Power distribution network high-resistance fault line selection method based on zero-sequence current integral offset

Abstract

The invention discloses a power distribution network high-resistance fault line selection method based on zero-sequence current integral offset, and belongs to the field of power system relay protection. The method comprises the following steps of: when a fault occurs, uploading collected zero-sequence current to a master station by measuring points distributed on each feeder line; the improved CEEMDAN is used for decomposing the zero sequence current; selecting IMF reconstruction fault waveforms meeting conditions; integral processing is carried out on the reconstructed waveform, and offset is calculated; and performing fault line selection according to the offset module value percentage.

Description

Technical Field

[0001] The invention relates to a method for selecting a high-resistance fault line in a distribution network based on zero-sequence current integral offset, and belongs to the technical field of power system relay protection. Background Art

[0002] In recent years, with the rapid development of renewable energy generation technologies, distributed power sources (such as photovoltaics and wind power) have been connected to distribution networks on a large scale, significantly changing the structure and operating characteristics of traditional distribution networks. The intermittent output of renewable energy, the high-frequency switching of power electronic equipment, and the widespread use of grid-connected inverters have led to increased distortion of current and voltage waveforms in distribution networks and a significant increase in high-frequency harmonic content. Furthermore, the decentralization of renewable energy access points complicates the distribution characteristics of fault currents, posing a significant challenge to traditional fault detection methods based on power frequency components. In particular, in high-impedance fault (HIF) scenarios, the fault current amplitude is weak and easily masked by harmonics and random fluctuations generated by renewable energy equipment, further increasing the difficulty of fault detection and line selection.

[0003] In addition, the measurement devices widely deployed in distribution networks are affected by environmental noise, electromagnetic interference, and communication transmission errors, and the collected zero-sequence current signals often contain a large amount of noise components. In the case of a high-resistance fault, the amplitude of the fault signal itself is small, close to or even lower than the amplitude of the measurement noise. As a result, traditional signal decomposition methods (such as empirical mode decomposition (EMD)) are prone to modal aliasing under noise interference, making it difficult to effectively extract fault features. Existing line selection methods are mostly based on steady-state power frequency components or single transient characteristics. In strong noise environments or non-stationary signal conditions caused by the access of new energy sources, their sensitivity and reliability are significantly reduced, which can easily lead to misjudgments or missed judgments, seriously affecting the efficiency of fault isolation and system recovery.

[0004] Therefore, there is an urgent need for a high-resistance fault line selection method that can overcome measurement noise interference and adapt to renewable energy access environments, thereby improving the accuracy and robustness of fault detection under complex operating conditions. This invention, through the coordinated optimization of signal decomposition-reconstruction-integral offset criteria, achieves precise high-resistance fault line selection in complex noise environments and renewable energy access scenarios, providing a reliable technical guarantee for the safe operation of distribution networks. Summary of the Invention

[0005] The purpose of the present invention is to address the defects of the existing technology in high-resistance fault detection, such as weak noise resistance, poor adaptability to new energy access scenarios and insufficient sensitivity, and to provide a distribution network high-resistance fault line selection method based on zero-sequence current integral offset. Through noise suppression, transient feature reconstruction and integral offset judgment, the method can achieve rapid and accurate identification of fault feeders under complex working conditions, providing reliable technical guarantee for the safe operation and maintenance of the distribution network.

[0006] The present invention adopts the following technical solutions.

[0007] The method for selecting high-resistance fault line in distribution network based on zero-sequence current integral offset includes:

[0008] Step 1: When a fault occurs, the measurement points distributed across the feeders upload the collected zero-sequence current to the master station;

[0009] Step 2: Use CEEMDAN to decompose the zero-sequence current and obtain multiple IMFs;

[0010] Step 3: Optimize each IMF and reconstruct the fault waveform;

[0011] Step 4: Integrate the reconstructed waveform and calculate the offset;

[0012] Step 5: Select the fault line based on the offset modulus percentage;

[0013] When a fault occurs in step 1, the measurement points distributed across the feeders upload the collected zero-sequence current to the master station:

[0014] When the busbar zero-sequence voltage exceeds the threshold, the fault recorders installed at each measuring point in the distribution network immediately activate and record the zero-sequence current for 10 power frequency cycles following the fault. The voltage threshold should be set based on the actual distribution network conditions. Typically, a fault is considered to have occurred when the busbar zero-sequence voltage exceeds 0.15 times its rated voltage. At this point, each measuring device will transmit zero-sequence current data.

[0015] In step 2, the improved CEEMDAN is used to decompose the zero-sequence current:

[0016] Step 2 includes:

[0017] Step 2.1, initialize the signal:

[0018] S0(t)=I(t)

[0019] Where I(t) is the zero-sequence current sampling signal, and S0(t) is the initialization signal, that is, the 0th layer residual signal.

[0020] Generate a noise operator:

[0021] W i (t)~N(0,σ 2 ),σ=ò·std(I)

[0022] Where W i (t) is the noise generation operator of the i-th group, ∈ is the noise intensity coefficient, and std(I) is the standard deviation of the original signal.

[0023] Step 2.2, first mode extraction:

[0024] The initialization signal S0(t) is divided into N groups, and random white noise W is added to each group. i (t), and get the mixed signal

[0025]

[0026] Using EMD method to analyze mixed signals Processing is performed to obtain the modal extraction operator

[0027]

[0028] Where EMD(·)1 means taking only the first-order IMF after EMD decomposition.

[0029] For N mode extraction operators Taking the average, we get the first intrinsic mode function IMF1(t):

[0030]

[0031] Step 2.3, residual recursion:

[0032] S k (t) = S k-1 (t)-IMF k (t)

[0033] Where S k (t) is the residual signal of the kth layer, that is, the remaining signal component after stripping off the extracted mode.

[0034] Step 2.4, deep modal mining:

[0035] In S k Continue to add white noise in (t) to obtain the k-th layer mixed signal:

[0036]

[0037] in, is the k-th layer modulated noise, After performing k-order EMD decomposition on the original noise, the sum of the first k-order IMFs is taken.

[0038] Generate k+1th order IMF:

[0039]

[0040] Termination detection:

[0041] If condition 1 is met:

[0042]

[0043] If yes, the decomposition will continue; otherwise, the decomposition will terminate.

[0044] At the same time, condition 2 is met:

[0045]

[0046] If yes, the decomposition will continue; otherwise, the decomposition will terminate.

[0047] Where, is the total energy of the original signal.

[0048] Step 3 optimizes each IMF and reconstructs the fault waveform:

[0049] Constraint 1, power frequency core bandwidth constraint:

[0050] f IMF ∈[40,60]Hz (±20% power frequency tolerance)

[0051] Constraint 2, dominant energy threshold constraint:

[0052] A IMF ≥0.3×A max

[0053] Where A IMF is the peak amplitude of IMF, A max The maximum peak amplitude among all IMFs.

[0054] All IMFs that simultaneously satisfy the above two constraints will reconstruct the zero-sequence current waveform:

[0055] I re (t) = ∑IMF obj

[0056] Where, I re (t) is the zero-sequence current waveform after reconstruction, IMF obj is the eigenmode function that satisfies the constraints.

[0057] Step 4 integrates the reconstructed waveform and calculates the offset:

[0058] Step 4 includes:

[0059] Step 4.1, reconstruct the zero sequence current I re (t) is integrated to obtain the integrated waveform K(t):

[0060]

[0061] Where t0 is the fault start time, and T is the power frequency period of the power grid, which is generally 20ms.

[0062] Step 4.2, calculate the offset:

[0063] Assume that the maximum value in the period [t0+7T, t0+8T] is M1 and the minimum value is m1, then the offset ΔQ'1 in this period is:

[0064]

[0065] Similarly, the offsets in the time periods [t0+8T, t0+9T] and [t0+9T, t0+10T] are ΔQ'2 and ΔQ'3 respectively.

[0066] Then the zero-sequence current offset ΔQ of the measuring point is:

[0067]

[0068] In step 5, the fault line is selected based on the offset modulus percentage:

[0069] Step 5 includes:

[0070] Step 5.1, calculate the offset modulus percentage:

[0071] If the feeder numbers are 1 to n, then the offset modulus percentage of each feeder is for:

[0072]

[0073] Step 5.2, fault line selection:

[0074] If there are at least 2 It is determined to be a bus fault; if max(|Q ％ |) except for the offset corresponding to Then determine max(|Q ％ The feeder corresponding to |) is the faulty feeder.

[0075] Principles of the method

[0076] High resistance fault equivalent circuit such as Figure 2 As shown. Among them, u f =-E f =-Usin(ωt+θ) is the virtual voltage of the fault phase, U is the virtual voltage amplitude of the fault phase, ω is the power frequency angular frequency, and θ is the initial phase angle of the fault. R≈3R f , R f is the transition resistance, i 0f is the zero sequence current at the fault point, i 0L is the zero sequence current of the arc suppression coil, C 0j is the zero-sequence capacitance of each feeder to ground, is the sum of the capacitance of the healthy feeder to ground and the zero-mode capacitance of the upstream fault feeder to ground, u0 is the busbar zero-sequence voltage, i 0j is the zero-sequence current at the outlet of feeder j, j=1,2,...,n-1, assuming that feeder numbered n is short-circuited, then i 0n is the zero-sequence current at the fault feeder outlet.

[0077] according to Figure 1 The equivalent circuit shown, combined with the initial fault conditions, can be derived as a second-order differential equation:

[0078]

[0079] Solving the differential equation given by formula (1), we can get two characteristic roots p1 and p2:

[0080]

[0081] Since R f Generally, it is several hundred to several thousand ohms. The system usually operates under overdamping conditions, so there is Therefore, formula (2) becomes:

[0082] p 1,2 =-λ±jω f (3)

[0083] in, is the attenuation factor, is the transient resonant frequency.

[0084] Further solving the differential equation given by formula (1), we can finally get the zero-sequence current of the arc suppression coil as:

[0085]

[0086] in:

[0087]

[0088] Formula (4) shows that the zero-sequence current of the arc suppression coil consists of two parts: one is the periodic component i that oscillates with equal amplitude at the power frequency ω. 0L_F , and secondly, according to the angular frequency ω f The transient component i of the oscillation decay 0L_T .

[0089] According to the zero-sequence current of the arc suppression coil in formula (4), the busbar zero-sequence voltage can be obtained:

[0090]

[0091] The zero-sequence current of the healthy feeder j can be obtained from formula (6):

[0092]

[0093] From equations (4) and (7), we can get the zero-sequence current of the fault feeder n:

[0094]

[0095] Integrate for a sound feeder:

[0096]

[0097] The waveform after integration is the bus voltage C 0j When τ is large enough, that is, when the system enters a steady state, the waveform oscillates sinusoidally around 0, so the offset is 0.

[0098] Integrate the faulted feeder:

[0099]

[0100] When τ is large enough, that is, when the system enters a steady state, the waveform is a superposition of a sine wave oscillating around 0 and a DC component. This DC component is caused by the zero-sequence current of the arc suppression coil and has nothing to do with transient parameters. Therefore, the offset of the fault feeder is the DC component D f :

[0101]

[0102] In summary, the integral offset of a healthy feeder is 0, while the integral offset of a faulty feeder is not 0. Therefore, the zero-sequence current integral offset can be used as a criterion for distinguishing a faulty feeder from a healthy feeder. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 Line selection algorithm flow chart

[0104] Figure 2 High resistance fault equivalent circuit

[0105] Figure 3 Radial Simulation Model of Distribution Network Based on PSCAD

[0106] Figure 4 Improved CEEMDAN decomposition waveform

[0107] Figure 5 Comparison of signal reconstructed waveform and original waveform

[0108] Figure 6 Multi-feeder waveform reconstruction

[0109] Figure 7 Schematic diagram of zero-sequence current integral offset of each feeder DETAILED DESCRIPTION

[0110] In order to verify the feasibility of this method, this paper uses PSCAD to conduct simulation tests and establishes a radial distribution network as shown in the following figure. Figure 3 As shown in the figure, the 110kV system is connected to the distribution network via a transformer with a transformation ratio of 110kV / 10.5kV. The high-voltage side of the transformer is connected in a star configuration, with the neutral point grounded via an arc suppression coil, overcompensated at 8% and an inductance of L = 0.46H. The distribution network feeder utilizes a hybrid cable and overhead line connection, totaling 68km. The cable is YJV22-3*400, with a total length of 34km, and the overhead line is JKLYJ-150, with a total length of 34km. At the end of the distribution network feeder, a 10kV / 0.4kV transformer is used to connect to the 1MW load. Furthermore, at terminals L2 and L4, 10kV / 0.69kV transformers are used to connect to photovoltaic power generation systems DG1 and DG2.

[0111] In order to verify the effectiveness of the improved CEEMD in reconstructing high-resistance fault signals in resonant grounding systems, the PSCAD electromagnetic transient simulation software was used. Figure 3 At point (a), a single-phase ground fault occurs with an initial phase angle of 45° and a transition resistance of 1800Ω. The measured zero-sequence current at the feeder outlet is as follows: Figure 4 As shown in the orange waveform, the fault signal is 5 power frequency cycles, i.e. 0.1s. In order to be closer to engineering practice, Gaussian white noise with a signal-to-noise ratio of 2.5dB is added to the original signal, as shown in Figure 5 As shown in the blue waveform. The improved CEEMDAN is used to decompose the zero-sequence current, and the final IMFs are as follows Figure 4 As shown, according to the optimization principle in step 3, IMF 5 and IMF 6 meet the requirements and are selected as the reconstructed signals. The final reconstructed waveform is as follows Figure 4 As shown in the orange waveform, the reconstructed waveform can effectively remove noise pollution, which is beneficial for subsequent processing steps.

[0112] exist Figure 3 At 3.5 km (a) from the start of feeder l1, a fault occurs on phase A with an initial phase angle of θ = 90° and a transition resistance of R f =2800Ω single-phase ground fault. Add noise with a signal-to-noise ratio of 2.5dB to the zero-sequence current measured at the beginning of each feeder to obtain the zero-sequence current containing noise, such as Figure 6 (a) is shown. The improved CEEMDAN is used to reconstruct the zero-sequence current of each feeder, and the results are as follows: Figure 6 (b) As shown. The reconstructed waveform of each feeder is integrated, and the offset of each feeder is finally Figure 7The offset modulus percentages of each feeder are 99.836%, 0.051%, 0.042%, 0.031%, and 0.040% respectively. According to the line selection principle: except for max(|Q ％ |) except for the offset corresponding to Therefore, feeder l1 is selected as the fault feeder, and the line selection is correct.

Claims

1. A method for selecting high-resistance fault lines in distribution networks based on zero-sequence current integral offset, characterized in that: The method comprises the following steps: Step 1: When a fault occurs, the measurement points distributed across the feeders upload the collected zero-sequence current to the master station; Step 2: Use CEEMDAN to decompose the zero-sequence current and obtain multiple IMFs; Step 3: Optimize each IMF and reconstruct the fault waveform; Step 4: Integrate the reconstructed waveform and calculate the offset; Step 5: Select the fault line based on the offset modulus percentage.

2. The method for selecting a high-resistance fault line in a distribution network based on zero-sequence current integral offset according to claim 1, characterized in that: When a fault occurs in step 1, the measurement points distributed across the feeders upload the collected zero-sequence current to the master station: When the busbar zero-sequence voltage value exceeds the limit, the fault recording device installed at each measuring point of the distribution network will start immediately to record the zero-sequence current within 10 power frequency cycles after the fault. The voltage limit value should be set according to the actual situation of the distribution network. Under normal circumstances, when the zero-sequence voltage of the busbar exceeds 0.15 times its rated voltage, it is considered that a fault has occurred. At this time, each measuring device will return the zero-sequence current data.

3. The method for selecting a high-resistance fault line in a distribution network based on zero-sequence current integral offset according to claim 1, characterized in that: In step 2, the improved CEEMDAN is used to decompose the zero-sequence current: Step 2 includes: Step 2.1, initialize the signal: S0(t)=I(t) Where I(t) is the zero-sequence current sampling signal, S0(t) is the initialization signal, that is, the 0th layer residual signal; Generate a noise operator: W i (t)~N(0,σ 2 ),σ=ò·std(I) Where W i (t) is the noise generation operator of group i, ∈ is the noise intensity coefficient, and std(I) is the standard deviation of the original signal; Step 2.2, first mode extraction: The initialization signal S0(t) is divided into N groups, and random white noise W is added to each group. i (t), and get the mixed signal Using EMD method to analyze mixed signals Processing is performed to obtain the modal extraction operator D1 (i) : Among them, EMD(·)1 means taking only the first-order IMF after EMD decomposition; For N mode extraction operators D1 (i) Taking the average, we get the first intrinsic mode function IMF1(t): Step 2.3, residual recursion: S k (t)=S k-1 (t)-IMF k (t) Where S k (t) is the residual signal of the kth layer, that is, the remaining signal component after stripping the extracted mode; Step 2.4, deep modal mining: In S k Continue to add white noise in (t) to obtain the k-th layer mixed signal: Among them, W~ i (k) (t) is the k-th layer modulated noise, W~ i (k) (t)=ò·EMD k (W i (t)), EMD k (·) is the sum of the first k-order IMFs after performing k-order EMD decomposition on the original noise; Generate k+1th order IMF: Termination detection: If condition 1 is met: Then continue to decompose, otherwise the decomposition is terminated; At the same time, condition 2 is met: Then continue to decompose, otherwise the decomposition is terminated; Where, is the total energy of the original signal.

4. The method for selecting a high-resistance fault line in a distribution network based on zero-sequence current integral offset according to claim 1, characterized in that: Step 3 optimizes each IMF and reconstructs the fault waveform: Constraint 1, power frequency core bandwidth constraint: f IMF ∈[40,60]Hz (±20% power frequency tolerance) Constraint 2, dominant energy threshold constraint: A IMF ≥0.3×A max Where A IMF is the peak amplitude of IMF, A max The maximum peak amplitude among all IMFs; All IMFs that simultaneously satisfy the above two constraints will reconstruct the zero-sequence current waveform: I re (t)=∑IMF obj Where, I re (t) is the zero-sequence current waveform after reconstruction, IMF obj is the eigenmode function that satisfies the constraints.

5. The method for selecting a high-resistance fault line in a distribution network based on zero-sequence current integral offset according to claim 1, characterized in that: Step 4 integrates the reconstructed waveform and calculates the offset: Step 4 includes: Step 4.1, reconstruct the zero-sequence current I re (t) is integrated to obtain the integrated waveform K(t): Where t0 is the fault start time, T is the power frequency period of the power grid, which is generally 20ms; Step 4.2, calculate the offset: Assume that the maximum value in the period [t0+7T, t0+8T] is M1 and the minimum value is m1, then the offset ΔQ1 in this period is: Similarly, the offsets in the time periods [t0+8T, t0+9T] and [t0+9T, t0+10T] are ΔQ'2 and ΔQ'3 respectively. Then the zero-sequence current offset ΔQ is:

6. The method for selecting a high-resistance fault line in a distribution network based on zero-sequence current integral offset according to claim 1, characterized in that: In step 5, the fault line is selected based on the offset modulus percentage: Step 5 includes: Step 5.1, calculate the offset modulus percentage: If the feeder numbers are 1 to n, then the offset modulus percentage of each feeder is for: Step 5.2, fault line selection: If there are at least 2 It is determined to be a bus fault; if max(|Q ％ |) except for the offset corresponding to Then determine max(|Q ％ The feeder corresponding to |) is the faulty feeder.

Images