Train converter open-circuit fault positioning method and system
By building an instant learning prediction model based on variable nearest neighbor sample set and a dynamic threshold fault detection function, the problem of difficult to diagnose open circuit faults of train converter is solved, rapid detection and precise positioning are achieved, and the reliability and safety of the system are improved. .
Patent Information
- Application Number
- CN202411109495.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-13
AI Technical Summary
The prior art is difficult to effectively and promptly diagnose open circuit failures of train converters, resulting in system control performance damage and potential safety risks.
By constructing a real-time learning prediction model based on variable nearest neighbor sample set, using dynamic threshold fault detection function based on Z score and open circuit fault sensitivity factor based on the maximum current, rapid detection and step-by-step precise positioning of open circuit faults of train converter are achieved by using dynamic threshold faults based on Z-score. .
It realizes rapid detection and precise positioning of open circuit faults of train converter, reduces calculation complexity, improves system reliability and safety, and is suitable for online immediacy requirements of dynamic industries.
Smart Images

Figure CN119001536A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of circuit fault diagnosis, and in particular relates to a method and system for locating an open-circuit fault of a train converter. Background Art
[0002] Train inverters have multiple responsibilities for energy conversion, power output control, and maintaining system stability. As electronic equipment development tends to be more integrated and lightweight, its application environment has become increasingly complex and harsh. In such an environment, the inverter has gradually become one of the most vulnerable parts among many components. Once the inverter fails, it will not only damage the control performance of the system, but also, in extreme cases, threaten the safety of the entire system or equipment. The faults of train inverters are mainly divided into two types: open circuit and short circuit. Once a short circuit fault occurs, the damage speed is extremely fast and the destructive power is huge, and it is usually necessary to rely on integrated hardware to prevent it. Open circuit faults are often difficult to detect in the early stages. If the system runs for a long time with an open circuit fault, it may cause chain damage to other components. Therefore, how to diagnose the open circuit fault of the train inverter in a timely manner is crucial to ensure the stable operation of the inverter and prevent potential systemic risks. Summary of the invention
[0003] The technical problem to be solved by the present invention is to overcome the deficiencies and defects mentioned in the above background technology and provide a method and system for locating open-circuit faults of train converters.
[0004] In order to solve the above technical problems, the technical solution proposed by the present invention is:
[0005] In a first aspect, the present application provides a method for locating an open circuit fault of a train converter, comprising:
[0006] S1: Based on the current sampling samples and historical sample sets of the three-phase output current of the train converter and the voltage of the intermediate DC link capacitor, the distance between the samples is calculated to determine the total number of elements of the variable neighbor sample set;
[0007] S2: constructing an instant learning prediction model of the three-phase output current of the train converter according to the total number of elements, and obtaining an output current residual value;
[0008] S3: constructing a dynamic threshold fault detection function based on a Z score according to the output current residual value;
[0009] S4: Determine whether a fault occurs according to the output current residual value and the dynamic threshold;
[0010] S5: If a detection fault occurs, the output current and capacitor voltage after the fault are collected to establish an open circuit fault sensitivity factor based on the maximum current value;
[0011] S6: constructing a preliminary fault location judgment function based on the sensitivity factor according to the open circuit fault sensitivity factor, and determining a preliminary location result of the train converter open circuit fault;
[0012] S7: Constructing an accurate fault location judgment function according to the preliminary location result to obtain an accurate location result of the train converter open circuit fault.
[0013] In a second aspect, the present invention further provides a train converter open circuit fault location system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the computer program.
[0014] Compared with the prior art, the present invention has the following beneficial effects:
[0015] The train converter open-circuit fault location method provided in the present application predicts the output phase current by constructing an instant learning prediction model based on a variable nearest neighbor sample set, and adopts a dynamic threshold fault detection function based on the Z score and an open-circuit fault sensitivity factor based on the current maximum value to achieve rapid fault detection and step-by-step precise location. This method is intuitive and easy to model, and does not require offline training. It shows good results and strong robustness for the online immediacy requirements of dynamic industrial detection. The design of step-by-step fault diagnosis with preliminary positioning first and then precise positioning also effectively reduces the complexity of calculation, while improving the reliability and safety of the system. Through this method, the fault detection and diagnosis process of the train converter becomes more efficient and accurate. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0017] Figure 1 It is one of the flow charts of a method for locating an open circuit fault of a train converter according to a preferred embodiment of the present invention;
[0018] Figure 2 is a main circuit topology diagram of a three-level inverter according to a preferred embodiment of the present invention;
[0019] Figure 3 This is a second flow chart of a train converter open circuit fault locating method according to a preferred embodiment of the present invention;
[0020] Figure 4 is S of the preferred embodiment of the present invention a1Open circuit fault detection result diagram;
[0021] Figure 5 is S of the preferred embodiment of the present invention b2 Figure 2. Results of preliminary location of open circuit fault.
[0022] Figure 6 is S of the preferred embodiment of the present invention b2 Result diagram of precise location of open circuit fault. DETAILED DESCRIPTION
[0023] In order to facilitate the understanding of the present invention, the present invention will be described more comprehensively and meticulously below in conjunction with the accompanying drawings and preferred embodiments of the present invention, but the protection scope of the present invention is not limited to the following specific embodiments.
[0024] Unless otherwise defined, all professional terms used below have the same meanings as those generally understood by those skilled in the art. The professional terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the scope of protection of the present invention.
[0025] Unless otherwise specified, various raw materials, reagents, instruments and equipment used in the present invention can be purchased from the market or prepared by existing methods.
[0026] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the usual meanings understood by persons with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate quantity restrictions, but indicate the existence of at least one. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.
[0027] See also Figure 1 The present application provides a method for locating an open-circuit fault of a train converter, comprising:
[0028] S1: Based on the current sampling samples and historical sample sets of the three-phase output current of the train converter and the intermediate DC link capacitor voltage, the distance between samples is calculated to determine the total number of elements of the variable neighbor sample set.
[0029] In this step, the current sample refers to the sampled data of the three-phase output current of the train converter and the intermediate DC link capacitor voltage at the current moment. The historical sample set includes a normal state historical sample subset and a fault state historical sample subset. The normal state historical sample subset refers to a sampled data set of not less than one complete current cycle in the normal state; the fault state historical sample subset refers to a sampled data set of not less than one complete current cycle in the single-tube open circuit fault state in the train converter. Among them, the sampled data includes the three-phase output current of the train converter and the intermediate DC link capacitor voltage.
[0030] S2: Based on the total number of elements, construct an instant learning prediction model for the three-phase output current of the train converter to obtain the output current residual value.
[0031] S3: constructing a dynamic threshold fault detection function based on a Z score according to the output current residual value.
[0032] S4: Determine whether a fault occurs according to the output current residual value and the dynamic threshold.
[0033] S5: If a detection fault occurs, the output current and capacitor voltage after the fault are collected, and an open circuit fault sensitivity factor based on the maximum current value is established.
[0034] S6: constructing a preliminary fault location judgment function based on the open circuit fault sensitivity factor according to the open circuit fault sensitivity factor, and determining a preliminary location result of the train converter open circuit fault.
[0035] S7: Constructing an accurate fault location judgment function according to the preliminary location result to obtain an accurate location result of the train converter open circuit fault.
[0036] The train converter open-circuit fault location method predicts the output phase current by constructing an instant learning prediction model based on a variable neighbor sample set. At the same time, a dynamic threshold fault detection function based on the Z score and an open-circuit fault sensitivity factor based on the current maximum value are used to achieve rapid fault detection and step-by-step precise location. This method is intuitive and easy to model, and does not require offline training. It shows good results and strong robustness for the online instant requirements of dynamic industrial detection. The design of step-by-step fault diagnosis with preliminary positioning first and then precise positioning also effectively reduces the complexity of calculation, while improving the reliability and safety of the system. With this method, the fault detection and diagnosis process of the train converter becomes more efficient and accurate.
[0037] This embodiment is described by taking a three-level inverter in a train converter as an example. The main circuit schematic diagram of the three-level inverter is shown in FIG. Figure 2 As shown, the DC side voltages U1 and U2 are 1800V, the DC side capacitors C1 and C2 are 16mF, and the motor parameters used are shown in Table 1.
[0038] Table 1 Motor parameters
[0039]
[0040] Optionally, the S1 includes:
[0041] S11: Calculate the distance between the sample at the qth sampling moment and the samples in the normal state historical sample subset. The formula is as follows:
[0042]
[0043] Where, d l [q] is the distance between the sample at the qth sampling moment and the lth sample in the normal state historical sample subset, l∈{1,2,...,L}, L is the total number of samples in the normal state historical sample subset; f d (·) is the calculation function of the distance between samples, i1[q] and i2[q] are the output currents of any two phases in the sample at the qth sampling moment, and is the output current of any two phases in the lth sample of the normal state historical sample subset, i1[q] and is the same phase output current sample, i2[q] and is the same-phase output current sample; u n [q] is the voltage of the nth capacitor in the sample at the qth sampling moment, It is the nth capacitor voltage in the lth sample of the normal state historical sample subset, where n=1, 2, ..., N, and N is the total number of intermediate DC link capacitors.
[0044] S12: Compare the sample sampled at the qth sampling moment with each sample in the normal state historical sample subset.
[0045] The distance between them constitutes a set, and the formula is:
[0046] D[q]={d1[q],...,d l [q],...,d L [q]};
[0047] Where D[q] is the set of distances between the sample at the qth sampling moment and the L samples in the normal state historical sample subset;
[0048] Arrange the elements in the set D[q] in ascending order from small to large to form an ascending distance set D′[q] that satisfies the following relationship:
[0049] D′[q]=sort(D[q])={d′1[q],...,d′ l[q],...,d′ L [q]};
[0050] In the formula, sort(·) is a function that arranges the elements of a set in ascending order, and d′ l [q] represents the value of the lth element after D[q] is arranged in ascending order. The elements in D′[q] satisfy the inequality: d′1[q]≤...≤d′ l [q]≤d′ l+1 [q]≤...≤d′ L [q];
[0051] For any two consecutive elements of the set D′[q], the difference between them is calculated to form a variable neighbor sample set E[q] that satisfies the following relationship:
[0052] E[q]={e1,...,e l ,…,e L-1},e l =d′ l+1 [q]-d′ l [q];
[0053] Where el represents the adjacent element d′ in the ascending distance set D′[q] l+1 [q], d′ l The difference of [q].
[0054] Specifically, in this embodiment, i1[q] and i2[q] are respectively a [q]、i b [q], N = 2, the intermediate DC link capacitor voltage includes u1[q], u2[q], the similarity measurement method between samples is Euclidean distance, and the specific calculation is
[0055] The formula is as follows:
[0056]
[0057] S13: Calculate the total number of elements in the variable neighbor sample set. The formula is:
[0058] k[q] = argmax(E[q]);
[0059] Where k[q] is the total number of elements in the variable neighbor sample set of the sample sampled at the qth sampling time, and argmax(·) means finding the element with the largest value in the set and returning the index l∈{1,2,...,L-1} of the element.
[0060] Optionally, the S2 includes:
[0061] S21: Establish an instant learning prediction model for the output phase current, expressed as:
[0062]
[0063] In the formula, is the predicted value of the output phase current i3[q] at the qth sampling moment, i3[q] is The true value of is the output phase current sampling value at the qth sampling moment, i3[q] is the output phase current of the three-phase current except i1[q] and i2[q], f r (·) is the random forest local prediction function, whose input includes the output phase current i1[q], i2[q], the intermediate DC link capacitor voltage u1[q], u2[q], ..., u n [q], …, u N [q], the total number of elements in the variable neighbor sample set of the sample at the qth sampling moment k[q], the normal state historical sample subset X normal .
[0064] S22: Calculate the predicted residual of the output phase current, the formula is:
[0065]
[0066] Where r3[q] is the predicted residual of the output phase current i3[q] at the qth sampling moment.
[0067] Specifically, in this embodiment, an instant learning prediction model is constructed for the output phase current of phase C. That is r3[q] is r c [q].
[0068] Optionally, the S3 includes:
[0069] S31: Construct the upper limit of the dynamic detection threshold based on the Z score, the formula is:
[0070] r 3_h [q]=μ3[q]+(z0*λ[q])σ3[q];
[0071] In the formula, r 3_h [q] is the upper limit of the dynamic detection threshold of r3[q], z0 is the initial threshold, λ[q] is the dynamic threshold adjustment factor at the qth sampling moment, μ3[q] and σ3[q] are the mean and standard deviation of r3[q] respectively. The calculation formula of μ3[q] is:
[0072] If t<0, r3[t]=0;
[0073] In the formula, t represents the sampling time, T1 is the dynamic detection threshold sliding window size, and its value is a positive integer. The calculation formula of σ3[q] is:
[0074]
[0075] S32: Construct the lower limit of the dynamic detection threshold based on the Z score, the formula is as follows:
[0076] r 3_lo [q]=μ3[q]-(z0*λ[q])σ3[q];
[0077] Where r 3_lo [q] is the lower limit of the dynamic detection threshold of r3[q].
[0078] S33: Construct a dynamic threshold adjustment factor, the formula is as follows:
[0079]
[0080] Where λ[q] is the dynamic threshold adjustment factor at the qth sampling moment, α is a constant greater than 0 and less than 1, w[q] is the cumulative number of samples of the continuous super-dynamic detection threshold of r3[t] counted at the qth sampling moment, and w th is the number of times r3[t] continuously exceeds the dynamic detection threshold, w th >1. The calculation formula of w[q] is:
[0081]
[0082] In the formula Indicates the number of times r3[t] exceeds the dynamic detection threshold consecutively minus 1.
[0083] S34: Construct a dynamic threshold fault detection function, the formula is as follows:
[0084]
[0085] Where R[q] is the fault detection function value at the qth sampling moment. When R[q] = 1, the system is judged to be faulty. When R[q] = 0, the system is judged to be in a normal state.
[0086] Optionally, the S5 includes:
[0087] S51: Collect the output phase current at the pth sampling time after the fault, and calculate the maximum value of the output phase current. The formula is:
[0088]
[0089] In the formula, is the maximum output phase current of the x(x∈{1,2,3}) phase bridge arm calculated in the sliding window T2 at the pth sampling moment (p∈[q,q+O]), O is the total number of output phase current samples collected after the fault, and its value is the integer value of the ratio of the output phase current cycle to the sampling interval, T2 is the sliding window size for calculating the maximum output phase current, and its value is not less than one output phase current cycle, i x [j] is the output phase current of the x-phase bridge arm at the j-th sampling moment.
[0090] Specifically, in this embodiment, the value of T2 is equal to one output phase current cycle:
[0091] S52: Calculate the minimum output phase current. The formula is:
[0092]
[0093] In the formula, It is the minimum output phase current of the x(x∈{1,2,3}) phase bridge arm calculated within the sliding window T2 at the p(p∈[q,q+O])th sampling moment.
[0094] S53: Construct an open circuit fault sensitivity factor based on the maximum output phase current. The formula is:
[0095]
[0096] In the formula, is the open-circuit fault sensitivity factor of the maximum output phase current of the x-phase bridge arm of the converter at the p-th (p∈[q,q+O]) sampling moment, and ξ is a non-zero infinitesimal positive number.
[0097] Optionally, the S6 includes:
[0098] S61: Construct a preliminary fault location judgment function, the formula is:
[0099]
[0100] In the formula, is the preliminary fault location judgment function at the pth sampling moment, β th_x is the threshold value of the maximum fault sensitivity factor of the output phase current of the x-phase bridge arm. It indicates that a fault has occurred in the inner power switch tube of the converter x-phase bridge arm; when This indicates that a fault has occurred in the outer power switch tube of the three-phase bridge arm of the converter.
[0101] S62: Determine the preliminary fault location result, the formula is:
[0102]
[0103] Where m[p] is the preliminary fault location result at the pth sampling moment, and its value is the number of the power switch tube in the converter, m[p]∈{1,2,…,M}, and M is the total number of power switch devices.
[0104] Specifically, in this embodiment, m[p]∈{1,2,…,12}. A subset of fault status history samples can be filtered, which can be expressed as follows:
[0105]
[0106] When an open-circuit fault occurs in the train's three-level inverter, the preliminary positioning results can determine whether the open-circuit fault is located on the inside or outside of the bridge arm. If it is on the inside, it can be directly located to the faulty phase, and m[p] has 2 different values; if it is on the outside, m[p] has 6 different values.
[0107] Optionally, S7 includes:
[0108] S71: Combine the preliminary fault location results and the subset of fault status historical samples to calculate the output phase current prediction residual after the fault. The formula is:
[0109]
[0110] In the formula, is the output phase current prediction residual after the fault at the pth sampling moment, is the predicted value of the output phase current after the fault at the p-th sampling moment, and i3[p] is the output phase current value after the fault of any phase in the sampling samples at the p-th sampling moment. The calculation formula is:
[0111]
[0112] Where i1[p] and i2[p] are the output phase current values after the fault of any two phases (except the phase where i3[p] is located) in the sampled samples at the pth sampling time, k[p] is the total number of elements in the variable neighbor sample set of the fault sampled samples at the pth sampling time, and u n [p] is the voltage of the nth capacitor of the fault sample at the pth sampling moment, X fault_m[p] is the fault status historical sample subset X fault The fault status sampling data of the power switch tube numbered m[p]. r f (·) is the random forest local prediction function under fault, whose input includes the output phase current i1[p], i2[p] under fault at the pth sampling moment, the intermediate DC link capacitor voltage u1[p], u2[p], ..., u n [p], ..., u N[p], the total number of elements in the variable neighbor sample set of the fault sample at the pth sampling moment k[p], the fault state historical sample subset X fault_m[p] .
[0113] Specifically, in this embodiment, the output phase current prediction residual after the fault is calculated for the C phase output phase current. That is
[0114] S72: Construct an accurate fault location judgment function based on the fault output phase current prediction residual, the formula is as follows:
[0115]
[0116] Where H[p] is the value of the post-fault precise positioning judgment function corresponding to the sample at the pth sampling time. is the p-th sampling time based on The fault detection function value of The calculation formula is:
[0117] In the formula, w m[p] [p] is the statistics at the pth sampling moment The cumulative number of samples of continuous ultra-dynamic detection threshold, w th_m[p] yes The number of consecutive super-dynamic detection thresholds, w th_m[p] >1. m[p] The calculation formula for [p] is:
[0118]
[0119] In the formula, express The number of times the dynamic detection threshold is exceeded consecutively minus 1. is the pth sampling time The upper limit of the dynamic detection threshold, is the pth sampling time Lower threshold of dynamic detection. The calculation formula is:
[0120]
[0121] In the formula, z 0_m[p] is the initial threshold, λ m[p] [p] is the dynamic threshold adjustment factor at the pth sampling moment, μ 3_m[p] [p] and σ 3_m[p] [p] are The mean and standard deviation of . The calculation formula is:
[0122]
[0123] In the formula, μ 3_m[p] The calculation formula for [p] is:
[0124] If t<0,
[0125] Where T3 is the dynamic detection threshold sliding window size, and its value is a positive integer. 3_m[p] [p]The calculation formula is:
[0126]
[0127] λ m[p] The calculation formula for [t] is:
[0128]
[0129] In the formula, α m[p] is a constant greater than 0 and less than 1.
[0130] The precise fault location judgment function value is judged and its stable value is calculated. The formula is:
[0131]
[0132] In the formula, H f is the stable value of the precise positioning judgment function, H[O] is the value of the precise positioning judgment function after the fault corresponding to the sample at the Oth sampling time, and is the final value of the precise positioning judgment function.
[0133] S73: Determine the accurate fault location result, the formula is:
[0134] m o =f o (H f );
[0135] In the formula, m o For accurate fault location results, its value is the number of the power switch tube in the converter, m o ∈{1,2,…,M},f o (·) is H f Mapped as a function of the power switch number in the converter.
[0136] It should be noted that, in this embodiment, H f The mapping function table of the power switch tube number in the converter can be constructed as shown in Table 2. The fault diagnosis flow chart is shown in Figure 3 shown.
[0137] Table 2 Relationship between accurate fault location judgment function value and fault location
[0138]
[0139] In this embodiment, S a1 Open circuit fault detection process and S b2 The open circuit fault diagnosis process is used as an example to illustrate the effectiveness of the method. First, the open circuit fault detection method is verified. a1 The open circuit fault detection result is as follows Figure 4 As shown. Figure 4 It can be seen that after the fault occurs at the 200th sample, the prediction residual of the output phase current increases rapidly after the 210th sample. The fault detection function value changes from 0 to 1 at the 214th sample, indicating that an open circuit fault is detected in the system.
[0140] Next, the open circuit fault diagnosis method is verified. b2 The preliminary location results of the open circuit fault are as follows: Figure 5 As shown. Figure 5 It can be seen that m[p] rises to values 6 and 7 at the 210th sample, respectively. This phenomenon indicates that an open circuit fault occurs in the power switch tube inside phase b. The result of the accurate fault location judgment function is as follows: Figure 6 As shown in the figure, the precise positioning judgment function value H[p] is stable at the value 64 at the 214th sample. By looking up Table 2, it can be seen that S b2 An open circuit fault has occurred.
[0141] In summary, this embodiment can detect and diagnose inverter open circuit faults, thereby improving the reliability and safety of the system.
[0142] The present application also provides a train converter open circuit fault location system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program. The train converter open circuit fault location system can implement various embodiments of the above train converter open circuit fault location method, and can achieve the same beneficial effects, which will not be described in detail here.
[0143] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.
Claims
1. A method for locating open circuit fault of train converter, characterized in that: include: S1: Based on the current sampling samples and historical sample sets of the three-phase output current of the train converter and the voltage of the intermediate DC link capacitor, the distance between the samples is calculated to determine the total number of elements of the variable neighbor sample set; S2: constructing an instant learning prediction model of the three-phase output current of the train converter according to the total number of elements, and obtaining an output current residual value; S3: constructing a dynamic threshold fault detection function based on a Z score according to the output current residual value; S4: Determine whether a fault occurs according to the output current residual value and the dynamic threshold; S5: If a detection fault occurs, the output current and capacitor voltage after the fault are collected to establish an open circuit fault sensitivity factor based on the maximum current value; S6: constructing a preliminary fault location judgment function based on the sensitivity factor according to the open circuit fault sensitivity factor, and determining a preliminary location result of the train converter open circuit fault; S7: Constructing an accurate fault location judgment function according to the preliminary location result to obtain an accurate location result of the train converter open circuit fault.
2. The train converter open circuit fault location method according to claim 1, characterized in that: The S1 includes: S11: Calculate the distance between the sample sampled at the qth sampling moment and the samples in the normal state historical sample subset, satisfying the following relationship: Where, d l [q] is the distance between the sample at the qth sampling moment and the lth sample in the normal state historical sample subset, l∈{1,2,...,L}, L is the total number of samples in the normal state historical sample subset; f d (·) is the calculation function of the distance between samples, i1[q] and i2[q] are the output currents of any two phases in the sample at the qth sampling moment, and is the output current of any two phases in the lth sample of the normal state historical sample subset, i1[q] and is the same phase output current sample, i2[q] and is the same-phase output current sample; u n [q] is the voltage of the nth capacitor in the sample at the qth sampling moment, is the nth capacitor voltage in the lth sample of the normal state historical sample subset, where n = 1, 2, ..., N, and N is the total number of intermediate DC link capacitors; S12: The distances between the sample sampled at the qth sampling moment and each sample in the normal state historical sample subset form a set that satisfies the following relationship: D[q]={d1[q],...,d l [q],...,d L [q]}; Where D[q] is the set of distances between the sample at the qth sampling moment and the L samples in the normal state historical sample subset; Arrange the elements in the set D[q] in ascending order from small to large to form an ascending distance set D′[q] that satisfies the following relationship: D′[q]=sort(D[q])={d′1[q],...,d′ l [q],...,d′ L [q]}; In the formula, sort(·) is a function that arranges the elements of a set in ascending order, and d′ l [q] represents the value of the lth element after D[q] is arranged in ascending order. The elements in D′[q] satisfy the inequality: d′1[q]≤...≤d′ l [q]≤d′ l+1 [q]≤...≤d′ L [q]; For any two consecutive elements of the set D′[q], the difference between them is calculated to form a variable neighbor sample set E[q] that satisfies the following relationship: E[q]={e1,…,e l ,…,e L-1 },e l =d′ l+1 [q]-d′ l [q]; In the formula, e l Represents the adjacent element d′ in the ascending distance set D′[q] l+1 [q], d′ l The difference of [q]; S13: Calculate the total number of elements in the variable neighbor sample set, satisfying the following relationship: k[q] = argmax(E[q]); Where k[q] is the total number of elements in the variable neighbor sample set of the sample sampled at the qth sampling time, and argmax(·) means finding the element with the largest value in the set and returning the index l∈{1,2,...,L-1} of the element.
3. The train converter open circuit fault location method according to claim 1, characterized in that: The S2 includes: S21: Establish an instant learning prediction model for the output phase current, satisfying the following relationship: In the formula, is the predicted value of the output phase current i3[q] at the qth sampling moment, i3[q] is The true value of is the output phase current sampling value at the qth sampling moment, i3[q] is the output phase current of the three-phase current except i1[q] and i2[q], f r (·) is the random forest local prediction function, whose input includes the output phase current i1[q], i2[q], the intermediate DC link capacitor voltage u1[q], u2[q], ..., u n [q], …, u N [q], k[q] represents the total number of elements in the variable neighbor sample set of the sample at the qth sampling moment, X normal Represents a subset of historical samples in normal state; S22: Calculate the predicted residual of the output phase current, satisfying the following relationship: Where r3[q] is the predicted residual of the output phase current i3[q] at the qth sampling moment.
4. The train converter open circuit fault location method according to claim 1, characterized in that: The S3 includes: S31: Construct an upper limit of the dynamic detection threshold based on the Z score, satisfying the following relationship: r 3_h [q]=μ3[q]+(z0*λ[q])σ3[q]; In the formula, r 3_h [q] is the upper limit of the dynamic detection threshold of r3[q], z0 is the initial threshold, λ[q] is the dynamic threshold adjustment factor at the qth sampling moment, μ3[q] and σ3[q] are the mean and standard deviation of r3[q] respectively, where μ3[q] satisfies the following relationship: Where t represents the sampling time, T1 is the dynamic detection threshold sliding window size, which is a positive integer, and σ3[q] satisfies the following relationship: S32: Construct a lower limit of the dynamic detection threshold based on the Z score, satisfying the following relationship: r 3_lo [q]=μ3[q]-(z0*λ[q])σ3[q]; Where r 3_lo [q] is the lower limit of the dynamic detection threshold of r3[q]; S33: Construct a dynamic threshold adjustment factor to satisfy the following relationship: Where λ[q] is the dynamic threshold adjustment factor at the qth sampling moment, α is a constant greater than 0 and less than 1, w[q] is the cumulative number of samples of the continuous super-dynamic detection threshold of r3[t] counted at the qth sampling moment, and w th is the number of times r3[t] continuously exceeds the dynamic detection threshold, w th >1, where w[q] is calculated as follows: In the formula Indicates the number of times r3[t] exceeds the dynamic detection threshold continuously minus 1; S34: Construct a dynamic threshold fault detection function, the formula is as follows: Where R[q] is the fault detection function value at the qth sampling moment.
5. The train converter open circuit fault location method according to claim 1, characterized in that: The S4 includes: When the fault detection function value at the qth sampling moment is equal to 1, it is determined that the system has a fault. When the fault detection function value at the qth sampling moment is equal to 0, it is determined that the system is in a normal state.
6. The train converter open circuit fault location method according to claim 1, characterized in that: The S5 includes: S51: Collect the output phase current at the pth sampling time after the fault, calculate the maximum value of the output phase current, and satisfy the following relationship: In the formula, is the maximum output phase current of the x(x∈{1,2,3}) phase bridge arm calculated in the sliding window T2 at the pth sampling moment (p∈[q,q+O]), O is the total number of output phase current samples collected after the fault, and its value is the integer value of the ratio of the output phase current cycle to the sampling interval, T2 is the sliding window size for calculating the maximum output phase current, and its value is not less than one output phase current cycle, i x [j] is the output phase current of the x-phase bridge arm at the j-th sampling moment; S52: Calculate the minimum output phase current to satisfy the following relationship: In the formula, is the minimum output phase current of the x(x∈{1,2,3}) phase bridge arm calculated within the sliding window T2 at the p(p∈[q,q+O])th sampling moment; S53: Construct an open circuit fault sensitivity factor based on the maximum output phase current, satisfying the following relationship: In the formula, is the open-circuit fault sensitivity factor of the maximum output phase current of the x-phase bridge arm of the converter at the p-th (p∈[q,q+O]) sampling moment, and ξ is a non-zero infinitesimal positive number.
7. The train converter open circuit fault location method according to claim 1, characterized in that: The S6 includes: S61: Construct a preliminary fault location judgment function that satisfies the following relationship: In the formula, is the preliminary fault location judgment function at the pth sampling moment, β th_x is the threshold value of the maximum fault sensitivity factor of the output phase current of the x-phase bridge arm. It indicates that a fault has occurred in the inner power switch tube of the converter x-phase bridge arm; when This indicates that a fault has occurred in the outer power switch tube of the three-phase bridge arm of the converter; S62: Determine the preliminary fault location result, which satisfies the following relationship: Where m[p] is the preliminary fault location result at the pth sampling moment, and its value is the number of the power switch tube in the converter, m[p]∈{1,2,…,M}, and M is the total number of power switch devices.
8. The train converter open circuit fault location method according to claim 1, characterized in that: The S7 includes: S71: Combine the preliminary fault location result and the fault status historical sample subset to calculate the output phase current prediction residual after the fault, satisfying the following relationship: In the formula, is the output phase current prediction residual after the fault at the pth sampling moment, is the predicted value of the output phase current after the fault at the p-th sampling moment, i3[p] is the output phase current value after the fault of any phase in the sampling sample at the p-th sampling moment; The calculation formula is as follows: Where i1[p] and i2[p] are the output phase current values of any two phases except the phase where i3[p] is located in the sampled samples at the pth sampling time after the fault, k[p] is the total number of elements in the variable neighbor sample set of the fault sampled samples at the pth sampling time, and u n [p] is the voltage of the nth capacitor of the fault sample at the pth sampling moment, X fault_m[p] is the fault status historical sample subset X fault The fault state sampling data of the power switch tube numbered m[p]; is the local prediction function of random forest under fault, whose input includes the output phase current i1[p], i2[p] under fault at the pth sampling moment, the intermediate DC link capacitor voltage u1[p], u2[p], ..., u n [p], ..., u N [p], the total number of elements in the variable neighbor sample set of the fault sample at the pth sampling moment k[p], the fault state historical sample subset X fault_m[p] ; S72: Construct an accurate fault location judgment function based on the fault output phase current prediction residual, satisfying the following relationship: Where H[p] is the value of the post-fault precise positioning judgment function corresponding to the sample at the pth sampling time. is the p-th sampling time based on The fault detection function value of The calculation formula is as follows: In the formula, w m[p] [p] is p Statistics of sampling time The cumulative number of samples of continuous ultra-dynamic detection threshold, w th_m[p] yes The number of consecutive super-dynamic detection thresholds, w th_m[p] >1,w m[p] The calculation formula for [p] is as follows: In the formula, express The number of times the dynamic detection threshold is exceeded consecutively minus 1. is the pth sampling time The upper limit of the dynamic detection threshold, is the pth sampling time Dynamic detection threshold lower limit, The calculation formula is: In the formula, z 0_m [ p ] is the initial threshold, λ m[p] [p] is p Dynamic threshold adjustment factor at sampling time, μ 3_m[p] [p] and σ 3_m[p] [p] are The mean and standard deviation of The calculation formula is: In the formula, μ 3_m[p] The calculation formula for [p] is: Where T3 is the dynamic detection threshold sliding window size, which is a positive integer, σ 3_m[p] [p]The calculation formula is: λ m[p] The calculation formula for [t] is: In the formula, α m[p] is a constant greater than 0 and less than 1; The precise fault location judgment function value is judged and its stable value is calculated to satisfy the following relationship: In the formula, H f is the stable value of the accurate fault location judgment function, H[O] is the value of the accurate fault location judgment function after the fault corresponding to the sample at the Oth sampling time, and is the final value of the accurate fault location judgment function; S73: Determine an accurate fault location result, satisfying the following relationship: m o =f o (H f ); In the formula, m o For accurate fault location results, its value is the number of the power switch tube in the converter, m o ∈{1,2,…,M},f o (·) is H f Mapped as a function of the power switch number in the converter.
9. A train converter open circuit fault location system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.
Citation Information
Patent Citations
Method and system for diagnosis of open circuit fault in three-level traction converter
CN108680818A
Two-level converter system switch device open-circuit fault diagnosis method and system
CN109375029A
Two-level three-phase voltage source inverter bridge arm open-circuit fault diagnosis method
CN111381188A
Composite open-circuit fault diagnosis method for NPC three-level inverter
CN113075585A
Fault diagnosis method for medium-voltage modular multilevel converter
CN113159119A
