Early fault monitoring method for roof cable of motor train unit

Abstract

The invention relates to the technical field of cable early-stage fault monitoring, and discloses a motor train unit roof cable early-stage fault monitoring method, which comprises the following steps: establishing an equivalent circuit model comprising an insulation carbonization resistor, an arc resistor and a grounding resistor, and quantifying the influence of carbonization damage on fault current transient behaviors; based on an equivalent circuit model, extracting a rising trend slope of a phase current amplitude, and distinguishing early fault and non-fault interference through positive and negative values of the rising trend slope; the nonlinear distortion characteristic of a phase voltage-phase current characteristic curve is utilized, and the closed area of the characteristic curve is used as an index for quantifying a fault state. By fusing a physical model and multi-feature signal analysis, the method achieves the dynamic monitoring of the whole period of insulation degradation, is especially suitable for the high-frequency cold and hot circulation and high-speed airflow environment of a motor train unit roof cable, remarkably improves the accuracy and reliability of fault diagnosis, and provides an effective technical support for the safe operation and maintenance of a high-speed rail.

Description

Technical Field

[0001] The present invention relates to the technical field of early fault monitoring of cables, and particularly to a method for early fault monitoring of overhead cables of EMUs. Background Art

[0002] With the rapid development of China's high-speed railways, a large number of overhead cables have been installed for the high-voltage power supply system of EMUs. Due to complex operating environments such as high-speed air flow impact, frequent mechanical vibration, and insulation material aging, the terminals of overhead cables are prone to failures. Overhead cable failures can be divided into permanent failures (such as conductor breakage and insulation breakdown) and early faults. In the actual operation and maintenance of EMUs, early fault detection usually follows traditional cable detection methods and diagnoses by extracting arc characteristics. However, different from ordinary cables, early faults of overhead cables involve the expansion of air gaps at the insulation interface and material fatigue deterioration, and existing research mainly focuses on arc discharge characteristics, making it difficult to accurately evaluate the fault development stage. Accurately detecting the fault state of overhead cables is crucial for the safe operation of high-speed railways, mainly for the following reasons:

[0003] 1) The overhead cables of EMUs are subjected to high-frequency thermal cycles (alternating between full-load operation during the day and outage maintenance at night) for a long time, resulting in the accumulation of thermo-mechanical stress in the insulation material. This intermittent impact load accelerates the expansion of air gap defects at the multi-layer dielectric interface of the cable terminal, and existing detection means are difficult to capture this progressive deterioration process.

[0004] 2) The insulation degradation of early faults of overhead cables is mainly manifested as interface delamination caused by stress concentration at the root of the petticoat, and the periodic deformation fatigue of silicone rubber materials under the action of high-speed air flow. This mechanical-electrothermal coupling deterioration process will gradually reduce the insulation strength. If different deterioration stages cannot be accurately identified, the cable terminals with potential hazards cannot be replaced in time, and ultimately sudden flashover accidents may occur.

[0005] Existing early fault detection methods are mainly based on the characteristics of air arcs, such as nonlinearity, harmonics, asymmetry, and randomness. Generally, these methods are mainly divided into signal processing methods based on system-recorded waveforms, artificial intelligence (AI)-based methods, and fault detection methods based on arc models.

[0006] Signal processing methods based on fault waveforms have been widely used in the early fault detection of cables. For example, early fault detection methods based on chaotic spread-spectrum sequences, using wavelet transform to extract the transient characteristics of fault signals, applying similarity functions to the voltage signals at the front end of fault lines to achieve fault detection, and using Kalman filters to analyze the characteristics of fault signals, etc.

[0007] In recent years, artificial intelligence (AI)-based methods have been increasingly used for early cable fault detection. For example, methods based on human concept learning (HLCL), detection methods using restricted Boltzmann machines and stacked autoencoders, or pre-trained models (PTM) to detect waveform anomalies, as well as detection methods based on cumulative sum and adaptive linear neurons.

[0008] Fault detection methods based on arc models utilize physical fault characteristics to establish equivalent circuits or mathematical models to identify early faults. These models include the Mayr, Cassie, Schwarz, and Hochrainer models, which are widely applied in high-voltage, medium-voltage, and low-voltage power systems. There are studies focused on developing arc fault models and non-arc interference models for detecting early cable faults. In addition, a high-resistance air arc model incorporating air pressure and arc gap length parameters has been introduced, or a more accurate non-linear arc high-impedance fault model based on the logarithmic arc model has been proposed.

[0009] However, existing detection methods do not consider the physical behavior of faults and only construct standards or models based on air arc characteristics for detection. These methods ignore the problem of the decrease in insulation resistivity, resulting in difficulty in determining the fault state and inability to achieve precise maintenance. Summary of the Invention

[0010] Aiming at the above problems, the purpose of the present invention is to provide an early fault monitoring method for the roof cables of EMUs, which uses the changing trend of phase current for early fault detection and determines the fault state through the non-linear distortion characteristics of the phase voltage-phase current characteristic curve. The technical solution is as follows:

[0011] An early fault monitoring method for the roof cables of EMUs, comprising the following steps:

[0012] Step 1: Establish an equivalent circuit model for early cable faults, where the equivalent circuit model includes a resistor R representing insulation carbonization damage c , a resistor R representing the cable arc resistance arc , and a resistor R representing the grounding resistance of the copper shielding layer con ; analyze the influence of the insulation carbonization resistance on the transient behavior of the fault current through the equivalent circuit model;

[0013] Step 2: Based on the equivalent circuit model, extract the phase current signal of the cable line, calculate the slope a1 of the amplitude change trend, and distinguish early faults from non-fault interferences through the positive and negative values of the slope a1;

[0014] Step 3: Determine the fault state through the non-linear distortion characteristics of the phase voltage-phase current characteristic curve, and use the enclosed area of the phase voltage-phase current characteristic curve as an index to quantify the fault state.

[0015] The beneficial effects of the present invention are as follows:

[0016] 1) By analyzing the physical evolution characteristics of the initial cable faults, a new initial cable fault model is established on the basis of the existing arc model based on diode resistance considering insulation degradation.

[0017] 2) The theoretical analysis of the transient equivalent circuit of the present invention reveals the mapping relationship between the fault state and the fault current. The phase current amplitude change trend is used for basic fault detection, and the accurate fault state determination is realized through the nonlinear distortion characteristics of the phase voltage-phase current (V-A) characteristic curve.

[0018] 3) The present invention can realize the monitoring of the early fault state of the cable, so as to realize the effective maintenance of the cable system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] FIG. 1(a) shows the typical structure of a medium-voltage railway power supply cable.

[0020] FIG. 1(b) shows a schematic diagram of the accumulation of carbonized particles caused by an arc fault in a railway power supply cable.

[0021] FIG. 1(c) shows the equivalent circuit model of an early fault of a railway power supply cable.

[0022] Figure 2 is the equivalent circuit of the initial fault of the railway power supply cable.

[0023] Figure 3 is a typical medium-voltage railway distribution system.

[0024] Figure 4 is the transient circuit when a fault occurs.

[0025] Figure 5 is the fault current waveform in the under-damped state.

[0026] Figure 6 is the fault current waveform in the over-damped state.

[0027] Figure 7 is the change trend of the fault current under different degrees of insulation carbonization; (a) R c = 20 kΩ, (b) R c = 2 kΩ, (c) R c = 500 Ω,

[0028] (d) R c = 100 Ω.

[0029] Figure 8 is the V-A characteristic curve of the cable during normal operation; (a) before phase correction, (b) after phase correction.

[0030] Figure 9 It is the V - A characteristic curve at the initial stage of cable fault.

[0031] Figure 10 It is the detection process of the early fault state of medium - voltage cables.

[0032] Figure 11 They are common fault - free disturbance models; (a) Load switching model, (b) Capacitor switching model.

[0033] Figure 12 It is the simulation model.

[0034] Figure 13 They are the simulation results of the change trend of phase current and the calculation results of parameter a1 under different disturbance types; (a) Initial cable fault, (b) Capacitor switch, (c) Load switching.

[0035] Figure 14 They are the V - A characteristic curve and the fault diagnosis results; (a) V - A characteristic curve, (b) Calculation results of A. Specific implementation manners

[0036] The following further elaborates on the present invention in detail in conjunction with the attached drawings and specific embodiments.

[0037] 1. Physical property analysis, circuit modeling and transient behavior analysis of early cable faults

[0038] A. Physical property analysis and circuit modeling of early cable faults

[0039] The typical structural section of a medium - voltage cable is shown in Figure 1(a). Its core components form a hierarchical structure consisting of a conductor core, a main insulation layer and a copper shielding layer. The XLPE insulation layer will undergo progressive deterioration due to electro - thermal stress during long - term operation, forming local defects. These defects may trigger early faults under the action of a continuous electric field, manifested as periodic arc discharge phenomena inside the insulating medium, as shown in Figure 1(b).

[0040] Compared with the arc - grounding faults with strong gas self - recovery characteristics in overhead lines, the arc discharge in the cable system has significant differences. Inside the cable insulation layer, the energy of arc discharge will form irreversible carbonization damage in the medium. Each discharge event will exacerbate the deposition of carbonization particles, as shown in Figure 1(b). This cumulative effect affects the insulation resistance, and the dynamic evolution of this carbonization degree will directly affect the insulation characteristics at the fault location.

[0041] Based on the above physical damage mechanism, a multi - parameter coupling model of cable faults as shown in Figure 1(c) is constructed in this paper. This model decomposes the fault point into three core impedance elements: the dynamic resistance R representing the carbonization damage area c , the resistance R describing the intrinsic characteristics of the arc arc, and the resistance R reflecting the contact state of the copper shielding layer con .

[0042] Regarding the non-linear characteristics of low-current arcs in medium-voltage systems, this model introduces a diode-modulated resistance architecture, as Figure 2 shown. This architecture precisely simulates the three core characteristics of the cable fault arc current through the combination of diodes D p and D n connected in parallel in the forward and reverse directions, adjustable DC power supplies V p and V n : 1) Discharge intermittency 2) Random volatility 3) Polarity asymmetry. The introduction of the current-limiting resistance R0 realizes the physical constraint of the arc current amplitude. Under the excitation of the power frequency phase voltage V ph , the total fault current if is composed of the carbonized channel current i c and the arc current i arc , and its mathematical representation is shown in Equation (1). When D p and D n conduct alternately in the positive / negative half cycles, the i arc component exhibits a pulsating characteristic; conversely, the fault current only conducts through the R c branch.

[0043]

[0044] B. Early fault transient analysis considering the dynamic development of cable faults

[0045] During the development of cable faults, the change in the equivalent resistance of the cable caused by the cracking and carbonization of the insulating material will change the electrical characteristics of the system, and thus have a significant impact on the fault transient process. In response to this phenomenon, it is necessary to focus on studying the impact of the impedance change at the carbonized part on the dynamic characteristics of the fault current.

[0046] Among the three grounding methods of the neutral point of the power system, due to the relatively large distributed capacitance to the ground in the medium-voltage power grid cable system, the capacitive current is relatively large when a single-phase ground fault occurs. Therefore, the arc suppression coil compensation method is commonly used, as Figure 3 shown. In the early fault stage, the grounding arc exhibits a high-resistance characteristic. After compensation by the arc suppression coil, the fault current is composed of the system's capacitive component i C to the ground and the inductive compensation component i L together. Figure 4 In , the fault equivalent potential source can be expressed as f The total impedance R p of the fault loop is jointly determined by the resistance of the XLPE insulation carbonized channel and the air arc impedance. In this equivalent model, R and L represent the total resistance and total inductance of the line modulus parameters respectively, C is the line capacitance to the ground, L cDenotes the voltage across the capacitor. In the early fault stage, the charging and discharging process of the distributed capacitor has high-frequency resonance characteristics. Due to the line compensation characteristics in the system, the inductive reactance value of the arc suppression coil is significantly higher than the capacitive reactance of the faulty phase of the cable, so the inductor L can be ignored p .

[0047] The fault current consists of two components: power frequency component and damped high-frequency component. Therefore, after the initial fault occurs, the overall current shows a decaying trend Figure 5 is the fault current waveform under the underdamped state. After the fault occurs, the overall current shows a slow rising trend, as Figure 6 shown in the simulation results

[0048] In addition, it is worth noting that the fault current shows obvious non-linear distortion, which is different from the trend Figure 5 shown. This indicates that there is a correlation between the non-linear distortion of the fault current and the insulation carbonization damage. Based on the model Figure 2 shown, by setting V p = 3 kV, V n = 2.5 kV, R0 = 300 Ω and R con = 5 Ω, the fault current waveforms under different insulation carbonization resistances are simulated. The results are as Figure 7 (a)-(d) shown. As R c decreases, the non-linear distortion decreases

[0049] Therefore, there are essential differences in the current transient characteristics under different fault conditions. Specifically, in the early stage of the early fault, the current shows overdamped characteristics, and its amplitude increases with time. In engineering practice, although there are technical obstacles to directly measuring the fault point current, the phase current of the line has an increasing characteristic similar to that of the fault current and can be used as an effective monitoring object. It is worth noting that the literature research shows that the phase current caused by non-fault disturbances (such as load, capacitor switching operations) shows a decaying characteristic, which is in sharp contrast to the increasing characteristic of the fault current and provides a criterion basis for fault identification. Further, by analyzing the non-linear distortion degree of the phase voltage-current characteristic curve, the accurate discrimination of the fault state can be realized. The change trend of the fault current under different insulation carbonization degrees

[0050] 2. Cable early fault detection method

[0051] Based on the above analysis, in order to achieve accurate fault state detection, the present invention proposes a fault state detection method based on the non-linear distortion characteristics of the V-A characteristic curve

[0052] A. Cable early fault detection method based on the change trend of phase current

[0053] According to the above analysis, the fault detection method needs to extract features from the development trend of the phase current waveform envelope. For this purpose, the following steps are required:

[0054] 1) First, since there is significant noise interference in on-site monitoring, it is necessary to denoise the current signal to remove high-frequency noise. In this embodiment, a 1 kHz low-pass filter is used for signal denoising.

[0055] 2) Next, extract the positive peak current i within three power frequency cycles after the fault occurs ph (t i ), where t i represents the peak time. Considering that the fault current signal shows an exponential decay form, in this embodiment, the logarithm of the current signal is taken, and then the slope a1 is obtained by linear fitting, as shown in Equation (2).

[0056] 3) Finally, construct an objective function as shown in Equation (3) to determine appropriate values of a0 and a1 such that the error between the observed value i ph (t i ) and the predicted value y(t i ) obtained from Equation (2) is minimized.

[0057] y(t i ) = ln|i ph (t i )| = a1t i + a0 (2)

[0058]

[0059] In Equation (3), the minimum S(a0, a1) can be obtained by separately setting and . Therefore, a1 can be calculated as:

[0060]

[0061] where,

[0062] According to Equation (4), the coefficient a1 can be calculated. By determining the sign of a1, we can determine whether the current trend is increasing or decaying, so as to distinguish the early fault of the cable from other system interference signals. When a1 > 0, it is determined as an early fault, and when a1 < 0, it is determined as a non-fault interference

[0063] B. Cable early fault state diagnosis method based on the non-linear characteristics of the V-A characteristic curve

[0064] As the severity of the cable fault increases, the non-linear distortion characteristics of the fault current weaken. Therefore, the non-linear characteristics of the phase voltage-phase current (V-A) characteristic curve gradually weaken, which can be used to determine the fault state. When the cable is operating normally, the V-A characteristic curve is as shown in Figure 8 (a). Due to the phase difference, the curves do not completely overlap. Therefore, phase correction is first performed by detecting and aligning the zero-crossing points of the voltage and current signals, so that the V-A curve exhibits linear characteristics, as shown in Figure 8 (b).

[0065] When an early fault occurs in the cable, the fault current shows non-linear distortion, resulting in the distortion of the V-A characteristic curve, as shown in Figure 9 . According to previous observations, as the severity of the fault increases, the non-linear distortion of the fault current weakens. Therefore, the V-A characteristic curve tends to be more linear, resulting in a decrease in the enclosed area of the V-A characteristic curve. Therefore, the present invention proposes to use the enclosed area of the V-A characteristic curve as an index for quantifying the fault state.

[0066] To calculate the enclosed area of the curve, the polygon area formula is used to perform numerical integration on the discrete data points. Specifically, given the phase voltage V(t) and phase current I(t) signals, the enclosed area A can be calculated by Equation (5).

[0067]

[0068] where V i and I i are the voltage and current values of the i-th data point respectively, and n is the total number of data points.

[0069] Set the area threshold A th , when A ≤ A th , it is determined that the insulation carbonization damage enters the severe stage.

[0070] In summary, the fault state detection method proposed by the present invention is as shown in Figure 10 .

[0071] 3. Simulation verification

[0072] In the power system, system interferences of non-fault types always occur, such as common load switching and capacitor switching. Load switching can be equivalently represented as an R-L series grounded circuit, as shown in Figure 11 . Capacitor switching can be equivalently represented as an R-C series grounded circuit. In the following sections, fault state detection is performed based on the values of the proposed parameters a1 and A in sequence.

[0073] A. Simulation model

[0074] A medium-voltage distribution system simulation model was developed, as shown inFigure 12 As shown, early cable faults, load switching, and capacitor bank switching are simulated respectively. The positive sequence parameters of the cable are: R+ = 0.193 Ω / km, L++ = 0.442 mH / km, C + = 0.143 nF / km; the zero sequence parameters are: R0 = 1.93 Ω / km, L0 = 5.48 mH / km, C0 = 14.3 nF / km; the arc suppression coil L = 350 mH. In the simulation, only one type of interference is allowed to occur at a time. Their parameter settings are as follows:

[0075] 1) Early cable fault: The early cable fault is simulated using the Figure 2 shown fault equivalent model. The parameters are V p

[0076] = 3 ± 10% kV, V n = 2.5 ± 10% kV, R0 = 300 Ω, and R con = 5 Ω. The carbonized resistor R c ranges from 500 Ω to 10 kΩ.

[0077] 2) Non-fault interference: Two types of non-fault interference are used in the simulation: load switching and capacitor bank switching. For load switching, the load is set to 2.19 MW. For capacitor bank switching, the capacity is 6.42 MVAR.

[0078] B. Analysis of simulation results

[0079] Figure 13 shows the change trends of phase currents under different types of interference ((a) early cable fault, (b) capacitor switching, (c) load switching) and the calculation results of parameter a1. It can be clearly observed that when an early fault occurs, the peak value of the phase current shows a slow upward trend, resulting in a positive a1 value. During capacitor switching and load switching, the peak value of the phase current gradually decreases, resulting in a negative a1 value. Therefore, by calculating parameter a1, early faults can be detected.

[0080] Figure 14 (a) shows the phase voltage - phase current (V - A) characteristic curves under different carbonized resistor Rc values. As the severity of the fault increases, the curve shrinks inward and the nonlinearity gradually weakens. The calculated parameter A is as Figure 14 (b) shown. It can be observed that as the severity of the fault increases, the value of parameter A gradually decreases. Therefore, parameter A can be effectively used to diagnose the severity of early cable faults.

[0081] In summary, compared with the existing methods, the method proposed in the present invention takes into account the deterioration of cable insulation. The main conclusions are as follows:

[0082] 1) First, the physical evolution characteristics of the cable initial fault were analyzed, and based on the existing arc model based on diode resistance considering insulation degradation, a new cable initial fault model was established.

[0083] 2) The theoretical analysis of the transient equivalent circuit reveals the mapping relationship between the fault state and the fault current. This analysis produces obvious physical characteristics: accurate fault state determination is achieved through the nonlinear distortion characteristics of the phase voltage - phase current (V - A) characteristic curve.

[0084] 3) Simulations and experiments verify the effectiveness of the proposed method. By applying this method to industry, the monitoring of the cable early fault state can be realized, thus achieving effective cable system maintenance.

Claims

1. A method for monitoring early faults of the roof cable of a train unit, characterized in that: The following steps are involved: Step 1: Establish an equivalent circuit model of early cable failure, which includes a resistor R that represents insulation carbonization damage. c , R, which characterizes the arc resistance of the cable arc And R, which represents the grounding resistance of the copper shield con ; Analyze the influence of insulation carbonization resistance on the transient behavior of fault current through the equivalent circuit model; Step 2: Based on the equivalent circuit model, extract the phase current signal of the cable line, calculate the slope a1 of its amplitude change trend, and distinguish early faults from non-fault interference by the positive and negative values ​​of the slope a1; Step 3: Determine the fault state through the nonlinear distortion characteristics of the phase voltage-phase current characteristic curve, and use the closed area of ​​the phase voltage-phase current characteristic curve as an indicator to quantify the fault state.

2. The method for monitoring early faults of the EMU roof cable according to claim 1 is characterized in that: The equivalent circuit model in step 1 uses a diode-based resistance model to simulate the cable arc resistance R arc The diode-based resistor model includes two reverse-parallel diodes D p and D n , adjustable DC voltage source V p and V n And the current limiting resistor R o ; Two DC voltage sources V p and V n Connected to two diodes D p and D n ; The current limiting resistor R0 is used to simulate the amplitude of the arc current; the fault current i f The expression is Among them, V ph Indicates the phase voltage of the cable line.

3. The method for monitoring early faults of the EMU roof cable according to claim 1, characterized in that: The step 2 specifically includes: Step 2.1: De-noising the current signal to remove high-frequency noise; Step 2.2: Extract the positive peak current i within three power frequency cycles after the fault occurs ph (t i ), where t i Indicates peak time; Step 2.3: Take the natural logarithm of the peak current to obtain the sequence ln|i ph (t i )|; Obtain the slope a1 using linear fitting; The fitting equation is: y(t i )=ln|i ph (t i )|=a1t i +a0 Among them, y(t i ) is the predicted value, a0 is the intercept; Step 2.4: Construct the following objective function to determine the appropriate values ​​of a0 and a1 so that the positive peak current i ph (t i ) and the predicted value y(t i ) has the smallest error; Step 2.5: Calculate the slope a1 according to the following formula: in, is the average value of the peak time of three power frequency cycles, is the average value of the three power frequency cycle prediction values, When the slope a1>0, it is judged as an early fault, and when the slope a1<0, it is judged as non-fault interference.

4. The method for monitoring early faults of the EMU roof cable according to claim 1, characterized in that: The step 3 specifically includes: Step 3.1: Perform phase correction on the phase voltage V(t) and phase current I(t) signals to eliminate phase offset by zero-crossing alignment. Make the VA curve present linear characteristics; Step 3.2: Discrete the data points {(V i ,I i )} to perform numerical integration, and the formula for calculating the closed area A is: Where V i and I i are the voltage and current values ​​of the i-th data point, respectively, and n is the total number of data points; Step 3.3: Set the area threshold A th , when A≤A th It is judged that the insulation carbonization damage has entered a serious stage.

Images