A method for identifying distribution network line parameters based on fault diagnosis

Abstract

The present invention provides a method for identifying distribution network line parameters based on fault diagnosis, comprising denoising and supplementing missing values ​​of real-time voltage and power signals; constructing a feature vector of historical measurement data; simultaneously normalizing the feature vector of historical measurement data and the feature vector of real-time measurement data; dividing the normalized historical measurement data vectors of the head and terminal points of the distribution network into a training set and a test set; training a GA-Elman neural network; inputting the test set into the trained GA-Elman neural network to achieve fault diagnosis; identifying the line parameters of the entire normally operating distribution network using the least squares method; and calculating the accuracy of the distribution network identification line parameters. The present invention solves the problem that the use of line parameters with errors for distribution network analysis applications without considering the impact of actual operating conditions and line faults on actual parameters will result in a large discrepancy between the calculated results and the actual values.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network fault diagnosis, and in particular relates to a distribution network line parameter identification method based on fault diagnosis. Background Art

[0002] With the access of large-scale distributed power sources, user-side energy storage, and electric vehicles to the distribution network, the operation of the distribution network has become more complicated, and faults occur frequently, requiring research on fault diagnosis technology. At the same time, accurate line parameters are very important for distribution network analysis applications such as power flow calculation, line loss calculation, and relay protection setting.

[0003] Traditional fault analysis systems and traditional distribution network fault diagnosis methods rely on a combination of relay protection and manual line inspections. Specifically, after a distribution line's relay protection device activates, power line inspectors are notified to locate the fault along the tripped distribution line. These traditional fault diagnosis methods suffer from significant delays and errors, making real-time performance and accuracy difficult to guarantee, hindering timely fault diagnosis and troubleshooting. Furthermore, the determination of distribution network field parameters often relies on empirical data and classical mathematical models, failing to consider the impact of actual operating conditions and line faults on these parameters. For example, line parameters are affected by factors such as distribution network operating conditions, line temperature, line environment, and overvoltage. This can lead to numerical errors between the line parameters stored in the distribution management system and the actual line parameters. This is especially true after a distribution network fault, where line parameters vary significantly. Using these erroneous line parameters for distribution network analysis can result in significant discrepancies between the calculated results and the actual values. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for identifying distribution network line parameters based on fault diagnosis.

[0005] The present invention provides a method for identifying distribution network line parameters based on fault diagnosis, comprising:

[0006] Collect real-time voltage and power signals at the beginning and end points of each line in the distribution network;

[0007] Perform denoising and missing value supplementation on real-time voltage and power signals to obtain real-time measurement data feature vectors;

[0008] Obtain historical voltage and power signals at the head and end points of the distribution network;

[0009] Construct feature vectors of historical measurement data;

[0010] Perform normalization processing on the historical measurement data feature vector and the real-time measurement data feature vector simultaneously;

[0011] Divide the normalized historical measurement data vector into multiple groups, and use the genetic algorithm to calculate the optimal weight of each group. Take the average value of the optimal weights of multiple groups as the initial weight of the GA-Elman neural network;

[0012] Divide the historical measurement data vector of the head and tail points of the distribution network after normalization into a training set and a test set; the historical measurement data vector is used as the input of the GA-Elman neural network, and the number corresponding to the fault composite determination result is used as the output of the GA-Elman neural network. Select the training set and combine it with the initial weight of the GA-Elman neural network to train the GA-Elman neural network;

[0013] Input the test set into the trained GA-Elman neural network to achieve fault diagnosis;

[0014] Draw the structure of the distribution network after fault diagnosis to obtain the distribution network structure diagram, which includes the fault node, the fault determination result and the fault area;

[0015] Use the least squares method to identify the line parameters of the entire distribution network operating normally;

[0016] Calculate the accuracy of the line parameters identified by the distribution network.

[0017] Further, the denoising and missing value supplementation processing of the real-time voltage signal and power signal to obtain the real-time measurement data feature vector includes:

[0018] Classify the measured data such as the collected voltage signal and power signal into three categories: active power, reactive power and voltage amplitude corresponding to different nodes;

[0019] Arrange the measured data such as active power, reactive power and voltage amplitude in ascending order in their respective categories;

[0020] Assume that the data anomaly point is x d , calculate the average value avg of all data;

[0021] Calculate the estimator y of the arithmetic mean and standard deviation;

[0022] Calculate the statistic g d :

[0023]

[0024] Compare g d with the g(a,n) obtained from the critical value table of the Grubbs test method. If g d < g(a,n), then the data point x d is not an outlier; if g d>g(a,n), then the data point x d is an outlier;

[0025] The same type mean interpolation method is used to determine the data category based on the data that has been classified in the denoising process, and the mean of the samples in each category is used to interpolate the missing values.

[0026] Furthermore, the normalization processing of the historical measurement data feature vector and the real-time measurement data feature vector simultaneously includes:

[0027] The historical measurement data feature vector and the real-time measurement data feature vector are normalized simultaneously according to the following formula:

[0028]

[0029] Among them, T qh 、 are respectively the normalized historical measurement data vector and real-time measurement data vector; U qh 、P qh , Q qh are the historically measured voltage vector, active power vector, and reactive power vector; U qc 、P qc , Q qc They are respectively the normalized real-time measured voltage vector, active power vector, and reactive power vector;

[0030] Furthermore, the composite fault determination result includes normal state, general fault, intermediate fault and severe fault.

[0031] Furthermore, the identification of line parameters of the entire normally operating distribution network using the least squares method includes:

[0032] Construct the distribution network line parameter identification equation, where the distribution network node s is the starting node; node m is the intermediate node; and the terminal node t k ,k=1,2,…,n;the power flow of the distribution network is from node s to the terminal node t k Outflow; the line parameters to be identified are resistance and reactance; for the intermediate node m to the terminal node t k Any line in satisfies the following formula at time i:

[0033]

[0034] in, is the voltage of node m at time i; is node t at time i k voltage; From the intermediate node m to the terminal node t at time i k The current on this line; From the intermediate node m to the end node t k The impedance on this line; From the intermediate node m to the end node t k The resistance in this line; From the intermediate node m to the end node t k The reactance on this line; j is the imaginary unit of reactance;

[0035] Will Disassembled into The transverse component perpendicular to the direction and with Horizontal longitudinal component transverse component With longitudinal component Expressed as:

[0036]

[0037] End node t k Active power at time i and reactive power for:

[0038]

[0039] and The relationship between the four is as follows:

[0040]

[0041] We can get:

[0042]

[0043] Where, i = 1, 2, ..., T; T is the time section of the measurement system; is the voltage at time i With current The angle between For the established node t k The voltage equation of

[0044] The real-time measurement data of the complete distribution network after normal operation or restoration of normal operation is input into the parameter identification and verification module, and the least squares method is used to identify the distribution network line parameters to obtain the line parameter identification results.

[0045] Furthermore, the distribution network identification line parameter accuracy includes:

[0046] The accuracy of distribution network identification line parameters is calculated according to the following formula:

[0047]

[0048] Where w is the accuracy of the distribution network identification line parameters; θ′ is the parameter identification result, and θ is the true value of the parameter.

[0049] The present invention provides a distribution network line parameter identification method based on fault diagnosis, comprising collecting real-time voltage signals and power signals at the head and end points of each line in the distribution network; performing denoising and missing value supplementation processing on the real-time voltage signals and power signals to obtain real-time measurement data feature vectors; obtaining historical voltage signals and power signals at the head and end points of the distribution network; constructing historical measurement data feature vectors; simultaneously normalizing the historical measurement data feature vectors and the real-time measurement data feature vectors; dividing the normalized historical measurement data vectors into multiple groups, using a genetic algorithm to calculate the optimal weight of each group, and using the average of the multiple groups of optimal weights as the initial weight of a GA-Elman neural network; and performing normalized historical measurement data on the head and end points of the distribution network. Historical measurement data vectors are divided into training sets and test sets; the historical measurement data vectors are used as the input of the GA-Elman neural network, and the numbers corresponding to the fault composite judgment results are used as the output of the GA-Elman neural network. The training set is selected and combined with the initial weights of the GA-Elman neural network to train the GA-Elman neural network; the test set is input into the trained GA-Elman neural network to realize fault diagnosis; the distribution network structure after fault diagnosis is drawn to obtain a distribution network structure diagram, which includes the faulty nodes, fault judgment results and the fault area; the least squares method is used to identify the line parameters of the entire normally operating distribution network; and the accuracy of the distribution network identification line parameters is calculated.

[0050] The present invention utilizes historical measurement data and combines multiple genetic algorithms to optimize the initial weights of the neural network to train and verify the Elman neural network distribution network fault diagnosis model. This not only overcomes the defect of the traditional Elman neural network that is prone to falling into local minima, but also improves the generalization ability and prediction accuracy. The present invention not only uses the Grubbs test method to denoise the measurement data, but also uses the same type mean interpolation method to process the missing values ​​of the measurement data, thereby improving data reliability and enhancing data integrity, thereby further improving the accuracy of the parameter identification result. The present invention uses multiple genetic algorithms to obtain the mean of the weights, avoiding the randomness caused by the algorithm's own limitations when obtaining the weights using a single genetic algorithm, and achieving the effect of optimizing the initial weights of the neural network. The present invention uses a combination of data denoising and data missing value interpolation to process the measurement data, remove outliers in the measurement data, and then applies the least squares method to distribution network parameter identification to solve the distribution network line parameters, thereby reducing the influence of outliers on the least squares parameter identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A flowchart of a method for identifying distribution network line parameters based on fault diagnosis provided by an embodiment of the present invention;

[0053] Figure 2 This is a diagram of the Elman neural network structure of an embodiment of the present invention;

[0054] Figure 3 A node sequential search graph using a traversal algorithm after fault diagnosis according to an embodiment of the present invention;

[0055] Figure 4 A simple network structure diagram of a line parameter identification equation in an embodiment of the present invention;

[0056] Figure 5 This is a voltage vector relationship diagram in line parameter identification according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0058] As described in the background art, traditional fault analysis systems and traditional distribution network fault diagnosis methods rely on the combination of relay protection and manual line patrol. That is, after the relay protection device of the distribution line operates, it notifies the power line patrol personnel to search for the location of the fault along the tripped distribution line. The traditional fault diagnosis method has a large time delay and error, and it is difficult to ensure real-time performance and accuracy, which is not conducive to the timely diagnosis and elimination of faults. At the same time, the determination of the inherent parameters of the distribution network site often relies on empirical data and classical mathematical model derivation, without considering the actual operating conditions and the impact of line faults on the actual parameters. For example, the line parameters are affected by factors such as the operating conditions of the distribution network, line temperature, line environment, and overvoltage, resulting in a certain error in the numerical value between the line parameters stored in the distribution management system and the actual line parameters. Especially when a fault occurs in the distribution network, the line parameters change greatly. Using the line parameters with errors for distribution network analysis applications will result in a large gap between the calculation results and the actual values.

[0059] Therefore, to solve the above problems, as Figure 1 shown, some embodiments of the present invention provide a method for identifying distribution network line parameters based on fault diagnosis, including:

[0060] Step S101, collect the real-time voltage signals and power signals at the beginning and end points of each line in the distribution network.

[0061] Step S102, perform denoising and missing value supplementation processing on the real-time voltage signals and power signals to obtain real-time measurement data feature vectors.

[0062] In this step, the measurement data such as the collected voltage signals and power signals are classified into three categories: active power, reactive power, and voltage amplitude corresponding to different nodes;

[0063] Arrange the measurement data such as active power, reactive power, and voltage amplitude in ascending order within their respective categories; assume that the data anomaly point is x d , calculate the average value avg of all data; calculate the estimator y of the arithmetic mean and standard deviation; calculate the statistic g d :

[0064]

[0065] Compare g d with the critical value g(a,n) obtained from the Grubbs test method table. If g d <g(a,n), then the data point x d is not an outlier; if g d >g(a,n), then the data point x d is an outlier.

[0066] The same type mean interpolation method is used to determine the data category based on the data that has been classified in the denoising process, and the mean of the samples in each category is used to interpolate the missing values.

[0067] Step S103: Acquire historical voltage signals and power signals of the head and terminal points of the distribution network.

[0068] Step S104: constructing a feature vector of historical measurement data.

[0069] Step S105 : normalizing the historical measurement data feature vector and the real-time measurement data feature vector simultaneously.

[0070] In this step, the processed signal feature vector is extracted, and the voltage and power signals of the distribution network head and terminal points recorded in the past are classified and organized into the historical measurement data feature vector T h =[U qh ,P qh ,Q qh ], and then the historical measurement data vector and the preprocessed measurement data feature vector T q =[U qc ,P qc ,Q qc ] for normalization.

[0071] The historical measurement data feature vector and the real-time measurement data feature vector are normalized simultaneously according to the following formula:

[0072]

[0073] Among them, T qh 、 are respectively the normalized historical measurement data vector and real-time measurement data vector; U qh 、P qh , Q qh are the historically measured voltage vector, active power vector, and reactive power vector; U qc 、P qc , Q qc They are respectively the normalized real-time measured voltage vector, active power vector, and reactive power vector;

[0074] Step S106 , dividing the normalized historical measurement data vectors into multiple groups, using a genetic algorithm to calculate the optimal weight of each group, and taking the average of the multiple groups of optimal weights as the initial weight of the GA-Elman neural network.

[0075] In this step, the advantages of the global search of the genetic algorithm are used to find the optimal neural networks of each layer of the Elman neural network. The genetic algorithm design steps are as follows:

[0076] Initialize the population, including the initial size M of the population and the crossover probability P j , mutation probability P b , maximum number of iterations Z, current number of iterations z.

[0077] The fitness function is selected based on the difference between the actual output and the expected output of the network:

[0078]

[0079]

[0080] Where, T i 、Y i They represent the actual output and expected output of the i-th training sample, and n represents the number of training samples.

[0081] Calculate and find the individual with the best fitness and perform multiple genetic iterations, find the individual with the worst fitness and eliminate it. After multiple iterative evolutions, when the set population genetic generation is reached, the optimal initial weights and thresholds of the Elman network are obtained.

[0082] After denoising and supplementing missing values, the measured data are aggregated and shuffled into 5 groups. The 5 groups of data are input into the genetic algorithm to find the optimal weights for each group. The optimal weights for each group are u1, u2, u3, u4, and u5. The average of these five groups of weights is taken as the final weight u z , which is determined as the initial weight of the subsequent Elman neural network:

[0083]

[0084] In step S107, the normalized historical measurement data vectors of the head and terminal points of the distribution network are divided into a training set and a test set; the historical measurement data vectors are used as the input of the GA-Elman neural network, and the numbers corresponding to the fault composite judgment results are used as the output of the GA-Elman neural network. The training set is selected and combined with the initial weights of the GA-Elman neural network to train the GA-Elman neural network.

[0085] In this step, a composite judgment is performed on the distribution network fault in combination with the historical measurement data vector. Different composite fault judgment results correspond to numbers 1 to 4. The specific strategy for composite fault judgment is as follows:

[0086] The distribution network operating status corresponding to the historical records is divided into four categories, namely normal state, general fault, intermediate fault and severe fault. The data of different operating states are classified by combining the historical measurement data sets and trigger conditions corresponding to these four states.

[0087] Normal state: When the voltage, active power, and reactive power of the distribution network are within the allowable range, small disturbances or slight load fluctuations do not affect the normal operation of the network, and the distribution network has a certain safety reserve, the measured data vector at this time becomes the input of the Elman neural network to determine whether the distribution network is in a normal state. The normal state corresponds to the output digital 1;

[0088] General fault: When the safety reserve coefficient of the distribution network decreases or the interference probability increases, the network security level gradually decreases. At this time, the measurement data vector becomes the input of the Elman neural network that determines that the distribution network is in a general fault state. The corresponding output of general fault is 2.

[0089] Intermediate fault: When the distribution network is affected by a large disturbance or an abnormal phenomenon occurs, the network voltage, active power, and reactive power exceed or fall below the allowable value. The measured data vector at this time becomes the input of the Elman neural network that determines that the distribution network is in a medium fault state. The corresponding output number is 3 for general faults.

[0090] Severe fault: When the distribution network is affected by a large disturbance, the fault cannot be eliminated, or the load fluctuates violently beyond the controllable range, the measurement data vector at this time becomes the input of the Elman neural network to determine whether the distribution network is in a severe fault state. A general fault corresponds to the output number 4.

[0091] Input the initial weight u of the Elman neural network z , select the training set to train the Elman neural network, the input of the Elman neural network is the normalized historical measurement data vector T qh , the output of the Elman neural network is the number 1 to 4 corresponding to the composite fault judgment result.

[0092] like Figure 2 As shown in the figure, the Elman neural network structure consists of four layers: input layer, hidden layer, receiving layer, and output layer. After the training set is input into the input layer, the hidden layer processes and transforms the signal. The receiving layer stores the output value of the previous hidden layer iteration and feeds it back to the hidden layer. The output layer performs weighted processing on the output signal of the hidden layer and then performs output operation. The expressions of the hidden layer, receiving layer, and output layer of the Elman neural network are:

[0093] 1) Hidden layer output:

[0094] x(l)=f(W1x c(l)+W2(u(l-1)+b1));

[0095] 2) Output of the receiving layer:

[0096] x c (l) = x(l-1);

[0097] 3) Output layer:

[0098] y(l)=g(W3x(l)+b2);

[0099] Among them, l is the number of network training times, u is the r-dimensional input vector, x is the n-dimensional hidden layer unit vector, y is the m-dimensional output vector, x c is the n-dimensional hidden layer vector, g(x) is the transfer function of the output layer, using the purelin function to perform a linear transformation on the hidden layer output, f(x) is the transfer function of the hidden layer, using the tansig function, W1, W2 and W3 are the weight matrices between the hidden layer and the receiving layer, between the input layer and the hidden layer, and between the hidden layer and the output layer, respectively, W1∈R n×n , W2∈R n×r , W3∈R m ×n .

[0100] The error back propagation algorithm is used to correct the weights. In the time period T, the weight correction value is:

[0101]

[0102] The weight matrix is ​​continuously updated until the error between the actual output and the expected output is less than the set target error. The target error function uses the square sum function:

[0103]

[0104] in: is the target input vector, and y(l) is the output vector.

[0105] Step S108: input the test set into the trained GA-Elman neural network to implement fault diagnosis.

[0106] In this step, the specific determination strategy for accuracy determination is:

[0107] Z = number of samples successfully identified / total number of samples;

[0108] When Z ≥ 95%, the trained neural network is effective and can be used directly;

[0109] When Z is less than 95%, the trained neural network is invalid or the effect is not obvious, and further training is required.

[0110] Step S109 , drawing the distribution network structure after fault diagnosis to obtain a distribution network structure diagram, which includes faulty nodes, fault determination results, and the fault area.

[0111] In this step, if Figure 3 As shown in the figure, for the distribution network that has completed fault diagnosis, a traversal algorithm is used, starting from the node at the head end of the distribution network line, and then sequentially searching from left to right to the next layer of nodes. Combined with the real-time fault diagnosis results, the block diagram drawing module is used to draw the structural diagram of the entire distribution network, and the fault node, fault judgment result and fault area are automatically calibrated.

[0112] The calibrated fault structure diagram and related parameters are input into the fault repair module, and the specific method for subsequent fault processing and repair is as follows:

[0113] Use automation devices and automated dispatching systems to take different actions on the distribution network under different fault judgment results.

[0114] In normal conditions, the distribution network is adjusted and controlled normally under dynamic balance to maintain stable network operation and keep the system in a safe operating state.

[0115] In the event of a general fault, the distribution network will be subject to operations such as load adjustment and change of operating mode to reduce the overall operating risk of the network, increase the safety margin, and restore the system to normal as soon as possible.

[0116] In the event of a medium-level fault, emergency correction and control measures are taken on the distribution network to selectively and quickly remove some faulty components or faulty areas, reduce the interference of the faulty parts on the entire network, and restore the network voltage and frequency to within the allowable range.

[0117] In the event of a serious fault, the distribution network will be decoupled, load shedding, line disconnection, and re-parallel operation of several small systems to transition the small systems to a parallel large system operation state, making full use of the means provided by the dispatching automation system to restore the entire network to normal operation.

[0118] Step S110 , using the least square method to identify the line parameters of the entire normally operating distribution network.

[0119] In this step, if Figure 4 and Figure 5 As shown in the figure, the distribution network line parameter identification equation is constructed, the distribution network node s is the starting node; the node m is the middle node; the terminal node t k ,k=1,2,…,n;the power flow of the distribution network is from node s to the terminal node t k Outflow; the line parameters to be identified are resistance and reactance; for the intermediate node m to the terminal node t kAny line in satisfies the following formula at time i:

[0120]

[0121] in, is the voltage of node m at time i; is node t at time i k voltage; From the intermediate node m to the terminal node t at time i k The current on this line; From the intermediate node m to the end node t k The impedance on this line; From the intermediate node m to the end node t k The resistance in this line; From the intermediate node m to the end node t k The reactance on this line; j is the imaginary unit of reactance;

[0122] Will Disassembled into The transverse component perpendicular to the direction and with Horizontal longitudinal component transverse component With longitudinal component Expressed as:

[0123]

[0124] End node t k Active power at time i and reactive power for:

[0125]

[0126] and The relationship between the four is as follows:

[0127]

[0128] We can get:

[0129]

[0130] Where, i = 1, 2, ..., T; T is the time section of the measurement system; is the voltage at time i With current The angle between For the established node t k The voltage equation of

[0131] The real-time measurement data of the complete distribution network after normal operation or restoration of normal operation is input into the parameter identification and verification module, and the least squares method is used to identify the distribution network line parameters to obtain the line parameter identification results.

[0132] Step S111, calculating the accuracy of distribution network identification line parameters.

[0133] In this step, the accuracy of the distribution network identification line parameters is calculated according to the following formula:

[0134]

[0135] Where w is the accuracy of the distribution network identification line parameters; θ′ is the parameter identification result, and θ is the true value of the parameter.

[0136] The present invention has been described in detail above with reference to specific embodiments and exemplary examples. However, these descriptions should not be construed as limiting the present invention. Those skilled in the art will appreciate that various equivalent substitutions, modifications, or improvements may be made to the technical solutions and implementations of the present invention without departing from the spirit and scope of the present invention, all of which fall within the scope of the present invention. The scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for identifying distribution network line parameters based on fault diagnosis, characterized in that: include: Collect real-time voltage and power signals at the beginning and end points of each line in the distribution network; Perform denoising and missing value supplementation on real-time voltage and power signals to obtain real-time measurement data feature vectors; Obtain historical voltage and power signals at the head and end points of the distribution network; Construct feature vectors of historical measurement data; Normalize the historical measurement data feature vector and the real-time measurement data feature vector simultaneously; The normalized historical measurement data vectors are divided into multiple groups, and the optimal weight of each group is calculated using a genetic algorithm. The average value of the optimal weights of multiple groups is used as the initial weight of the GA-Elman neural network. The normalized historical measurement data vectors of the distribution network's head and terminal points are divided into a training set and a test set. The historical measurement data vectors serve as the input of the GA-Elman neural network, and the corresponding numerical values ​​of the composite fault determination results serve as the output of the GA-Elman neural network. The training set is selected and combined with the initial weights of the GA-Elman neural network to train the GA-Elman neural network. Input the test set into the trained GA-Elman neural network to achieve fault diagnosis; The distribution network structure after fault diagnosis is drawn to obtain a distribution network structure diagram, which includes the fault node, fault judgment result and fault location area; The least square method is used to identify the line parameters of the entire normally operating distribution network; Calculate the accuracy of distribution network identification line parameters.

2. The method for identifying distribution network line parameters according to claim 1, characterized in that: The real-time voltage signal and power signal are subjected to denoising and missing value supplementation processing to obtain a real-time measurement data feature vector, including: The collected voltage signal, power signal and other measurement data are classified into three categories: active power, reactive power and voltage amplitude corresponding to different nodes; Arrange the measured data such as active power, reactive power and voltage amplitude in ascending order within their respective categories; Calculate the average value avg of all data; Compute the estimates y of the arithmetic mean and standard deviation; Calculate the statistic g d : Compare g d with g(a,n) obtained from the critical value table of Grubbs' test. If g d < g(a,n), then the data point x d is not an outlier; if g d > g(a,n), then the data point x d is an outlier. The same type mean interpolation method is used to determine the data category based on the data that has been classified in the denoising process, and the mean of the samples in each category is used to interpolate the missing values.

3. The method for identifying distribution network line parameters according to claim 1, characterized in that: The simultaneous normalization of the historical measurement data feature vector and the real-time measurement data feature vector includes: The historical measurement data feature vector and the real-time measurement data feature vector are normalized simultaneously according to the following formula: Among them, T qh 、T q* are respectively the normalized historical measurement data vector and real-time measurement data vector; U qh 、P qh , Q qh are the historically measured voltage vector, active power vector, and reactive power vector; U qc 、P qc , Q qc are the normalized real-time measured voltage vector, active power vector, and reactive power vector; U q* =100kV, P q* =1000MVA, Q q* =1000MVA.

4. The method for identifying distribution network line parameters according to claim 1, wherein: The composite fault determination results include normal status, general fault, intermediate fault and severe fault.

5. The method for identifying distribution network line parameters according to claim 1, characterized in that: The method of using the least squares method to identify line parameters of the entire normally operating distribution network includes: Construct the distribution network line parameter identification equation, where the distribution network node s is the starting node; node m is the intermediate node; and the terminal node t k ,k=1,2,…,n;the power flow of the distribution network is from node s to the terminal node t k Outflow; the line parameters to be identified are resistance and reactance; for the intermediate node m to the terminal node t k Any line in satisfies the following formula at time i: in, is the voltage of node m at time i; is node t at time i k voltage; From the intermediate node m to the terminal node t at time i k The current on this line; From the intermediate node m to the end node t k The impedance on this line; From the intermediate node m to the end node t k The resistance in this line; From the intermediate node m to the end node t k The reactance on this line; j is the imaginary unit of reactance; Will Disassembled into The transverse component perpendicular to the direction and with Horizontal longitudinal component transverse component With longitudinal component Expressed as: End node t k Active power at time i and reactive power for: and The relationship between the four is as follows: We can get: Where i = 1, 2, ..., T; T is the time section of the measurement system, is the voltage at time i With current The angle between is the voltage equation established about node tk; The real-time measurement data of the complete distribution network after normal operation or restoration of normal operation is input into the parameter identification and verification module, and the least squares method is used to identify the distribution network line parameters to obtain the line parameter identification results.

6. The method for identifying distribution network line parameters according to claim 1, characterized in that: The distribution network identification line parameter accuracy includes: The accuracy of distribution network identification line parameters is calculated according to the following formula: Where w is the accuracy of the distribution network identification line parameters; θ′ is the parameter identification result, and θ is the true value of the parameter.

Images