Wind power inverter fault diagnosis method

Abstract

The invention discloses a fault diagnosis method for a wind power inverter, and belongs to the technical field of fault diagnosis. The method comprises the steps that a hybrid logic dynamic model is established, decoupling is conducted through matrix transformation, the model is decomposed into two subsystems, and a finite time observer is established; and defining a fault diagnosis variable and a fault diagnosis adaptive threshold, performing fault diagnosis, and then performing fault positioning. According to the invention, the sensor fault and the switch tube fault are separated through matrix transformation, so that the inverter switch tube fault and the sensor fault can be diagnosed at the same time; the accuracy and robustness of fault diagnosis are improved by adopting a residual error and a self-adaptive threshold value; and defining a switch tube fault phasing characteristic value, a positioning characteristic value and a sensor fault positioning characteristic value through the characteristics of the sliding window accumulated value and the sensor fault estimated value, and carrying out combined diagnosis on the sensor and the switch tube of the wind power grid-connected inverter.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault diagnosis, and particularly relates to a wind power inverter fault diagnosis method. BACKGROUND

[0002] A wind power grid-connected inverter realizes power conversion through pulse width modulation control of power switching tubes (such as IGBT, MOSFET), and a sensor network (including current sensors, position sensors, etc.) provides real-time feedback for closed-loop control. However, under complex working conditions, the system reliability faces two key challenges: the long-term current / voltage stress impact and temperature cycle effect of power devices caused by high-frequency switching action, which easily causes open-circuit, short-circuit or aging degradation faults. Research shows that more than 38% of wind power system downtime accidents are directly related to switching tube failure (IEEE Trans. Power Electron., 2021). The traditional method based on current waveform symmetry analysis has insufficient sensitivity to early intermittent faults, and the scheme using additional voltage sensors will significantly increase the system cost and complexity. The signal drift, sticking and other faults of the current sensor will cause the magnetic field orientation control to be inaccurate, and then cause torque pulsation or even system collapse. Existing researches mostly use hardware redundancy or observer compensation strategies, but it is difficult to distinguish the coupling effect of sensor faults and switching tube faults. Current inverter fault diagnosis techniques mostly only target switching tube faults and can be roughly divided into the following categories:

[0003] 1. The core idea of the model-based fault diagnosis method is to obtain prior information through the establishment of a dynamic system analytical model, and to construct a residual signal by comparing the system measurement signal, and to judge the fault by residual evaluation and decision. The literature“A Fast and Robust Open-Switch Fault Diagnosis Method for Variable-Speed PMSM System”(“Variable-Speed PMSM System Open-Switch Fault Diagnosis Method”)(X. Zhou, J. Sun, P. Cui, Y. Lu, M. Lu and Y. Yu) published in IEEE in 2021 aims at power switching tube open-circuit fault, and proposes a method based on a differential current observer, and designs an adaptive threshold to realize fault diagnosis. The literature“Current residual vector-based open-switch fault diagnosis of inverters in PMSM drive systems”(“Permanent Magnet Synchronous Motor Inverter Open-Circuit Fault Diagnosis Based on Current Residual Vector”)(An Q T, Sun L, Zhao K, et al) published in IEEE in 2015 establishes a hybrid model of the inverter and the permanent magnet synchronous motor, constructs a motor stator current residual signal, and diagnoses the inverter open-circuit fault.

[0004] 2. Data-based fault diagnosis methods do not require a precise mathematical model of the diagnostic object, nor do they require the addition of new hardware circuits such as sensors. Instead, they only require preprocessing of the fault data using wavelet transforms, Fourier transforms, and other methods to extract fault features. Faults are then identified using diagnostic network algorithms. Typical methods include artificial intelligence algorithms such as neural networks, support vector machines, fuzzy logic, clustering, and deep learning. The 2020 IET paper, "Fault diagnosis for open-circuit faults in NPC inverter based on knowledge-driven and data-driven approaches" (Kou L, Liu C, GW Cai, et al), uses fast Fourier transforms to extract features from current data for a three-phase inverter system, and finally uses a novel random vector function chain network for fault diagnosis.

[0005] 3. Signal-based methods typically extract fault symptoms and obtain fault indicators by processing the current and voltage signals in the time and frequency domains. The 2022 IEEE paper "Multiple open-circuit faults diagnosis in six-phase induction motor drives using stator current analysis" (H. Mesai-Ahmed, I. Jlassi, AJM Cardoso, and A. Bentaallah) proposed a method for detecting IGBT open-circuit faults in voltage source inverter systems using average current measurement.

[0006] In summary, the existing technologies for wind power inverter fault diagnosis have the following deficiencies:

[0007] 1. For wind power inverters, only open circuit fault diagnosis of power switch tubes is often considered, while joint diagnosis of switch tube and current sensor faults is rarely considered.

[0008] 2. Existing fault diagnosis methods have problems such as high signal processing complexity, long diagnosis cycle, poor robustness and large computational complexity. Summary of the Invention

[0009] The technical problem to be solved by the present application is the problems existing in the prior art. Specifically, the present application provides a wind power inverter fault diagnosis method, which decouples the original system into two subsystems through matrix transformation, and each of the two subsystems contains sensor faults and switch tube faults. Meanwhile, the subsystem containing only sensor faults is augmented to obtain an augmented subsystem. Then, a finite-time observer is designed for the two subsystems to realize accurate estimation of the system state and sensor faults. Meanwhile, a residual norm design detection threshold and detection variable based on the finite-time observer of the two subsystems are proposed to distinguish between sensor faults and switch tube faults. For switch tube faults, a switch tube fault positioning method based on sliding window cumulative amount is proposed. The switch tube fault and sensor fault are decoupled for joint diagnosis, the complexity and redundancy of the fault diagnosis method are reduced, and the robustness of the fault diagnosis is improved.

[0010] To achieve the above object, the present application provides a wind power inverter fault diagnosis method, the faults including switch tube faults and sensor faults, the topology structure of the wind power inverter including a DC side power supply, a main inverter circuit, three same inductors, three same resistors, a grid-connected side three-phase voltage source, three same sensors and a control module; the main inverter circuit including three-phase bridge arms, each phase bridge arm including an upper bridge arm switch tube and a lower bridge arm switch tube; the fault diagnosis method including the following steps:

[0011] Step 1, sampling and establishing a hybrid logic dynamic model of the wind power inverter;

[0012] Step 2, establishing an inverter system state equation containing switch tube faults and sensor faults, and decoupling, decomposing and augmenting through matrix transformation to obtain a subsystem 1 containing only switch tube open circuit faults and a subsystem 2 containing only sensor faults;

[0013] Step 3, building a finite-time observer for the subsystem 1 and the subsystem 2 respectively, and obtaining a k-phase sensor fault estimation value

[0014] Step 4, obtaining two subsystem residual matrices based on the finite-time observer, then calculating a switch tube fault detection variable η a , a sensor fault detection variable η s , a switch tube fault detection threshold Tha and a sensor fault detection threshold Ths;

[0015] Step 5, if η a <Tha and η s <Ths, it is determined that the inverter is working normally, and the fault diagnosis is ended, otherwise, step 6 is entered;

[0016] Step 6: Based on the switch fault residual and the k-phase current cumulative value obtained through the sliding window, a fixed threshold of the switch is given to diagnose and locate the switch fault; based on the k-phase sensor fault estimation value The sensor fault classification variable is calculated by the three-phase output current, and the sensor fault location threshold is given to perform sensor fault diagnosis and location.

[0017] Preferably, the process of implementing step 1 is as follows:

[0018] Sampling the three-phase output current i of the wind power inverter a ,i b ,i c And the three-phase output voltage u of the wind power inverter a ,u b ,u c ; Sample the voltage U of the DC power supply dc ;

[0019] A hybrid logic dynamic model of wind power inverter is established, and its expression is:

[0020]

[0021]

[0022]

[0023] Among them, L is the inductance value, R is the resistance value, δ a , δ b , δ c is the three-phase hybrid logic dynamic switching function; the k-phase hybrid logic dynamic switching function δ is introduced k , k represents the phase sequence, k = a, b, c.

[0024] Preferably, the implementation process of step 2 is as follows:

[0025] Step 2.1: Establish the inverter system state equation containing switch tube fault and sensor fault, and record it as the inverter fault state equation, which is expressed as:

[0026]

[0027] y=Cx+Ff s

[0028] Among them, y is the output signal of the inverter system, x is the state variable, x=[i a i b i c ] T , is the derivative of the state variable x, u is the grid-connected voltage matrix, δ is a mixed logic dynamic switching function matrix, A is a first coefficient matrix, B1 is a second coefficient matrix, B2 is a third coefficient matrix, C is a fourth coefficient matrix, C = diag[1 1 1], D is a switching tube fault coefficient matrix, F is a sensor fault coefficient matrix, M is a disturbance coefficient matrix, f a is a switching tube fault signal, f s is a sensor fault signal, d is a disturbance signal;

[0029] Step 2.2, a first non-singular transformation matrix T and a second non-singular transformation matrix S are introduced to decouple the inverter fault state equation by matrix transformation, and the following equation is obtained:

[0030]

[0031] Wherein, A1, A2, A3, A4 are respectively the 1st decomposition sub-matrix, the 2nd decomposition sub-matrix, the 3rd decomposition sub-matrix, and the 4th decomposition sub-matrix of the first coefficient matrix A, B 11 , B 12 are respectively the 1st decomposition sub-matrix and the 2nd decomposition sub-matrix of the second coefficient matrix B1, B 21 , B 22 are respectively the 1st decomposition sub-matrix and the 2nd decomposition sub-matrix of the third coefficient matrix B2, D1 is a decomposition sub-matrix of the switching tube fault coefficient matrix D, M1, M2 are respectively the 1st decomposition sub-matrix and the 2nd decomposition sub-matrix of the disturbance coefficient matrix M, C1, C4 are respectively the 1st decomposition sub-matrix and the 2nd decomposition sub-matrix of the fourth coefficient matrix C, and F2 is a decomposition sub-matrix of the sensor fault coefficient matrix F;

[0032] Step 2.3, the matrix transformation value of the state variable x and the matrix transformation value of the output signal y are introduced, and the original fault state equation is decomposed into a subsystem 1 containing only switching tube open circuit fault and a primary subsystem 2 containing only sensor fault, wherein, x1, x2 are respectively the first decomposition state variable and the second decomposition state variable of the state variable x; y1, y2 are respectively the first decomposition signal and the second decomposition signal of the output signal y;

[0033] The expressions of the subsystem 1 and the primary subsystem 2 are respectively:

[0034]

[0035] y1 = C1x1

[0036]

[0037] y2 = C4x2 + F2f s

[0038] wherein, is the derivative of the first decomposed state variable x1 of the state variable x, is the derivative of the second decomposed state variable x2 of the state variable x;

[0039] Augmenting the primary subsystem 2, we obtain the subsystem 2, whose expression is:

[0040]

[0041] wherein, is the augmentation of the second decomposed state variable x2 of the state variable x, and is denoted as the state variable second decomposed variable augmentation is the derivative of , and ω is the augmentation of the sensor fault and the disturbance quantity, is the derivative of the sensor fault signal f s , is the augmented coefficient matrix 1, is the augmented coefficient matrix 2, is the augmented coefficient matrix 3, is the augmented coefficient matrix 4, N is the augmented system sensor fault and disturbance coefficient matrix, L3 is a unit matrix, is the augmented coefficient matrix 5,

[0042] Preferably, the two observers in step 3 are respectively: a finite time observer designed for the subsystem 1 and denoted as observer 1, and a finite time observer designed for the subsystem 2 and denoted as observer 2, whose expressions are respectively:

[0043]

[0044] wherein, L1 is the observer gain matrix 1, L2 is the observer gain matrix 2, α is the observer nonlinear gain 1, β is the observer nonlinear gain 2, k1 is the observer nonlinear gain 3, and k2 is the observer nonlinear gain 4, is the estimated value of the first decomposed state variable of the state variable x, is the derivative of , is the estimated value of the second decomposed state variable x2 of the state variable x, is the derivative of ; is the estimation value of the state variable second decomposition variable augmentation , is the estimation value of ;

[0045] The estimation value of the state variable second decomposition variable augmentation is obtained in the finite time observer of subsystem 2 is the estimation matrix of the sensor fault variable f s , wherein, is the three-phase sensor fault variable estimation value;

[0046] The k-phase sensor fault estimation value

[0047] Preferably, the implementation process of step 4 is as follows:

[0048] Step 4.1, the two subsystem residual matrices include a first residual matrix e1 and a second residual matrix e2, e 1_a , e 1_b , e 1_c is a three-phase residual containing only switch tube faults, e 2_a , e 2_b , e 2_c is a three-phase residual containing only sensor faults;

[0049] The calculation formula of the switch tube fault detection variable η a , the sensor fault detection variable η s is respectively:

[0050] Step 4.2, introduce k-phase switch tube fault residual e 1_k and k-phase sensor fault residual e 2_k , k=a, b, c; calculate k-phase switch tube fault detection threshold Tha k and k-phase sensor fault detection threshold Ths k , the calculation formula is respectively:

[0051]

[0052]

[0053] wherein, γ0, θ0 are respectively the first exponential convergence rate and the second exponential convergence rate of the error dynamic, λ k , μ krespectively, k-phase switch fault residual error e 1_k value at initial time and k-phase sensor fault residual error e 2_k value at initial time, t is current cycle time, τ is integral variable;

[0054] switch fault detection threshold Tha and sensor fault detection threshold Ths are calculated,

[0055] Preferably, the implementation process of step 6 is as follows:

[0056] Step 6.1, if η a ≥ Tha and η s < Ths, it is determined that a switch fault occurs, and step 6.2 is entered; if η a < Tha and η s ≥ Ths, it is determined that a sensor fault occurs, and step 6.3 is entered;

[0057] Step 6.2, switch fault diagnosis and positioning

[0058] The sliding window length L s is defined, L s = T m × f, wherein f is the sampling rate, and T m is the current cycle; each window corresponds to a complete current cycle T m , the k-phase current cumulative value Hk is accumulated through the sliding window, l is any sample obtained once, l = 1, 2, … L s ;

[0059] First, the phase positioning of the switch fault is performed, and the switch phase variable p k is set, p k = e 1_k . The rule is: if p k > Tha k , it is determined that the k-phase switch fault occurs

[0060] Then, a fixed switch threshold T0 is given, and the fault switch positioning is performed, and the rule is:

[0061] If H k < -T0, it is determined that the upper bridge switch of the k-phase bridge arm has an open circuit fault;

[0062] If H k > +T0, it is determined that the lower bridge switch of the k-phase bridge arm has an open circuit fault;

[0063] Step 6.3, sensor fault diagnosis and positioning

[0064] The sensor fault positioning threshold Ts , T s i max = i max ×5%, wherein i k is the amplitude of the output current;

[0065] First, sensor fault localization is performed, with the rules being:

[0066] If and , then it is determined that the A-phase sensor has a fault;

[0067] If and , then it is determined that the B-phase sensor has a fault;

[0068] If and , then it is determined that the C-phase sensor has a fault;

[0069] Then, a first sensor fault classification variable m k and a second sensor fault classification variable n k are introduced, n k = sign(|i k |- T s ), sign being a sign function, and sensor fault classification is determined according to the following rules:

[0070] If m k = 0 and n k = 1, then it is determined that the k-phase sensor has a stuck fault;

[0071] If m k = 0 and n k = -1, then it is determined that the k-phase sensor has an open circuit fault;

[0072] If m k ≠ 0 and n k = 1, then it is determined that the k-phase sensor has a gain fault.

[0073] Compared with the prior art, the present application has the following beneficial effects:

[0074] 1. In the matrix transformation of the present application, the sensor faults and switch tube faults of a wind power inverter system are decoupled, precise estimation of the system state and sensor faults is achieved, and joint diagnosis of the switch tube and sensor is achieved.

[0075] 2. In the joint diagnosis method of the present application, the residual norm of two subsystems is used to set the sensor fault detection threshold and the switch tube fault detection threshold, respectively, so that the fault categories can be more accurately distinguished.

[0076] 3. In the fault positioning method of the application, the sliding window cumulative value is used to determine the fault switch position, which is more rapid and accurate in determining the fault position. BRIEF DESCRIPTION OF DRAWINGS

[0077] Figure 1 is the topology diagram of the wind power inverter in the embodiment.

[0078] Figure 2 is the flow chart of the combined fault diagnosis method of the application.

[0079] Figure 3 is the switch V a1 When an open circuit fault occurs, the a-phase output current i a and the estimated value Three-phase sensor fault estimated value Sensor fault detection variable η s and sensor fault detection threshold Ths, switch fault detection variable η a and switch fault detection threshold Tha.

[0080] Figure 4 is the switch V a1 When a fault occurs, the switch phasing variable p k , fixed threshold T0, k-phase current cumulative value H k and k-phase switch fault detection threshold Tha k of the simulation waveform diagram;

[0081] Figure 5 is the switch V a1 and the switch V b1 When a fault occurs, the a-phase output current i a and the estimated value b-phase output current i b and the estimated value Three-phase sensor fault estimated value Sensor fault detection variable η s and sensor fault detection threshold Ths, switch fault detection variable η a and switch fault detection threshold Tha.

[0082] Figure 6 is the switch V a1 and the switch V b1 When a fault occurs, the open circuit fault switch phasing variable p k , fixed threshold T0, k-phase current cumulative value H k and k-phase switch fault detection threshold Tha k of the simulation waveform diagram.

[0083] Figure 7 When the a-phase sensor in this embodiment has an open circuit fault, the a-phase output current i a and estimated values Sensor fault detection variable η s And sensor fault detection threshold Ths, switch tube fault detection variable η a And the simulation waveform of the switch tube fault detection threshold Tha.

[0084] Figure 8 is the sensor fault location value M when the a-phase sensor in this embodiment has an open circuit fault k and sensor fault location threshold T s , sensor failure first classification variable m k and sensor failure second classification variable n k .

[0085] Figure 9 In this embodiment, when the a-phase sensor is stuck, the a-phase output current i a and estimated values Sensor fault detection variable η s And sensor fault detection threshold Ths, switch tube fault detection variable η a And the simulation waveform of the switch tube fault detection threshold Tha.

[0086] Figure 10 In this embodiment, when the a-phase sensor is stuck, the sensor fault location value M k and sensor fault location threshold T s , sensor failure first classification variable m k and sensor failure second classification variable n k .

[0087] Figure 11 The sensor S in this embodiment a When an offset fault occurs, the output current i of phase a a and estimated values Sensor fault detection variable η s And sensor fault detection threshold Ths, switch tube fault detection variable η a And the simulation waveform of the switch tube fault detection threshold Tha.

[0088] Figure 12 is the sensor fault location value M when the phase a sensor in this embodiment has an offset fault k and sensor fault location threshold T s , sensor failure first classification variable m k and sensor failure second classification variable nk . DETAILED DESCRIPTION

[0089] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0090] Figure 1 is the topology diagram of the wind power inverter in the embodiment of the present invention, Figure 1 It can be seen that the topology involved in the wind power inverter includes a DC side power supply, a main inverter circuit, three identical sensors, three identical inductors, three identical resistors, a grid-connected three-phase voltage source and a control module; the main inverter circuit includes a three-phase bridge arm, and each phase bridge arm includes an upper bridge arm switch tube and a lower bridge arm switch tube;

[0091] The main inverter circuit is connected in parallel between the positive and negative busbars of the DC power supply. The three-phase output terminals of the main inverter circuit are connected in series with an inductor and a resistor respectively, and then connected to the three-phase voltage of the grid-connected side. The three single-phase sensors are respectively denoted as S a , S b , S c , and the detection ends are respectively connected between the three-phase output ends of the main inverter circuit and the inductor, the output ends are connected to the input ends of the control module, and the output ends of the control module are respectively connected to the switch tubes of the three-phase bridge arms.

[0092] exist Figure 1 Up, V a1 , V a2 ,V b1 , V b2 , V c1 , V c2 is the switch tube on the three-phase bridge arm, a, b, c are the phase sequence, 1 is the upper bridge arm, 2 is the lower bridge arm, D a1 ,D a2 , D b1 , D b2 , D c1 , D c2 It is a diode connected in anti-parallel to the corresponding switch tube. dc =800V.

[0093] Figure 2 This is a flow chart of the fault diagnosis method of the present invention. As can be seen from the figure, the faults described in the present invention include switch tube faults and sensor faults, and include the following steps:

[0094] Step 1: Establish a hybrid logic dynamic model of the wind power inverter.

[0095] In this embodiment, the process of implementing step 1 is as follows:

[0096] Sampling the three-phase output current i of the wind power inverter a ,ib , i c and the three-phase output voltage u of the wind power inverter a , u b , u c ; sampling the voltage U of the DC side power supply dc ;

[0097] A hybrid logic dynamic model of the wind power inverter is established, and the expression is as follows:

[0098]

[0099]

[0100]

[0101] wherein L is the inductance, R is the resistance, δ a , δ b , δ c is a three-phase hybrid logic dynamic switching function; a k-phase hybrid logic dynamic switching function δ k is introduced, wherein k represents a phase sequence, and k=a, b, c.

[0102] A k-phase output current i k is introduced, and a k-phase hybrid logic dynamic switching function δ k is introduced, wherein k represents a phase sequence, and k=a, b, c.

[0103] Step 2, the inverter system state equation containing the switching tube fault and the sensor fault is established, and is decoupled, decomposed and augmented through matrix transformation, to obtain a subsystem 1 containing only the switching tube open circuit fault and a subsystem 2 containing only the sensor fault.

[0104] In the embodiment, the implementation process of step 2 is as follows:

[0105] Step 2.1, the inverter system state equation containing the switching tube fault and the sensor fault is established, and is recorded as the inverter fault state equation, and the expression is as follows:

[0106]

[0107] y=Cx+Ff s

[0108] wherein y is the output signal of the inverter system, x is a state variable, x=[i a i b i c ] T , is the derivative of the state variable x, u is the grid-connected side voltage matrix, δ is a hybrid logic dynamic switching function matrix, A is a first coefficient matrix, B1 is a second coefficient matrix, B2 is a third coefficient matrix, C is a fourth coefficient matrix, C = diag[1 1 1], D is a switch tube fault coefficient matrix, F is a sensor fault coefficient matrix, M is a disturbance coefficient matrix, f a is a switch tube fault signal, f s is a sensor fault signal, d is a disturbance signal;

[0109] Step 2.2, a first non-singular transformation matrix T and a second non-singular transformation matrix S are introduced to decouple the inverter fault state equation by matrix transformation, and the following equation is obtained:

[0110]

[0111] Wherein, A1, A2, A3, A4 are respectively the first decomposition sub-matrix, the second decomposition sub-matrix, the third decomposition sub-matrix and the fourth decomposition sub-matrix of the first coefficient matrix A, B 11 , B 12 are respectively the first decomposition sub-matrix and the second decomposition sub-matrix of the second coefficient matrix B1, B 21 , B 22 are respectively the first decomposition sub-matrix and the second decomposition sub-matrix of the third coefficient matrix B2, D1 is a decomposition sub-matrix of the switch tube fault coefficient matrix D, M1, M2 are respectively the first decomposition sub-matrix and the second decomposition sub-matrix of the disturbance coefficient matrix M, C1, C4 are respectively the first decomposition sub-matrix and the second decomposition sub-matrix of the fourth coefficient matrix C, and F2 is a decomposition sub-matrix of the sensor fault coefficient matrix F;

[0112] Step 2.3, the matrix transformation value of the state variable x and the matrix transformation value of the output signal y are introduced, and the original fault state equation is decomposed into a subsystem 1 containing only switch tube open circuit fault and a primary subsystem 2 containing only sensor fault, wherein, x1, x2 are respectively the first decomposition state variable and the second decomposition state variable of the state variable x; y1, y2 are respectively the first decomposition signal and the second decomposition signal of the output signal y;

[0113] The expressions of the subsystem 1 and the primary subsystem 2 are respectively:

[0114]

[0115] y1 = C1x1

[0116]

[0117] y2=C4x2+F2f s

[0118] Where, is the derivative of the first decomposition state variable x1 of the state variable x, is the derivative of the second decomposition state variable x2 of the state variable x;

[0119] Subsystem 2 is obtained by augmenting the primary subsystem 2, and its expression is:

[0120]

[0121] in, It is the augmentation of the second decomposition state variable x2 of the state variable x, and is recorded as the state variable second decomposition variable augmentation for The derivative of ω is the augmentation of sensor fault and disturbance, is the sensor fault signal f s The derivative of is the augmented coefficient matrix 1, is the augmented coefficient matrix 2, is the augmented coefficient matrix 3, is the augmented coefficient matrix 4, N is the augmented system sensor fault and disturbance coefficient matrix, I3 is the unit matrix, is the augmented coefficient matrix 5,

[0122] Step 3: Build finite-time observers for subsystem 1 and subsystem 2 respectively, and obtain the k-phase sensor fault estimation value through the observer

[0123] In this embodiment, the two observers in step 3 are: a finite-time observer designed for subsystem 1 and denoted as observer 1, and a finite-time observer designed for subsystem 2 and denoted as observer 2, and their expressions are:

[0124]

[0125] Among them, L1 is the observer gain matrix 1, L2 is the observer gain matrix 2, α is the observer nonlinear gain 1, β is the observer nonlinear gain 2, k1 is the observer nonlinear gain 3, k2 is the observer nonlinear gain 4, is the estimated value of the first decomposition state variable of the state variable x, for derivative of is an estimated value of the second decomposition state variable x2 of the state variable x, derivative of is an estimated value of the second decomposition variable augmented state variable x2 of the state variable x, is an estimated value of

[0126] is an estimated value of the second decomposition variable augmented state variable x2 of the state variable x through the finite-time observer of the subsystem 2 is an estimated matrix of the sensor fault variable f s wherein, is a three-phase sensor fault variable estimated value;

[0127] introducing a k-phase sensor fault estimated value

[0128] Step 4, based on the finite-time observer, two subsystem residual matrices are obtained, and then a switch tube fault detection variable η is calculated a , a sensor fault detection variable η s , a switch tube fault detection threshold Tha and a sensor fault detection threshold Ths.

[0129] In this embodiment, the implementation process of Step 4 is as follows:

[0130] Step 4.1, the two subsystem residual matrices include a first residual matrix e1 and a second residual matrix e2, e 1_a , e 1_b , e 1_c is a three-phase residual containing only switch tube faults, e 2_a , e 2_b , e 2_c is a three-phase residual containing only sensor faults;

[0131] The calculation formula of the switch tube fault detection variable η a , the sensor fault detection variable η s is respectively:

[0132] Step 4.2, introduce k-phase switch tube fault residuals e 1_k and k-phase sensor fault residuals e 2_k , k=a, b, c; calculate k-phase switch tube fault detection threshold Tha k and k-phase sensor fault detection threshold Ths k , whose calculation formula is respectively:​​​​​

[0133]

[0134] where γ0, θ0are the first and second exponential convergence rates of the error dynamics, respectively, λ k , μ k are the values of the k-phase switch fault residual e 1_k and the k-phase sensor fault residual e 2_k at the initial time, t is the current cycle time, and τ is the integration variable;

[0135] The switch fault detection threshold Thaand the sensor fault detection threshold Thsare calculated,

[0136] Step 5, if η a < Thaand η s < Ths, it is determined that the inverter is working normally, and the fault diagnosis is ended, otherwise, step 6 is entered.

[0137] Step 6, according to the switch fault residual and the k-phase current cumulative value obtained through the sliding window, a switch fixed threshold is given, and switch fault diagnosis and positioning are performed; according to the k-phase sensor fault estimation value and the sensor fault classification variable calculated through the three-phase output current, a sensor fault positioning threshold is given, and sensor fault diagnosis and positioning are performed.

[0138] In the embodiment, the implementation process of step 6 is as follows:

[0139] Step 6.1, if η a ≥ Thaand η s < Ths, it is determined that a switch fault occurs, and step 6.2 is entered; if η a < Thaand η s ≥ Ths, it is determined that a sensor fault occurs, and step 6.3 is entered.

[0140] Step 6.2, switch fault diagnosis and positioning

[0141] The sliding window length L s is defined, L s = T m × f, where f is the sampling rate, and T m is the current cycle; a complete current cycle T m is allowed to correspond to each window, and the k-phase current cumulative value H k is accumulated through the sliding window, l is an arbitrary sample, l = 1, 2,... L s ;

[0142] First, the phase location of the switch tube fault is performed, and the switch tube phase variable p is set k , p k =e 1_k The rule is: if p k >Tha k , it is determined that the k-phase switch tube is faulty

[0143] Then, a fixed threshold value T0 of the switch tube is given to locate the faulty switch tube. The rule is:

[0144] If H k <-T0, it is determined that the upper bridge arm switch tube of the k-phase bridge arm has an open circuit fault;

[0145] If H k >+T0, it is determined that the lower bridge arm switch tube of the k-phase bridge arm has an open circuit fault;

[0146] Step 6.3, sensor fault diagnosis and location

[0147] Introducing sensor fault location threshold T s , T s =i max ×5%, where i max is the amplitude of the output current;

[0148] First, locate the sensor fault. The rules are as follows:

[0149] like and It is determined that the A phase sensor is faulty;

[0150] like and It is determined that the B phase sensor is faulty;

[0151] like and It is determined that the C phase sensor is faulty;

[0152] Then introduce the first classification variable m of sensor failure k and sensor failure second categorical variable n k , n k =sign(|i k |-T s ), sign is the sign function, and the sensor fault category is determined according to the following rules:

[0153] If m k =0,n k =1, it is determined that the k-phase sensor is stuck;

[0154] If m k =0,nk =-1, it is determined that the k-phase sensor has an open circuit fault;

[0155] If m k ≠0, n k =1, it is determined that the k-phase sensor has a gain failure.

[0156] In order to demonstrate the beneficial effects of the present invention, a simulation of the present invention was performed.

[0157] Figure 3 is the switch tube V in this embodiment a1 When an open circuit fault occurs, the output current of phase a is i a and estimated values Three-phase sensor fault estimation value and sensor fault detection variable η s , sensor fault detection variable η s And sensor fault detection threshold Ths, switch tube fault detection variable η a And the simulation waveform of the switch tube fault detection threshold Tha. Figure 3 It can be seen that after 0.061s, i a There have been major changes, unchanged, η s Does not exceed the threshold Ths,η a Exceeds the threshold Tha.

[0158] Figure 4 The switch tube V in the embodiment of the present invention a1 When an open circuit fault occurs, the switching tube phase variable p k , fixed threshold T0, k-phase current cumulative value H k And the k-phase switch fault detection threshold Tha k The simulation waveform diagram; It can be seen from the figure that after 0.062s, the fault phase variable p of phase a a Greater than the adaptive threshold Tha a , the other two phases do not exceed the fault location variable H a <-T0, it is judged that the switch tube on phase A is faulty.

[0159] Figure 5 is the switch tube V in this embodiment a1 With the switch tube V b1 When a fault occurs, the output current of phase a is i a and estimated values Phase b output current i b and estimated values Three-phase sensor fault estimation value Sensor fault detection variable η s And sensor fault detection threshold Ths, switch tube fault detection variable η aand the simulation waveform chart of the switching tube fault detection threshold value Tha. From Figure 5 It can be seen that after 0.061s, i a has a great change, is unchanged, i b has no great change, and after 0.133s, i b has a great change, is unchanged, i s is not more than the threshold value Ths, i a is more than the threshold value Tha.

[0160] Figure 6 is the simulation waveform chart of the switching tube V a1 and the switching tube V b1 , the switching tube phasing variable p k , the fixed threshold value T0, the k-phase current cumulative value H k , and the k-phase switching tube fault detection threshold value Tha k in this embodiment. From Figure 6 It can be seen that after 0.062s, the fault phasing variable p a of the a-phase is greater than the threshold value Tha a , and then after 0.133s, the fault phasing variable p b of the b-phase is greater than the threshold value Tha b , H a <-T0, H b <-T0, and it is judged that the switching tube on the a-phase is faulty and the switching tube on the b-phase is faulty.

[0161] Figure 7 is the simulation waveform chart of the a-phase sensor open circuit fault, the a-phase output current i a and the estimated value , the sensor fault detection variable η s , the sensor fault detection threshold value Ths, the switching tube fault detection variable η a , and the switching tube fault detection threshold value Tha in this embodiment. From Figure 7 It can be seen that after 0.11s, i a has a great change, is unchanged, i s is more than the threshold value Ths, i a is not more than the threshold value Tha, and it is judged that the sensor is faulty.

[0162] Figure 8 is the simulation waveform chart of the a-phase sensor open circuit fault, the sensor fault positioning variable M k , the sensor fault positioning threshold value T s , the sensor fault first classification variable m k , and the sensor fault second classification variable n kis the simulation waveform diagram of the a-phase sensor fault estimation value is the simulation waveform diagram of the a-phase sensor fault estimation value Figure 8 It can be seen that after 0.11s, the a-phase sensor fault estimation value is not 0, the rest of the b-phase and c-phase sensor fault estimation values are 0, it is judged that the a-phase sensor is faulty, m a = o, n a = -1, and it is judged that it is an open circuit fault.

[0163] Figure 9 is the simulation waveform diagram of the a-phase output current i a and the estimation value of the sensor fault detection variable η s and the sensor fault detection threshold value Ths, the switch tube fault detection variable η a and the switch tube fault detection threshold value Tha. It can be seen that after 0.112s, i a undergoes a large change, η s remains unchanged, η a exceeds the threshold value Ths, and η k does not exceed the threshold value Tha, and it is judged that the sensor is faulty.

[0164] Figure 10 is the simulation waveform diagram of the sensor fault positioning variable M s and the sensor fault positioning threshold value T k , the sensor fault first classification variable m k and the sensor fault second classification variable n a . It can be seen that after 0.115s, is not 0, the rest of the two phases are 0, it is judged that the a-phase sensor is faulty, m a = 0, n a = 1, and it is judged that the sensor is stuck.

[0165] Figure 11 is the simulation waveform diagram of the a-phase output current i a and the estimation value of the sensor fault detection variable η s and the sensor fault detection threshold value Ths, the switch tube fault detection variable η a and the switch tube fault detection threshold value Tha when the sensor S a experiences a shift fault in this embodiment. It can be seen that after 0.115s, i s undergoes a large change, η a remains unchanged, η k exceeds the threshold value Ths, and η s does not exceed the threshold value Tha, and it is judged that the sensor is faulty.a If the value does not exceed the threshold Tha, it is determined that the sensor has an offset failure.

[0166] Figure 12 is the sensor fault location value M when the phase a sensor in this embodiment has an offset fault k and sensor fault location threshold T s , sensor failure first classification variable m k and sensor failure second classification variable n k The simulation waveform of . Figure 12 It can be seen that after 0.115s, If it is not 0 and the other two phases are 0, it is judged that the sensor of phase a is faulty. a ≠0, n a =1, it is judged that the sensor has an offset failure.

Claims

1. A method for diagnosing faults in a wind turbine inverter, wherein the faults include switch tube faults and sensor faults. The topology of the wind turbine inverter includes a DC power supply, a main inverter circuit, three identical inductors, three identical resistors, a grid-connected three-phase voltage source, three identical sensors, and a control module. The main inverter circuit includes three-phase bridge arms, each of which includes an upper bridge arm switch tube and a lower bridge arm switch tube. The method is characterized in that: The fault diagnosis method comprises the following steps: Step 1: sampling and establishing a wind power inverter hybrid logic dynamic model; Step 2: Establish the inverter system state equation containing switch tube faults and sensor faults, and decouple, decompose, and augment through matrix transformation to obtain subsystem 1 containing only switch tube open circuit faults and subsystem 2 containing only sensor faults; Step 3: Build finite time observers for subsystem 1 and subsystem 2 respectively, and obtain the k-phase sensor fault estimation value through the observer Step 4: Based on the finite time observer, the residual matrices of the two subsystems are obtained, and then the switch tube fault detection variable η is calculated. a , sensor fault detection variable η s , switch tube fault detection threshold Tha and sensor fault detection threshold Ths; Step 5: If η a <Tha and η s <Ths, the inverter is judged to be operating normally and the fault diagnosis is completed, otherwise it goes to step 6; Step 6: Based on the switch fault residual and the k-phase current cumulative value obtained through the sliding window, a fixed threshold of the switch is given to diagnose and locate the switch fault; based on the k-phase sensor fault estimation value The sensor fault classification variable is calculated by the three-phase output current, and the sensor fault location threshold is given to perform sensor fault diagnosis and location.

2. A wind power inverter fault diagnosis method according to claim 1, characterized in that: The process of implementing step 1 is as follows: Sampling the three-phase output current i of the wind power inverter a ,i b ,i c And the three-phase output voltage u of the wind power inverter a ,u b ,u c ; Sample the voltage U of the DC power supply dc ; A hybrid logic dynamic model of wind power inverter is established, and its expression is: Among them, L is the inductance value, R is the resistance value, δ a , δ b , δ c is the three-phase hybrid logic dynamic switching function; the k-phase hybrid logic dynamic switching function δ is introduced k , k represents the phase sequence, k = a, b, c.

3. A wind power inverter fault diagnosis method according to claim 2, characterized in that: The implementation process of step 2 is as follows: Step 2.1: Establish the inverter system state equation containing switch tube fault and sensor fault, and record it as the inverter fault state equation, which is expressed as: y=Cx+Ff s Among them, y is the output signal of the inverter system, x is the state variable, x=[i a i b i c ] T , is the derivative of the state variable x, is the grid-connected voltage matrix, δ is the hybrid logic dynamic switch function matrix, A is the first coefficient matrix, B1 is the second coefficient matrix, B2 is the third coefficient matrix, C is the fourth coefficient matrix, C=diag[111], D is the switch tube fault coefficient matrix, F is the sensor fault coefficient matrix, M is the disturbance coefficient matrix, f a is the switch failure signal, f s is the sensor fault signal, d is the disturbance signal; Step 2.2: Introduce the first non-singular transformation matrix T and the second non-singular transformation matrix S, perform matrix transformation decoupling on the inverter fault state equation, and obtain: Among them, A1, A2, A3, and A4 are the first decomposition submatrix, the second decomposition submatrix, the third decomposition submatrix, and the fourth decomposition submatrix of the first coefficient matrix A, respectively. 11 , B 12 are the first decomposition submatrix and the second decomposition submatrix of the second coefficient matrix B1, respectively. 21 , B 22 are the first and second decomposition submatrices of the third coefficient matrix B2, respectively. D1 is the decomposition submatrix of the switch tube fault coefficient matrix D. M1 and M2 are the first and second decomposition submatrices of the disturbance coefficient matrix M, respectively. C1 and C4 are the first and second decomposition submatrices of the fourth coefficient matrix C, respectively. F2 is the decomposition submatrix of the sensor fault coefficient matrix F. Step 2.3, introduce the matrix transformation value of the state variable x And the matrix transformation value of the output signal y The original fault state equation is decomposed into a subsystem 1 containing only the switch tube open circuit fault and a primary subsystem 2 containing only the sensor fault, where: x1, x2 are the first decomposition state variable and the second decomposition state variable of the state variable x; y1, y2 are the first decomposition signal and the second decomposition signal of the output signal y respectively; The expressions of subsystem 1 and primary subsystem 2 are: y1=C1x1 y2=C4x2+F2f s Where, is the derivative of the first decomposition state variable x1 of the state variable x, is the derivative of the second decomposition state variable x2 of the state variable x; Subsystem 2 is obtained by augmenting the primary subsystem 2, and its expression is: in, It is the augmentation of the second decomposition state variable x2 of the state variable x, and is recorded as the state variable second decomposition variable augmentation for The derivative of ω is the augmentation of sensor fault and disturbance, is the sensor fault signal f s The derivative of is the augmented coefficient matrix 1, is the augmented coefficient matrix 2, is the augmented coefficient matrix 3, is the augmented coefficient matrix 4, N is the augmented system sensor fault and disturbance coefficient matrix, I3 is the unit matrix, is the augmented coefficient matrix 5, 4. A wind power inverter fault diagnosis method according to claim 3, characterized in that: The two observers in step 3 are: the finite-time observer designed for subsystem 1 and denoted as observer 1, and the finite-time observer designed for subsystem 2 and denoted as observer 2, and their expressions are: Among them, L1 is the observer gain matrix 1, L3 is the observer gain matrix 2, α is the observer nonlinear gain 1, β is the observer nonlinear gain 2, k1 is the observer nonlinear gain 3, k2 is the observer nonlinear gain 4, is the estimated value of the first decomposition state variable of the state variable x, for The derivative of is the estimated value of the second decomposition state variable x2 of the state variable x, for The derivative of Augment the second decomposition variable of the state variable The estimated value of for estimated value of; The second decomposition variable augmentation of the state variable is obtained from the finite time observer of subsystem 2 Estimated value of is the sensor fault quantity f s The estimation matrix of in, is the estimated value of the three-phase sensor fault quantity; Introducing k-phase sensor fault estimation 5. A wind power inverter fault diagnosis method according to claim 4, characterized in that: The implementation process of step 4 is as follows: Step 4.1: The two subsystem residual matrices include a first residual matrix e1 and a second residual matrix e2. e 1_a , e 1_b , e 1_c is the three-phase residual error including only the switch failure, e 2_a , e 2_b , e 2_c is the three-phase residual including only sensor fault; The switch tube fault detection variable η a , sensor fault detection variable η s The calculation formulas are: Step 4.2: Introduce the k-phase switch fault residual e 1_k and k-phase sensor fault residual e 2_k , k=a,b,c; Calculate the k-phase switch fault detection threshold Tha k and k-phase sensor fault detection threshold Ths k , and the calculation formulas are: Among them, γ0, θ0 are the first exponential convergence rate and the second exponential convergence rate of error dynamics, λ k , μ k are the k-phase switch fault residuals e 1_k The value at the initial moment and the k-phase sensor fault residual e 2_k The value at the initial moment, t is the current cycle time, and τ is the integral variable; Calculate the switch tube fault detection threshold Tha and sensor fault detection threshold Ths, 6. A wind power inverter fault diagnosis method according to claim 5, characterized in that: The implementation process of step 6 is as follows: Step 6.1, if η a ≥Tha and η s <Ths, it is determined that a switch failure has occurred and the process goes to step 6.2; if η a <Tha and η s ≥Ths, it is determined that a sensor failure has occurred and the process goes to step 6.3; Step 6.2, switch tube fault diagnosis and location Define the sliding window length L s , L s =T m ×f, where f is the sampling rate, T m is the current cycle; let each window correspond to a complete current cycle T m , accumulate the k-phase current cumulative value Hk through the sliding window, l is the sample obtained from any sampling, l = 1, 2, ...L s ; First, the phase location of the switch tube fault is performed, and the switch tube phase variable p is set k , p k =e 1_k The rule is: if p k >Tha k , it is determined that the k-phase switch tube is faulty Then, a fixed threshold value T0 of the switch tube is given to locate the faulty switch tube. The rule is: If H k <-T0, it is determined that the upper bridge arm switch tube of the k-phase bridge arm has an open circuit fault; If H k >+T0, it is determined that the lower bridge arm switch tube of the k-phase bridge arm has an open circuit fault; Step 6.3, sensor fault diagnosis and location Introducing sensor fault location threshold T s , T s =i max ×5%, where i max is the amplitude of the output current; First, locate the sensor fault. The rules are as follows: like and It is determined that the A phase sensor is faulty; like and It is determined that the B phase sensor is faulty; like and It is determined that the C phase sensor is faulty; Then introduce the sensor fault first classification variable mk and sensor fault second classification variable n k , n k =sign(|i k |-T s ), sign is the sign function, and the sensor fault category is determined according to the following rules: If m k =0,n k =1, it is determined that the k-phase sensor is stuck; If m k =0,n k =-1, it is determined that the k-phase sensor has an open circuit fault; If m k ≠0, n k =1, it is determined that the k-phase sensor has a gain failure.

Images