the defect severity probability P (Defect Severity|Defect Type) at that moment, and further obtain the fault type probability P (Fault Type|Defect Severity). Step 6: Use inference algorithms to infer the Bayesian network and obtain the probability of occurrence of each fault type; Step 7: Based on the reasoning results, further determine the severity of the fault. Use conditional probability P(FaultSeverity|FaultType) to estimate the severity of the fault under a given fault type.

[0018] This invention employs a Bayesian network approach to establish a connection between a database of equipment defects and a database of equipment faults. By calculating conditional probabilities, it assesses the severity of equipment defects and faults, and further determines the likelihood of a defect escalating into a fault. This not only helps reduce computational load but also ensures real-time and rapid defect assessment, and helps prevent equipment faults in power distribution networks.

[0019] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A real-time method for determining the fault level of a distribution network based on Bayesian network analysis; characterized in that, Includes the following steps: Step 1: Establish a standard defect library: Select features such as defective equipment ID, defective equipment name, and defect occurrence time to model equipment defects, forming a defect model R1 in the standard defect library, as shown in the following formula: R1={Q1(ID_defect,t1),Q2(n1,f1)} The defect model R1 of the standard defect library is divided into Q1 and Q2 parts; In part Q1, ID_defect is the defective device ID, and t1 represents the time when the device defect occurred. In part Q2, n1 represents the name of the device with the defect, and f1 represents the name of the defect type. Step 2: Establish an equipment fault database: Extract faulty equipment ID, faulty equipment name, fault type, and other features from a large original fault feature database to model and analyze equipment faults, forming a fault model R2 for the equipment fault database, as shown in the following formula: R2={Q3(ID_fault,t2),Q4(n2,f2)} The fault model R2 of the equipment fault library is divided into Q3 and Q4 parts; In Q3, ID_fault represents the faulty device ID. The unique device ID can be used to link devices that have had defects and faults, i.e., ID_fault = ID_defect. In part Q4, t2, n2, and f2 represent the time of equipment failure, the name of the faulty equipment, and the name of the fault type, respectively. Step 3: Define conditional probabilities for defect severity and fault type respectively: The conditional probability of defect severity is defined as: P(Defect Severity|Defect Type), which is the probability distribution of defect severity given a defect type. The conditional probability of a fault type is defined as: P(Fault Type|Defect Severity), which is the probability of a fault type occurring given the severity of the defect. Step 4: Using the large amount of historical data collected, calculate the conditional probabilities using the maximum likelihood estimation method to complete the parameter learning of the Bayesian network. Step 5: Input the defect information detected at a certain moment into the Bayesian network constructed according to the above steps to obtain the defect severity probability P (Defect Severity|Defect Type) at that moment, and further obtain the fault type probability P (Fault Type|Defect Severity). Step 6: Use an inference algorithm to infer the Bayesian network and obtain the probability of occurrence of each fault type; use the conditional probability P(Fault Severity|Fault Type) to estimate the severity of the fault under a given fault type.

Images