CN103018592A - Fault diagnosis method of traction transformer based on model diagnosis - Google Patents

Abstract

The invention is a fault diagnosis method of a traction transformer based on model diagnosis. The method uses the knowledge of traction transformer structure model to establish a two-layer structure model of traction transformer, uses voltage transformer and current transformer to measure the actual traction transformer operating parameters, and inputs the parameters into the diagnosis system for diagnosis. The present invention divides the diagnosis into two parts of off-line diagnosis and online diagnosis at the same time, improves the diagnosis speed, and finally diagnoses the fault element and the fault type through consistency diagnosis and retrospective diagnosis. The simulation results show that the method can quickly and accurately diagnose the faults of traction transformer winding short circuit, open circuit, and indirect ground fault, and it is a new fault diagnosis method for traction transformers.

Description

Fault Diagnosis Method of Traction Transformer Based on Model Diagnosis

technical field

The invention belongs to the technical field of fault diagnosis of a traction power supply system, in particular to a fault diagnosis method for a traction transformer based on model diagnosis.

Background technique

The traction power supply system is the power supply system of the railway. Once a part of it fails and the power supply is interrupted, it will not only affect the normal operation of the trains on the power supply line, but may even trigger a chain reaction in the current high-density and high-flow railway transportation system, causing The poor operation of many railway lines caused significant economic losses. As an important part of the traction power supply system, the safe operation of the traction transformer is directly related to the reliable and stable operation of the traction power supply system, and then to the normal and orderly operation of the railway. Therefore, the normal operation of the traction transformer is an important guarantee for the normal operation of the traction power supply system and the railway system. In order to ensure the smooth operation of railways, traction transformers are required to have extremely high operational availability, reliability and safety. However, due to equipment manufacturing issues and operating environment and operating time issues, traction transformers inevitably have some failures. Therefore, in order to ensure the normal operation of the traction transformer, it is necessary to grasp the operation status of the traction transformer in real time. When a minor fault occurs and it is operating in an abnormal state, it must be discovered and dealt with in time to avoid further development of the fault and damage to the equipment and cause power interruption; At the same time, when a serious fault occurs and the power supply is interrupted, it is necessary to find out the cause of the fault as soon as possible and take effective measures to restore the power supply as much as possible, so as to shorten the power outage time and reduce the economic loss caused by the power outage.

At present, the diagnosis method of traction transformer mainly has two aspects: the first one is to observe a single parameter such as temperature, ultrasonic wave, moisture content of insulating paper, etc., and diagnose the transformer by analyzing the change of parameters, but this method can only diagnose one aspect The second is the artificial intelligence method, such as the application of expert system and artificial neural network, but it is difficult to obtain experience knowledge of expert system, artificial neural network needs a large amount of sample data, and the development of diagnostic system The time is longer.

The method of Model-based Diagnosis (MBD) was proposed in the 1970s. MBD uses the principle structure and knowledge of the diagnostic object without the accumulation process of knowledge. At the same time, system modeling and system diagnosis The reasoning is two parts completely separated, with good independence and portability. The literature "Liu Zhigang, Zhong Wei, Deng Yunchuan, Qu Changjun. Model-Based Diagnosis Method for Traction Substation Faults. Proceedings of the Chinese Society for Electrical Engineering, 2010, 30(34): 36-41" uses MBD for fault diagnosis of traction substations, but in The equipment diagnosis of the traction transformer adopts a simple equivalent component method, which cannot diagnose the specific faults of the traction transformer, but can only diagnose the faults of the circuits on both sides of the traction transformer; in addition, the diagnosis method is only single-layer modeling and direct reasoning However, it cannot effectively and comprehensively diagnose the faults of the traction transformer. In order to change the difficulties encountered in traction transformer fault diagnosis and make up for the deficiencies of expert systems, it is of great practical significance to use MBD method for traction transformer modeling and transformer fault diagnosis.

Contents of the invention

The purpose of the present invention is to provide a fault diagnosis method for traction transformers based on model diagnosis to solve the shortcomings of conventional diagnosis methods for traction transformers. In order to achieve the above purpose, the technical solution adopted by the present invention is a model-based diagnosis method for traction transformer fault diagnosis, which provides real-time fault diagnosis for the traction transformer, which is the key equipment in the traction power supply system of the railway power supply system. The method includes the following means:

Step 1: According to the structure of the traction transformer, establish an abstract model of the traction transformer structure and construct a diagnosis system. The specific method is: use the phase voltage and phase current of the transformer as system model variables to describe the connection relationship of each winding; adopt a layered structure abstract model: The first layer establishes a small component model to describe the normal behavior of small components; the second layer establishes a large component model to describe the fault behavior of small components and the fault behavior between small components; the layered model converts the topology fault between small components into The internal failure of large components is conducive to the analysis of failures.

Step 2: Search the abstract model of the traction transformer offline to obtain all the analytical redundant relations of the diagnostic system and their associated minimum candidate conflict sets;

Step 3: Measure the actual fault voltage and current data of the traction transformer through the voltage transformer and current transformer;

Step 4: Obtain the minimum conflict set from the minimum candidate conflict set: bring the fault state information of the traction transformer measured by the current and voltage transformers into the analytical redundancy relationship, and check whether the analytical redundancy relationship is satisfied; if not, the system prediction and actual If there is a deviation between the observed values, then the supporting environment for parsing the redundant relationship is a minimum conflict set;

Step 5: From the obtained minimum conflict set, use the collision set calculation method in the field of artificial intelligence to find the minimum collision set of the minimum conflict set, and obtain all the minimum candidate diagnoses;

Step 6: Perform fault matching on the components in the minimum candidate diagnosis, and finally obtain the specific faulty components and their fault types.

Due to the accuracy of modeling, some of the analytical redundancy relations under normal operation observations do not hold. A permissible error value is set to eliminate misjudgments caused by accuracy issues.

In the process of fault matching, the faults of candidate components are sorted by using the probability of component failure, and the fault mode with the highest failure probability is prioritized for fault matching. If the matching is unsuccessful, the fault with a small failure probability is considered.

The beneficial effects of the present invention are:

1. The present invention uses a model-based method to directly perform online diagnosis of traction transformer faults, which meets real-time requirements and obtains objective and accurate diagnosis results.

2. The present invention applies the model structure knowledge of the traction transformer, which overcomes the shortcomings of the traditional expert system diagnosis method, such as difficulties in collecting empirical knowledge.

Description of drawings

Figure 1 Actual Diagnosis Process

Fig. 2 Flowchart of traction transformer fault diagnosis based on model diagnosis

Figure 3 Equivalent circuit diagram of traction transformer

Detailed ways

The present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments.

Figure 1 is the process of fault diagnosis for the actual traction transformer, and its specific process is shown in Figure 2.

Fig. 2 is a standard flowchart of traction transformer fault diagnosis based on model diagnosis. In Fig. 2, the traction transformer fault diagnosis method based on model diagnosis includes:

Step 1: According to the structure of the traction transformer, an abstract model of the traction transformer structure is established to obtain a diagnosis system;

The phase voltage and phase current of the transformer are used as system model variables to describe the connection relationship of each winding. Abstract model with layered structure: the first layer establishes small component models to describe the normal behavior of small components; the second layer establishes large component models to describe the fault behavior of small components and the fault behavior between small components. The hierarchical model converts topological faults among small components into internal faults of large components, which is beneficial to fault analysis. The traction transformer can be represented by abstract components as COMPS={T1_AB, T1-BC, T1, T21_TRF1, T22TRF2, T21, T22}, and the meanings represented by the symbols of each abstract component are shown in Table 1.

Table 1 The meanings of the symbols in the component composition

Figure 3 is the single-phase equivalent circuit diagram of the VX traction transformer, which is the equivalent diagram converted to the primary side, where A and B represent the A-phase and B-phase of the power line respectively, and T, F and N represent the catenary and the positive feeder respectively and the earth. First of all, for the single-phase transformer shown in Figure 3, there are the following observation variables: {VA, VB, VAB, IA, IB, VTR, VRF, VFT, IT, IR, IF}, VA, VB represent the incoming line AB incoming line Phase voltage, VAB, VTR, VRF, VFT are the line voltage between the respective phases, IA, IB are the current values flowing into the transformer at the incoming terminal, IT, IR, IF are the current values flowing out of the transformer at the outgoing terminal.

Then, when the transformer is in normal operation, the voltage and current constraints for the entire transformer are:

$\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; IA</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; IB</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; IT</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; IF</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; IR</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; IA</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; IT</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; IF</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; IT</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; tr</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; VTR</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

$\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; VAB</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; IA</span>}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; ab</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; tn</span>}}<span\; class="notranslate">\; *</span>\frac{\mathrm{<span\; class="notranslate">\; IT</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; VTR</span>}\\ \mathrm{<span\; class="notranslate">\; VAB</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; IA</span>}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; ab</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; nf</span>}}<span\; class="notranslate">\; *</span>\frac{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; IF</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; VRF</span>}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

For T1_AB, there is also a voltage relation:

VAB＝VA-VB (3)

For T21_TRF1, there is also a voltage relationship:

0＝VTR+VRF+VFT (4)

The above formulas (1) to (4) are the relationship models of the transformer in normal operation, and are used in the modeling of the first layer of single components. Applying the normal model relationship in the consistency-based diagnosis method identifies component failures, but the specific failure type cannot be identified. Therefore, in the second-level large component set, the fault types of various components are introduced, and the specific component fault types can be diagnosed through fault matching through abductive diagnosis.

Common transformer faults include phase-to-phase short-circuit, ground short-circuit, inter-turn short-circuit, etc. Some can establish direct fault model relational expressions, such as phase-to-phase short-circuit to ground short-circuit, etc. For example, when phase A is short-circuited to ground, VA=0, IA+IB≠0; if phase A is short-circuited, VAB=0, IA=IB. Some faults, such as inter-turn short circuit and other fault features, are not obvious, so it is impossible to conduct retrospective diagnosis by establishing a fault model. Only component faults can be judged through consistency diagnosis, and the specific fault type cannot be judged.

Step 2: Search the abstract model of the traction transformer offline to obtain all the analytical redundancy relations of the system and their associated minimum candidate conflict sets;

The arrangement of the monitoring devices of the traction transformer remains unchanged, and some fixed analytical redundant relations containing only observable quantities can be generated through the information of the measuring devices. Using the search algorithm, the normal model established by the system is searched offline to obtain the analytic redundancy relationship, and then its minimum support environment (that is, the minimum conflict set candidate) can be obtained. A total of 8 candidate candidates for all analytic redundancy relationships of traction transformers and corresponding minimum conflict sets are obtained through offline search. For example: MinCSC3={{T21_TRF1}, 0=-IT1-IR1-IF1}, that is, {T21_TRF1} is the minimum candidate conflict set, and its corresponding analytical redundancy relationship is "0=-IT1-IR1-IF1".

Step 3: Measure the actual fault voltage and current data of the traction transformer through the voltage transformer and current transformer;

The voltage transformer and current transformer of the system observation device are mainly used to observe the system and provide system status information for the diagnosis of the diagnostic system. The failure of the voltage transformer and current transformer is not considered here. Therefore, assuming that the measuring devices are working normally, the measured data is accurate.

Step 4: Bring the fault data into the parsing redundancy relationship, and obtain the minimum conflict set from the minimum candidate conflict set;

Assuming that the traction transformer is short-circuited to ground at the outlet terminal of T1 and RF2 is short-circuited at the secondary side of T2, because R is the rail, that is, the F-winding is short-circuited to ground, the measured value of the voltage and current transformer of the traction transformer under this fault is obtained through simulation as Table 2 and Table 3 are shown.

Table 2 Measured values of voltage transformers under fault conditions

Table 3 Measured values of current transformers under fault conditions

The measured values obtained in Table 2 and Table 3 are brought into the analytic redundancy relation of the minimum candidate conflict set, and the magnitude of the obtained vector value is taken to obtain the absolute residual. Since the comparison between the absolute residuals has no credibility, the relative residuals are introduced, which is the ratio of the absolute residuals to the maximum magnitude of the variable in the analytical redundancy relationship. The calculation results are shown in Table 4.

Table 4 Residuals for parsing redundant relations

Due to the accuracy of modeling, some of the analytical redundancy relations under normal operation observations do not hold, and an allowable error value is set to eliminate misjudgments caused by accuracy issues. Here, the allowable relative residual is set to 0.2, and the minimum conflict set MinCs={MinCSC1, MinCSC4}, namely MinCsC={{T1_AB}, {T22_TRF2}} is obtained from the minimum candidate conflict set with a relative residual greater than 0.2 in Table 4.

Step 5: From the obtained minimum conflict set, use the collision set calculation method in the field of artificial intelligence to find the minimum collision set of the minimum conflict set, and obtain all the minimum candidate diagnoses;

From the minimum collision set MinCs={MinCSC1, MinCSC4}, namely MinCsC={{T1_AB}, {T22_TRF2}}, then obtain the minimum collision set MinHs={T1_AB, T22_TRF2}

So far, through the normal model and consistency reasoning diagnosis of system components, the faulty components {T1_AB, T22_TRF2} can be obtained.

Step 6: Perform fault matching on the components in the minimum candidate diagnosis, and finally obtain the specific faulty components and their fault types.

In order to quickly find out the specific fault type of the faulty component, according to the actual engineering data, it can be assumed that the probability of short-circuit fault is 0.6, the probability of ground fault is 0.4, and the probability of disconnection fault is 0.2.

In the process of matching candidate diagnoses by abductive reasoning, according to the magnitude of the failure probability, the failure mode with the highest failure probability is firstly matched, and if the matching is unsuccessful, then the failure with a lower failure probability is considered. Based on the previous failure probability assumptions, the failure probability of the failure components is sorted, and the top five types with the highest failure probability are shown in Table 5.

Table 5 Various failure modes and their qualitative failure probabilities

Referring to the qualitative probability of each failure mode given in Table 5, the failure mode is preferentially selected: {T1_AB, {groundA}}, {T22_TRF2, {groundF}} for matching, and the maximum allowable residual is set to 0.2. Finally, the matching is successful and the failure type is determined It is {T1_AB, {groundA}}, {T22_TRF2, {groundF}}, so it can be determined that T1_AB, phase A is ground fault, and T22_TRF2, F is grounded, that is, FR is short-circuited.

Claims (3)

1. a kind of traction transformer fault diagnosis method based on model diagnosis, to traction power supply system is key equipment traction transformer in the power supply system of railway to provide real-time fault diagnosis, it is characterized in that, described method comprises following means: Step 1: According to the structure of the traction transformer, establish an abstract model of the traction transformer structure and construct a diagnosis system. The specific method is: use the phase voltage and phase current of the transformer as system model variables to describe the connection relationship of each winding; adopt a layered structure abstract model: The first layer establishes a small component model to describe the normal behavior of small components; the second layer establishes a large component model to describe the fault behavior of small components and the fault behavior between small components; Step 2: Search the abstract model of the traction transformer offline to obtain all the analytical redundant relations of the diagnostic system and their associated minimum candidate conflict sets; Step 3: Measure the actual fault voltage and current data of the traction transformer through the voltage transformer and current transformer; Step 4: Obtain the minimum conflict set from the minimum candidate conflict set: bring the fault state information of the traction transformer measured by the current and voltage transformers into the analytical redundancy relationship, and check whether the analytical redundancy relationship is satisfied; if not, the system prediction and actual If there is a deviation between the observed values, then the supporting environment for parsing the redundant relationship is a minimum conflict set; Step 5: From the obtained minimum conflict set, use the collision set calculation method in the field of artificial intelligence to find the minimum collision set of the minimum conflict set, and obtain all the minimum candidate diagnoses; Step 6: Perform fault matching on the components in the minimum candidate diagnosis, and finally obtain the specific faulty components and their fault types. 2. According to the method described in claim 1, it is characterized in that, when the abstract model of the traction transformer structure is established, due to the accuracy of the modeling, the existing part of the analytical redundancy relationship under normal operation observation does not hold, by setting An allowable error value to exclude false positives caused by precision problems. 3. The method according to claim 1, characterized in that, in the process of fault matching, the faults of the candidate components are sorted by using the probability of component failure, and the failure mode with the highest failure probability is prioritized for fault matching. If the matching is unsuccessful, then consider the fault with a small fault probability.

Images