Modular multilevel converter open-circuit fault diagnosis method based on DTW algorithm
Through dynamic time alignment algorithm and reconstructed switch function calculation, the problems of high hardware cost, large calculation amount and high misdiagnosis rate in open circuit fault diagnosis of modular multi-level converter are solved, and fast and accurate fault diagnosis is achieved. , improves the robustness and application prospects of diagnosis.
Patent Information
- Application Number
- CN202411214332.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-31
- Publication Date
- 2024-12-06
AI Technical Summary
The existing modular multi-level converter open circuit fault diagnosis methods have problems such as high hardware cost, strong mathematical model dependence, large calculation amount and high misdiagnosis rate, especially in terms of rapid diagnosis and robustness.
Using a method based on dynamic time alignment (DTW) algorithm, the fault position is detected by calculating the distance of the capacitance voltage of the submodule, the fault position is determined by using the difference between the fault submodule and the normal submodule, and the capacitor voltage is calculated by reconstructing the switching function. Estimates reduce hardware cost and calculation amount, reduce dependence on system parameters, and improve diagnostic robustness.
It realizes fast and accurate fault diagnosis, reduces hardware cost and calculation amount, avoids misdiagnosis, and improves the robustness and application prospects of diagnosis.
Smart Images

Figure CN119087294A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for diagnosing an open circuit fault of a modular multi-level converter, and more particularly to a method for diagnosing an open circuit fault of a modular multi-level converter based on a DTW algorithm. Background Art
[0002] Modular Multilevel Converter (MMC) has become one of the most promising topologies in medium and high voltage applications. Its main application areas include high voltage DC transmission systems, battery energy storage systems, power electronic transformers, motor drives, synchronous static compensators, etc. MMC has the advantages of modularity, easy expansion, low switching frequency and good harmonic performance. However, it has a large number of submodules, each of which contains two insulated gate bipolar transistors (IGBTs). The existence of these submodules also increases the possibility of switch failure. Open circuit failure will not immediately damage the submodule, but will cause waveform distortion, overcurrent and overvoltage problems. If remedial measures are not taken in time, it may cause secondary failure of the entire MMC. Therefore, how to quickly diagnose the open circuit failure of the submodule is of great significance to improve the reliability of the MMC.
[0003] Traditional MMC open circuit fault diagnosis methods can be divided into hardware-based methods, mathematical model-based methods and artificial intelligence-based methods. Hardware-based methods have the advantages of fast diagnostic speed, short positioning time and low computational complexity, which are crucial to improving the stability and reliability of the system. However, the installation of sensors will also increase the complexity and maintenance cost of the system. The method based on mathematical models needs to establish a corresponding mathematical model based on the observed quantity. The diagnostic results are affected by the system parameters, and misdiagnosis occurs with the change of working conditions. The applicability and robustness are poor. The method based on artificial intelligence processes and analyzes the sampled data to extract the features related to the fault, and uses the trained model or algorithm to classify and locate the extracted features. However, this method requires a large number of data samples, the algorithm complexity is high and the computational complexity is large. Summary of the invention
[0004] The technical problem to be solved by the present invention is to provide a modular multi-level converter open circuit fault diagnosis method based on DTW algorithm which does not require additional hardware circuits, has a small number of mathematical models to establish, and does not require a large number of data samples in order to overcome the shortcomings of the prior art.
[0005] The technical solution adopted by the present invention is: a modular multi-level converter open circuit fault diagnosis method based on DTW algorithm, comprising the following steps:
[0006] 1) In the current control cycle, a sliding window with a length of L is designed to obtain capacitor voltage samples of the modular multilevel converter;
[0007] 2) Take the capacitor voltages of two submodules in the same bridge arm, ensure that their sequence lengths are equal and both are L, and calculate the distance D of the capacitor voltages of the two submodules based on the dynamic time normalization algorithm j,xy ;
[0008] 3) Determine the location of the faulty submodule
[0009] When the capacitor voltage in the bridge arm is in a completely linear relationship, the distance D based on the dynamic time normalization algorithm is 0. When the distance is in a completely inverse linear relationship, the value of D is not 0. The closer to 0, the stronger the correlation between the capacitor voltages. When the value of D is not 0, the location of the submodule is the fault location.
[0010] 4) Calculate the upper tube of the submodule and the lower tube of the submodule S 2 Switching function S during fault jx_S1fault , S jx_S2fault , and thus calculate the estimated capacitor voltage value U based on the reconstruction switch function when the upper tube of the submodule and the lower tube of the submodule fail sm_jx_S1fault and U sm_jx_S2fault ;
[0011] 5) Real-time comparison of the difference ΔU between the estimated value of the capacitor voltage of the upper tube and the lower tube of the submodule and the measured capacitor voltage sm_jx_S1 and ΔU sm_jx_S2 , if ΔU sm_jx_S1 <ΔU th , then start the timer T 1 , if T 1 Greater than the set time threshold T th , it can be determined that the upper tube of the xth submodule is faulty. sm_jx_S2 Less than the set capacitor voltage threshold ΔU th , then start the timer T 1 , if T 1 >T th , it can be determined that the lower tube of the x-th submodule is faulty.
[0012] The modular multi-level converter open circuit fault diagnosis method based on DTW algorithm of the present invention detects the position of the faulty submodule by calculating the DTW distance of the capacitor voltage of the submodule in the same bridge arm, determines the fault position by using the difference between the faulty submodule and the normal submodule, and calculates the estimated value of the capacitor voltage of the submodule based on the reconstructed switch function, and locates the faulty switch tube according to the duration of the actual value of the capacitor voltage tracked by the estimated value. Compared with the fault diagnosis method based on the hardware method, the method of the present invention does not need to add additional hardware circuits, which reduces the cost of the fault diagnosis device; compared with the fault diagnosis method based on the mathematical model method, the number of mathematical models established is small, and after determining the faulty submodule, it is not necessary to construct the mathematical models of the upper and lower bridge arms, which reduces the dependence of the diagnosis algorithm on the bridge arm inductance L, and only needs to construct two mathematical models for the faulty submodule, avoiding the operation of the remaining normal submodules, greatly reducing the amount of calculation, the diagnosis result is not affected by the system parameters, avoiding the misdiagnosis caused by interference, and has high robustness; compared with the fault diagnosis method based on the artificial intelligence method, it does not require a large number of data samples, the algorithm is relatively simple and the amount of calculation is small. The proposed method has a good application prospect. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1a It is the flow path when the submodule is in the on state and the bridge arm current is greater than 0;
[0014] Figure 1b It is the flow path when the submodule is in the on state and the bridge arm current is less than 0;
[0015] Figure 1c It is the flow path when the submodule is in the cut-off state and the bridge arm current is greater than 0;
[0016] Figure 1d It is the flow path when the submodule is in the cut-off state and the bridge arm current is less than 0;
[0017] Figure 2a It is the flow path when the tube on the submodule fails;
[0018] Figure 2b It is the flow path when the lower tube of the submodule fails;
[0019] Figure 3 It is the DTW regularized path graph;
[0020] Figure 4 It is a flow chart of the modular multi-level converter open circuit fault diagnosis method based on the DTW algorithm of the present invention. DETAILED DESCRIPTION
[0021] The following is a detailed description of the modular multi-level converter open circuit fault diagnosis method based on the DTW algorithm of the present invention in conjunction with the embodiments and drawings.
[0022] The flow path diagram of the MMC submodule under normal conditions is as follows Figure 1a-1b As shown in the figure, the open circuit fault situation is as follows Figure 2a-2b As shown. When the submodule upper tube S 1 When an open circuit fault occurs, the submodule current flow path is Figure 1b becomes Figure 2a ; When the submodule lower tube S 2 When an open circuit fault occurs, the submodule current flow path is Figure 1c becomes Figure 2b By comparison, it can be determined that when S 1 or S 2 In case of an open circuit fault, the submodule output voltage and charge and discharge status will change.
[0023] The modular multi-level converter open circuit fault diagnosis method based on DTW algorithm of the present invention is as follows: Figure 4 As shown, the following steps are included:
[0024] 1) In the current control cycle, in order to obtain the capacitor voltage samples of the modular multilevel converter, a sliding window with a length of L is designed
[0025] Taking the upper bridge arm of phase a as an example, according to Figure 3 Calculate the DTW distance D between the capacitor voltages of two submodules in the same bridge arm a,12 , D a,13 , D a,14 First, a sliding window with a basic period is designed to obtain capacitor voltage samples. The expression of the window length L is:
[0026]
[0027] Where f is the output frequency, T s is the sampling period;
[0028] 2) Take the capacitor voltages of two submodules in the same bridge arm, ensure that their sequence lengths are equal and both are L, and calculate the distance D between the capacitor voltages of the two submodules based on the dynamic time warping algorithm (DTW) j,xy
[0029] First, determine the distance metric d(U sm_jx,w , U sm_jy,v ), whose expression is:
[0030]
[0031] Where U sm_jx,w represents the value of the capacitor voltage of the xth submodule of phase j at the wth sampling point, U sm_jy,v represents the value of the capacitor voltage of the yth submodule in the jth phase at the vth sampling point; d(Usm_jx,w , U sm_jy,v ) represents U sm_jx The wth sampling point in U sm_jy The distance metric of the vth sampling point in;
[0032] The cumulative distance c(w,v) is calculated based on the distance metric. The value of the sum of the minimum value of the three cumulative distances c(w-1,v), c(w,v-1) and c(w-1,v-1) and the distance metric of the current search position is divided into four cases, namely:
[0033]
[0034] Finally, the DTW (based on dynamic time warping algorithm) distance between the capacitor voltages of the two submodules is expressed as:
[0035] D j,x y =d(U sm_jx,L ,U sm_jy,L )+min{c(L-1,L),c(L,L-1),c(L-1,L-1)} (4)
[0036] Wherein, c(LI,L) represents the cumulative distance between the capacitor voltage of the first submodule at the LI sampling point and the capacitor voltage of the second submodule at the L sampling point, c(L,LI) represents the cumulative distance between the capacitor voltage of the first submodule at the L sampling point and the capacitor voltage of the second submodule at the LI sampling point, and c(LI,LI) represents the cumulative distance between the capacitor voltage of the first submodule at the LI sampling point and the capacitor voltage of the second submodule at the LI sampling point;
[0037] 3) Determine the location of the faulty submodule
[0038] When the capacitor voltage in the bridge arm is in a completely linear relationship, the distance D based on the dynamic time normalization algorithm is 0. When the distance D is in a completely inverse linear relationship, the value of D is not 0. The closer to 0, the stronger the correlation between the capacitor voltages. When the value of D is not 0, the location of the submodule is the fault location.
[0039] Taking the number of bridge arm submodules N = 4 on phase a as an example, the DTW distances in the same bridge arm are compared with the threshold D th For comparison, if D a,12 ≈0, D a,13 ≈0, D a,14 ≈0, then there is no fault; if D a,12 >D th , D a,13 >D th , D a,14 >D th , then a phase SM 1 Fault; if Da,12 >D th , D a,13 ≈0, D a,14 ≈0, then the a-phase SM 2 Fault; if D a,12 ≈0, D a,13 >D th , D a,14 ≈0, then the a-phase SM 3 Fault; if D a,12 ≈0, D a,13 ≈0, D a,14 >D th , then a phase SM 4 Fault;
[0040] 4) Calculate the upper tube S of the submodule 1 , submodule lower tube S 2 Switching function S during fault jx_S1fault , S jx_S2fault , thus calculating the S based on the submodule 1 , submodule lower tube S 2 The estimated capacitor voltage U calculated by reconstructing the switching function during a fault sm_jx_S1fault and U sm_jx_S2fault ;
[0041] After the faulty submodule location is determined, the estimated value of the capacitor voltage U is calculated by reconstructing the switching function. sm_jx_S1fault , U sm_jx_S2fault This is used to locate faults in the upper tube and lower tube of the submodule.
[0042] When the upper tube of the submodule or the lower tube of the submodule fails, the voltage of the submodule capacitor changes. At this time, the switching function S of the upper tube of the submodule and the lower tube of the submodule is redefined. jx _ S1open , S jx _ S2open is the reconstructed switching function, which is obtained by the following formula:
[0043]
[0044] In the formula, S jx Represented as a switching function, i arm is the submodule input current.
[0045] According to the modified switching function, the switching function S of the upper tube of the final submodule and the lower tube of the submodule is obtained. jx_S1fault / S jx_S2fault ,as follows:
[0046] S jx_S1fault =S jx -S jx_S1open (7)
[0047] S jx_S2fault =S jx +S jx_S2open (8)
[0048] Finally, the capacitor voltage estimate U is calculated based on the reconstructed switching function when the upper tube and lower tube of the submodule fail. sm_jx_S1fault , U sm_jx_S2fault It is expressed as:
[0049]
[0050] Where, C is the DC side capacitance of the submodule; U sm_jx_S1fault (k+1) and U sm_jx_S1fault (k) is the time after the faulty submodule x is detected, based on the submodule upper tube S at time k+1 and k respectively. 1 The estimated value of the capacitor voltage calculated by the reconstructed switching function during the fault; U sm_jx_S2fault (k+1) and U sm_jx_S2fault (k) is the time after the faulty submodule x is detected, based on the submodule downpipe S at time k+1 and k respectively. 2 The estimated value of the capacitor voltage calculated by the reconstructed switching function at the time of the fault; i arm (k) is the bridge arm current value at time k, S jx (k), S jx _ S1open (k) and S jx _ S2open (k) are the submodule switch functions before and after modification at time k.
[0051] 5) Real-time comparison of the difference ΔU between the estimated value of the capacitor voltage of the upper tube and the lower tube of the submodule and the measured capacitor voltage sm_jx_S1 and ΔU sm_jx_S2 , which is expressed as follows:
[0052] ΔU sm_jx_S1 =|U sm_jx_S1fault -U sm_jx | (11)
[0053] ΔU sm_jx_S2 =|U sm_jx_S2fault -U sm_jx | (12)
[0054] Where U sm_jx_S1fault , U sm_jx_S2fault is the estimated value of the capacitor voltage when the upper tube of the submodule or the lower tube of the submodule fails, U sm_jx is the initial value of the capacitor voltage of the x-th submodule of phase j.
[0055] The real-time comparison difference is compared with the set threshold ΔU th Compare. If ΔUsm_jx_S1 <ΔU th , then start the timer T 1 , if T 1 Greater than the set time threshold T th , it can be determined that the upper tube of the xth submodule is faulty. sm_jx_S2 Less than the set capacitor voltage threshold ΔU th , then start the timer T 1 , if T 1 >T th , it can be determined that the lower tube of the x-th submodule is faulty.
Claims
1. A modular multilevel converter open circuit fault diagnosis method based on DTW algorithm, It is characterized in that The steps include: 1) In the current control cycle, a sliding window with a length of L is designed to obtain capacitor voltage samples of the modular multilevel converter; 2) Take the capacitor voltages of two submodules in the same bridge arm, ensure that their sequence lengths are equal and both are L, and calculate the distance D of the capacitor voltages of the two submodules based on the dynamic time normalization algorithm j,xy ; 3) Determine the location of the faulty submodule When the capacitor voltage in the bridge arm is in a completely linear relationship, the distance D based on the dynamic time normalization algorithm is 0. When the distance is in a completely inverse linear relationship, the value of D is not 0. The closer to 0, the stronger the correlation between the capacitor voltages. When the value of D is not 0, the location of the submodule is the fault location. 4) Calculate the upper tube of the submodule and the lower tube of the submodule S 2 Switching function S during fault jx_S1fault , S jx_S2fault , and thus calculate the estimated capacitor voltage value U based on the reconstruction switch function when the upper tube of the submodule and the lower tube of the submodule fail sm_jx_S1fault and U sm_jx_S2fault ; 5) Real-time comparison of the difference ΔU between the estimated value of the capacitor voltage of the upper tube and the lower tube of the submodule and the measured capacitor voltage sm_jx_S1 and ΔU sm_jx_S2 , if ΔU sm_jx_S1 <ΔU th , then start the timer T 1 , if T 1 Greater than the set time threshold T th , it can be determined that the upper tube of the xth submodule is faulty. sm_jx_S2 Less than the set capacitor voltage threshold ΔU th , then start the timer T 1 , if T 1 >T th , it can be determined that the lower tube of the x-th submodule is faulty.
2. The method for diagnosing open-circuit faults of modular multilevel converters based on DTW algorithm according to claim 1, It is characterized in that The sliding window of length L described in step 1) is obtained by the following formula: Where f is the output frequency, T s is the sampling period.
3. The method for diagnosing open-circuit faults of modular multilevel converters based on DTW algorithm according to claim 1, It is characterized in that The distance D between the capacitor voltages of the two submodules described in step 2) based on the dynamic time normalization algorithm j,xy for: D j,xy =d(U sm_jx,L ,U sm_jy,L )+min{c(L-1,L),c(L,L-1),c(L-1,L-1)} Where U sm_jx,w represents the value of the capacitor voltage of the xth submodule of phase j at the wth sampling point, U sm_jy,v represents the value of the capacitor voltage of the y submodules in the j phase at the vth sampling point; d(U sm_jx,w , U sm_jy,v ) indicates U sm_jx The wth sampling point in U sm_jy c(w,v) represents the distance metric of the vth sampling point in ; c(w,v) represents the cumulative distance, which is the sum of the distance metric of the current search position and the minimum value of the three cumulative distances c(w-1,v), c(w,v-1) and c(w-1,v-1); w, v are the numbers of the sampling points, where w,v=1,2,…,L; c(LI,L) represents the cumulative distance between the capacitor voltage of the first submodule at the LIth sampling point and the capacitor voltage of the second submodule at the Lth sampling point, c(L,LI) represents the cumulative distance between the capacitor voltage of the first submodule at the Lth sampling point and the capacitor voltage of the second submodule at the LIth sampling point, and c(LI,LI) represents the cumulative distance between the capacitor voltage of the first submodule at the LIth sampling point and the capacitor voltage of the second submodule at the LIth sampling point.
4. The method for diagnosing open-circuit faults of modular multilevel converters based on DTW algorithm according to claim 1, It is characterized in that Step 4) includes: When the upper tube of the submodule or the lower tube of the submodule fails, the voltage of the submodule capacitor changes. At this time, the switching function S of the upper tube of the submodule and the lower tube of the submodule is redefined. jx_S1open , S jx_S2open is the reconstructed switching function, which is obtained by the following formula: In the formula, S jx Represented as a switching function, i arm is the submodule input current; According to the modified switching function, the switching function S of the upper tube of the final submodule and the lower tube of the submodule is obtained. jx_S1fault / S jx_S2fault ,as follows: S jx_S1fault =S jx -S jx_S1open S jx_S2fault =S jx +S jx_S2open Finally, the capacitor voltage estimate U is calculated based on the reconstructed switching function when the upper tube and lower tube of the submodule fail. sm_jx_S1fault / U sm_jx_S2fault It is expressed as: Where, C is the DC side capacitance of the submodule; U sm_jx_S1fault (k+1) and U sm_jx_S1fault (k) is the estimated value of the capacitor voltage calculated based on the reconstructed switching function when the submodule upper tube fails at time k+1 and k respectively after the faulty submodule x is detected; U sm_jx_S2fault (k+1) and U sm_jx_S2fault (k) is the estimated value of the capacitor voltage calculated based on the reconstructed switching function when the submodule under the faulty tube is faulty at time k+1 and k respectively after the faulty submodule x is detected; i arm (k) is the bridge arm current value at time k, S jx (k), S jx _ S1open (k) and S jx _ S2open (k) are the submodule switch functions before and after modification at time k.
5. The method for diagnosing open-circuit faults of modular multilevel converters based on DTW algorithm according to claim 1, It is characterized in that The difference ΔU between the estimated value of the capacitor voltage of the upper tube of the submodule and the lower tube of the submodule and the measured capacitor voltage sm_jx_S1 and ΔU sm_jx_S2 , which is expressed as follows: ΔU sm_jx_S1 =|U sm_jx_S1fault -U sm_jx | ΔU sm_jx_S2 =|U sm_jx_S2fault -U sm_jx | Where U sm_jx_S1fault , U sm_jx_S2fault is the estimated value of the capacitor voltage when the upper tube of the submodule or the lower tube of the submodule fails, U sm_jx is the initial value of the capacitor voltage of the x-th submodule of phase j.
