A method for selecting single-phase ground fault lines in power grid based on transient energy and cross-correlation distance

Abstract

The present invention discloses a single-phase grounding fault line selection method for a distribution network based on transient energy and cross-correlation distance. Firstly, the fault characteristics of transient zero-sequence current are analyzed. Then, in order to completely retain the transient characteristics of the fault, a fundamental component elimination method is adopted to remove the fundamental component in the transient current. Then, based on the transient current after removing the fundamental component, three fault criteria are generated: kurtosis value, transient energy and cross-correlation distance between lines. Finally, in order to effectively integrate the three fault criteria, a Copula function distribution is introduced, and by calculating the joint probability distribution, the fault possibility of each line is quantitatively evaluated, and the line with the greatest fault possibility is determined as the fault line.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and more specifically, relates to a single-phase grounding fault line selection method for a distribution network based on transient energy and cross-correlation distance. Background Art

[0002] Among all power system faults, single-phase grounding faults account for more than 80%. When a single-phase grounding fault occurs, the installed relay protection device cannot be started due to the weak fault characteristics, which brings considerable difficulties to the detection of single-phase grounding faults and identification of faulty lines. In recent years, many single-phase grounding fault detection technologies have been proposed, but the detection results of these methods are still unsatisfactory. Therefore, in order to maintain the stability and safety of the distribution network, it is of great significance to accurately detect single-phase grounding faults.

[0003] At present, fault detection is mainly carried out by extracting the characteristics of transient zero-sequence current, including wavelet transform, empirical mode decomposition, traveling wave method and optimization algorithm. However, these methods have problems such as complex function selection, mode aliasing and weak anti-noise ability in practical applications. At the same time, although artificial intelligence-based detection technology has high accuracy under certain conditions, its disadvantage of relying on a large number of fault samples limits its actual promotion.

[0004] In order to solve the above problems, a detection method that can accurately capture fault characteristics under various grounding resistance conditions is urgently needed. The new method needs to take into account the fault feature extraction capability under complex network conditions and improve the reliability and adaptability of detection.

[0005] The present invention proposes a fault line detection method based on transient energy and cross-correlation distance. The method removes the fundamental component and combines transient feature fusion to construct a detection index including kurtosis value, transient energy and cross-correlation distance. In addition, the introduction of Copula function distribution to fuse multi-index information effectively improves the robustness of detection. Simulation and field tests verify the applicability and effectiveness of the method, which has important theoretical and practical significance for improving the accuracy of ground fault detection in power systems. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art. After analyzing the fault characteristics, a new detection method is proposed. The fundamental component in the fault current is removed by introducing the fundamental component removal technology, and the three fault indicators are fused by using the Copula function distribution, which effectively improves the accuracy and reliability of fault detection and can adapt to different fault conditions.

[0007] To achieve the above-mentioned object of the invention, the present invention is intended to be implemented through the following steps:

[0008] (1) Fault feature analysis;

[0009] Based on the neutral point non-effectively grounded distribution network, the zero-sequence equivalent circuit model is adopted. Figure 1 As shown, the analysis process of low-resistance and high-resistance grounding faults is simplified, and the simplified circuit diagram is as follows Figure 2 and Figure 3 shown.

[0010] (1.1) When a low-resistance ground fault occurs, the currents of the faulty line and the healthy line are calculated as follows:

[0011]

[0012] Among them U m is the phase voltage amplitude, C z is the sum of the equivalent capacitance of the line, ω0 is the fundamental frequency, δ=R f / (2L), is the fault angle,

[0013]

[0014] (1.2) When a high-resistance ground fault occurs, the zero-sequence current consists of a fundamental component and a transient component. The currents of the fault line and the healthy line are calculated as follows:

[0015] i 0n =i 0n_FFC +i 0n_TS +i 0n_SSC

[0016] i 0k =i 0k_FFC +i 0k_TS +i 0k_SSC

[0017] where i 0n_FFC and i 0k_FFC is the fundamental component, i 0n_TS and i 0k_TS is the transient component.

[0018]

[0019] in,

[0020]

[0021] Where δ = 1 / (2R f C z ),

[0022] (2) Fundamental component removal method

[0023] (2.1), frequency offset: multiply the analytical signal by Eliminating the influence of the fundamental component, the currents of the fault line and the healthy line under the above high-resistance grounding fault condition can be reorganized as:

[0024]

[0025] i 0n =x RB cosω0t-x IB sinω0t

[0026] +x RT (t)cosω f tx IT (t)sinω f t

[0027] in,

[0028] (2.2) Signal transformation: Apply Hilbert transform to the transient zero-sequence current to generate analytical signals of the fault line and the healthy line, and take the real part, which is calculated as follows:

[0029]

[0030] (3) Fault index construction: The present invention proposes three key fault indexes from the perspectives of kurtosis value, energy characteristics and waveform similarity to describe fault characteristics. The following are the specific steps of index construction:

[0031] (3.1) The change of zero-sequence current at the time of fault is approximately regarded as an impulse signal. The kurtosis can reflect the degree of mutation of an impulse signal. The kurtosis value is calculated as follows:

[0032]

[0033] Among them, k n is the kurtosis value of the nth feeder, i zj is the current amplitude of the nth feeder at time j, for i zj The mean of .

[0034] (3.2) Transient energy as the second fault indicator: Use the first-order cumulative generation operator to smooth the signal, reduce noise interference, and calculate the transient energy:

[0035]

[0036] E mean =(E1+E2+L+E n ) / n

[0037] E(j)=[E1 / E mean E2 / E mean LE n / E mean ]

[0038] The energy of the jth line can be calculated by The sum of the squares is obtained.

[0039] (3.3), the cross-correlation distance index, as the third fault index:

[0040] (3.3.1) Based on the polarity of the fault signal, the bubble sorting method is introduced to enhance the difference between the healthy line and the faulty line. The application of the bubble sorting method is as follows: When a single-phase grounding fault occurs, if the transient zero-sequence current i zj The initial extreme value γ of (t) j , then we can use γ j The sign of i zj (t) is reordered, and its expression is as follows:

[0041] S j (t) = [i 0j (1),i 0j (2),i 0j (3),L,i 0j (r)]

[0042] The sorting principle is as follows: when γ j >0 hours:

[0043] i 0j (1)≥i 0j (2)≥i 0j (3)≥L≥i 0j (r)

[0044] When γ j <0 hours:

[0045] i 0j (1)≤i 0j (2)≤i 0j (4)≤L≤i 0j (r)

[0046] Where r is the number of sampling points. In fact, γ j Indicates the current flow direction when a single-phase grounding fault occurs. In other words, the γ j Symbols and Health Lines of γ j The signs are opposite.

[0047] (3.3.2) Calculate the cross-correlation distance matrix between the i-th feeder and the j-th feeder:

[0048]

[0049] At this time D ij is an n×n matrix with all diagonal elements being 0.

[0050] Finally, remove the maximum value D in each row ij,max and minimum value D ij,min , define the comprehensive cross-correlation distance matrix CD j as follows:

[0051]

[0052] (4) Fault detection method fusion: The present invention combines three indicators and integrates them through the Copula function:

[0053] The Copula function is a multivariate cumulative distribution function, that is, the marginal distribution of multiple variables is characterized as a joint distribution of multivariate random variables through Sklar's theorem. The present invention adopts a semi-parametric Copula model to quantitatively represent the possibility of faulty lines, integrates the above three indicators, and evaluates the failure probability of each line.

[0054] (4.1) Integrate fault indicators and determine the fault line:

[0055] F X (x1,L,x N )=C X [F1(X1),L,F N (X N )]

[0056] In the formula, C X (·) is the Copula function, which is essentially [0,1] n A joint distribution function on .

[0057] (4.2) Select the appropriate semi-parametric model through the Bayesian information criterion:

[0058]

[0059] Among them, c k (·) represents the Copula function C X The density of (·), d1,L,d k Fitting values ​​to variables To fit the marginal distribution of the parameters, the model that minimizes the Bayesian information criterion is the optimal Copula function. The boundary conditions for parameter estimation are: In this case, the parameter θ of the coupled model is approximated by the maximum semiparametric logarithm:

[0060]

[0061] In a single-phase grounding fault, the probability of feeder failure can be converted into the joint probability distribution of the above three fault indicators, and the line with the highest probability is regarded as the fault line.

[0062] The present invention achieves its purpose through the following steps: first, the fault characteristics of transient zero-sequence current are analyzed; then, in order to completely retain the transient characteristics of the fault, the fundamental component in the transient current is removed by the fundamental component elimination method; then, based on the transient current after removing the fundamental component, three fault criteria are generated: kurtosis value, transient energy, and cross-correlation distance between lines. In order to effectively integrate these three fault criteria, the Copula function distribution is introduced, and the fault probability of each line is quantitatively evaluated by calculating the joint probability distribution. Finally, the line with the greatest fault probability is determined as the fault line.

[0063] In addition, the single-phase grounding fault line detection method based on transient energy and cross-correlation distance of the present invention has the following significant advantages:

[0064] 1. Efficiently capture fault characteristics: By removing the fundamental component and simplifying the transient zero-sequence current, the two major fault indicators of kurtosis value, transient energy and cross-correlation distance are calculated, and a detection criterion that can adapt to various fault conditions is constructed.

[0065] 2. Fault indicator fusion optimization: The Copula function distribution is introduced to integrate fault indicators. Its joint probability distribution can quantitatively reflect the fault possibility of each line and effectively improve the detection accuracy.

[0066] 3. Verification of wide applicability: The proposed fault line detection method has been implemented and verified in both the grid-connected mode and the island mode of the distribution network, proving its feasibility and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is the zero-sequence equivalent circuit adopted by the present invention;

[0068] Figure 2 It is to analyze the zero-sequence equivalent circuit under the simplified low-resistance fault condition of ground fault;

[0069] Figure 3 It is to analyze the zero-sequence equivalent circuit under the simplified high-resistance fault condition of ground fault;

[0070] Figure 4 It is a simulation model; DETAILED DESCRIPTION

[0071] The specific implementation of the present invention is described below in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0072] Example

[0073] In a novel single-phase grounding fault line selection method for distribution network based on transient energy and cross-correlation distance of the present invention, four steps are mainly included: S1, fault feature analysis, S2, fundamental component removal, S3, fault index construction, S4, fault detection method fusion;

[0074] In this embodiment, a typical radial distribution network is first selected as the first embodiment scenario: a typical radial distribution network model containing 4 lines and 2 distributed power sources is built in PSCAD, such as Figure 4 As shown below, we combine Figure 4 The model elaborates on the above four steps, including the following steps:

[0075] S1. Fault feature analysis;

[0076] In this embodiment, when a single ground fault occurs at line n of the compensated distribution network, its simplified zero-sequence equivalent circuit is as follows: Figure 1 As shown. Petersen coils are used in the network to compensate for the adverse effects on the fault current. When a low-resistance ground fault occurs, Figure 1 The circuit in can be simplified to Figure 2 In the circuit; when the grounding resistance R f When it is larger, a high resistance fault occurs. Figure 1 The circuit in can be simplified to Figure 3 The circuit in.

[0077] S1.1. Calculate the current of the faulty line and the healthy line in the case of a low-resistance ground fault as follows:

[0078]

[0079] S1.2. Calculate the current of the faulty line and the healthy line in case of high-resistance grounding fault as follows:

[0080] i 0n =i 0n_FFC +i 0n_TS

[0081] i 0k =i 0k_FFC +i 0k_TS

[0082]

[0083] S2. Fundamental component removal method

[0084] S2.1, multiply the analytical signal by Eliminate the influence of fundamental component

[0085]

[0086] i 0n =x RB cosω0t-x IB sinω0t

[0087] +x RT (t)cosω f tx IT (t)sinω f t

[0088] S2.2. Apply Hilbert transform to the transient zero-sequence current of the line and take the real part, and calculate as follows:

[0089]

[0090] S3. Fault index construction: The present invention proposes two key fault indicators from the perspectives of energy characteristics and waveform similarity to describe fault characteristics. The following are the specific steps for indicator construction:

[0091] S3.1. Calculate the kurtosis value:

[0092]

[0093] Among them, k n is the kurtosis value of the nth feeder, i zj is the current amplitude of the nth feeder at time j, for i zj The mean of .

[0094] S3.2. Calculate transient energy:

[0095]

[0096] S3.3. Calculate the cross-correlation distance index:

[0097] S3.3.1. Based on γ j The sign of i zj (t) is reordered, and its expression is as follows:

[0098] S j (t) = [i 0j (1),i 0j (2),i 0j (3),L,i0j (r)]

[0099] The sorting principle is as follows: when γ j >0 hours:

[0100] i 0j (1)≥i 0j (2)≥i 0j (3)≥L≥i 0j (r)

[0101] When γ j <0 hours:

[0102] i 0j (1)≤i 0j (2)≤i 0j (4)≤L≤i 0j (r)

[0103] Where r is the number of sampling points. In fact, γ j Indicates the current flow direction when a single-phase ground fault occurs. In other words, the γ j Symbols and Health Lines of γ j The signs are opposite.

[0104] S3.2.2. Calculate the cross-correlation distance matrix between the ith feeder and the jth feeder:

[0105]

[0106] At this time D ij is an n×n matrix with all diagonal elements being 0.

[0107] Finally, remove the maximum value D in each row ij,max and minimum value D ij,min , define the comprehensive cross-correlation distance matrix CD j as follows:

[0108]

[0109] S4. Fault detection method fusion: The present invention combines three indicators and integrates them through Copula function distribution:

[0110] S4.1. Application of Copula function distribution: The transient energy index E(j) is used as a positive evaluation index and is calculated as follows:

[0111] E mean =(E1+E2+L+E n ) / n

[0112] E(j)=[E1 / E mean E2 / E mean LEn / E mean ]

[0113] The energy of the jth line can be calculated by The sum of the squares is obtained.

[0114] Cross-correlation distance index CD j Is used as a negative evaluation indicator and is calculated as follows:

[0115]

[0116] S4.2. Integrate fault indicators and determine the fault line:

[0117] S4.2.1. Calculate the Copula function:

[0118] F X (x1,L,x N )=C X [F1(X1),L,F N (X N )]

[0119] In the formula, C X (·) is the Copula function, which is essentially [0,1] n A joint distribution function on .

[0120] S4.2.2. Selecting an appropriate semiparametric model using the Bayesian information criterion:

[0121]

[0122] Among them, c k (·) represents the Copula function C X The density of (·), d1,L,d k Fitting values ​​to variables To fit the marginal distribution of the parameters, the model that minimizes the Bayesian information criterion is the optimal Copula function. The boundary conditions for parameter estimation are: In this case, the parameter θ of the coupled model is approximated by the maximum semiparametric logarithm:

[0123]

[0124] In a single-phase grounding fault, the probability of feeder failure can be converted into the joint probability distribution of the above three fault indicators, and the line with the highest probability is regarded as the fault line.

[0125] In this embodiment, a typical radial distribution network model with 6 feeders and 2 distributed generation sources is established in PSCAD, as shown in FIG. Figure 4As shown. DG1 is a wind farm and DG2 is a photovoltaic farm. Both are connected to the radial distribution network through switch relays S1 and S2, and the distributed power control strategy considers the impact of low voltage ride-through. The line parameters are shown in Table 1. The arc suppression coil adopts 10% overcompensation mode, so the arc suppression coil parameters are inductance L p =0.3481H, resistance R p =20Ω; sampling frequency is 10kHz.

[0126] Table 1 shows the main parameters of the distribution network

[0127]

[0128] Table 1

[0129] Table 2 Results of fault line selection in the scenario of wind farm and photovoltaic power plant being connected together.

[0130]

[0131] Table 2

[0132] Since kurtosis, transient energy and cross-correlation distance are typical difference characteristics between faulty lines and sound lines, by integrating these characteristics, the comprehensive fault degree can be effectively estimated based on the Copula theory. In order to illustrate the effectiveness and applicability of the single fault grounding line selection method proposed by the present invention, a typical radial distribution network model is used to analyze its line selection results, as shown in Table 2, which verifies the applicability and robustness of the proposed method under various network conditions and fault scenarios. Regardless of low-resistance faults or high-resistance faults, the detection method shows high accuracy and efficiency, indicating that it has a wide range of practical application potential.

Claims

1. A method for selecting a single-phase ground fault line in a distribution network based on transient energy and cross-correlation distance, comprising the following steps: (1) Calculation of current of each line in the distribution network under different fault characteristics; (1.1) When a low-resistance ground fault occurs, the currents of the faulty line and the healthy line are calculated as follows: Among them, U m is the phase voltage amplitude, C z is the sum of the equivalent capacitance of the line, ω0 is the fundamental frequency, δ=R f / (2L), is the fault angle, (1.2) When a high-resistance ground fault occurs, the zero-sequence current consists of a fundamental component and a transient component. The currents of the fault line and the healthy line are calculated as follows: i 0n =i 0n_FFC +i 0n_TS +i 0n_SSC i 0k =i 0k_FFC +i 0k_TS +i 0k_SSC Among them, i 0n_FFC and i 0k_FFC is the fundamental component, i 0n_TS and i 0k_TS is the transient component; in, Where, δ = 1 / (2R f C z ), (2) Remove the fundamental component; (2.1), frequency offset: multiply the analytical signal by Eliminating the influence of the fundamental component, the currents of the fault line and the healthy line under the above high-resistance grounding fault condition can be reorganized as: i 0n =x RB cosω0t-x IB sinω0t+x RT (t)cosω f t-x IT (t)sinω f t in, (2.2) Signal transformation: Apply Hilbert transform to the transient zero-sequence current to generate analytical signals of the fault line and the healthy line, and take the real part, which is calculated as follows: (3) Constructing failure indicators; (3.1) Calculate the kurtosis value of the zero-sequence current at the time of fault: Among them, k n is the kurtosis value of the nth feeder, i zj is the current amplitude of the nth feeder at time j, for i zj The mean of (3.2) Calculate transient energy: THE mean \(E1+E2+L+E n ) / n E(j)=[E1 / E mean E2 / E mean LE n / AND mean ] The energy of the jth line can be calculated by The sum of the squares is obtained; (3.3) Calculate the cross-correlation distance index: (3.3.1) Based on the polarity of the fault signal, the bubble sorting method is introduced. The application of the bubble sorting method is as follows: When a single-phase grounding fault occurs, if the transient zero-sequence current i zj The initial extreme value γ of (t) j , then we can use γ j The sign of i zj (t) is reordered, and its expression is as follows: S j (t)=[i 0j (1),i 0j (2),i 0j (3),L,i 0j (r)] The sorting principle is as follows: when γ j >0 hours: i 0j (1)≥i 0j (2)≥i 0j (3)≥L≥i 0j (r) When γ j <0 hours: i 0j (1)≤i 0j (2)≤i 0j (4)≤L≤i 0j (r) Where r is the number of sampling points; In fact, γ j Indicates the current flow direction when a single-phase grounding fault occurs, that is, the γ of the fault line j Symbols and Health Lines of γ j The signs are opposite; (3.3.2) Calculate the cross-correlation distance matrix between the i-th feeder and the j-th feeder: At this time D ij is an n×n matrix with all diagonal elements being 0; Finally, remove the maximum value D in each row ij,max and minimum value D ij,min , define the comprehensive cross-correlation distance matrix CD j as follows: (4) Fault detection fusion through Copula function: (4.1) Integrate fault indicators and determine the fault line: F X (x1,L,x N )=C X [F1(X1),L,F N (X N )] In the formula, C X (·) is the Copula function, which is essentially [0,1] n A joint distribution function on ; (4.2) Select the appropriate semi-parametric model through the Bayesian information criterion: Among them, c k (·) represents the Copula function C X The density of (·), d1,L,d k Fitting values ​​to variables To fit the marginal distribution of the parameters, the model that minimizes the Bayesian information criterion is the optimal Copula function. The boundary conditions for parameter estimation are: In this case, the parameter θ of the coupled model is approximated by the maximum semiparametric logarithm: In a single-phase grounding fault, the probability of feeder failure can be converted into the joint probability distribution of the above three fault indicators, and the line with the highest probability is regarded as the fault line.

Images