A short-circuit fault area positioning method for grid-connected mode alternating current microgrid

Abstract

The application provides a short-circuit fault area positioning method for grid-connected mode AC micro-grid, which is used for positioning the short-circuit fault area of the grid-connected mode AC micro-grid with different capacities. The method comprises an approximate model establishing method and a fault area positioning method. A rapid calculation model for positioning the short-circuit fault area is constructed by an OLS-RBF neural network algorithm. When the current wavelet energy spectrum value measured at a detection point of the micro-grid is greater than a fault detection threshold, it is determined that a short-circuit fault occurs on the feeder. The first peak value of the fault current wavelet energy spectrum waveform, the micro-grid capacity at the moment and the detected fault initial phase angle are substituted into a fault area positioning formula corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, so that the short-circuit fault area of the micro-grid is determined. The application is suitable for the short-circuit fault area of the grid-connected mode AC micro-grid with different capacities, and can make the grid-connected mode AC micro-grid reliably operate and realize selective protection.

Description

Technical Field

[0001] This invention relates to the field of power grid operation and maintenance technology, and in particular to a method for locating short-circuit fault areas in grid-connected AC microgrids. Background Technology

[0002] With the increasing proportion of distributed generation capacity in power systems, microgrids are considered an effective means to improve the overall efficiency of energy utilization. If a short-circuit fault occurs on a microgrid line and is not cleared quickly, the distributed generation in the microgrid will cause stability problems. Therefore, accurate regional location of short-circuit faults in microgrids is an important direction for microgrid protection. Some scholars have applied wavelet energy spectrum algorithms to the early detection and location of short-circuit faults in microgrids, but the following problems exist in practical research: 1. Insufficient accuracy of regional location when the microgrid feeder length is long; 2. The fault area location method has limitations in adapting to capacity changes.

[0003] The study found that the size of the first peak of the fault current wavelet energy spectrum waveform is E4. max1 It decreases as the fault distance d increases, for different initial phase angles θ of the fault. c E4 max1 There is a different linear relationship between θ and d. Existing fault region localization methods require obtaining a specific θ. c Next E4 max1 The relationship curve between d and d, and the conversion factor K. r Normalization is performed, and the normalized curve relationship is piecewise fitted using the least squares method to obtain different θ values. c Next E4 max1 The formula relating θ to d. When the microgrid line is long, the same θ c K below r The differences between the values ​​are significant, and the calculated fault distance d is insufficient to meet the accuracy requirements for fault area location. Furthermore, when considering the impact of capacity variations, it is necessary to obtain values ​​for different capacities and different θ values. c Next E4 max1 The relationship curve between the fault distance d and d greatly increases the complexity of the fitting curve calculation process, thus limiting the practical application of this method. Since the RBF neural network performs excellently in handling multi-dimensional nonlinear problems, this scheme chooses to use the OLS-RBF neural network algorithm to calculate the fault distance d and the DG capacity S. DG Initial phase angle θ of the fault c E4 max1 An analytical relationship is established between them, and a method for locating short-circuit fault areas in AC microgrids with different capacity grid-connected modes is proposed, providing an effective method for reliable operation and selective protection of AC microgrids under grid-connected modes. SUMMARY

[0004] The application provides a short-circuit fault area positioning method for grid-connected mode AC micro-grid, which is suitable for short-circuit fault area positioning of grid-connected mode AC micro-grid with different capacities, and can make the grid-connected mode AC micro-grid reliably operate and realize selective protection.

[0005] The application adopts the following technical scheme.

[0006] The application provides a short-circuit fault area positioning method for grid-connected mode AC micro-grid, which is suitable for short-circuit fault area positioning of grid-connected mode AC micro-grid with different capacities, and can make the grid-connected mode AC micro-grid reliably operate and realize selective protection.

[0007] In the approximate model establishing method, the RBF neural network algorithm is composed of input layer, hidden layer and output layer neurons; wherein, the input layer is used to receive training set data of the training neural network; the implicit layer maps the low-dimensional linear non-separable problem of the input layer to the high-dimensional space through the radial basis function, so that it is linearly separable, and the mapping process of the radial basis function is expressed as formula one

[0008]

[0009] φ i is the base function of the i th hidden layer neuron, c i is the center point of the i th hidden layer neuron, σ i is the center point width of the i th hidden layer neuron, ||x-c i || represents the Euclidean distance between the sample point and the center point; when the sample point corresponding to the fault current wavelet energy spectrum value is closer to the center point c i , the output of the implicit layer is larger. The RBF neural network is a local approximation neural network, and the Gaussian function in the sample space is only non-zero value in a limited range; the last layer is the output layer, and the linear weight connection is used between the output layer and the implicit layer, which is expressed as formula

[0010]

[0011] y is the actual value of the prediction point, is the output value of the RBF neural network, ε is the error between the actual value and the predicted value, for the sampling points in the training set, the value of this item is 0, λ i is the weight value;

[0012] The OLS-RBF neural network selects the center point of the radial basis function by the orthogonal least squares method, and uses the obtained center point and the set center point width to determine the weight value connecting the hidden layer to the output layer; from formula one and formula two,

[0013] The radial basis function center point c i , the radial basis function center point width σ i and the output weight λ i determine the accuracy of the neural network model.

[0014] In the micro-grid, for different fault initial phase angles θ c , the first peak value E4 max1 of the fault current wavelet energy spectrum waveform decreases with the increase of the distance d from the short-circuit point to the detection point, and the positioning method approximates the fault distance d by E4' max1 measured at the detection point, to realize the identification of the short-circuit fault area.

[0015] In training the OLS-RBF neural network algorithm, the distributed power capacity S DG , the fault initial phase angle θ c , and the first peak value E4 max1 of the fault current wavelet energy spectrum waveform are used as variables, and the distance d from the short-circuit point to the detection point is used as the target function; when a single-phase ground fault occurs, the required number of E4' max1 values are obtained as sample points, and most of the sample points are selected as the training set, and the rest of the sample points are the test set; the radial basis function center point c i , the output weight λ i and the set center point width obtained by training are substituted into formula two to obtain the functional relationship between the distance d from the short-circuit point to the detection point and the distributed power capacity S DG , the fault initial phase angle θ c , and the first peak value E4 max1 of the fault current wavelet energy spectrum waveform, as shown in the following formula three, i.e. the fault area positioning formula

[0016]

[0017] In the micro-grid, under different distributed power capacities, the maximum value of the current wavelet energy spectrum under non-fault operation is smaller than that under fault operation, and the current size under normal operation is different when the power capacity is different;

[0018] In the fault area location method, the fault judgment threshold is set to N times the value of the current wavelet energy spectrum during normal operation. When a short-circuit fault is judged to have occurred in the feeder, the microgrid protection system substitutes the magnitude of the first peak of the fault current wavelet energy spectrum waveform, the current microgrid capacity, and the detected initial phase angle of the fault into the fault area location formula corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, thereby realizing the area location of short-circuit faults in microgrids under different capacities.

[0019] In the fault area location method, the fault judgment threshold is set to five times the current wavelet energy spectrum value when the microgrid is operating normally.

[0020] The fault area location method specifically includes the following steps;

[0021] Step S1: After receiving the current signal, the protection system detects and updates the zero-crossing information of the current signal, and at the same time performs wavelet energy spectrum transformation on the current signal at the fourth scale to obtain the wavelet energy spectrum value of the current.

[0022] Step S2: Compare the wavelet energy spectrum value of the current with the fault judgment threshold. When the wavelet energy spectrum value of the current is less than the fault judgment threshold, it is considered that no short circuit fault has occurred on the feeder of the microgrid, and return to the previous step to continue to perform zero-crossing detection and wavelet energy spectrum transformation on the current. When the wavelet energy spectrum value of the current is greater than the fault judgment threshold, it is considered that a short circuit fault has occurred on the feeder of the microgrid.

[0023] The fault area location method also includes a short-circuit fault type identification method, which uses the current and its wavelet energy spectrum value to determine the fault type. The protection system needs to detect the initial phase angle of the short-circuit fault and identify the short-circuit fault type. The detection of the initial phase angle of the short-circuit fault is based on the time difference between the moment the short-circuit fault occurs and the moment the current crosses zero. That is, the ratio of the time difference between the two to 10ms is equal to the ratio of the initial phase angle of the fault to 180 degrees.

[0024] In the short-circuit fault type identification method, the protection system determines the ground fault by detecting residual current. Specifically: if the current wavelet energy spectrum value of only one phase containing residual current is greater than the fault determination threshold, it is a single-phase ground fault; if the current wavelet energy spectrum value of only two phases containing residual current is greater than the fault determination threshold, it is a two-phase ground fault; if the current wavelet energy spectrum value of only two phases without residual current is greater than the fault determination threshold, it is a two-phase short circuit; if the current wavelet energy spectrum value of all three phases without residual current is greater than the fault determination threshold, it is a three-phase short circuit.

[0025] The protection system delays 1ms after determining the fault initial phase angle and the fault type by the short-circuit fault type identification method to detect the first peak value of the wavelet energy spectrum waveform of the fault current, then substitutes the first peak value of the wavelet energy spectrum waveform of the fault current, the micro-grid capacity at this time and the detected fault initial phase angle into the fault area positioning formula three corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, so as to determine the area where the short-circuit fault is located, then disconnects the circuit breaker of the corresponding area, and completes the positioning of the short-circuit fault area and eliminates the short-circuit fault area.

[0026] The present application has the advantages of:

[0027] 1. The regional positioning method using the wavelet energy spectrum characteristics combined with the OLS-RBF neural network algorithm solves the problem of the decline of the regional positioning accuracy when the feeder length is long.

[0028] 2. The method can realize the adaptive adjustment when the capacity changes, and realizes the rapid and accurate positioning of the short-circuit fault area of the grid-connected mode alternating current micro-grid under different capacities. BRIEF DESCRIPTION OF DRAWINGS

[0029] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0030] Figure 1 is an OLS-RBF neural network model flowchart; Figure 1 Figure 2 is a fault area positioning flowchart.

[0031] Figure 2 Figure 1 is an OLS-RBF neural network model flowchart; DETAILED DESCRIPTION

[0032] As shown in the figure, a grid-connected mode alternating current micro-grid short-circuit fault area positioning method is used for positioning the short-circuit fault area of the grid-connected mode alternating current micro-grid under different capacities, which comprises an approximate model establishment method and a fault area positioning method, and a rapid calculation model for the short-circuit fault area positioning is constructed by the OLS-RBF neural network algorithm, when the wavelet energy spectrum value of the current measured at the detection point of the micro-grid is greater than the fault detection threshold, it is determined that a short-circuit fault occurs on the feeder, the first peak value of the wavelet energy spectrum waveform of the fault current, the micro-grid capacity at this time and the detected fault initial phase angle are substituted into the fault area positioning formula corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, so as to determine the area where the short-circuit fault of the micro-grid occurs.

[0033] ​The approximate model establishing method, the RBF neural network algorithm is composed of input layer, hidden layer and output layer neurons; wherein, the input layer is used for receiving training set data of training neural network; the implicit layer maps the low-dimensional linear inseparable problem of the input layer to high-dimensional space through radial basis function, makes it linearly separable, the mapping process of the radial basis function is expressed as formula one

[0034]

[0035] φ i is the base function of the i th hidden layer neuron, c i is the center point of the i th hidden layer neuron, σ i is the center point width of the i th hidden layer neuron, ||x-c i || represents the Euclidean distance between the sample point and the center point; when the sample point corresponding to the fault current wavelet energy spectrum value is closer to the center point c i , the output of the hidden layer is larger. The RBF neural network is a local approximation neural network, the Gaussian function in the sample space is only non-zero value in a limited range; the last layer is the output layer, the linear weight connection is used between the output layer and the hidden layer, which is expressed as formula two

[0036]

[0037] y is the actual value of the prediction point, is the output value of the RBF neural network, ε is the error between the actual value and the predicted value, the value of the item is 0 for the sampling point in the training set, λ i is the weight value;

[0038] The OLS-RBF neural network selects the center point of the radial basis function through the orthogonal least square method, and determines the weight value connecting the hidden layer to the output layer using the obtained center point and the set center point width; from formula one and formula two,

[0039] the radial basis function center point c i , the radial basis function center point width σ i and the output weight value λ i determine the accuracy of the neural network model.

[0040] In the microgrid, for different fault initial phase angles θ c , the first peak value E4 max1 of the fault current wavelet energy spectrum waveform decreases with the increase of the distance d from the short-circuit point to the detection point, the positioning method reflects the fault distance d by E4' max1 measured at the detection point, and realizes the identification of the short-circuit fault area.

[0041] In the training of the OLS-RBF neural network algorithm, the distributed power capacity S DG , the fault initial phase angle θ c , the first peak value E4 of the fault current wavelet energy spectrum waveform, and the distance d from the short-circuit point to the detection point are used as variables, and the distance d from the short-circuit point to the detection point is used as the target function. max1 When a single-phase ground fault occurs, a required number of E4 max1 values are obtained as sample points, and most of the sample points are selected as a training set, and the rest of the sample points are a test set. i The radial basis function center point c i , the output weight λ DG , and the set center point width are substituted into Formula Two to obtain the functional relationship between the distance d from the short-circuit point to the detection point and the distributed power capacity S c , the fault initial phase angle θ max1 , and the first peak value E4 of the fault current wavelet energy spectrum waveform, as shown in the following Formula Three, that is, the fault area positioning formula

[0042]

[0043] In the microgrid, under different distributed power capacities, the maximum value of the current wavelet energy spectrum value under non-fault operation is less than the current wavelet energy spectrum value under fault operation, and the current size under normal operation is different when the power capacity is different.

[0044] In the fault area positioning method, the fault judgment threshold is set to N times the current wavelet energy spectrum value under normal operation; after judging that a short-circuit fault occurs in the feeder, the protection system of the microgrid substitutes the first peak value of the fault current wavelet energy spectrum waveform, the microgrid capacity at this time, and the detected fault initial phase angle into the fault area positioning formula corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, thereby realizing the positioning of the short-circuit fault area of the microgrid under different capacities.

[0045] In the fault area positioning method, the fault judgment threshold is set to five times the current wavelet energy spectrum value under normal operation of the microgrid.

[0046] The fault area positioning method specifically includes the following steps.

[0047] Step S1, after obtaining the current signal, the protection system detects and updates the zero-crossing point information of the current signal, and performs wavelet energy spectrum transformation under the fourth scale to obtain the wavelet energy spectrum value of the current;

[0048] Step S2: Compare the wavelet energy spectrum value of the current with the fault judgment threshold. When the wavelet energy spectrum value of the current is less than the fault judgment threshold, it is considered that no short circuit fault has occurred on the feeder of the microgrid, and return to the previous step to continue to perform zero-crossing detection and wavelet energy spectrum transformation on the current. When the wavelet energy spectrum value of the current is greater than the fault judgment threshold, it is considered that a short circuit fault has occurred on the feeder of the microgrid.

[0049] The fault area location method also includes a short-circuit fault type identification method, which uses the current and its wavelet energy spectrum value to determine the fault type. The protection system needs to detect the initial phase angle of the short-circuit fault and identify the short-circuit fault type. The detection of the initial phase angle of the short-circuit fault is based on the time difference between the moment the short-circuit fault occurs and the moment the current crosses zero. That is, the ratio of the time difference between the two to 10ms is equal to the ratio of the initial phase angle of the fault to 180 degrees.

[0050] In the short-circuit fault type identification method, the protection system determines the ground fault by detecting residual current. Specifically: if the current wavelet energy spectrum value of only one phase containing residual current is greater than the fault determination threshold, it is a single-phase ground fault; if the current wavelet energy spectrum value of only two phases containing residual current is greater than the fault determination threshold, it is a two-phase ground fault; if the current wavelet energy spectrum value of only two phases without residual current is greater than the fault determination threshold, it is a two-phase short circuit; if the current wavelet energy spectrum value of all three phases without residual current is greater than the fault determination threshold, it is a three-phase short circuit.

[0051] After determining the initial phase angle and fault type using the short-circuit fault type identification method, the protection system delays for 1ms to detect the first peak value of the wavelet energy spectrum waveform of the fault current. Then, it substitutes the first peak value of the wavelet energy spectrum waveform of the fault current with the current microgrid capacity and the detected initial phase angle of the fault into the fault area location formula three corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, thereby determining the area where the short-circuit fault is located. Then, it disconnects the circuit breaker in the corresponding area to complete the location of the short-circuit fault area and eliminate the short-circuit fault area.

[0052] In this example, when training the neural network, a European 400V standard low-voltage microgrid is used as an example. To ensure the accuracy of the model, S is taken as... DG ∈{63,74.5,86,97.5,109,120.5,132}, θ c Given d ∈ {0, 30, 60, 90, 120, 150} and d ∈ {70, 140, 280, 490, 560}, we obtain a total of 7 × 6 × 5 = 210 E4' values ​​when a single-phase-to-ground short circuit occurs. max1 The total number of sample points is 210. 180 sample points are randomly selected as the training set for the OLS-RBF neural network algorithm, and the remaining 30 sample points are used as the test set.

Claims

1. A method for short-circuit fault area location of grid-connected AC microgrid, for locating short-circuit fault area of grid-connected AC microgrid with different capacity, characterized in that: The method comprises an approximate model establishing method and a fault area positioning method, a rapid calculation model for short-circuit fault area positioning is constructed by an OLS-RBF neural network algorithm, when the current wavelet energy spectrum value measured at a micro-grid detection point is greater than a fault detection threshold, it is determined that a short-circuit fault occurs on a feeder, the first peak value of the fault current wavelet energy spectrum waveform, the micro-grid capacity at this time and the fault initial phase angle detected are substituted into a fault area positioning formula corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, so that the area where the short-circuit fault of the micro-grid occurs is determined. In the approximate model establishing method, the RBF neural network algorithm is composed of input layer, hidden layer and output layer neurons; wherein, the input layer is used to receive the training set data for training the neural network; the hidden layer maps the low-dimensional linearly inseparable problem of the input layer to a high-dimensional space through a radial basis function, so that it is linearly separable, and the mapping process of the radial basis function is expressed as formula one. φ i is the basis function of the i-th hidden layer neuron, c i is the center point of the i-th hidden layer neuron, σ i is the center point width of the i-th hidden layer neuron, ||x-c i || denotes the Euclidean distance between the sample point and the center point; When the sample point corresponding to the fault current wavelet energy spectrum value is closer to the center point c i , the output of the hidden layer is larger. The RBF neural network is a local approximation neural network. The Gaussian function in the sample space is only non-zero within a limited range. The last layer is the output layer. The output layer and the hidden layer are connected using linear weights. y is the actual value of the prediction point, y is the output value of the RBF neural network, ε is the error between the actual value and the predicted value, and the value of this item is 0 for the sampling points in the training set, λ i is the weight; The OLS-RBF neural network selects the center point of the radial basis function by the orthogonal least squares method, and uses the obtained center point and the set center point width to determine the weight value for connecting the hidden layer to the output layer. As can be seen from formula one and formula two, radial basis function center point c i radial basis function center point width σ i and output weight λ i determine the accuracy of the neural network model.

2. The method of claim 1, wherein the method is characterized by: In the micro-grid, for different fault initial phase angle θ c , the first peak value size E4 of the fault current wavelet energy spectrum waveform max1 decreases with the increase of the distance d from the short-circuit point to the detection point, and the positioning method realizes the identification of the short-circuit fault area by detecting E4' max1 , which reflects the fault distance d and is approximated.

3. The method of claim 2, wherein: In the training of OLS-RBF neural network algorithm, the distributed power capacity S DG , the fault initial phase angle θ c , the first peak value E4 of the fault current wavelet energy spectrum waveform max1 are taken as variables, and the distance d from the short-circuit point to the detection point is taken as the target function. When a single-phase ground fault occurs, the required number of E4 max1 values are obtained as sample points, and most of the sample points are selected as the training set, and the remaining sample points are the test set. The center point c of the radial basis function obtained during training i Output weight λ i Substituting the set center point width into Formula 2, we obtain the distance d from the short-circuit point to the detection point and the distributed power supply capacity S. DG Initial phase angle θ of the fault c The first peak value E4 of the fault current wavelet energy spectrum waveform max1 The functional relationship between them is shown in Formula 3 below, which is the fault area location formula.

4. The method of claim 3, wherein: In the micro-grid, under different distributed power capacities, the maximum value of the current wavelet energy spectrum value under the non-fault operation is less than the current wavelet energy spectrum value under the fault operation, and the current size under the normal operation is different when the power capacity is different. In the fault area positioning method, the fault judgment threshold is set as N times of the current wavelet energy spectrum value under the normal operation; after judging that a short-circuit fault occurs on the feeder, the first peak value of the fault current wavelet energy spectrum waveform, the micro-grid capacity at this time and the fault initial phase angle detected are substituted into the fault area positioning formula corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, so that the area positioning of the micro-grid short-circuit fault under different capacities is realized.

5. The method of claim 4, wherein: In the fault area positioning method, the fault judgment threshold is set as five times of the current wavelet energy spectrum value under the normal operation of the micro-grid.

6. The method of claim 4, wherein: The fault area positioning method specifically comprises the following steps. Step S1, after obtaining the current signal, the protection system detects and updates the zero-crossing point information of the current signal, and performs wavelet energy spectrum transformation under the fourth scale to obtain the current wavelet energy spectrum value; Step S2, the current wavelet energy spectrum value is compared with the fault judgment threshold, when the current wavelet energy spectrum value is less than the fault judgment threshold, it is considered that the short-circuit fault does not occur on the feeder of the micro-grid, and the zero-crossing point detection and wavelet energy spectrum transformation of the current are continued; when the current wavelet energy spectrum value is greater than the fault judgment threshold, it is considered that the short-circuit fault occurs on the feeder of the micro-grid.

7. The method of claim 6, wherein the method is characterized by: The fault area positioning method further comprises a short-circuit fault type identification method, which utilizes the current and its wavelet energy spectrum value to determine the fault type, and the protection system needs to detect the initial phase angle of the short-circuit fault and identify the short-circuit fault type, and the initial phase angle of the short-circuit fault is referenced by the time difference between the time when the short-circuit fault occurs and the time when the current zero-crosses, that is, the ratio of the time difference between the two to 10 ms is equal to the ratio of the fault initial phase angle to 180 degrees.

8. The method of claim 7, wherein the method is characterized by: In the short-circuit fault type identification method, the protection system judges the ground fault by detecting the residual current; specifically, if the residual current is contained and only one-phase current wavelet energy spectrum value is greater than the fault determination threshold, it is a single-phase ground short-circuit; if the residual current is contained and only two-phase current wavelet energy spectrum values are greater than the fault determination threshold, it is a two-phase ground short-circuit; if the residual current is not contained and only two-phase current wavelet energy spectrum values are greater than the fault determination threshold, it is a two-phase inter-phase short-circuit; and if the residual current is not contained and three-phase current wavelet energy spectrum values are greater than the fault determination threshold, it is a three-phase short-circuit.

9. The method of claim 8, wherein: After the protection system determines the fault initial phase angle and the fault type by the short-circuit fault type identification method, it delays 1 ms to detect the first peak value of the wavelet energy spectrum waveform of the fault current; then, the first peak value of the wavelet energy spectrum waveform of the fault current, the micro-grid capacity at this time and the detected fault initial phase angle are substituted into the fault area positioning formula three corresponding to the fault type to calculate the distance from the short-circuit point to the detection point, so as to judge the area where the short-circuit fault is located, then disconnect the circuit breaker of the corresponding area, complete the positioning of the short-circuit fault area and exclude the short-circuit fault area.

Images