Real-time diagnosis method for commutation failure of phase-controlled converter based on dynamic time warping

Abstract

The present invention discloses a real-time diagnosis method for commutation failure of a phase-controlled converter based on dynamic time warping, which relates to the field of phase-controlled converters, and comprises: designing a time series template S0 covering the trigger angle change of the phase-controlled converter, with a length of n and a Euclidean distance of L; using n collected data as input samples Si, aligning the starting point and the end point of S0 with Si; using an n#imgabs0#n matrix M to represent the Euclidean distance between each point of S0 and Si, using an Itakura window to constrain the matrix M, and determining a search range H; calculating the value of the matrix M within the search range H; setting different weights for data at different positions of the time series, and searching for the matrix coordinate point from (x1, y1) to (x n ,y n ) of the shortest path L#imgabs1#; the margin of the Euclidean distance L of S0 is set to λ, if L#imgabs2# is greater than λL, it is judged that a commutation failure occurs, otherwise it is considered that the system is working normally. The present invention improves the accuracy and real-time performance of diagnosis.

Description

Technical Field

[0001] The present invention relates to the field of phase-controlled converters, and in particular to a real-time diagnosis method for commutation failure of a phase-controlled converter based on dynamic time warping (DTW). Background Art

[0002] Fusion magnet power supply is one of the key subsystems of magnetic confinement nuclear fusion device, providing necessary engineering means for the generation, confinement, maintenance and heating of plasma. At present, fusion magnet power supply mainly adopts phase-controlled converter scheme to achieve high power and high current output. However, commutation failure is an inherent defect of phase-controlled converter. Continuous commutation failure will cause emergency shutdown or even loss of control of the power supply, and then cause plasma hard rupture, threatening the safety of the device. If the fault can be diagnosed in real time and the control parameters can be adjusted when the commutation failure of the phase-controlled converter occurs for the first time, the commutation failure can be effectively avoided and the system can operate normally. Therefore, the real-time and accuracy of commutation failure fault diagnosis are required to be high.

[0003] Dynamic time warping is widely used in the field of real-time diagnosis because of its good robustness and nonlinear alignment characteristics. It can adjust the distance measurement method according to different application scenarios. Dynamic time warping methods mainly include local dynamic time warping, global dynamic time warping, window dynamic time warping, etc. Among them, the window dynamic time warping method includes fixed window dynamic time warping and dynamic window dynamic time warping. In some occasions with certain requirements for accuracy and calculation amount, the dynamic time warping method has obvious effects and is widely used. However, its calculation amount is large, the matrix elements involved in the calculation are more complex, and the accuracy is reduced due to sensitivity to outliers. Since the trigger angle of the fusion magnet power system changes dynamically, when the dynamic time warping method is used for commutation failure fault diagnosis, the designed template should be compatible with different working conditions such as rapid real-time changes of the trigger angle under the inverter state and stable operation of the trigger angle. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a real-time diagnosis method for commutation failure of a phase-controlled converter based on dynamic time warping, which completes the judgment of the fault location and improves the accuracy of the algorithm and the real-time performance of fault judgment.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A real-time diagnosis method for commutation failure of a phase-controlled converter based on dynamic time warping comprises the following steps:

[0007] Step 1: Design a voltage waveform with a time series length of n as the time series template S0, S0=(y 1 ~y n), and calculate the Euclidean distance L of the time series template S0; y 1 ~y n is the n data in the time series template S0;

[0008] Step 2: Take the n sampling point data of the current time series as the input sample Si, Si=(x 1 ~x n ), align the starting and ending points of the time series template S0 with the input sample Si, and perform a monotonicity check on the time series template S0 and the input sample Si to confirm that there is no cross correspondence, where x 1 ~x n is the n sampling point data of the input sample Si, i represents the i-th input sample;

[0009] Step 3: Input sample Si=(x 1 ~x n ) as the horizontal coordinate, the time series template S0=(y 1 ~y n ) as the ordinate to establish a coordinate system, n The matrix M represents the Euclidean distance between the time series template S0 and each point of the input sample Si. Itakura window is used to calculate the n n matrix M is constrained to determine the search range H;

[0010] Step 4: Calculate only n n Matrix element values ​​of matrix M within the search range H;

[0011] Step 5: In the search range H, n Set different weights for the matrix element values ​​in the matrix M and find the coordinate point (x 1 ,y 1 ) to the coordinate point (x n ,y n ) The shortest path L with the smallest sum of matrix element values ​​in n matrix M ;

[0012] Step 6: Set the margin of the Euclidean distance L of the time series template S0 to ;

[0013] Step 7: If the shortest path L If it is greater than λL, it is judged that a commutation failure occurs, otherwise it is considered that the system is working normally.

[0014] Preferably, the step 1 comprises:

[0015] Step 1.1 Set the trigger angle range: Commutation failure occurs in the inverter state of the phase-controlled converter, and the trigger angle range is π / 2~5π / 6;

[0016] Step 1.2 Calculate the trigger angle α under two different working voltages according to the Euclidean distance formula 1 and the trigger angle α 2 The distance L' between:

[0017] ;

[0018] in, is the sampling point phase angle, U is the input voltage, is the angular frequency, t is the time;

[0019] Step 1.3 Obtain the universal Euclidean distance L u :

[0020] ;

[0021] Among them, the average line voltage U d for:

[0022] ;

[0023] Among them, α is the trigger angle;

[0024] Step 1.4 General Euclidean distance L u The functional relationship that is positively correlated with the input voltage U is:

[0025] ;

[0026] Among them, the coefficient for:

[0027] ;

[0028] Step 1.5 Set the trigger angle α 1 Take π / 2, the trigger angle α 2 Take π / 2 and set the trigger angle α 2 Increase by π / 18 until the trigger angle α 2 Take 5π / 6, calculate and draw the first curve, and then change the trigger angle α 1 Take π / 2+π / 18, the trigger angle α 2 Take π / 2 and set the trigger angle α 2 Increase by π / 18 until the trigger angle α 2 Take 5π / 6, calculate and draw the second curve, and so on, calculate and draw 13 curves to show the coefficients at each trigger angle , and obtain the coefficient graph;

[0029] Step 1.6: According to the coefficient diagram, the ninth curve is obtained, and the corresponding maximum value of the Euclidean distance does not exceed 6U, that is, the trigger angle α is taken. 1 The time series of the voltage waveform with an angle of 130° is the best time series template;

[0030] Step 1.7 According to the corresponding Euclidean distance Range: , the Euclidean distance of the best time series template is 6U.

[0031] Preferably, step 5 includes: the weights corresponding to the first n / 3 matrix elements are q1, the weights corresponding to the middle n / 3 to 5n / 6 matrix elements are q2, and the weights corresponding to the last 5n / 6 to n matrix elements are q3.

[0032] Preferably, in step 3, n nThe element M in the matrix M mj Calculated according to the following formula:

[0033] ;

[0034] Among them, x m Represents the mth sampling point data in the input sample Si, y m Represents the mth data in the time series template S0, and j is the index value.

[0035] Preferably, in step 3, the search range H is limited to a parallelogram with slopes of 1 / 2 and 2, and the path is planned from the coordinate point (x 1 ,y 1 ) to the coordinate point (x n ,y n ) ends, the four coordinate points of the parallelogram in the search range H are (x 1 ,y 1 ), (x n / 3 ,y 2n / 3 ), (x 2n / 3 ,y n / 3 ), (x n ,y n ).

[0036] Preferably, in step 6, λ is 1.5.

[0037] Preferably, q1 is 0.5, q2 is 0.8, and q3 is 1.3.

[0038] The beneficial effects of the present invention are:

[0039] The present invention calculates the changes of each trigger angle in the full range of inversion, and can achieve accurate real-time diagnosis results of commutation failure faults for stable systems with unchanged trigger angles, such as constant trigger angle operation in high-voltage direct current transmission, and dynamic systems with real-time changes in trigger angles, such as fusion magnet power supply systems using phase-controlled rectification schemes. In addition, by applying different weights at different sequence positions, the calculation process is optimized, and the sensitivity of the algorithm to abnormal values ​​of the phase-controlled converter is effectively improved. The fault location can be judged more quickly, thereby improving the accuracy of the algorithm and the real-time nature of fault judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of a three-phase fully controlled converter unit.

[0041] Figure 2 It is a schematic diagram of the voltage waveform in the active inverter state.

[0042] Figure 3 It is the Euclidean distance coefficient diagram under different trigger angles.

[0043] Figure 4 This is a diagram of the Itakura constraint window.

[0044] Figure 5a It is a diagram of variable weight.

[0045] Figure 5b It is a schematic diagram of a variable weight window.

[0046] Figure 6 It is a flow chart of the real-time diagnosis method of phase-controlled converter commutation failure based on dynamic time warping of the present invention.

[0047] Figure 7a This is the single-phase full-controlled rectifier circuit single-phase attenuation 20% fault detection waveform.

[0048] Figure 7b This is the waveform diagram of the voltage change when a single-phase fully-controlled rectifier circuit decays by 20%.

[0049] Figure 7c This is the waveform diagram of the fault current change when the single-phase fully controlled rectifier circuit decays by 20%.

[0050] Figure 8a This is the waveform diagram of single-phase full-controlled rectifier circuit single-phase attenuation 40% fault detection.

[0051] Figure 8b This is the waveform diagram of the voltage change when a single-phase fully-controlled rectifier circuit decays by 40%.

[0052] Figure 8c This is the waveform diagram of the fault current change when the single-phase fully controlled rectifier circuit decays by 40%.

[0053] Figure 9a This is the waveform diagram of single-phase full-controlled rectifier circuit single-phase attenuation 60% fault detection.

[0054] Figure 9b This is the voltage change waveform of a single-phase fully controlled rectifier circuit with a single-phase attenuation of 60%.

[0055] Fig.9c This is the waveform diagram of the fault current change when the single-phase fully controlled rectifier circuit decays by 60%.

[0056] Fig.10a It is the single-phase attenuation 80% fault detection waveform of a single-phase fully controlled rectifier circuit.

[0057] Fig.10b This is the waveform diagram of the voltage change when a single-phase fully-controlled rectifier circuit decays by 80%.

[0058] Fig.10c This is the waveform diagram of the fault current change when the single-phase fully controlled rectifier circuit decays by 80%.

[0059] Fig.11a This is the waveform diagram of the single-phase attenuation 20% fault detection of the three-phase fully controlled rectifier circuit.

[0060] Fig.11b This is the waveform diagram of the voltage change when a single-phase three-phase fully controlled rectifier circuit decays by 20%.

[0061] Fig.11c This is the waveform diagram of the fault current change when a single-phase decays by 20% in a three-phase fully controlled rectifier circuit.

[0062] Fig.12a This is a graph of pulse loss diagnosis results.

[0063] Figure 12b This is the pulse loss voltage waveform change diagram.

[0064] Fig.12c This is the pulse loss current waveform change diagram. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0066] The EAST (Experimental Advanced Superconducting Tokamak) poloidal field power supply system adopts a phase-controlled rectification scheme. Each power supply consists of four three-phase fully controlled converter units. The topology of a single three-phase fully controlled converter unit is as follows: Figure 1 As shown, the first thyristor V T1 and the fourth thyristor V T4 Series, the third thyristor V T3 With the sixth thyristor V T6 , the fifth thyristor V T5 With the second thyristor V T2 Series connection, load inductance and loop resistance are connected in series, and the above components are connected in series and then in parallel. a , U b , U c One end of each is connected to the ground point O, and the other end is connected to the load inductor L and then to the first thyristor V T1 and the fourth thyristor V T4 、The third thyristor V T3 With the sixth thyristor V T6 、The fifth thyristor V T5 With the second thyristor V T2 Each switch device (i.e., thyristor) is controlled by a control signal to turn on and off. Their function is to control the direction and magnitude of the current and to achieve the switching operation through the trigger pulse. That is, the present invention adopts a rectifier bridge in a bridge structure, and every two switch devices form a bridge arm. There are three bridge arms in total, which are respectively connected to the three phases of the three-phase power supply. The switch devices at the upper and lower ends of each bridge arm complement each other, that is, when one is turned on, the other is turned off, forming a six-pulse rectifier.

[0067] like Figure 6 As shown, the real-time diagnosis method of phase-controlled converter commutation failure based on dynamic time warping of the present invention comprises the following steps:

[0068] Step 1: Design a voltage waveform with a time series length of n as the time series template S0, S0=(y 1 ~y n ), and calculate the Euclidean distance L of the time series template S0; y 1 ~y n is the n data in the time series template S0; that is Figure 6 The sampling in the signal data sequence that needs to be diagnosed; Step 2, take the n sampling point data of the current time series as the input sample Si, Si=(x 1 ~x n), align the starting and ending points of the time series template S0 with the input sample Si, and perform a monotonicity check on the time series template S0 and the input sample Si to confirm that there is no cross correspondence, where x 1 ~x n is the n sampling point data of the input sample Si, i represents the i-th input sample;

[0069] Step 3: Input sample Si=(x 1 ~x n ) as the horizontal coordinate, the time series template S0=(y 1 ~y n ) as the ordinate to establish a coordinate system, n The matrix M represents the Euclidean distance between the time series template S0 and each point of the input sample Si. Itakura window is used to calculate the n n matrix M is constrained to determine the search range H;

[0070] Step 4: Calculate only n n Matrix element values ​​of matrix M within the search range H;

[0071] Step 5: In the search range H, n Set different weights for the matrix element values ​​in the matrix M and find the coordinate point (x 1 ,y 1 ) to the coordinate point (x n ,y n ) The shortest path L with the smallest sum of matrix element values ​​in n matrix M ;

[0072] Step 6: Set the margin of the Euclidean distance L of the time series template S0 to ;

[0073] Step 7: If the shortest path L Greater than λ , it is judged that a commutation failure occurs, otherwise it is considered that the system is working normally.

[0074] Specifically, the step 1 includes:

[0075] Using the DTW method for commutation failure fault diagnosis requires first finding a suitable template. Commutation failure occurs in the inverter state of the thyristor converter. The maximum range of the trigger angle is within π / 2~π. Considering that the inverter angle of the actual system cannot reach π, the upper limit of the inverter angle can be adjusted according to different systems. Here, the trigger angle range is π / 2~5π / 6. The voltage waveform in the active inverter state is as follows: Figure 2 As shown. Among them, u dis the output voltage, ω is the angular frequency, and t is the time. Figure 2 It can be seen that the shapes and amplitudes of the waveforms corresponding to different trigger angles are very different. As the trigger angle increases, the amplitude of the output voltage increases.

[0076] The sampling frequency is 10kHz, the six-pulse cycle is 20ms, each pulse is 3.3ms, corresponding to 33 sampling points :

[0077] ;

[0078] Expandable ;

[0079] Calculate the line voltage U x for:

[0080] ;

[0081] Among them, α is the trigger angle, β is the sampling point phase angle, U is the input voltage, is the angular frequency, and t is the time. According to the six-pulse voltage waveform characteristics, each sampling point is calculated The corresponding line voltage value.

[0082] In order to make the time series template compatible with the stable system with constant trigger angle and the dynamic system with real-time change of trigger angle in the full range of inverter angle, it is necessary to calculate the trigger angle α under two different working voltages according to the Euclidean distance formula. 1 , and the trigger angle α 2 The distance L' between:

[0083] ;

[0084] The average value of the line voltage under the two trigger angles can be removed, and the universal Euclidean distance L u The formula is:

[0085] ;

[0086] Average line voltage U d for:

[0087] ;

[0088] Set the coefficient k to:

[0089] ;

[0090] The calculated universal Euclidean distance L u The functional relationship that is positively correlated with the input voltage U is:

[0091] ;

[0092] Trigger angle α 1 Take π / 2, the trigger angle α 2 Take π / 2 and set the trigger angle α 2 Increase by π / 18 until the trigger angle α 2 Take 5π / 6, calculate and draw the first curve, and then change the trigger angle α 1 Take π / 2+π / 18, the trigger angle α 2 Take π / 2 and set the trigger angle α 2 Increase by π / 18 until the trigger angle α 2 Take 5π / 6, calculate and draw curve 2, and so on, calculate and draw 13 curves to show the coefficient changes at each trigger angle, and obtain the coefficient diagram, such as Figure 3 As shown. Figure 3 The time series corresponding to the trigger angle with the minimum Euclidean distance in the whole range of trigger angles is determined as the best time series template. The ninth curve obtained according to the coefficient diagram is the corresponding maximum value of the Euclidean distance, which does not exceed 6U, that is, the trigger angle α is taken. 1 The time series of the voltage waveform with an angle of 130° is the best time series template.

[0093] The corresponding Euclidean distance The range is:

[0094] ;

[0095] Therefore, the Euclidean distance of the optimal time series template is 6U.

[0096] Specifically, in step 3, Si=(x 1 ~x n ) as the horizontal coordinate, S0=(y 1 ~y n ) as the ordinate to establish a coordinate system, and the element M in the matrix M mj Calculated according to the following formula:

[0097] ;

[0098] Among them, x m Represents the mth sampling point data in the input sample Si, y m Represents the mth data in the time series template S0, and j is the index value.

[0099] Limit the search range H to the parallelogram with slopes 1 / 2 and 2, and plan the path from the coordinate (x 1 ,y 1 ) to the coordinate point (x n ,yn ) ends, such as Figure 4 As shown, the four coordinate points of the parallelogram of the search range H are (x 1 ,y 1 ), (x n / 3 ,y 2n / 3 ), (x 2n / 3 ,y n / 3 ), (x n ,y n ).

[0100] Specifically, in step 5, the first 1 / 3 of the input sample Si contains the lowest amount of outlier information and is assigned a low weight. The middle 1 / 2 of the time series reflects the similarity of part of the entire path and is assigned a medium weight. The tail 1 / 6 of the one-dimensional time series has a high amount of outlier information and is assigned a high weight. The weight size is designed according to the actual test effect, such as Figure 5a , Figure 5b As shown, Figure 5a is a diagram of variable weights, Figure 5b is a schematic diagram of a variable weight window, where n is the length of the acquisition sequence, is the i-th sampling point of the acquisition time series, is the i-th data of the template. Within the search range, different weights are set for data at different positions in the time series. Figure 5a As shown in the figure, the weight corresponding to the first n / 3 matrix elements is q1, the weight corresponding to the middle n / 3~5n / 6 matrix elements is q2, and the weight corresponding to the last 5n / 6~n matrix elements is q3. According to the actual detection results, q1, q2, and q3 are set, q1 is 0.5, q2 is 0.8, and q3 is 1.3.

[0101] In actual engineering, various external interferences such as noise are considered, and the margin of the Euclidean distance L of the time series template S0 is set to λ. λ can be adjusted according to the actual test results. Here, it is selected as 1.5. If the corresponding Euclidean distance L Greater than λ , it is judged that a commutation failure occurs, otherwise it is considered that the system is working normally.

[0102] like Figures 7a to 12a The following are the diagnostic results of the method, including single-phase fading 20% ​​diagnostic result diagram, single-phase fading 40% diagnostic result diagram, single-phase fading 60% diagnostic result diagram, single-phase fading 80% diagnostic result diagram, three-phase fading 20% ​​diagnostic result diagram, and pulse loss diagnostic result diagram. The horizontal axis is the sampling point, and the vertical axis is the dynamic time warping distance. The curve represents the DTW distance, and the straight line represents the fault detection result.

[0103] like Figure 7b to Figure 12bAs shown, they are single-phase attenuation 20% voltage waveform change diagram, single-phase attenuation 40% voltage waveform change diagram, single-phase attenuation 60% voltage waveform change diagram, single-phase attenuation 80% voltage waveform change diagram, three-phase attenuation 20% voltage waveform change diagram, pulse loss voltage waveform change diagram, the horizontal axis is the sampling point, and the vertical axis is the voltage value.

[0104] like Figure 7c to Figure 12c As shown, they are the current waveform change diagrams of single-phase attenuation 20%, single-phase attenuation 40%, single-phase attenuation 60%, single-phase attenuation 80%, three-phase attenuation 20%, and pulse loss current waveform change diagrams. The horizontal axis is the sampling point and the vertical axis is the current value. Under normal operating conditions, the dynamic time regularization distance value of the voltage signal is stable, conforming to the multi-pulse law, and no abnormal values ​​appear within a certain amplitude range. Figure 7b and Figure 7c It can be seen that 0.15 seconds after the fault was injected, the DC side current fluctuated, but there was no obvious commutation failure in the DC side voltage. However, the waveform had lost the multi-pulse regularity. Figure 7a It can be seen that the present invention determines that commutation failure may occur after 95 sampling points. Figure 8b and Figure 8c It can be seen that when the AC side voltage decays by 40%, a commutation failure fault has already occurred. 0.15 seconds after the fault is injected, the waveforms of the first and second commutation cycles are distorted, and the third commutation fails. Figure 8a It can be seen that the present invention determines that commutation failure may occur after 20 sampling points, that is, the waveform distortion is found during the first commutation. Figures 9a to 11a The commutation failure judgment cases when the AC side voltage decays by 60%, 80% and the three-phase decays by 20% are shown respectively. Figure 7a resemblance. Figures 9b to 11b The voltage waveform changes when the AC side voltage is attenuated by 60%, 80% and the three-phase voltage is attenuated by 20% are shown respectively. Figures 9c to 11c The current waveform changes when the AC side voltage is attenuated by 60%, 80% and the three-phase attenuation is 20% are shown respectively. Fig.12a As shown in the figure, after the fault occurs, the voltage fluctuates significantly and the commutation failure is determined after 38 sampling points. Figure 12b and 12c The voltage and current waveforms change for pulse loss fault.

Claims

1. A real-time diagnosis method for commutation failure of a phase-controlled converter based on dynamic time warping, characterized in that: The steps include: Step 1: Design a voltage waveform with a time series length of n as the time series template S0, S0=(y1~y n ), and calculate the Euclidean distance L of the time series template S0; y1~y n The n data in the time series template S0 include: Step 1.1 Set the trigger angle range: Commutation failure occurs in the inverter state of the phase-controlled converter, and the trigger angle range is π / 2~5π / 6; Step 1.2 Calculate the distance L' between the trigger angle α1 and the trigger angle α2 under two different working voltages according to the Euclidean distance formula: ； in, is the sampling point phase angle, U is the input voltage, is the angular frequency, t is the time; Step 1.3 Obtain the universal Euclidean distance L u : ； Among them, the average line voltage U d for: ； Among them, α is the trigger angle; Step 1.4 General Euclidean distance L u The functional relationship that is positively correlated with the input voltage U is: ； Among them, the coefficient for: ； Step 1.5 Set the trigger angle α1 to π / 2, the trigger angle α2 to π / 2, and increase the trigger angle α2 by π / 36 until the trigger angle α2 is 5π / 6. Calculate and draw the first curve. Then set the trigger angle α1 to π / 2+π / 36, the trigger angle α2 to π / 2, and increase the trigger angle α2 by π / 36 until the trigger angle α2 is 5π / 6. Calculate and draw the second curve. And so on. Calculate and draw 13 curves to display the coefficients at each trigger angle. , and obtain the coefficient graph; Step 1.6: The ninth curve is obtained according to the coefficient diagram, and the corresponding maximum value of the Euclidean distance does not exceed 6U, that is, the time series of the voltage waveform with a trigger angle α1 of 130° is taken as the best time series template; Step 1.7 According to the corresponding Euclidean distance Range: , the Euclidean distance of the best time series template is 6U; Step 2: Take the n sampling points of the current time series as the input sample Si, Si=(x1~x n ), align the starting and ending points of the time series template S0 with the input sample Si, and perform a monotonicity check on the time series template S0 and the input sample Si to confirm that there is no cross correspondence, where x1~x n is the n sampling point data of the input sample Si, i represents the i-th input sample; Step 3: Input sample Si=(x1~x n ) as the horizontal coordinate, the time series template S0=(y1~y n ) as the ordinate to establish a coordinate system, n The matrix M represents the Euclidean distance between the time series template S0 and each point of the input sample Si. Itakura window is used to calculate the n n matrix M is constrained to determine the search range H; n nThe element M in the matrix M mj Calculated according to the following formula: ； Among them, x m Represents the mth sampling point data in the input sample Si, y m Represents the mth data in the time series template S0, j is the index value; Limit the search range H to the parallelogram with slopes 1 / 2 and 2, and plan the path from the coordinate point (x1, y1) to the coordinate point (x n ,y n ) ends, the four coordinate points of the parallelogram in the search range H are (x1, y1), (x n / 3 ,y 2n / 3 ), (x 2n / 3 ,y n / 3 ), (x n ,y n ); Step 4: Calculate only n n Matrix element values ​​of matrix M within the search range H; Step 5: In the search range H, n Set different weights for the matrix element values ​​in the matrix M and find the coordinate point (x1, y1) to the coordinate point (x n ,y n ) The shortest path with the smallest sum of matrix element values ​​in n matrix M , including: the weight corresponding to the first n / 3 matrix elements is q1, the weight corresponding to the middle n / 3~5n / 6 matrix elements is q2, and the weight corresponding to the last 5n / 6~n matrix elements is q3; Step 6: Set the margin of the Euclidean distance L of the time series template S0 to ; Step 7: If the shortest path Greater than λ , it is judged that a commutation failure occurs, otherwise it is considered that the system is working normally.

2. The real-time diagnosis method for phase-controlled converter commutation failure based on dynamic time warping according to claim 1 is characterized in that: In step 6, λ is 1.

5.

3. The real-time diagnosis method for phase-controlled converter commutation failure based on dynamic time warping according to claim 1 is characterized in that: q1 is 0.5, q2 is 0.8, and q3 is 1.3.

Images