CN111579941A - Online Rail Insulation Defect Diagnosis Method Based on Feature Extraction of Full Line Rail Potential - Google Patents

Abstract

The invention discloses an on-line rail insulation defect diagnosis method based on the extraction of potential features of the whole rail, belonging to the application field of urban rail transit fault detection. Feature extraction, insulation diagnosis is performed by matching the conditional features to the reference group; finally, probability statistics are performed for the diagnosis results in a judgment cycle, and finally the defect location is realized by a given limit value. The invention can not only diagnose the defect situation quickly, but also accurately locate the defect position, so as to provide reference for stray current protection.

Description

Online Rail Insulation Defect Diagnosis Method Based on Feature Extraction of Full Line Rail Potential

technical field

The invention belongs to the application field of urban rail transit fault detection. Specifically, it relates to an on-line rail insulation defect diagnosis method based on the extraction of potential features of the whole rail.

Background technique

In 2019, the number of urban rail transit cities in mainland China increased to more than 40, with a total operating mileage of more than 6,700 kilometers. Urban rail transit is generally powered by 1500V or 750V DC. Most urban rail systems use steel rails as return rails. The train draws current from the catenary through the pantograph for the locomotive to consume energy, and the current returns to the negative pole of the traction power supply station from the rail. Due to the existence of the rail-to-ground transition resistance, the current will leak to the ground when it flows in the rail, and this part of the current is called stray current. The stray current flows into the ground, inevitably flows in the metal structure in the ground, and finally flows back to the negative electrode of the traction station through the transition resistance and other channels.

The electric current flows from the cathode area to the anode area in the underground pipeline. After the oxidation reaction occurs in the anode area, the metal dissolves, and the pipe wall becomes thinner or even perforated. Usually there is spontaneous corrosion of the pipeline, and the corrosion rate is slow. Due to the interference of stray current, electrochemical corrosion occurs. When the current value is large, the corrosion rate is fast and the perforation time is short. The current flows in the soil near the substation, and is left inside the transformer through the transformer grounded at the neutral point, causing the transformer to have a DC bias. After the occurrence of DC bias, the harmonics generated in the excitation current of the transformer will cause the transformer to be abnormal, causing problems such as temperature rise, vibration, and noise. Larger stray current disturbances cause additional economic losses.

The stray current protection design adopts the principle of "mainly blocking, supplemented by exhaust, combined with prevention and exhaust, and strengthen monitoring". Blocking: Reduce the current flowing into the ground from the source. Reduce the distance between the traction stations; reduce the longitudinal resistance of the rail; increase the transition resistance of the rail to the ground; improve the voltage level. Drainage: reduce the current flowing into the ground through drainage measures. Set up a drain system to create a loop for stray current to flow back to the negative pole of the traction station. Test: monitor the line condition to provide the basis for the operation and maintenance of stray current.

The important parameters that affect the distribution of subway stray current are the rail longitudinal resistance and the rail-to-ground transition resistance. After the subway is completed and enters the operation stage, with the accumulation of time, the accumulation of line pollution will reduce the transition resistance of the rail to the ground, thereby increasing the stray current. "CJJ49-1992 Subway Stray Current Corrosion Protection Technical Regulations" stipulates that the transition resistance value between the subway track that is also used for backflow and the main structure of the tunnel (or the ground) (measured according to the block section and converted to 1km long resistance value), not less than 15Ω·km for new lines, and not less than 3Ω·km for operating lines. The monitoring of rail-to-ground transition resistance is particularly important. If the transition resistance does not meet the requirements, it is regarded as a defect in the insulation. At this time, the current leakage of the rail to the ground is aggravated, resulting in pipeline corrosion and DC bias.

The existing line transition resistance detection method adopts the method in Appendix B.2 of GB28026.2-2018. During the detection, it is affected by the upstream and downstream flow line and the flow line between peers, and the detection length generally does not exceed 400m. The single detection efficiency is low, and the number of detections increases when the line is long, which consumes a lot of manpower and material resources. Transition resistance detection of operating lines can only be carried out in the early hours of the morning. In view of the above situation, the research of an online insulation defect diagnosis technology is very necessary. When the rail-to-ground insulation changes in a certain section, the distribution of rail-to-ground potential across the entire line will change significantly. The most direct characterization factor of the insulation situation should be the distribution of the rail potential across the line.

SUMMARY OF THE INVENTION

The purpose of the invention is to monitor the rail potential in the urban rail power supply system, and to record and analyze the data online in time, so as to realize the diagnosis of the line insulation and provide a reference for the stray current protection.

To this end, the present invention provides an on-line rail insulation defect diagnosis method based on the extraction of full-line rail potential features, comprising the following steps:

Step 1: Data acquisition and storage. The real-time rail potential data information of each measurement point is obtained through the rail potential measurement device of each measurement point, and the information including the departure interval, load current and related equipment status is obtained through information interaction. Store the collected data in a real-time database; convert it into data identifiable by the monitoring system through the server's own protocol; the system processes the data, providing historical data management, data query (including various feature quantities), statistical calculation, insulation diagnosis, And send control commands such as alarm to the system.

Step 2: Manage the historical data, select the stage of the initial operation of the line and the line condition is good, process the historical data of this stage, and form different reference groups.

For different departure intervals _{th in the historical data, when ZT h =} 00 in the calculation _{period, the short-term average value AVE ih and} The maximum value of rail potential MAX _ih , the minimum value of short-term rail potential MIN _ih , and the short-term absolute average value ABV _ih are used as different reference groups. The short-term feature quantities here are all based on the corresponding departure interval as the calculation period. The calculation formula is as follows :

MAX _ih =max(I _ihn ,n=1,2,3....N),i=1,2,...I (3)

MIN _ih =min(I _ihn ,n=1,2,3....N),i=1,2,...I (4)

Among them, n represents the nth data in a calculation cycle, N represents the total number of such data in the calculation cycle; q represents the qth traction station on the line, Q represents the total number of traction stations on the line; i represents the line th i measurement points, I represents the total number of line measurement points.

Step 3: Extract feature quantities from real-time data, and perform insulation diagnosis by matching conditional features to reference groups.

3.1 Real-time data processing:

Taking the operating time within 24 hours as the diagnosis period, and with different departure intervals t _s in the period, when ZT _s = 00 in the period, the short-term average load current AVI _qs of each traction station q is calculated as the condition quantity for matching the reference value; The short-term average AVE _is of the rail potential at each measurement point i, the short-term maximum rail potential MAX _is , the short-term rail potential minimum MIN _is , and the arithmetic mean ABV _is of the short-term absolute values are used as the characteristic quantities of insulation diagnosis, Calculated as follows:

MAX _is =max(I _isn ,n=1,2,3....N),i=1,2,...I (8)

MIN _is =min(I _isn ,n=1,2,3....N),i=1,2,...I (9)

3.2 Match the reference group, take _ts = t _h , ZT _s = 00, the smallest difference between the AVI _qs and AVI _qh calculated in real time as the matching condition for matching the reference group, and E is the minimization target, where E is expressed as follows:

3.3 Data comparison: Compare the calculated feature quantities of each measurement point with the matched reference groups, and calculate the deviation ΔU _ixj and the offset rate σ _i as follows:

ΔU _ixj =U _ixj-me -U _ix-ref ,i=1,2,...I (12)

Among them, j represents the jth comparison in the diagnosis cycle, j=1, 2, 3...; x represents the selected diagnostic feature quantity.

3.4 Abnormal diagnosis: for different measurement points i, when the deviation rate σ _i ≤ u _x , u _x is the limit of the deviation rate; it is considered that the rail potential characteristic x of the jth comparison measurement point i is abnormal, and the abnormal times k _{ix =} k _ix +1.

3.5 Single diagnosis of suspected defects: in the jth comparison, when analyzing different diagnostic feature x, if any measurement point diagnostic feature x is abnormal, compare the ΔU _ixj of different measurement points i; if it is relatively short If it is the maximum value of rail potential or the short-term absolute average value, the maximum ΔU _ixj is taken as the abnormal comparison result of this index x; if the short-term rail potential is the minimum value, the smallest ΔU _ixj is taken as the abnormality of this index x. Comparison result; the measurement point i pointed to by the abnormal comparison result is the suspected defect measurement point in the j- _th comparison and diagnosis of the characteristic quantity x, then the suspected defect statistic of the diagnosis characteristic quantity x of this measurement point is +1.

3.6 Single diagnosis of defects based on comprehensive diagnostic feature quantities: In the j-th comparison, according to the difference in the selected diagnostic feature quantities, the measurement point i pointed to by the abnormal comparison result of the diagnostic feature quantity x satisfies:

When the arithmetic mean of the maximum value, the minimum value and the absolute value is taken as the diagnostic feature quantity, there is at least the same i that the comparison results of the two indicators point to, and the statistic of suspicious defects at the measurement point is n _i +1.

When taking the maximum value and the minimum value as the diagnostic indicators, there are two possibilities. The i pointed to by the maximum value and the i pointed to by the minimum value are the same i, and n _i +1; the i pointed to by the maximum value and the i pointed to by the minimum value are not The same i, will be different n _i +1.

When taking the maximum value or the minimum value as the diagnostic index, the abnormal result of the index points to i, n _i +1.

3.7 The next group of data processing and comparison in the diagnosis period: update j=j+1, and repeat steps 3.1 to 3.6 until the diagnosis period ends.

Step 4: Carry out probability statistics on the diagnosis results, and finally realize the defect location through the given limit value.

The number of comparisons in each diagnostic cycle is J, and the matrix [K] represents the number of abnormality, the matrix [M] represents the number of suspected defects, and the matrix [N] represents the number of defects; the initial matrix elements are all 0; where [K], [M ] The middle row represents the measurement point, the column represents the compared characteristic quantity; the [N] middle row represents the measurement point;

与缺陷比率

The abnormal ratio of each measurement point in the statistical diagnosis cycle

to defect ratio

When β _i ≥ β _u , i=1, . . . 10, i is the positioned defective measurement point. If all β _i <β _u , there is no defective point; where β _u is the rated defect ratio.

Further, the relevant equipment is the rail potential limiter OVPD and the drain device; when the equipment state is open, it is expressed as "0"; when the equipment state is closed, it is expressed as "1". The device status is represented by the status value ZT as shown in Table 1:

Table 1 Equipment Status Table

Historical/real-time device status ZT_h/ZT_s OVPD, drainage device are disconnected 00 OVPD open, drain closed 01 OVPD closed, drain open 10 OVPD, drainage device are closed 11

Further, there are various calculation cycles in the diagnosis cycle, and the data should be exchanged with PSCADA real-time information. The total number of data in one diagnosis cycle is J groups, that is, the number of comparisons is J times.

Further, the diagnostic feature quantities are the short-term rail potential maximum value, the short-term rail potential minimum value, and the short-term absolute average value.

Further, u _x has different values according to different lines, and no matter which diagnostic feature quantity is taken, it is negative.

The beneficial technical effect of the present invention compared with the prior art is:

1. Using the online diagnosis method, the defect position can be quickly located.

2. The stray current is closely related to the transition resistance. When a certain part of the insulation has problems, the stray current increases. Through this method, defects can be found in time, which can provide reference for stray current protection.

Description of drawings

Figure 1 is a schematic diagram of the invention data collection.

Figure 2 is a schematic flow chart of the present invention.

Detailed ways

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

The flow chart of the online rail insulation defect diagnosis method based on the feature extraction of the full line rail potential of the present invention is shown in FIG. According to the data acquisition device of each measurement point in Figure 1, the data is collected and transmitted to the monitoring system, and other real-time information, such as train departure interval time, equipment status, etc., is obtained from the PASCDA system.

According to the method of the present invention, the historical data and real-time data are processed to calculate the short-term average load current of each traction station, the short-term average value of the rail potential of each measuring point, the short-term rail potential maximum value, the short-term rail potential minimum value, Arithmetic mean of short-term absolute values, etc. After matching the reference group, the deviation value and deviation rate are calculated, and a single defect diagnosis is carried out for different diagnostic feature quantities. After the diagnosis cycle is over, statistics are performed. Defect diagnosis is performed with the defect ratio as the limit value.

Specific examples are:

Reference value acquisition:

The historical data processing results when ZT _h = 00 are shown in Table 2:

Table 2 Detailed list of reference groups

data processing:

A group of real-time data processing results when ZT _s = 00 are shown in Table 3.

Table 3 Real-time data processing results

Insulation Diagnostics:

1. Match the reference group. Since t _s =150 s = t _h1 , and the E value of the real-time data and the reference group is calculated, E ₁ =0.4<E2=9.57e+04<E3=1.72e+05<E3=2.14e+05. Match to reference group 1.

2. Data comparison. The calculation results of deviation value and deviation rate are shown in Table 4:

Table 4 Calculation results of deviation value and deviation rate

3. Abnormal diagnosis. This line takes u _x =-20%, then after this comparison k ₃₁ +1, k ₃₃ +1.

4. One-time diagnosis of suspected defects. m ₃₁ +1, m ₃₃ +1.

5. Comprehensive diagnostic feature quantity for single-time defect diagnosis. Cases of n _i under different comparison rules:

(1) Comprehensive three indicators, short-term maximum rail potential, short-term minimum rail potential, short-term absolute average rail potential: n ₃ +1;

(2) Comprehensive short-term maximum rail potential and short-term minimum rail potential: n ₃ +1.

(3) When only the maximum rail potential is used as the diagnostic feature quantity, n ₃ +1; when only the minimum rail potential is used as the diagnostic feature quantity, n _i +0.

6. Statistics of single diagnosis in one day. The comprehensive maximum rail potential and minimum rail potential are taken as diagnostic feature quantities, and the statistical results are as follows:

7. The total number of comparisons is J=4. After data statistics, the diagnosis result is that the rail potential of station 1, station 2, station 3 and station 4 is abnormal. Station 3 has a maximum anomaly ratio of 100%. The defect ratios β ₃ =100% and β ₄ =50% are located at measurement point 3 and measurement point 4 .

The invention utilizes the characteristics of the change of the rail potential to realize the diagnosis of the rail-ground transition resistance insulation defect, makes up for the gap of the online insulation diagnosis method, and the diagnosis indirectly reflects the stray current situation, which can be used as a reference for the stray current protection.

Claims (5)

1. The online steel rail insulation defect diagnosis method based on the whole-line steel rail potential feature extraction is characterized by comprising the following steps of:

step 1: data acquisition and storage: acquiring real-time steel rail potential data information of each measuring point through a steel rail potential measuring device of each measuring point, and acquiring departure intervals, load current and related equipment states through information interaction; storing the acquired data to a real-time database;

step 2: managing historical data, selecting a stage with initial line operation and good line condition, processing the historical data of the stage to form different reference groups:

for different departure intervals t in historical data_hSatisfy ZT in calculation cycle_hWhen the load current is 00 hours, a plurality of groups of short-time average load currents AVI are arranged_qhShort-time average AVE of rail potential of each corresponding measuring point_ihShort-time rail potential maximum MAX_ihMinimum value MIN of short-time rail potential_ihShort term absolute average ABV_ihAs different reference groups, the short-time characteristic quantities here all take the corresponding departure intervals as calculation periods, and the calculation formula is as follows:

MAX_ih＝max(I_ihn,n＝1,2,3....N),i＝1,2,...I (3)

MIN_ih＝min(I_ihn,n＝1,2,3....N),i＝1,2,...I (4)

wherein N represents the nth data in a calculation period, and N represents the total number of the data in the calculation period; q represents the Q-th traction station on the line, and Q represents the total number of the traction stations on the line; i represents the ith measuring point of the line, and I represents the total number of the measuring points of the line;

and step 3: extracting feature quantity of the real-time data, and performing insulation diagnosis on a reference group through condition feature matching:

3.1 real-time data processing:

taking the operating time within 24h as a diagnosis period, and carrying out different departure intervals t in the period_sSatisfy the ZT in the cycle_sWhen the load current exceeds 00, the average load current AVI of each traction station q is calculated_qsAs a condition quantity matching the reference value; average value AVE of rail potential at each measurement point i_isShort-time rail potential maximum MAX_isMinimum value MIN of short-time rail potential_isABV, arithmetic mean of absolute values in short time_isAs the characteristic amount of the insulation diagnosis, the calculation formula is as follows:

MAX_is＝max(I_isn,n＝1,2,3....N),i＝1,2,...I (8)

MIN_is＝min(I_isn,n＝1,2,3....N),i＝1,2,...I (9)

3.2 matching reference groups, t_s＝t_h、ZT_sAVI calculated in real time 00 ═ 00_qsAnd AVI_qhIs used as a matching condition for matching the reference group, and E is a minimization target, wherein E is expressed as follows:

3.3 data comparison: comparing the characteristic quantity calculated by each measuring point with the matched reference group, and calculating the deviation delta U_ixjAnd offset ratio sigma_iThe following were used:

ΔU_ixj＝U_ixj-me-U_ix-ref,i＝1,2,...I (12)

wherein j represents the jth comparison within the diagnostic cycle, j being 1,2, 3.; x represents the selected diagnostic characteristic quantity;

3.4 abnormality diagnosis: for different measurement points i, when the offset ratio σ_i≤u_xWhen u is turned on_xIs an offset rate limit; considering that the j-th comparison measurement point i has abnormal steel rail potential characteristic quantity x and abnormal times k_ix＝k_ix+1；

3.5 Single diagnosis of suspected defects: in the j comparison, when different diagnosis characteristic quantities x are analyzed, if any measurement point diagnosis characteristic quantity x is abnormal, comparing the delta U of different measurement points i_ixj(ii) a If the maximum value or the absolute average value of the short-time rail potential is compared, the maximum delta U is taken_ixjIs the abnormal ratio of the index x of this timeComparing the results; if the rail potential minimum value is short, the minimum delta U is taken_ixjIs the abnormal comparison result of the index x; if the abnormal comparison result points to the measurement point i as the suspicious defect measurement point when the diagnostic feature quantity x is compared for the jth time, the suspicious defect statistic m of the diagnostic feature quantity x of the measurement point is determined_ix+1；

3.6 the comprehensive diagnosis characteristic quantity is used for defect single diagnosis: during the jth comparison, aiming at different selected diagnosis characteristic quantities, the measurement point i pointed by the abnormal comparison result of the diagnosis characteristic quantity x satisfies the following conditions:

when the arithmetic mean value of the maximum value, the minimum value and the absolute value is taken as the diagnosis characteristic quantity, at least two indexes exist and point to the same i after the comparison result of the two indexes, and the suspicious defect statistic n of the measuring point_i+1；

When the maximum value and the minimum value are taken as diagnosis indexes, two possibilities exist, i pointed by the maximum value and i pointed by the minimum value are the same i, n_i+ 1; i pointed by the maximum value is not the same as i pointed by the minimum value, and n is different_i+1；

When the maximum value or the minimum value is taken as a diagnosis index, the index abnormal result points to i, n_i+1；

3.7 the next set of data processing and comparison during the diagnostic period: updating j to j +1, and repeating the steps 3.1 to 3.6 until the diagnosis period is ended;

and 4, step 4: carrying out probability statistics on the diagnosis result, and finally realizing defect positioning through a given limit value:

the comparison frequency of each diagnosis period is J, a matrix [ K ] represents the abnormal frequency, a matrix [ M ] represents the suspicious defect frequency, and a matrix [ N ] represents the defect frequency; the initial matrix elements are all 0; wherein, the middle lines of [ K ] and [ M ] represent measuring points, and the columns represent compared characteristic quantities; the [ N ] middle row represents a measurement point;

counting the abnormal ratio of each characteristic quantity of each measuring point in the diagnosis period

To the defect ratio of each measuring point

When β_i≥β_uI of 10 is a measured point of the defect located, if all β are present_i＜β_uThere are no defect points β_uIs the nominal defect rate.

2. The online steel rail insulation defect diagnosis method based on the whole-line steel rail potential feature extraction is characterized in that the related equipment is a steel rail potential limiter (OVPD) and a drainage device; when the device state is off, it is denoted as "0"; when the device state is closed, it is denoted as "1".

3. The online steel rail insulation defect diagnosis method based on the whole-line steel rail potential feature extraction is characterized in that multiple calculation periods exist in the diagnosis period, data are interacted with PSCADA in real time, the total number of data in one diagnosis period is J groups, and the comparison times are J times.

4. The online steel rail insulation defect diagnosis method based on all-line steel rail potential feature extraction is characterized in that the diagnosis feature quantity is a short-time steel rail potential maximum value, a short-time steel rail potential minimum value and a short-time absolute average value.

5. The on-line steel rail insulation defect diagnosis method based on all-line steel rail potential feature extraction as claimed in claim 1, wherein u is_xDifferent values are taken according to different lines, and the diagnostic characteristic quantity is negative no matter what is taken.

Images