A fast line selection method for small current grounding faults

Abstract

The present invention discloses a rapid line selection method for low-current grounding faults, belonging to the field of power fault processing technology. It solves the problems of low line selection accuracy and slow line selection speed in the existing single-phase grounding fault line selection technology. The present invention includes the following steps: Step 1: Collecting electrical information; Step 2: Performing signal processing to extract steady-state and transient fault characteristics; Step 3: Obtaining harmonic amplitudes; Calculating the transient zero-sequence active power integral component; Obtaining line energy based on variational modal decomposition; Normalizing the data to obtain a sample data set; Step 4: Using the sample data set obtained in Step 3 for training, obtaining the trained optimal neural network model, and inputting test samples into the trained model to determine whether a low-current grounding fault has occurred. The present invention improves the computational efficiency of low-current grounding fault assessment and analysis in the distribution network, improves the efficiency of rapid line selection for low-current grounding faults, and enhances the support capacity of distribution network operations.

Description

Technical Field

[0001] The present invention relates to the technical field of power fault processing, and in particular to a fast line selection method for a small current grounding fault. Background Art

[0002] As a network directly distributing electricity to consumers, the safety and reliability of the distribution network plays a vital role in the production and development of the national economy. Currently, low-current grounding systems are the predominant grounding method in distribution networks. Fault line selection in these systems has always been a challenging problem in power systems. Installing a fault line selection device for low-current grounding systems at the station end utilizes transient signal characteristics to select lines, ensuring the reliability of the distribution network. However, the short duration of transient processes makes fault signal extraction difficult. Line selection methods that integrate multiple fault characteristics can overcome the shortcomings of single-line selection methods. After obtaining system data characteristics at the time of the fault, existing data optimization algorithms such as BP neural networks and support vector machines suffer from slow iteration speeds and are prone to falling into local optimality, resulting in low line selection accuracy and slow speed. Low line selection accuracy and slow speed can cause the system to remain in a faulty state for extended periods. The ground current generated by single-phase grounding can form arcs at the grounding point. If not cleared promptly, this can damage equipment, causing economic losses and even safety incidents. Therefore, there is an urgent need to improve the accuracy and speed of line selection. Summary of the Invention

[0003] In view of the problems existing in the prior art, the present invention provides a fast line selection method for a small current grounding fault, which aims to solve the problems of low line selection accuracy and slow line selection speed in the prior art for single-phase grounding fault line selection.

[0004] The technical solution adopted in the present invention is as follows:

[0005] A fast line selection method for a small current grounding fault includes the following steps:

[0006] Step 1: After a low-current grounding system fault occurs, use a fault acquisition device to record various post-fault electrical information. Extracting the zero-sequence current and busbar zero-sequence voltage signals for each line is as follows: After a low-current grounding system fault occurs, use a fault acquisition device to record various post-fault electrical information, including the steady-state zero-sequence current signals of each branch feeder, as well as the transient zero-sequence current signals and busbar zero-sequence voltage signals.

[0007] Step 2: After obtaining the zero-sequence current and zero-sequence voltage signals, signal processing should be performed to extract the steady-state and transient fault characteristics of the line.

[0008] Step 3: Perform fast Fourier transform on the extracted steady-state component of zero-sequence current to obtain the amplitude of the fifth harmonic component I of the jth line in steady state j250HZ The ratio of the fifth harmonic component amplitude of each line to the sum of the fifth harmonic component amplitudes of all lines is taken as the fault feature F1.

[0009]

[0010] In formula (1), I j,250HZ is the amplitude of the fifth harmonic component of line j, It is the sum of the amplitudes of the fifth harmonic components of all lines.

[0011] The zero-sequence voltage and zero-sequence current components in transient state are multiplied and then integrated to amplify the characteristic value, which is helpful to distinguish the characteristic value difference between the fault line and the normal line. The calculation formula is as follows:

[0012]

[0013] In formula (2), V(t) is the zero-sequence voltage at time t, I(t) is the zero-sequence active current at time t, and T represents the integration period, which is taken as one cycle after the fault.

[0014] Furthermore, the fault characteristic data is normalized as follows: Assuming that the transient zero-sequence active power integral component of the j-th line is P j , the ratio of this value to the sum of the integral components of all lines is taken as the fault characteristic F2, and the calculation formula is:

[0015]

[0016] In formula (3), P j is the transient zero-sequence active power integral value of line j, It is the sum of the integral values ​​of the transient zero-sequence active power of all lines.

[0017] To obtain line energy based on VMD, with the decomposition number K set to 2, VMD decomposition is performed on the extracted line transient zero-sequence current signal. The line energy is calculated after VMD decomposition. The ratio of this line energy value to the sum of all line energy values ​​is used as the fault signature value F3.

[0018] Furthermore, based on VMD to obtain line energy, the extracted transient zero-sequence current signal is decomposed by VMD. When the decomposition number K is 2, the zero-sequence current is decomposed into two IMF functions, which are recorded as d 1(t) , d 2(t) , the frequency band energy of the j-th IMF component is:

[0019]

[0020] In formula (4), k is the number of sampling points, n is the length of the sampling data, and j = 1, 2. The energy value corresponding to the line (outlet l) is obtained by adding the frequency band energies of the two IMF components.

[0021] Add the frequency band energies of the two IMF components to obtain the energy value corresponding to the line (outlet l):

[0022] E l =E1+E2 (5)

[0023] Furthermore, the ratio of the energy value of line 1 to the sum of the energy values ​​of all lines is used as the energy proportion value, which is used as the third fault characteristic value F3 when the line fails:

[0024]

[0025] In formula (6), E l is the energy value of line l, is the sum of the energy values ​​of all lines.

[0026] Step 4: Process the fault feature quantities F1, F2, and F3 according to the optimized BP neural network and select the line.

[0027] Furthermore, before using the whale algorithm to optimize the BP neural network, the whale algorithm itself is optimized to enhance its global search diversity, including chaos mapping and inertia weight optimization. The optimization steps are as follows:

[0028] The cubic map chaos operation is used as the chaos mapping to improve the initial population of the system. The expression of the cubic map chaos mapping is:

[0029]

[0030] In formula (7), ρ is the control parameter, which is set to 1; y k is the whale population sequence after the kth iteration. Next, the inertia weight formula is introduced to balance the global search capability and the local search capability of the whale algorithm in the later stage. The inertia weight formula is:

[0031] ω(t)=ω min +(ω max -ω min )×m×exp(-t / M) (8)

[0032] In formula (8), ω(t) is the inertia weight value after t iterations, m is the adjustment coefficient, and ω min and ω max are the initial minimum and maximum weights respectively, and M is the maximum number of iterations.

[0033] Furthermore, the optimized whale algorithm (CI-WOA) is used to optimize the BP neural network. By optimizing the whale algorithm, the initial weights and thresholds of the BP neural network (CI-WOA-BP) are optimized, ensuring that the BP neural network has the optimal initial weights and thresholds. This can address the problem of the BP neural network easily falling into local optimality. The fault characteristics obtained above are then input into the optimized BP neural network as training sample data, and the optimal CI-WOA-BP neural network model is obtained through training. The characteristic state of the line to be tested is then fed into the optimized CI-WOA-BP neural network model to determine whether a low-current ground fault has occurred.

[0034] In summary, the beneficial effects of the present invention are as follows:

[0035] The CI-WOA-BP neural network model of the present invention can overcome the shortcomings and limitations of the previous single BP neural network. Its optimized algorithm can solve the problems that the BP neural network is prone to falling into local optimality and slow iteration speed. After a fault occurs in a low-current grounding system, it can be used to improve the accuracy and speed of fault line selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention will now be described by way of example with reference to the accompanying drawings, in which:

[0037] Figure 1 Schematic diagram of the method flow of the present invention;

[0038] Figure 1: Ground fault in the line; 2: Extracting zero-sequence current and bus zero-sequence voltage signals from each line; 3: Extracting steady-state transient fault features; 4: Obtaining the fifth harmonic amplitude of the line using FFT; 5: Obtaining the transient zero-sequence active power integral; 6: Obtaining line energy based on VMD; 7: Obtaining a sample set through data normalization; 8: Training set; 9: Chaotic mapping equation; 10: Nonlinear adaptive weight; 11: Optimizing the initial population and inertia weight of the WOA algorithm (CI-WOA); 12: CI-WOA optimizing the BP neural network; 13: Training to obtain the optimal CI-WOA-BP neural network model; 14: Testing set; 15: Testing the trained model; 16: Determining whether a small current ground fault has occurred. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0040] In the description of the embodiments of the present application, it should be noted that the terms "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings, or are the orientation or positional relationship in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, they cannot be understood as limiting the present application. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0041] The following combination Figure 1 The present invention is described in detail.

[0042] Example:

[0043] A fast line selection method for a small current ground fault of the present invention is implemented as follows:

[0044] like Figure 1 As shown in the figure, 1 is a ground fault in the line, and 2 is the extraction of zero-sequence current signals of each line and zero-sequence voltage signals of the bus: when a ground fault is detected in the distribution network, the zero-sequence current signals of each line and the zero-sequence voltage signals of the bus should be collected immediately using the acquisition device.

[0045] 3 is the steady-state transient fault feature extraction: After collecting the zero-sequence current and zero-sequence voltage signals, these signals should be processed to obtain the line characteristic value input. There are three characteristic inputs obtained, namely the line's fifth harmonic amplitude ratio, the transient zero-sequence active power integral ratio, and the VMD-based line energy ratio. At the same time, the line characteristic value output is set to 0 or 1, where the output 0 indicates that the line is a normal line, and 1 indicates that the line is a fault line, which is convenient for subsequent line selection.

[0046] 4 is FFT to obtain the fifth harmonic amplitude of the line: Use the acquisition device to collect the steady-state zero-sequence current signals of each line after the fault, perform FFT transformation on each signal, and obtain the fifth harmonic amplitude. At the same time, the ratio of the fifth harmonic component amplitude of each line to the sum of the fifth harmonic component amplitudes of all lines is used as the first fault feature F1:

[0047]

[0048] In formula (1), I j,250HZ is the amplitude of the fifth harmonic component of line j, It is the sum of the amplitudes of the fifth harmonic components of all lines.

[0049] 5 is to obtain the transient zero-sequence active power integral: after the transient zero-sequence current of each line and the zero-sequence voltage of the bus are extracted by the acquisition device, the zero-sequence current of each line is multiplied by the zero-sequence voltage of the bus and then integrated, that is:

[0050]

[0051] In formula (2), V(t) is the zero-sequence voltage at time t, I(t) is the zero-sequence active current at time t, and T represents the integration period, which is taken as one cycle after the fault.

[0052] In this way, the transient zero-sequence active power integral value of each line after the fault is obtained, and the ratio of the transient zero-sequence active power integral value of each line to the sum of the transient zero-sequence active power integral values ​​of all lines is used as the second fault characteristic value F2:

[0053]

[0054] In formula (3), P j is the transient zero-sequence active power integral value of line j, It is the sum of the integral values ​​of the transient zero-sequence active power of all lines.

[0055] 6 is to obtain line energy based on VMD: after decomposing the zero-sequence current into two IMF functions, denoted as d 1(t) , d 2(t) , the frequency band energy of the j-th IMF component is:

[0056]

[0057] In formula (4), k is the number of sampling points, n is the length of the sampling data, and j = 1, 2. The energy value corresponding to the line (outlet l) is obtained by adding the frequency band energies of the two IMF components:

[0058] E l =E1+E2 (5)

[0059] The ratio of the energy value of line 1 to the sum of the energy values ​​of all lines is taken as the energy weight value, which is used as the third fault characteristic value F3 when the line fails:

[0060]

[0061] In formula (6), E l is the energy value of line l, is the sum of the energy values ​​of all lines.

[0062] 7 is data normalization processing to obtain a sample set: After the above method obtains the data features, due to the differences between the various data features, the data needs to be normalized. The normalization benchmark is the sum of the corresponding feature values ​​of each line.

[0063] 8 is the training set: After the data set is normalized, a part of the normalized data set should be divided into a training set. The amount of data in the training set accounts for 80% of all data and is used for subsequent neural network training to train and obtain the optimal model for line selection.

[0064] 9 is the chaotic mapping equation: This invention uses an improved whale algorithm to optimize the initial weights and thresholds of the BP neural network. Chaotic variables are used to generate a large number of diverse chaotic initial populations. Then, populations with better fitness values ​​are selected from these populations as the initial populations of the whale algorithm to improve search efficiency. In the chaotic mapping formula, the cubic map chaotic operation formula is selected to improve the initial population of the whale algorithm. The expression of the cubic map chaotic mapping is:

[0065]

[0066] In formula (7), ρ is the control parameter, which is set to 1; y k is the whale population sequence after the kth iteration,

[0067] 10 is the nonlinear inertia adaptive weight: For the whale algorithm, the inertia weight has a great impact on the convergence speed and global optimization ability. Secondly, the inertia weight formula is introduced to balance the global search ability and the local search ability of the whale algorithm in the later stage. The inertia weight formula is:

[0068] ω(t)=ω min +(ω max -ω min )×m×exp(-t / M) (8)

[0069] In formula (8), ω(t) is the inertia weight value after t iterations, m is the adjustment coefficient, and ω min and ω maxare the initial minimum and maximum weights, respectively, and M is the maximum number of iterations. From formula (8), we can see that in the early stage of the iteration, as the number of iterations t remains at a small value, the adaptive weight ω remains at a high value, which can increase the global search ability of the algorithm; as t increases in the later stage, ω gradually decreases, which can improve the local optimization ability of the algorithm.

[0070] Figure 11 is to optimize the initial population and inertia weight of the WOA algorithm (CI-WOA): The initial population and inertia weight of the WOA algorithm are optimized using the chaos mapping equation and inertia weight, resulting in the optimized whale algorithm (CI-WOA).

[0071] 12. CI-WOA optimizes BP neural network: The optimized whale algorithm (CI-WOA) is used to optimize the initial weights and thresholds of the BP neural network to form a CI-WOA-BP model, which accelerates the training accuracy and speed of the neural network.

[0072] 13 is to obtain the optimal CI-WOA-BP neural network model: the training set in the sample is input into the CI-WOA-BP neural network for training, and the optimal network model is obtained through the training convergence of the neural network.

[0073] 14 is the test set: the normalized data set is divided into a training set and a test set. The training set has been used for the previous neural network training, and the remaining data features are used as the test set.

[0074] 15 is to test the trained model: the test set is input into the optimal CI-WOA-BP neural network model obtained by the previous training for testing to evaluate the reliability of the previously optimized network model.

[0075] 16 is to determine whether a small current ground fault occurs: the characteristic state of the line to be tested is sent to the optimized CI-WOA-BP neural network model to determine whether a small current ground fault occurs.

[0076] The above-described embodiments merely represent specific implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of protection of the present application. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the technical concept of the present application, and all such variations and improvements fall within the scope of protection of the present application.

Claims

1. A fast line selection method for a small current ground fault, characterized in that: The following steps are involved: Step 1: When a fault occurs in the low-current grounding system, use the fault collection device to record the electrical information of each line after the fault in various situations; Step 2: After obtaining the zero-sequence current and zero-sequence voltage signals of each line based on step 1, perform signal processing to extract steady-state and transient fault characteristics; Step 3: Use FFT transform to extract the amplitude of the fifth harmonic component of each line, and use the ratio of the fifth harmonic component amplitude to the sum of the fifth harmonic component amplitudes of all lines as the fault feature F1; Obtain the transient zero-sequence active power integral component, that is, multiply the zero-sequence voltage and zero-sequence current components during the transient state and then integrate them; the ratio of this value to the sum of the integral components of all lines is used as the fault characteristic value F2; Perform VMD decomposition on the extracted transient zero-sequence current signal. When the number of decompositions K is 2, the transient zero-sequence current signal of the i-th line is decomposed by VMD to calculate the energy ratio of the i-th line to the total energy of all lines, and obtain the fault feature F3. Normalize the fault feature quantities F1, F2, and F3 to obtain a sample data set; Step 4: Optimize the CI-WOA-BP neural network model parameters and train it using the sample data set obtained in step 3 to obtain the optimized CI-WOA-BP neural network model. Send the line characteristic state to be tested into the optimized CI-WOA-BP neural network model to determine whether a small current ground fault occurs. Specifically, first, use the cubic map chaos operation as the chaos map to improve the initial population method of the whale algorithm. The expression of the cubic map chaos map is: In formula (7), ρ is the control parameter, which is set to 1; y k is the whale population sequence after the kth iteration. Next, the inertia weight formula is introduced to balance the global search capability and the local search capability of the whale algorithm in the later stage. The inertia weight formula is: ω(t)=ω min +(ω max -oh min )×m×exp(-t / M) (8) In formula (8), ω(t) is the inertia weight value after t iterations, m is the adjustment coefficient, and ω min and ω max are the initial minimum and maximum weights respectively, and M is the maximum number of iterations; It can be seen from formula (8) that in the early stage of iteration, the weight value ω(t) is large, which accelerates the convergence speed in the early stage; in the late stage of iteration, the weight value ω(t) is small, which increases the search accuracy in the late stage; The optimized whale algorithm (CI-WOA) is used to optimize the BP neural network. The initial weights and thresholds of the BP neural network are continuously optimized by optimizing the whale algorithm, so that the BP neural network has the optimal initial weights and thresholds (CI-WOA-BP). Then, part of the acquired fault characteristic values ​​are input into the optimized BP neural network as training sample data for training, and the optimal CI-WOA-BP neural network model is obtained through training. The characteristic state of the line to be tested is sent to the optimized CI-WOA-BP neural network model to determine whether a small current grounding occurs.

2. The method for rapid line selection for a small current ground fault according to claim 1, characterized in that: The step 1 is specifically as follows: the acquisition device records the electrical information after various fault conditions, including the zero-sequence current signal of each branch feeder in steady state, as well as the zero-sequence current signal and bus zero-sequence voltage signal in transient state.

3. The method for rapid line selection for a small current grounding fault according to claim 1, characterized in that: The specific acquisition process of the fault characteristic quantity F1 in step 3 is: using the acquisition device to collect the zero-sequence current steady-state signal of each line after the fault, performing FFT transformation on each signal to obtain the fifth harmonic amplitude, and at the same time taking the ratio of the fifth harmonic component amplitude of each line to the sum of the fifth harmonic component amplitudes of all lines as the fault characteristic quantity F1.

4. A fast line selection method for a small current grounding fault according to claim 1 or 3, characterized in that: The specific mathematical expression of the fault characteristic value F1 is: In formula (1), I j,250HZ is the amplitude of the fifth harmonic component of line j, It is the sum of the amplitudes of the fifth harmonic components of all lines.

5. The method for rapid line selection for a small current grounding fault according to claim 1, characterized in that: The specific acquisition process of the fault characteristic value F2 in step 3 is as follows: after the transient zero-sequence current of each line and the zero-sequence voltage of the bus are extracted by the acquisition device, the zero-sequence current and zero-sequence voltage of each line are multiplied and then integrated to obtain the transient zero-sequence active power integral value of the line; the ratio of the transient zero-sequence active power integral value of each line to the sum of the transient zero-sequence active power integral values ​​of all lines is used as the fault characteristic value F2.

6. A fast line selection method for a small current grounding fault according to claim 5, characterized in that: The specific mathematical expression for multiplying and integrating the zero-sequence active current and zero-sequence voltage of each line is: In formula (2), V(t) is the zero-sequence voltage at time t, I(t) is the zero-sequence active current at time t, and T represents the integration period, which is taken as one cycle after the fault.

7. A fast line selection method for a small current grounding fault according to claim 1 or 5, characterized in that: The specific mathematical expression of the fault characteristic value F2 is: In formula (3), P j is the transient zero-sequence active power integral value of line j, It is the sum of the integral values ​​of the transient zero-sequence active power of all lines.

8. The method for rapid line selection for a small current grounding fault according to claim 1, characterized in that: The specific acquisition process of the feature quantity F3 in step 3 is as follows: perform VMD decomposition on the extracted transient zero-sequence current signal. When the decomposition number K is 2, the zero-sequence current is decomposed into two IMF functions, which are denoted as d1(t) and d2(t). The frequency band energy of the j-th IMF component is: In formula (4), k is the number of sampling points, n is the length of the sampling data, and j = 1, 2. The energy value corresponding to the line is obtained by adding the frequency band energies of the two IMF components. THE l =E1+E2 (5) The ratio of the energy value of each line to the sum of the energy values ​​of all lines is taken as the energy proportion value, which is used as the fault characteristic value F3 when the line fails.

9. A fast line selection method for a small current grounding fault according to claim 1 or 8, characterized in that: The specific mathematical expression of the fault characteristic value F3 is: In formula (6), E l is the energy value of line l, is the sum of the energy values ​​of all lines.

Images