CN108562821B - Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax - Google Patents

Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax Download PDF

Info

Publication number
CN108562821B
CN108562821B CN201810432276.5A CN201810432276A CN108562821B CN 108562821 B CN108562821 B CN 108562821B CN 201810432276 A CN201810432276 A CN 201810432276A CN 108562821 B CN108562821 B CN 108562821B
Authority
CN
China
Prior art keywords
fault
data
regression model
softmax regression
softmax
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810432276.5A
Other languages
Chinese (zh)
Other versions
CN108562821A (en
Inventor
李辉
陈江波
陈浩然
戴敏
陈维江
吕军
宁昕
田野
陈金猛
方茂欢
蒋元宇
邱进
马硕
翟文鹏
蔡胜伟
郭慧浩
邵苠峰
尹晶
费烨
何妍
陈程
杜砚
程婷
王华云
郑蜀江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
State Grid Jiangxi Electric Power Co
China Electric Power Research Institute Co Ltd CEPRI
Original Assignee
State Grid Jiangxi Electric Power Co
China Electric Power Research Institute Co Ltd CEPRI
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by State Grid Jiangxi Electric Power Co, China Electric Power Research Institute Co Ltd CEPRI filed Critical State Grid Jiangxi Electric Power Co
Priority to CN201810432276.5A priority Critical patent/CN108562821B/en
Publication of CN108562821A publication Critical patent/CN108562821A/en
Application granted granted Critical
Publication of CN108562821B publication Critical patent/CN108562821B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/08—Locating faults in cables, transmission lines, or networks
    • G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
    • G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/08—Locating faults in cables, transmission lines, or networks
    • G01R31/088—Aspects of digital computing

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Theoretical Computer Science (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Remote Monitoring And Control Of Power-Distribution Networks (AREA)

Abstract

The invention discloses a method and a system for determining single-phase earth fault line selection of a power distribution network based on Softmax, wherein the method comprises the following steps: acquiring fault data of each preset node in a plurality of preset nodes of a line in a power distribution network; processing the fault data to obtain fault characteristic data, and establishing a fault identification and fault characteristic database; dividing the fault identification and data in the fault feature database into a training set and a testing set according to fault types, and performing line selection identification training by using a Softmax regression model and the training set according to preset iteration times to determine optimal parameters so as to determine a Softmax regression model corresponding to the optimal parameters; and predicting the fault type and position according to the real-time fault data of the power grid by using the Softmax regression model corresponding to the optimal parameters so as to determine the unidirectional earth fault route selection of the power distribution network. The invention trains and verifies the Softmax regression model by using data, improves the data processing capacity and the line selection accuracy, and promotes the research and development and application of a novel power distribution terminal.

Description

Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax
Technical Field
The invention relates to the technical field of risk analysis and control of a power distribution network, in particular to a method and a system for determining single-phase earth fault line selection of the power distribution network based on Softmax.
Background
The problem of low-current grounding is a worldwide problem, and the current fault identification mainly has the following difficulties: the low-current single-phase grounding has small fault current and is difficult to detect; secondly, the system structure is complex, the neutral point contact modes are different, the fault positions are different, the fault types are different, the caused fault characteristics are also different, and the detected threshold value is difficult to set; thirdly, fault characteristic signals are refracted and reflected in the network, are mutually overlapped and interfered, and are not easy to separate meaningful characteristic quantities.
Meanwhile, the main reasons for the poor effectiveness of the existing methods are the following three points: (1) specific fault data analysis is not carried out, only short-circuit current is set, the small-current grounding current value is small, and the small-current grounding is difficult to judge by simply limiting the short-circuit current; (2) the line selection is carried out only based on a certain characteristic signal, and because different small-current ground faults are diversified due to different grounding modes and different grounding resistances of a neutral point, and a single characteristic signal cannot describe the fault characteristics of all types, the effectiveness of the line selection method based on a characteristic signal is limited; (3) because the ground fault simulation can not be carried out on a large number of actual distribution networks, the line selection method based on the multi-signal data characteristic quantity is usually carried out in a dynamic simulation mode, but the digital simulation modeling can not completely and accurately simulate the condition of real distribution network equipment, influence factors such as transient characteristics can be ignored, and meanwhile, due to the limitation of simulation step length, the frequency range of the provided transient signal is limited to a certain extent, so that the application effect of the provided method in an actual system is limited.
The Softmax regression algorithm is a typical multi-classification problem regression algorithm, is widely applied to the field of machine learning and is used for multi-classification purposes such as hand-written number recognition, animal and plant classification and the like, and an application example of the algorithm for fault diagnosis of a hydraulic pump is provided. However, the Softmax regression algorithm is not applied to the field of power systems. In the field of power systems, in recent years, a large number of recording-wave type fault indicators are configured in each province of China, so that a large number of real power distribution network fault data are obtained, and meanwhile, the data processing and analyzing capacity is insufficient.
At present, if a single-phase grounding line selection strategy based on real effective data is to be proposed, the following problems need to be solved: firstly, obtaining effective waveform data in a fault occurrence state, wherein the data is required to be matched with the type and the fault occurrence position of the fault, position information in a line selection strategy is a line number, and the sampling rate of the recorded data waveform is required to be high enough and contains fault data characteristics; secondly, the recorded data waveform can not be directly used for fault identification, and preprocessing is needed to separate out characteristic quantity strongly related to the fault data characteristic, and fault identification is carried out once; and thirdly, an application method of the fault characteristics is provided, generally, the fault characteristics and the fault position information are in a nonlinear corresponding relation, and classification is carried out through a regression algorithm, so that the purpose of effectively selecting a fault line is achieved.
Therefore, due to the above various defects and shortcomings, in order to provide a more effective route selection strategy and fully utilize fault data in a real power distribution network, a single-phase ground fault route selection method for a power distribution network is needed to solve the problem that a fault route cannot be accurately determined.
Disclosure of Invention
The invention provides a method and a system for determining single-phase earth fault line selection of a power distribution network based on Softmax, and aims to solve the problem that a faulted route cannot be accurately determined.
In order to solve the above problem, according to an aspect of the present invention, there is provided a method for determining single-phase ground fault line selection of a power distribution network based on Softmax, the method comprising:
step 1, acquiring fault data of each preset node in a plurality of preset nodes of a line in a power distribution network, wherein the fault data comprises: three-phase voltage data, three-phase current data, phase data of voltage, phase information of current and absolute time information of fault occurrence;
step 2, processing the fault data to obtain fault characteristic data, and establishing a fault identification and fault characteristic database according to the fault characteristic data, the fault type and the fault position identification, wherein the fault characteristic data comprises a plurality of characteristic quantities;
step 3, dividing the fault identification and data in the fault feature database into a training set and a testing set according to fault types, and performing line selection identification training by using a Softmax regression model and the training set according to preset iteration times to determine optimal parameters so as to determine the Softmax regression model corresponding to the optimal parameters;
and 4, predicting the fault type and position according to the real-time fault data of the power grid by using the Softmax regression model corresponding to the optimal parameters so as to determine the unidirectional earth fault line selection of the power distribution network.
Preferably, the fault data at each preset node is collected by using a recording type fault indicator.
Preferably, the processing the fault data to obtain fault feature data includes:
processing the fault data, and synthesizing to obtain a fault state quantity, wherein the fault state quantity comprises: frequency variation, voltage variation, current variation, phase angle variation, power variation, no power variation, frequency variation with power, current harmonic distortion, voltage harmonic distortion, power factor, and harmonic components of voltage and current;
and processing the fault state quantity according to a rough set theory, removing the fault state quantity with low correlation degree, and acquiring fault characteristic data.
Preferably, the determining the optimal parameter by performing line selection identification training by using a Softmax regression model and a training set according to the preset iteration number to determine the Softmax regression model corresponding to the optimal parameter includes:
determining a cost function of the Softmax regression algorithm, wherein the cost function is as follows:
Figure BDA0001653717410000031
wherein, the samples of the Softmax regression model are from k classes, and m samples are total, then the training set formed by the samples is { (x)(1),y(1)),(x(2),y(2)),…,(x(m),y(m)) Therein of
Figure BDA0001653717410000032
Labeling: y is(t)E {1,2, …, k }, for a given input x, the hypothesis function estimates for each class j a probability value p (y j | x) for estimating the probability of each classification result of x occurring;
and performing line selection identification training on the cost function by adopting a gradient descent method according to preset iteration times, determining an optimal parameter, and acquiring a Softmax regression model corresponding to the optimal parameter.
Preferably, wherein the method further comprises:
predicting the fault type and the fault position corresponding to the fault feature data of the sample in the test set by using the Softmax regression model corresponding to the optimal parameters;
calculating the prediction accuracy, and judging whether the prediction accuracy is greater than or equal to a preset accuracy threshold value; wherein the content of the first and second substances,
if the prediction accuracy is greater than or equal to the preset accuracy threshold, directly determining the single-phase earth fault route selection of the power distribution network by using a Softmax regression model corresponding to the determined optimal parameters;
and if the prediction accuracy is smaller than the preset accuracy threshold, increasing the iteration times according to the preset iteration step length, returning to the step 3, performing line selection identification training by using the Softmax regression model and the training set to determine the optimal parameters, and determining the Softmax regression model corresponding to the optimal parameters.
According to another aspect of the invention, a system for determining single-phase earth fault line selection of a power distribution network based on Softmax is provided, the system comprising:
the fault data acquisition unit is used for acquiring fault data of each preset node in a plurality of preset nodes of a line in a power distribution network, wherein the fault data comprises: three-phase voltage data, three-phase current data, phase data of voltage, phase information of current and absolute time information of fault occurrence;
the database establishing unit is used for processing the fault data, acquiring fault characteristic data, and establishing a fault identification and fault characteristic database according to the fault characteristic data, the fault type and the fault position identification, wherein the fault characteristic data comprises a plurality of characteristic quantities;
the model determining unit is used for dividing the fault identification and the data in the fault feature database into a training set and a testing set according to fault types, performing line selection identification training by using a Softmax regression model and the training set according to preset iteration times to determine optimal parameters, and determining the Softmax regression model corresponding to the optimal parameters;
and the fault line selection determining unit is used for predicting the fault type and the fault position according to the real-time fault data of the power grid by using the Softmax regression model corresponding to the optimal parameters so as to determine the unidirectional earth fault line selection of the power distribution network.
Preferably, the fault data at each preset node is collected by using a recording type fault indicator.
Preferably, the fault data acquiring unit processes the fault data to acquire fault feature data, and includes:
processing the fault data, and synthesizing to obtain a fault state quantity, wherein the fault state quantity comprises: frequency variation, voltage variation, current variation, phase angle variation, power variation, no power variation, frequency variation with power, current harmonic distortion, voltage harmonic distortion, power factor, and harmonic components of voltage and current;
and processing the fault state quantity according to a rough set theory, removing the fault state quantity with low correlation degree, and acquiring fault characteristic data.
Preferably, the determining unit, according to the preset iteration number, performs line selection identification training by using a Softmax regression model and a training set to determine an optimal parameter, so as to determine the Softmax regression model corresponding to the optimal parameter, includes:
determining a cost function of the Softmax regression algorithm, wherein the cost function is as follows:
Figure BDA0001653717410000051
wherein, the samples of the Softmax regression model are from k classes, and m samples are total, then the training set formed by the samples is { (x)(1),y(1)),(x(2),y(2)),…,(x(m),y(m)) Therein of
Figure BDA0001653717410000052
Labeling: y is(t)E {1,2, …, k }, for a given input x, the hypothesis function estimates for each class j a probability value p (y j | x) for estimating the probability of each classification result of x occurring;
and performing line selection identification training on the cost function by adopting a gradient descent method according to preset iteration times, determining an optimal parameter, and acquiring a Softmax regression model corresponding to the optimal parameter.
Preferably, wherein the system further comprises: an accuracy verification unit for verifying the accuracy of the measurement data,
the Softmax regression model is used for predicting the fault type and the fault position corresponding to the fault feature data of the sample in the test set by using the Softmax regression model corresponding to the optimal parameters;
the device is used for calculating the prediction accuracy and judging whether the prediction accuracy is greater than or equal to a preset accuracy threshold value; wherein the content of the first and second substances,
if the prediction accuracy is greater than or equal to the preset accuracy threshold, directly determining the single-phase earth fault route selection of the power distribution network by using a Softmax regression model corresponding to the determined optimal parameters;
and if the prediction accuracy is smaller than the preset accuracy threshold, increasing the iteration times according to the preset iteration step length, returning to the model determination unit, and performing line selection identification training by using the Softmax regression model and the training set to determine the optimal parameter so as to determine the Softmax regression model corresponding to the optimal parameter.
The invention provides a method and a system for determining single-phase earth fault line selection of a power distribution network based on Softmax, which are used for acquiring fault data of lines in the power distribution network; processing the fault data to obtain fault characteristic data and establishing a fault identification and fault characteristic database; dividing the fault identification and data in the fault feature database into a training set and a testing set according to fault types, and performing line selection identification training by using a Softmax regression model and the training set according to preset iteration times to determine a Softmax regression model corresponding to optimal parameters; and finally, predicting the fault type and position according to the real-time fault data of the power grid by using a Softmax regression model corresponding to the optimal parameters so as to determine the unidirectional earth fault line of the power distribution network. The method utilizes real data, extracts characteristic quantity related to fault positions through data processing, introduces a Softmax regression model into the field of single-phase grounding data identification, trains and verifies the Softmax regression model by using the data, develops a new path for identifying faults, improves data processing capacity and line selection accuracy, provides a new thought for verification of a novel power distribution terminal protection algorithm, threshold setting and action logic strategy, and promotes research, development and application of the novel power distribution terminal.
Drawings
A more complete understanding of exemplary embodiments of the present invention may be had by reference to the following drawings in which:
fig. 1 is a flow chart of a method 100 for determining single-phase ground fault line selection of a power distribution network based on Softmax according to an embodiment of the present invention;
FIG. 2 is a diagram of a rough set model according to an embodiment of the present invention;
fig. 3 is a diagram of a typical overhead line grid structure of a power distribution network according to an embodiment of the invention;
FIG. 4 is a data separation plot resulting from 500 iterations according to an embodiment of the present invention;
FIG. 5 is a data separation plot resulting from 1000 iterations according to an embodiment of the present invention; and
fig. 6 is a schematic structural diagram of a system 600 for determining single-phase ground fault line selection of a power distribution network based on Softmax according to an embodiment of the present invention.
Detailed Description
The exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, however, the present invention may be embodied in many different forms and is not limited to the embodiments described herein, which are provided for complete and complete disclosure of the present invention and to fully convey the scope of the present invention to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the accompanying drawings is not intended to be limiting of the invention. In the drawings, the same units/elements are denoted by the same reference numerals.
Unless otherwise defined, terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Further, it will be understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense.
Fig. 1 is a flow chart of a method 100 for determining single-phase ground fault line selection of a power distribution network based on Softmax according to an embodiment of the present invention. As shown in fig. 1, the method for determining the single-phase earth fault line selection of the power distribution network based on Softmax provided by the embodiment of the invention utilizes real data, extracts the characteristic quantity related to the fault position through processing the data, introduces the Softmax regression model into the field of single-phase earth data identification, trains and verifies the Softmax regression model by using the data, develops a new path for identifying faults by one, improves the data processing capability and the line selection accuracy, provides a new idea for the verification of a novel power distribution terminal protection algorithm, threshold setting and action logic strategy, and promotes the research, development and application of a novel power distribution terminal. The method 100 for determining the single-phase earth fault route selection of the power distribution network based on Softmax provided by the embodiment of the invention starts from step 101, and obtains fault data of each preset node in a plurality of preset nodes of a route in the power distribution network in step 101, wherein the fault data comprises: three-phase voltage data, three-phase current data, phase data of voltage, phase information of current, and absolute time information of occurrence of a fault.
Preferably, the fault data at each preset node is collected by using a recording type fault indicator.
Preferably, in step 102, the fault data is processed to obtain fault feature data, and a fault identifier and fault feature database is established according to the fault feature data, the fault type, and the fault location identifier, where the fault feature data includes a plurality of feature quantities.
Preferably, the processing the fault data to obtain fault feature data includes: processing the fault data, and synthesizing to obtain a fault state quantity, wherein the fault state quantity comprises: frequency variation, voltage variation, current variation, phase angle variation, power variation, no power variation, frequency variation with power, current harmonic distortion, voltage harmonic distortion, power factor, and harmonic components of voltage and current;
and processing the fault state quantity according to a rough set theory, removing the fault state quantity with low correlation degree, and acquiring fault characteristic data.
In the embodiment of the present invention, recorded failure data is processed, and the change amount Δ f of the frequency is synthesizediChange in voltage Δ ViRate of change of frequency (Δ f/Δ t)iRate of change of voltage (Δ V/Δ t)iRate of change of current (Δ I/Δ t)iRate of change of phase angle
Figure BDA0001653717410000081
Rate of change with power (Δ P/Δ t)iRate of change without power (Δ Q/Δ t)iRate of change of frequency with power (Δ f/Δ P)iElectric powerCurrent harmonic distortion CTHD, voltage harmonic distortion VTHD, power factor
Figure BDA0001653717410000082
And state quantities such as harmonic components of the voltage and current. Then, the state quantity with low correlation degree, namely the fault characteristic data, is removed according to the rough set theory.
In the embodiment of the present invention
Figure BDA0001653717410000083
X is fault information, and U is a fault feature set. RBAn equivalence relation for U is determined based on prior knowledge, i.e. known fault location information, U/B ═ { E ═ E1,E2,…,EeIndicates the degree of correspondence of each kind of failure information with the failure location. X is in relation to RBThe lower approximation set and the upper approximation set of (2) are respectively defined as follows:
Figure BDA0001653717410000084
wherein the upper approximation set represents according to the prior knowledge RBJudging a set of objects in U which are or may be X, and expressing the lower approximate set according to the prior knowledge RBAnd judging a set consisting of the objects which are affirmatively belonging to the X in the U.
It is clear that,
Figure BDA0001653717410000085
the upper and lower approximation set of X talks the domain U and is divided into three disjoint areas, namely the positive domain POSRB(X), boundary Domain BNDRB(X) and negative field NEGRB(X) wherein, in the above-mentioned formula,
POSRB(X)＝ BR(X)，
Figure BDA0001653717410000086
Figure BDA0001653717410000087
FIG. 2 is a diagram of a rough set model according to an embodiment of the present invention. As shown in fig. 2, a positive domain, a boundary domain and a negative domain of X are visually displayed, the feature data falling into the positive domain has high correlation with the fault location, the boundary domain displays that some feature data are correlated with the fault location, and the feature data in the negative domain is completely uncorrelated with the fault location. And (4) combining the residual state quantity, namely fault characteristic data, with fault type and fault position information to establish a fault identification and fault characteristic database. And dividing different types of faults into two groups, wherein one group is used for fault identification algorithm training, and the other group is used for verifying the effectiveness of the algorithm.
Preferably, in step 103, the fault identifier and the data in the fault feature database are divided into a training set and a test set according to the fault type, and line selection identification training is performed by using a Softmax regression model and the training set according to a preset iteration number to determine an optimal parameter, so as to determine the Softmax regression model corresponding to the optimal parameter.
Preferably, the determining the optimal parameter by performing line selection identification training by using a Softmax regression model and a training set according to the preset iteration number to determine the Softmax regression model corresponding to the optimal parameter includes:
determining a cost function of the Softmax regression algorithm, wherein the cost function is as follows:
Figure BDA0001653717410000091
wherein, the samples of the Softmax regression model are from k classes, and m samples are total, then the training set formed by the samples is { (x)(1),y(1)),(x(2),y(2)),…,(x(m),y(m)) Therein of
Figure BDA0001653717410000092
Labeling: y is(t)E {1,2, …, k }, for a given input x, the assumption function estimates a probability value p for each class j (y j | x),to estimate the probability of occurrence of each classification result of x;
and performing line selection identification training on the cost function by adopting a gradient descent method according to preset iteration times, determining an optimal parameter, and acquiring a Softmax regression model corresponding to the optimal parameter.
In an embodiment of the present invention, determining a Softmax regression model corresponding to an optimal parameter includes:
assuming that the samples of the Softmax regression model are from k classes, and there are m samples in total, the training set consisting of these samples is { (x)(1),y(1)),(x(2),y(2)),…,(x(m),y(m))}. Wherein
Figure BDA0001653717410000093
Labeling: y is(t)E {1,2, …, k }. For a given input x, the hypothesis function estimates for each class j a probability value p (y j x) for estimating the probability of occurrence of each classification result for x, and thus the hypothesis function will output a k-dimensional vector (vector element sum of 1) to represent the probabilities of the k estimates. Let us assume a function hθ(x) The form is as follows:
Figure BDA0001653717410000101
wherein
Figure BDA0001653717410000102
And the model parameters are given to each characteristic quantity by theta so as to represent the capability of the characteristic quantity for judging the fault position. p (y)(i)＝j|x(i)；θj) Represents a sample x(i)Probability of belonging to class j.
Figure BDA0001653717410000103
Is to normalize the probability distribution and make the sum of all probabilities 1. For convenience of representation, 1{ · } is used as an indicative function, i.e., 1{ true } ═ 1, and 1{ false } ═ 0. Then, the cost function of the Softmax regression algorithm can be defined as:
Figure BDA0001653717410000104
the Softmax function accumulates k possible classes, i.e. the probability of classifying x into class J in the Softmax regression is:
Figure BDA0001653717410000105
the class corresponding to the maximum probability is the classification class of x.
In practical applications, weight attenuation is usually added to the cost function to solve the numerical problem caused by the parameter redundancy of the Softmax regression, and then the cost function becomes:
Figure BDA0001653717410000111
with the weight attenuation term, the cost function becomes a strict convex function, so that a unique solution can be obtained, and the problem of parameter redundancy is effectively solved.
Partial derivatives are taken of J (θ):
Figure BDA0001653717410000112
the updating process of theta can be obtained according to the gradient descent method:
Figure BDA0001653717410000113
where α is the learning step size, so we can obtain:
Figure BDA0001653717410000114
iterating theta approaching to the optimal solution by a gradient descent method, and then substituting all theta values back to the hypothesis function hθ(x) InAnd solving the probability of all position identifications expressed by a plurality of characteristic quantity sets under the fault event to obtain a trained Softmax regression model.
Preferably, in step 104, predicting the fault type and position according to the real-time fault data of the power grid by using a Softmax regression model corresponding to the optimal parameters, so as to determine the unidirectional ground fault route selection of the power distribution network.
Preferably, wherein the method further comprises:
predicting the fault type and the fault position corresponding to the fault feature data of the sample in the test set by using the Softmax regression model corresponding to the optimal parameters;
calculating the prediction accuracy, and judging whether the prediction accuracy is greater than or equal to a preset accuracy threshold value; wherein the content of the first and second substances,
if the prediction accuracy is greater than or equal to the preset accuracy threshold, directly determining the single-phase earth fault route selection of the power distribution network by using a Softmax regression model corresponding to the determined optimal parameters;
and if the prediction accuracy is smaller than the preset accuracy threshold, increasing the iteration times according to the preset iteration step length, returning to the step 3, performing line selection identification training by using the Softmax regression model and the training set to determine the optimal parameters, and determining the Softmax regression model corresponding to the optimal parameters.
In the embodiment of the invention, sample data in a test set is substituted into a hypothesis function to calculate the probability of the class to which the sample data belongs, the class with the highest probability is selected as a prediction result, and the accuracy is verified. The lines which are possible to have faults are numbered and are marked as 0-n, the Softmax model is subjected to regression training by using a training database of the fault characteristics corresponding to each number, the Softmax regression model corresponding to the optimal parameters is obtained by taking 100 times as a basic iteration number, and the trained model is tested by using test data. And if the accuracy is lower than 95%, increasing the iteration times for 100 times, continuously utilizing the Softmax regression model and the training set to perform line selection identification training to determine the optimal parameters, and determining the Softmax regression model corresponding to the optimal parameters until the accuracy is higher than 95%.
The following specifically exemplifies embodiments of the present invention
Fig. 3 is a diagram of a typical overhead line grid structure of a power distribution network according to an embodiment of the present invention. As shown in fig. 3, the grid structure of a typical 10kV power distribution network is provided, wherein the neutral point grounding mode is direct grounding, and the fault grounding type is metallic grounding. The method comprises the steps of arranging 12 wave recording type fault indicators at each node position shown in a line, namely nodes 646, 645, 632, 633, 634, 611, 684, 652, 671, 680, 692 and 675, collecting voltage, current, phase and time information, keeping the synchronism of time records when time is conducted among the fault indicators, and respectively arranging single-phase earth faults and synchronously collecting fault data on lines 0-9 shown in red letters in figure 3.
Synthesizing the acquired fault data to obtain the frequency variation delta fiChange in voltage Δ ViRate of change of frequency (Δ f/Δ t)iRate of change of voltage (Δ V/Δ t)iRate of change of current (Δ I/Δ t)iRate of change of phase angle
Figure BDA0001653717410000121
Rate of change with power (Δ P/Δ t)iRate of change without power (Δ Q/Δ t)iRate of change of frequency with power (Δ f/Δ P)iCurrent harmonic distortion CTHD, voltage harmonic distortion VTHD, power factor
Figure BDA0001653717410000122
And 3, 5 harmonic components of the voltage current. Then, processing the data through a rough set theory, and determining fault characteristic data comprises: variation amount of voltage Δ ViRate of change of current (Δ I/Δ t)iRate of change of voltage (Δ V/Δ t)iAnd rate of change of phase angle
Figure BDA0001653717410000123
There are 4 feature quantities with high correlation.
After the characteristic quantity is determined, training the cost function by adopting a gradient descent method to obtain an optimal parameter, and determining a Softmax regression model corresponding to the optimal parameter. And then, verifying the accuracy by using the sample data in the test set.
In the embodiment of the invention, the preset accuracy threshold is 95%, the prediction results are divided into 10 types, which correspond to the corresponding feeder lines shown in fig. 3, and the types are 0-9 respectively. The verification data separation point diagram obtained after 500 iterations is shown in fig. 4, and the verification data separation point diagram obtained after 1000 iterations is shown in fig. 5, and it can be seen from the diagrams that more overlaps are formed in fig. 4, less overlaps are formed in fig. 5, the fault positions corresponding to different feature data can be classified by increasing the iteration times, the fault position type cannot be accurately judged by the feature data at the overlapping position, and the fault position type can be accurately judged if the data are completely separated. If the judgment accuracy of fig. 4 is 97% and is greater than the preset accuracy threshold 95%, the single-phase earth fault line selection of the power distribution network can be determined directly by using the Softmax regression model corresponding to the determined optimal parameters.
After the Softmax regression model corresponding to the optimal parameters is determined, predicting the fault type and the fault position according to real-time fault data of the power grid by using the Softmax regression model corresponding to the optimal parameters, and determining the unidirectional earth fault route selection of the power distribution network.
Fig. 6 is a schematic structural diagram of a system 600 for determining single-phase ground fault line selection of a power distribution network based on Softmax according to an embodiment of the present invention. As shown in fig. 6, a system 600 for determining a single-phase ground fault line selection of a power distribution network based on Softmax according to an embodiment of the present invention includes: a fault data acquisition unit 601, a database establishment unit 602, a model determination unit 603, and a fault line selection determination unit 604. Preferably, in the fault data obtaining unit 601, the fault data of each preset node in a plurality of preset nodes of a line in the power distribution network is obtained, where the fault data includes: three-phase voltage data, three-phase current data, phase data of voltage, phase information of current, and absolute time information of occurrence of a fault.
Preferably, the fault data at each preset node is collected by using a recording type fault indicator.
Preferably, in the database establishing unit 602, the fault data is processed to obtain fault feature data, and a fault identifier and fault feature database is established according to the fault feature data, the fault type, and the fault location identifier, where the fault feature data includes a plurality of feature quantities.
Preferably, the fault data acquiring unit processes the fault data to acquire fault feature data, and includes:
processing the fault data, and synthesizing to obtain a fault state quantity, wherein the fault state quantity comprises: frequency variation, voltage variation, current variation, phase angle variation, power variation, no power variation, frequency variation with power, current harmonic distortion, voltage harmonic distortion, power factor, and harmonic components of voltage and current;
and processing the fault state quantity according to a rough set theory, removing the fault state quantity with low correlation degree, and acquiring fault characteristic data.
Preferably, in the model determining unit 603, the fault identifier and the data in the fault feature database are divided into a training set and a testing set according to the fault type, and line selection identification training is performed by using a Softmax regression model and the training set according to a preset iteration number to determine an optimal parameter, so as to determine the Softmax regression model corresponding to the optimal parameter.
Preferably, the determining unit, according to the preset iteration number, performs line selection identification training by using a Softmax regression model and a training set to determine an optimal parameter, so as to determine the Softmax regression model corresponding to the optimal parameter, includes:
determining a cost function of the Softmax regression algorithm, wherein the cost function is as follows:
Figure BDA0001653717410000141
wherein, the samples of the Softmax regression model are from k classes, and m samples are total, then the training set formed by the samples is { (x)(1),y(1)),(x(2),y(2)),…,(x(m),y(m)) Therein of
Figure BDA0001653717410000142
Labeling: y is(t)E {1,2, …, k }, for a given input x, the hypothesis function estimates for each class j a probability value p (y j | x) for estimating the probability of each classification result of x occurring;
and performing line selection identification training on the cost function by adopting a gradient descent method according to preset iteration times, determining an optimal parameter, and acquiring a Softmax regression model corresponding to the optimal parameter.
Preferably, in the fault route selection determining unit 604, the type and the position of the fault are predicted by using a Softmax regression model corresponding to the optimal parameter according to real-time fault data of the power grid, so as to determine the unidirectional ground fault route selection of the power distribution network.
Preferably, wherein the system further comprises: an accuracy verification unit for verifying the accuracy of the measurement data,
predicting the fault type and the fault position corresponding to the fault feature data of the sample in the test set by using the Softmax regression model corresponding to the optimal parameters; calculating the prediction accuracy, and judging whether the prediction accuracy is greater than or equal to a preset accuracy threshold value; if the prediction accuracy is greater than or equal to the preset accuracy threshold, determining the single-phase earth fault route selection of the power distribution network by directly using a Softmax regression model corresponding to the determined optimal parameters; and if the prediction accuracy is smaller than the preset accuracy threshold, increasing the iteration times according to the preset iteration step length, returning to the model determination unit, and performing line selection identification training by using the Softmax regression model and the training set to determine the optimal parameter so as to determine the Softmax regression model corresponding to the optimal parameter.
The system 600 for determining the single-phase ground fault line selection of the power distribution network based on Softmax in the embodiment of the present invention corresponds to the method 100 for determining the single-phase ground fault line selection of the power distribution network based on Softmax in another embodiment of the present invention, and is not described herein again.
The invention has been described with reference to a few embodiments. However, other embodiments of the invention than the one disclosed above are equally possible within the scope of the invention, as would be apparent to a person skilled in the art from the appended patent claims.
Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to "a/an/the [ device, component, etc ]" are to be interpreted openly as referring to at least one instance of said device, component, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not have to be performed in the exact order disclosed, unless explicitly stated.

Claims (10)

1. A method for determining single-phase earth fault line selection of a power distribution network based on Softmax is characterized by comprising the following steps:
step 1, acquiring fault data of each preset node in a plurality of preset nodes of a line in a power distribution network, wherein the fault data comprises: three-phase voltage data, three-phase current data, phase data of voltage, phase information of current and absolute time information of fault occurrence;
step 2, processing the fault data to obtain fault characteristic data, and establishing a fault identification and fault characteristic database according to the fault characteristic data, the fault type and the fault position identification, wherein the fault characteristic data comprises a plurality of characteristic quantities;
step 3, dividing the fault identification and data in the fault feature database into a training set and a testing set according to fault types, and performing line selection identification training by using a Softmax regression model and the training set according to preset iteration times to determine optimal parameters so as to determine the Softmax regression model corresponding to the optimal parameters;
and 4, predicting the fault type and position according to the real-time fault data of the power grid by using the Softmax regression model corresponding to the optimal parameters so as to determine the unidirectional earth fault line selection of the power distribution network.
2. The method of claim 1, wherein fault data at each predetermined node is collected using a logging-type fault indicator.
3. The method of claim 1, wherein the processing the fault data to obtain fault signature data comprises:
processing the fault data, and synthesizing to obtain a fault state quantity, wherein the fault state quantity comprises: frequency variation, voltage variation, current variation, phase angle variation, power variation, no power variation, frequency variation with power, current harmonic distortion, voltage harmonic distortion, power factor, and harmonic components of voltage and current;
and processing the fault state quantity according to a rough set theory, removing the fault state quantity with low correlation degree, and acquiring fault characteristic data.
4. The method of claim 1, wherein performing line selection identification training by using a Softmax regression model and a training set according to a preset iteration number to determine an optimal parameter so as to determine the Softmax regression model corresponding to the optimal parameter, comprises:
determining a cost function of the Softmax regression algorithm, wherein the cost function is as follows:
Figure FDA0001653717400000021
wherein, the samples of the Softmax regression model are from k classes, and m samples are total, then the training set formed by the samples is { (x)(1),y(1)),(x(2),y(2)),…,(x(m),y(m)) Therein of
Figure FDA0001653717400000022
Labeling: y is(t)E {1,2, …, k }, for a given input x, the hypothesis function estimates for each class j a probability value p (y j | x) for estimating the probability of each classification result of x occurring;
and performing line selection identification training on the cost function by adopting a gradient descent method according to preset iteration times, determining an optimal parameter, and acquiring a Softmax regression model corresponding to the optimal parameter.
5. The method of claim 1, further comprising:
predicting the fault type and the fault position corresponding to the fault feature data of the sample in the test set by using the Softmax regression model corresponding to the optimal parameters;
calculating the prediction accuracy, and judging whether the prediction accuracy is greater than or equal to a preset accuracy threshold value; wherein the content of the first and second substances,
if the prediction accuracy is greater than or equal to the preset accuracy threshold, directly determining the single-phase earth fault route selection of the power distribution network by using a Softmax regression model corresponding to the determined optimal parameters;
and if the prediction accuracy is smaller than the preset accuracy threshold, increasing the iteration times according to the preset iteration step length, returning to the step 3, performing line selection identification training by using the Softmax regression model and the training set to determine the optimal parameters, and determining the Softmax regression model corresponding to the optimal parameters.
6. A system for determining single-phase earth fault line selection of a power distribution network based on Softmax, the system comprising:
the fault data acquisition unit is used for acquiring fault data of each preset node in a plurality of preset nodes of a line in a power distribution network, wherein the fault data comprises: three-phase voltage data, three-phase current data, phase data of voltage, phase information of current and absolute time information of fault occurrence;
the database establishing unit is used for processing the fault data, acquiring fault characteristic data, and establishing a fault identification and fault characteristic database according to the fault characteristic data, the fault type and the fault position identification, wherein the fault characteristic data comprises a plurality of characteristic quantities;
the model determining unit is used for dividing the fault identification and the data in the fault feature database into a training set and a testing set according to fault types, performing line selection identification training by using a Softmax regression model and the training set according to preset iteration times to determine optimal parameters, and determining the Softmax regression model corresponding to the optimal parameters;
and the fault line selection determining unit is used for predicting the fault type and the fault position according to the real-time fault data of the power grid by using the Softmax regression model corresponding to the optimal parameters so as to determine the unidirectional earth fault line selection of the power distribution network.
7. The system of claim 6, wherein fault data at each predetermined node is collected using a logging-type fault indicator.
8. The system according to claim 6, wherein the fault data acquiring unit processes the fault data to acquire fault feature data, and includes:
processing the fault data, and synthesizing to obtain a fault state quantity, wherein the fault state quantity comprises: frequency variation, voltage variation, current variation, phase angle variation, power variation, no power variation, frequency variation with power, current harmonic distortion, voltage harmonic distortion, power factor, and harmonic components of voltage and current;
and processing the fault state quantity according to a rough set theory, removing the fault state quantity with low correlation degree, and acquiring fault characteristic data.
9. The system according to claim 6, wherein the model determining unit performs line selection identification training by using a Softmax regression model and a training set according to a preset iteration number to determine an optimal parameter, so as to determine the Softmax regression model corresponding to the optimal parameter, and the determining unit includes:
determining a cost function of the Softmax regression algorithm, wherein the cost function is as follows:
Figure FDA0001653717400000041
wherein, the samples of the Softmax regression model are from k classes, and m samples are total, then the training set formed by the samples is { (x)(1),y(1)),(x(2),y(2)),…,(x(m),y(m)) Therein of
Figure FDA0001653717400000042
Labeling: y is(t)E {1,2, …, k }, for a given input x, the hypothesis function estimates for each class j a probability value p (y j | x) for estimating the probability of each classification result of x occurring;
and performing line selection identification training on the cost function by adopting a gradient descent method according to preset iteration times, determining an optimal parameter, and acquiring a Softmax regression model corresponding to the optimal parameter.
10. The system of claim 6, further comprising: an accuracy verification unit for verifying the accuracy of the measurement data,
the Softmax regression model is used for predicting the fault type and the fault position corresponding to the fault feature data of the sample in the test set by using the Softmax regression model corresponding to the optimal parameters;
the device is used for calculating the prediction accuracy and judging whether the prediction accuracy is greater than or equal to a preset accuracy threshold value; wherein the content of the first and second substances,
if the prediction accuracy is greater than or equal to the preset accuracy threshold, directly determining the single-phase earth fault route selection of the power distribution network by using a Softmax regression model corresponding to the determined optimal parameters;
and if the prediction accuracy is smaller than the preset accuracy threshold, increasing the iteration times according to the preset iteration step length, returning to the model determination unit, and performing line selection identification training by using the Softmax regression model and the training set to determine the optimal parameter so as to determine the Softmax regression model corresponding to the optimal parameter.
CN201810432276.5A 2018-05-08 2018-05-08 Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax Active CN108562821B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810432276.5A CN108562821B (en) 2018-05-08 2018-05-08 Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810432276.5A CN108562821B (en) 2018-05-08 2018-05-08 Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax

Publications (2)

Publication Number Publication Date
CN108562821A CN108562821A (en) 2018-09-21
CN108562821B true CN108562821B (en) 2021-09-28

Family

ID=63538006

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810432276.5A Active CN108562821B (en) 2018-05-08 2018-05-08 Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax

Country Status (1)

Country Link
CN (1) CN108562821B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110659405A (en) * 2019-09-25 2020-01-07 南京源堡科技研究院有限公司 Network information acquisition method based on cloud environment
CN111160241B (en) * 2019-12-27 2022-08-12 华中科技大学 Power distribution network fault classification method, system and medium based on deep learning
CN111598166B (en) * 2020-05-18 2023-10-17 国网山东省电力公司电力科学研究院 Single-phase earth fault classification method and system based on principal component analysis and Softmax function
CN111796161B (en) * 2020-05-27 2023-06-13 山西浩然机电设备工程有限公司 Fault detection system and fault detection method for overhead cable
CN112924813B (en) * 2021-01-28 2022-09-16 国网浙江省电力有限公司绍兴供电公司 Power distribution network short-circuit fault monitoring method and device based on electrical data
CN116227538B (en) * 2023-04-26 2023-07-11 国网山西省电力公司晋城供电公司 Clustering and deep learning-based low-current ground fault line selection method and equipment
CN118150943A (en) * 2024-05-10 2024-06-07 国网山西省电力公司太原供电公司 Distribution network ground fault positioning method, system, electronic equipment and storage medium

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103760464A (en) * 2014-01-07 2014-04-30 河南理工大学 Small current grounding system fault line selecting method based on analytic graph solving and SVM
EP2752674A1 (en) * 2013-01-03 2014-07-09 ABB Technology AG A detection method of a ground fault in an electric power distribution network
CN104297635A (en) * 2014-10-14 2015-01-21 河南理工大学 Fault line selection method for distribution network on basis of atom sparse decomposition and extreme learning machine
CN105759167A (en) * 2016-01-28 2016-07-13 江苏省电力公司南京供电公司 Wavelet neural network-based distribution network single-phase short circuit line selection method
CN106291234A (en) * 2016-07-29 2017-01-04 武汉大学 A kind of transmission line of electricity internal fault external fault based on convolutional neural networks judges and fault phase-selecting method
CN106951900A (en) * 2017-04-13 2017-07-14 杭州申昊科技股份有限公司 A kind of automatic identifying method of arrester meter reading
CN107766879A (en) * 2017-09-30 2018-03-06 中国南方电网有限责任公司 The MLP electric network fault cause diagnosis methods of feature based information extraction
CN107909118A (en) * 2017-12-11 2018-04-13 北京映翰通网络技术股份有限公司 A kind of power distribution network operating mode recording sorting technique based on deep neural network
CN107947156A (en) * 2017-11-24 2018-04-20 国网辽宁省电力有限公司 Based on the electric network fault critical clearing time method of discrimination for improving Softmax recurrence

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2752674A1 (en) * 2013-01-03 2014-07-09 ABB Technology AG A detection method of a ground fault in an electric power distribution network
CN103760464A (en) * 2014-01-07 2014-04-30 河南理工大学 Small current grounding system fault line selecting method based on analytic graph solving and SVM
CN104297635A (en) * 2014-10-14 2015-01-21 河南理工大学 Fault line selection method for distribution network on basis of atom sparse decomposition and extreme learning machine
CN105759167A (en) * 2016-01-28 2016-07-13 江苏省电力公司南京供电公司 Wavelet neural network-based distribution network single-phase short circuit line selection method
CN106291234A (en) * 2016-07-29 2017-01-04 武汉大学 A kind of transmission line of electricity internal fault external fault based on convolutional neural networks judges and fault phase-selecting method
CN106951900A (en) * 2017-04-13 2017-07-14 杭州申昊科技股份有限公司 A kind of automatic identifying method of arrester meter reading
CN107766879A (en) * 2017-09-30 2018-03-06 中国南方电网有限责任公司 The MLP electric network fault cause diagnosis methods of feature based information extraction
CN107947156A (en) * 2017-11-24 2018-04-20 国网辽宁省电力有限公司 Based on the electric network fault critical clearing time method of discrimination for improving Softmax recurrence
CN107909118A (en) * 2017-12-11 2018-04-13 北京映翰通网络技术股份有限公司 A kind of power distribution network operating mode recording sorting technique based on deep neural network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
超高压输电线路的潜供电弧频谱特性;林莘 等;《高电压技术》;20090831;第35卷(第8期);全文 *

Also Published As

Publication number Publication date
CN108562821A (en) 2018-09-21

Similar Documents

Publication Publication Date Title
CN108562821B (en) Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax
CN110082640B (en) Distribution network single-phase earth fault identification method based on long-time memory network
CN109359271B (en) Transformer winding deformation degree online detection method based on logistic regression
CN102789545A (en) Method for predicating remaining life of turbine engine based on degradation model matching
CN107132454A (en) The abnormal quick determination method of power network based on random matrix spectral radius method
Chang et al. Wire mismatch detection using a convolutional neural network and fault localization based on time–frequency-domain reflectometry
CN106447040A (en) Method for evaluating the health state of mechanical equipment based on heterogeneous multi-sensor data fusion
CN109100627A (en) A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode
CN108802535B (en) Screening method, main interference source identification method and device, server and storage medium
CN108152676B (en) Power distribution network fault positioning method and system based on intelligent agent
CN114414942A (en) Power transmission line fault identification classifier, identification method and system based on transient waveform image identification
CN108154271A (en) A kind of surface air temperature method of quality control based on spatial coherence and surface fitting
CN111678699B (en) Early fault monitoring and diagnosing method and system for rolling bearing
CN112098889A (en) Single-phase earth fault positioning method based on neural network and feature matrix
CN103267652B (en) Intelligent online diagnosis method for early failures of equipment
CN112684295A (en) Power distribution network fault line selection method and system under high permeability based on similarity separation degree
CN110108971B (en) Arc detection method for tree grounding fault of overhead bare conductor
CN113516406B (en) High-speed rail line rainfall measurement point arrangement method based on real-time observation and analysis
CN111999591B (en) Method for identifying abnormal state of primary equipment of power distribution network
CN108052953B (en) Sample expansion method based on feature correlation
CN114330120A (en) 24-hour PM prediction based on deep neural network2.5Method of concentration
CN106646138B (en) Distribution net work earthing fault localization method based on the conversion of more sample frequency wavelet character energy
CN116466408B (en) Artificial neural network superbedrock identification method based on aeromagnetic data
Papadopoulos et al. Online parameter identification and generic modeling derivation of a dynamic load model in distribution grids
CN109784777B (en) Power grid equipment state evaluation method based on time sequence information fragment cloud similarity measurement

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant