CN108535572B - Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics - Google Patents

Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics Download PDF

Info

Publication number
CN108535572B
CN108535572B CN201810443022.3A CN201810443022A CN108535572B CN 108535572 B CN108535572 B CN 108535572B CN 201810443022 A CN201810443022 A CN 201810443022A CN 108535572 B CN108535572 B CN 108535572B
Authority
CN
China
Prior art keywords
zero sequence
current
voltage
vector
representing
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
CN201810443022.3A
Other languages
Chinese (zh)
Other versions
CN108535572A (en
Inventor
王旭红
胡劼睿
刘星
李�浩
徐佳夫
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.)
Changsha University of Science and Technology
Original Assignee
Changsha University of Science and Technology
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 Changsha University of Science and Technology filed Critical Changsha University of Science and Technology
Priority to CN201810443022.3A priority Critical patent/CN108535572B/en
Publication of CN108535572A publication Critical patent/CN108535572A/en
Application granted granted Critical
Publication of CN108535572B publication Critical patent/CN108535572B/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
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
    • G01R19/25—Arrangements for measuring currents or voltages or for indicating presence or sign thereof using digital measurement techniques

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Remote Monitoring And Control Of Power-Distribution Networks (AREA)

Abstract

The invention discloses a metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics, which are used for acquiring three-phase voltage in a voltage circuit of a high-voltage metering system, three-phase current in a current circuit and neutral current; obtaining a zero-sequence voltage component and a zero-sequence current component by vector superposition; extracting amplitude and phase of fundamental wave zero sequence component; and establishing a deep belief network classifier by using the amplitude and the phase of the fundamental voltage, the amplitude and the phase of the fundamental zero-sequence point current, the absolute value of the difference between the fundamental zero-sequence current and the neutral current as input variables and various faults as output variables to realize the classification and identification of the faults. The invention can accurately extract the fault characteristics of the secondary circuit of the high-voltage metering system, accurately classify and identify various faults, so as to effectively remove the faults and monitor the secondary circuit in real time, thereby ensuring the safety, stability and accuracy of electric energy metering.

Description

Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics
Technical Field
The invention relates to a secondary circuit state monitoring and fault diagnosis technology of a high-voltage metering system, in particular to a metering system secondary circuit monitoring method and a metering system secondary circuit monitoring device based on fundamental wave zero sequence characteristics, which are used for state monitoring and fault diagnosis of a voltage circuit and a current circuit of the high-voltage metering system.
Background
With the development of economy in China, the power consumption is increased rapidly, and the requirements on power supply and distribution stability are higher and higher. The high-voltage electric power metering system is an important link for electric energy metering, and the safety and stability of the system are very important, so that the real-time monitoring of the states of the current and the voltage of the secondary loop of the high-voltage metering system is of great significance. The traditional voltage loss and current loss alarm device starts from the definition of voltage loss and current loss, realizes the monitoring of the voltage and current states, and brings about a plurality of bad conditions such as false alarm due to the change of the definition of voltage loss and current loss. These adverse conditions lead to the following problems: 1) the inaccurate electric energy metering is caused, and property loss is brought to users and power supply companies; 2) causing malfunction of the fault removal device and causing interference to the normal operation of the metering secondary circuit; 3) faults cannot be timely eliminated, and the stable operation of a metering secondary circuit is seriously threatened. Therefore, aiming at the problems that special faults cannot be found, misjudgment cannot be found, real-time monitoring capability is poor and the like of the conventional voltage and current loss alarm device, a proper method is researched and related devices are designed, so that the real-time monitoring of the voltage and current states is realized, the accuracy of identifying the voltage and current loss faults is improved, and the voltage and current loss alarm device has important significance for maintaining the safety and stability of a secondary circuit of a high-voltage metering system.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: aiming at the problems in the prior art, the invention provides a metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics.
In order to solve the technical problems, the invention adopts the technical scheme that:
a metering system secondary circuit monitoring method based on fundamental wave zero sequence characteristics comprises the following implementation steps:
1) aiming at a monitored high-voltage metering system, acquiring three-phase voltage U in a voltage loop of the high-voltage metering systema、Ub、UcObtaining three-phase current I in current loopa、Ib、IcAnd a neutral current In；
2) Will three-phase voltage Ua、Ub、UcVector superposition is carried out to obtain a zero sequence voltage component 3U containing a large amount of harmonic waves and white noise0Will make three-phase current Ia、Ib、IcVector superposition is carried out to obtain a zero sequence current component 3I containing a large amount of harmonic waves and white noise0；
3) Respectively according to the zero sequence voltage component 3U0And zero sequence current component 3I0Extracting fundamental wave zero sequence component to obtain amplitude 3U of fundamental wave zero sequence voltage01And phase
Figure GDA0002594694490000011
Amplitude 3I of fundamental wave zero sequence current01And phase
Figure GDA0002594694490000012
4) Calculating fundamental wave zero sequence current and neutral current InThe absolute value of the difference between the zero sequence current and the neutral current is obtainede；
5) The amplitude of the fundamental wave zero sequence voltage is 3U01Phase of fundamental wave zero sequence voltage
Figure GDA0002594694490000021
Amplitude 3I of fundamental wave zero sequence current01Phase of fundamental wave zero sequence current
Figure GDA0002594694490000022
And the absolute value I of the difference between the fundamental zero-sequence current and the neutral currenteAnd inputting the machine learning classification model which is used as an input vector to finish training, wherein the machine learning classification model is trained to comprise a mapping relation between the input vector and the secondary circuit fault which is used as an output result, and finally obtaining a monitoring result of the secondary circuit of the monitored high-voltage metering system.
Preferably, the detailed step of extracting the fundamental zero-sequence component in step 3) includes:
3.1) the value of the initialization iteration number k is 0, and the original zero sequence voltage component 3U is input0Or zero sequence current component 3I0The original zero sequence voltage component is 3U0Or zero sequence current component 3I0Randomly generating an initial sample and randomly sampling;
3.2) calculating the original zero sequence voltage component 3U0Or zero sequence current component 3I0A state value and a particle weight of the generated particle;
3.3) establishing a generalized regression neural network GRNN, and carrying out 3U (zero sequence) on the original zero-sequence voltage component through the generalized regression neural network GRNN0Or zero sequence current component 3I0Optimizing and adjusting the state value of the system;
3.4) calculating the effective particle number;
3.5) judging whether the number of the effective particles is less than a threshold value of the set number of the effective particles, if so, resampling, and then skipping to execute the next step; otherwise, directly skipping to execute the next step;
3.6) carrying out zero sequence voltage or zero sequence current state estimation and calculating likelihood probability density to obtain a zero sequence voltage component 3U corresponding to the current optimal state estimation value0Or zero sequence current component 3I0；
3.7) calculating the accumulated log-likelihood ratio and the maximum accumulated log-likelihood ratio;
3.8) judging whether the maximum accumulated log-likelihood ratio exceeds a judgment threshold value, if so, judging that the zero sequence voltage component 3U corresponding to the current optimal state estimation value is zero0Or zero sequence current component 3I0Saving, and then skipping to execute the next step; otherwise, directly skipping to execute the next step;
3.9) updating the state, and adding 1 to the iteration number k;
3.10) judging whether the iteration is finished or not, and if not, skipping to execute the step 3.2); otherwise, skipping to execute the next step;
3.11) amplitude of output fundamental wave zero sequence voltage 3U01And phase
Figure GDA0002594694490000026
Or amplitude of fundamental zero-sequence current 3I01And phase
Figure GDA0002594694490000023
Preferably, the detailed steps of step 3.3) include:
3.3.1) establishing a generalized regression neural network GRNN, wherein an input vector of the generalized regression neural network GRNN is defined as
Figure GDA0002594694490000024
The target vector is defined as zkTraining a generalized regression neural network GRNN according to the formula (1);
Figure GDA0002594694490000025
formula (A), (B) and1) in (1),
Figure GDA0002594694490000031
indicates the initial predicted value, YiDependent variable value of training sample, zkIn order to be the target vector,
Figure GDA0002594694490000032
as elements in the input vector, σ represents the width coefficient of the gaussian function, also called smoothing factor, and n represents the sample capacity;
3.3.2) constructing an n-dimensional vector
Figure GDA0002594694490000033
j△<L (j ═ 1,2, …, n/2), where
Figure GDA0002594694490000034
J delta is an element in the input vector, represents an adjustment value of each element of the input vector, a parameter L represents a defined adjustment range, and n represents a vector dimension;
3.3.3) vector n dimensions according to equation (2)
Figure GDA0002594694490000035
As an input vector, a predicted value obtained by training equation (1)
Figure GDA0002594694490000036
Further training a generalized regression neural network GRNN as a dependent variable value of a training sample;
Figure GDA0002594694490000037
in the formula (2), the reaction mixture is,
Figure GDA0002594694490000038
indicating the predicted value after the conversion and,
Figure GDA0002594694490000039
denotes the initial predicted value, zkIn order to be the target vector,
Figure GDA00025946944900000310
as elements in the input vector, σ represents the width coefficient of the gaussian function, also called smoothing factor, and n represents the sample capacity;
3.3.4) output vector z through generalized regression neural network GRNNkIndication of (2), the sample
Figure GDA00025946944900000311
Is optimized point
Figure GDA00025946944900000312
The substitution is carried out by the following steps,
Figure GDA00025946944900000313
where j Δ represents the adjustment value for each element of the input vector,
Figure GDA00025946944900000314
represented as elements in the input vector.
Preferably, the machine learning classification model in step 5) is a deep belief network classifier, and the training step of the deep belief network classifier includes:
s1) selecting fundamental wave zero sequence voltage, amplitude and phase of current of a secondary loop of a high-voltage metering system when normal, voltage loss, current loss and neutral line break as sample data and characteristic variables, wherein the voltage loss comprises reverse polarity connection of a voltage transformer, single-phase voltage break and internal faults of the voltage transformer, the current loss comprises reverse polarity connection of the current transformer, single-phase current break of the current loop and short circuit of a measuring coil, and the sample data is divided into a training set according to a certain proportion after standardized processing;
s2) encoding the state of the secondary circuit of the high-pressure metering system;
s3) establishing a secondary circuit fault classification and identification model based on the deep belief network classifier;
s4) initializing the parameters of the fault classification and identification model to make the parameters into a group of small random values which obey Gaussian distribution;
s5) selecting unlabeled samples in the training set, and pre-training the limited Boltzmann machine layer at the bottom of the model of the secondary circuit fault classification and recognition model through a contrast divergence algorithm;
s6) optimizing the whole network by BP algorithm by using label samples in the training set, and finishing the training of the secondary loop fault classification and recognition model based on the deep belief network classifier.
Preferably, the detailed step of step S5) includes:
s5.1) initializing the initial state v of the visible layer unit of the secondary loop fault classification and identification model0＝x0Initializing W, a and b as random comparative values obeying Gaussian distribution, and setting the maximum training iteration times of each limited Boltzmann machine layer; wherein v is0Representing the initial state vector, x, of the visible layer element0Representing training samples, W representing a connection weight matrix, a representing a bias vector of a visible layer, and b representing a bias vector of a hidden layer;
s5.2) calculating formula (3) for all hidden units of secondary circuit fault classification and identification model according to condition distribution P (h)0j|v0) Middle extraction h0～P(h0|v0) Wherein h is0jRepresenting the initial state value, v, of the jth neuron of the hidden layer0Represents the initial state vector of the visible layer element, h0An initial state vector representing the hidden layer;
Figure GDA0002594694490000041
in the formula (3), h0jRepresenting the initial state value, v, of the jth neuron of the hidden layer0Representing the initial state vector of the visible layer element, bjBias value, v, representing the jth neuron of the hidden layer0iRepresents the initial state value, W, of the ith neuron of the visible layer unitijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, n representing the number of visible layer nodes, and sigma () being a sigmoid function;
s5.3) calculating an expression (4) for all visible units of the secondary circuit fault classification and identification model according to the condition distribution P (v)1i|h0) Middle extract v1～P(v1|h0) Wherein v is1iRepresents the state value, v, of the ith neuron of the visible layer unit after 1 Gibbs sampling1Represents the state vector of the visible layer unit after 1 Gibbs sampling, h0An initial state vector representing the hidden layer;
Figure GDA0002594694490000042
in the formula (4), v1iRepresents the state value h of the ith neuron of the visible layer unit after 1 Gibbs sampling0Initial state vector representing the hidden layer, aiRepresents the bias value, h, of the ith neuron of the visible layer0jRepresents the initial state value, W, of the jth neuron in the hidden layerijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, wherein m represents the number of hidden layer nodes, and sigma () is a sigmoid function;
s5.4) calculating all hidden units of the secondary circuit fault classification and identification model according to the formula (5);
Figure GDA0002594694490000043
in the formula (5), h1jRepresents the state value, v, of the jth neuron of the hidden layer cell after 1 Gibbs sampling0Representing the initial state vector of the visible layer element, bjBias value, v, representing the jth neuron of the hidden layer1iRepresents the state value of the ith neuron of the visible layer unit after 1 Gibbs sampling, WijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, wherein n represents the number of visible layer nodes, and sigma () is a sigmoid function;
s5.5) updating parameters of the secondary circuit fault classification and identification model according to the formula (6);
Figure GDA0002594694490000044
in the formula (6), W represents a connection weight matrix, and a represents a visible layerB denotes a bias vector of the hidden layer, p denotes a learning rate, h denotes a learning rate0Initial state vector, v, representing the hidden layer0Represents the initial state vector of the visible layer element,
Figure GDA0002594694490000051
representing the transpose of the initial state vector of the visible layer unit, v1Representing the state vector of the visible layer cell after 1 gibbs sampling,
Figure GDA0002594694490000052
representing the transpose of the visible layer cell state vector after 1 gibbs sampling.
The present invention also provides a fundamental zero sequence signature-based metering system secondary loop monitoring device comprising computer equipment programmed to perform the steps of the fundamental zero sequence signature-based metering system secondary loop monitoring method of the present invention.
The metering system secondary circuit monitoring method based on the fundamental wave zero sequence characteristics has the following advantages: the invention discloses a metering system secondary circuit monitoring method based on fundamental wave zero sequence characteristics, which respectively obtains three-phase voltage in a voltage circuit of a high-voltage metering system, three-phase current in a current circuit and neutral current, obtains zero sequence voltage components and zero sequence current components containing a large amount of harmonic waves and white noise by vector superposition, extracts amplitude and phase of fundamental wave zero sequence components by using an important sample adjustment particle filter algorithm based on a neural network, takes the amplitude and phase of fundamental wave voltage, the amplitude and phase of fundamental wave zero sequence point current and the absolute value of fundamental wave zero sequence current and neutral current difference as input variables, takes various faults as output variables, establishes a high-voltage metering system secondary circuit identification mechanism based on a deep belief network, and the identification mechanism comprises mapping relations between five input variables and various faults as output variables, and finally realizes classification and identification of the faults, compared with the existing no-voltage no-current alarm device, the device has the advantages of accurately identifying the fault, accurately positioning the fault type, effectively identifying the fault of the neutral line, avoiding false alarm and the like. The invention can effectively remove harmonic waves and white noise in the zero-sequence component and realize the extraction of the fundamental zero-sequence component; the general defects of the existing alarm device (the failure of identifying the neutral line disconnection fault, the easy occurrence of false alarm and the poor real-time monitoring capability) can be overcome, the fault can be accurately alarmed, and the state of the metering secondary circuit can be monitored in real time, so that the safety and the stability of a high-voltage metering system can be guaranteed.
The metering system secondary circuit monitoring device based on the fundamental wave zero sequence characteristic is a device formed by program units completely corresponding to the metering system secondary circuit monitoring method based on the fundamental wave zero sequence characteristic, so that the metering system secondary circuit monitoring device based on the fundamental wave zero sequence characteristic has the advantages of the metering system secondary circuit monitoring method based on the fundamental wave zero sequence characteristic, and is not repeated herein.
Drawings
FIG. 1 is a schematic diagram of a basic flow of a method according to an embodiment of the present invention.
FIG. 2 is a three-phase voltage waveform of a voltage loop in an embodiment of the present invention.
Fig. 3 is a three-phase current waveform of the current loop in the embodiment of the invention.
Fig. 4 is a flow chart of fundamental zero-sequence component extraction based on improved particle filtering in the embodiment of the present invention.
Fig. 5 is an original waveform of zero sequence voltage of the voltage loop in the embodiment of the present invention.
Fig. 6 is an original waveform of zero sequence current of the current loop in the embodiment of the present invention.
Fig. 7 shows the amplitude of the fundamental zero-sequence voltage of the method according to the embodiment of the invention.
Fig. 8 shows the phase of the fundamental zero-sequence voltage of the method according to the embodiment of the invention.
Fig. 9 shows the amplitude of the fundamental zero-sequence current of the method according to the embodiment of the invention.
Fig. 10 shows the phase of the fundamental zero-sequence current of the method according to the embodiment of the invention.
Fig. 11 is a flow chart of the fault diagnosis of the secondary circuit of the high-voltage metering system in the embodiment of the invention.
Fig. 12 is a schematic diagram of a secondary circuit fault diagnosis model of the high-pressure metering system in the embodiment of the invention.
Detailed Description
As shown in fig. 1, the implementation steps of the metering system secondary circuit monitoring method based on the fundamental zero sequence feature in this embodiment include:
1) aiming at a monitored high-voltage metering system, acquiring three-phase voltage U in a voltage loop of the high-voltage metering systema、Ub、UcObtaining three-phase current I in current loopa、Ib、IcAnd a neutral current In；
2) Will three-phase voltage Ua、Ub、UcVector superposition is carried out to obtain a zero sequence voltage component 3U containing a large amount of harmonic waves and white noise0Will make three-phase current Ia、Ib、IcVector superposition is carried out to obtain a zero sequence current component 3I containing a large amount of harmonic waves and white noise0；
3) Respectively according to the zero sequence voltage component 3U0And zero sequence current component 3I0Extracting fundamental wave zero sequence component to obtain amplitude 3U of fundamental wave zero sequence voltage01And phase
Figure GDA0002594694490000061
Amplitude 3I of fundamental wave zero sequence current01And phase
Figure GDA0002594694490000062
4) Calculating fundamental wave zero sequence current and neutral current InThe absolute value of the difference between the zero sequence current and the neutral current is obtainede(writable as function expression Ie＝|3I01-In|)；
5) The amplitude of the fundamental wave zero sequence voltage is 3U01Phase of fundamental wave zero sequence voltage
Figure GDA0002594694490000063
Amplitude 3I of fundamental wave zero sequence current01Phase of fundamental wave zero sequence current
Figure GDA0002594694490000064
And fundamental zero sequence current and neutralAbsolute value of line current difference IeAnd inputting the machine learning classification model which is used as an input vector to finish training, wherein the machine learning classification model is trained to comprise a mapping relation between the input vector and the secondary circuit fault which is used as an output result, and finally obtaining a monitoring result of the secondary circuit of the monitored high-voltage metering system.
In the embodiment, step 1) measures three-phase voltage U of a voltage loop of the high-voltage metering system through a DL-PT202H1 current type voltage transformera、Ub、UcMeasuring three-phase current I of current loop of high-voltage metering system by HBC-LSP closed-loop Hall current sensora、Ib、IcAnd a neutral current In(ii) a Aiming at the acquired signals, the acquired voltage and current are controlled in the range in which a reprocessor can normally work through circuits such as RC filtering, proportional operational amplifier, voltage raising and the like, and the processed voltage and current are converted into analog quantities through A/D conversion and then input into a central processing unit. The three-phase voltage waveforms acquired in the embodiment are shown in fig. 2, wherein (a) represents the three-phase voltage waveforms when the voltage circuit normally operates, (b) represents the three-phase voltage waveforms when the voltage circuit is disconnected in a single phase, (c) represents the three-phase voltage waveforms when the voltage transformer has an internal fault, and (d) represents the three-phase voltage waveforms when the polarity of the voltage transformer is reversed; the acquired three-phase current waveforms are shown in fig. 3, wherein (a) represents the three-phase current waveforms when the current loop normally operates, (b) represents the three-phase current waveforms when the current loop is disconnected from a single phase, (c) represents the three-phase current waveforms when the polarity of the current transformer is reversed, and (d) represents the three-phase current waveforms when the current loop measuring coil is short-circuited.
As shown in fig. 4, the detailed steps of extracting the fundamental zero-sequence component in step 3) include:
3.1) the value of the initialization iteration number k is 0, and the original zero sequence voltage component 3U is input0Or zero sequence current component 3I0The original zero sequence voltage component is 3U0Or zero sequence current component 3I0Randomly generating an initial sample and randomly sampling;
in this embodiment, the measurement equation ZkAs shown in the following formula:
Figure GDA0002594694490000071
in the above formula, k represents the number of sampling cycles, TsRepresenting the sampling period, ωnDenotes the angular velocity of the nth harmonic, τ denotes the decay time constant, AnDenotes the amplitude of the nth harmonic wave, phinIndicating the initial phase of the nth harmonic, A0Representing the amplitude of the attenuated DC component, HkRepresenting a measurement matrix, xkRepresents a state quantity, vkMeans zero mean and variance RkN represents the number of included harmonics. The equation of state is shown below:
Figure GDA0002594694490000072
in the above formula, x2nRepresenting the state quantity at time k, k representing the number of sampling cycles, n being the harmonic order, A0Representing the amplitude, w, of the attenuated DC componentk-1Means zero mean and Q variancek-1The process noise of (1). When k is 0, the prior probability density distribution function P (x)0) Randomly generating samples according to P (x)0) Distributed sampling to obtain
Figure GDA0002594694490000073
And performing state prediction on the zero sequence voltage component and the zero sequence current component, and sampling according to a state equation of the zero sequence voltage component and the zero sequence current component:
Figure GDA0002594694490000074
wherein
Figure GDA0002594694490000075
Representing the state value of the ith particle at time k, q representing the important density function, xkWhich represents the state quantity at time k,
Figure GDA0002594694490000076
indicates the beginningThe state value of the ith particle at the starting time, k represents the sampling time, z0The initial time measurement value is shown, and N represents the number of particles.
3.2) calculating the original zero sequence voltage component 3U0Or zero sequence current component 3I0A state value and a particle weight of the generated particle;
the calculation function expression of the particle weight in this embodiment is as follows:
Figure GDA0002594694490000077
in the above formula, the first and second carbon atoms are,
Figure GDA0002594694490000078
represents the weight value of the ith particle at time k,
Figure GDA0002594694490000079
represents the weight value of the ith particle at the time k-1, p represents the prior distribution density, zkThe measured value at the time k is shown,
Figure GDA00025946944900000710
representing the state value of the ith particle at time k-1,
Figure GDA00025946944900000711
the state value of the ith particle at time k is shown, and N is the number of particles. The calculated weight of the particles needs to be normalized by the following formula:
Figure GDA0002594694490000081
in the above formula, the first and second carbon atoms are,
Figure GDA0002594694490000082
the weight value of the ith particle at time k is represented, and N represents the number of particles.
3.3) establishing a generalized regression neural network GRNN, and carrying out 3U (zero sequence) on the original zero-sequence voltage component through the generalized regression neural network GRNN0Or zero sequence current component 3I0Optimizing and adjusting the state value of the system;
3.4) calculating the effective particle number;
3.5) judging whether the number of the effective particles is less than a threshold value of the set number of the effective particles, if so, resampling, and then skipping to execute the next step; otherwise, directly skipping to execute the next step;
3.6) carrying out zero sequence voltage or zero sequence current state estimation and calculating likelihood probability density to obtain a zero sequence voltage component 3U corresponding to the current optimal state estimation value0Or zero sequence current component 3I0；
When estimating the state of the zero-sequence voltage or the zero-sequence current in the embodiment, the optimal state estimation value is shown as the following formula;
Figure GDA0002594694490000083
in the above formula, the first and second carbon atoms are,
Figure GDA0002594694490000084
which represents an estimate of the particle at time k,
Figure GDA0002594694490000085
represents the weight value of the ith particle at time k, N represents the number of particles,
Figure GDA0002594694490000086
indicating the state quantity of the ith particle at time k.
According to the optimal state estimation value, the amplitude of the fundamental wave zero-sequence voltage or the fundamental wave zero-sequence current under the optimal state estimation value can be calculated:
Figure GDA0002594694490000087
in the above formula, A1Representing the amplitude, x, of the fundamental zero-sequence voltage/current under the optimal state estimation1An estimated value, x, representing the 1 st state quantity2An estimated value of the 2 nd state quantity is shown.
According to the optimal state estimation value, the phase of the fundamental wave zero-sequence voltage or the fundamental wave zero-sequence current under the optimal state estimation value can be calculated:
Figure GDA0002594694490000088
in the above formula, phi1Representing the phase, x, of the fundamental zero-sequence voltage/current at the optimum state estimate1An estimated value, x, representing the 1 st state quantity2An estimated value of the 2 nd state quantity is shown.
3.7) calculating the accumulated log-likelihood ratio and the maximum accumulated log-likelihood ratio;
3.8) judging whether the maximum accumulated log-likelihood ratio exceeds a judgment threshold value, if so, judging that the zero sequence voltage component 3U corresponding to the current optimal state estimation value is zero0Or zero sequence current component 3I0Saving, and then skipping to execute the next step; otherwise, directly skipping to execute the next step;
3.9) updating the state, and adding 1 to the iteration number k;
3.10) judging whether the iteration is finished or not, and if not, skipping to execute the step 3.2); otherwise, skipping to execute the next step;
3.11) amplitude of output fundamental wave zero sequence voltage 3U01And phase
Figure GDA00025946944900000917
Or amplitude of fundamental zero-sequence current 3I01And phase
Figure GDA00025946944900000918
In this embodiment, the detailed steps of step 3.3) include:
3.3.1) establishing a generalized regression neural network GRNN, wherein an input vector of the generalized regression neural network GRNN is defined as
Figure GDA0002594694490000091
The target vector is defined as zkTraining a generalized regression neural network GRNN according to the formula (1);
Figure GDA0002594694490000092
in the formula (1), the reaction mixture is,
Figure GDA0002594694490000093
indicates the initial predicted value, YiDependent variable value of training sample, zkIn order to be the target vector,
Figure GDA0002594694490000094
as elements in the input vector, σ represents the width coefficient of the gaussian function, also called smoothing factor, and n represents the sample capacity;
3.3.2) constructing an n-dimensional vector
Figure GDA0002594694490000095
Wherein
Figure GDA0002594694490000096
J delta is an element in the input vector, represents an adjustment value of each element of the input vector, a parameter L represents a defined adjustment range, and n represents a vector dimension;
3.3.3) vector n dimensions according to equation (2)
Figure GDA0002594694490000097
As an input vector, a predicted value obtained by training equation (1)
Figure GDA0002594694490000098
Further training a generalized regression neural network GRNN as a dependent variable value of a training sample;
Figure GDA0002594694490000099
in the formula (2), the reaction mixture is,
Figure GDA00025946944900000910
indicating the predicted value after the conversion and,
Figure GDA00025946944900000911
denotes the initial predicted value, zkIn order to be the target vector,
Figure GDA00025946944900000912
as elements in the input vector, σ represents the width coefficient of the gaussian function, also called smoothing factor, and n represents the sample capacity;
3.3.4) output vector z through generalized regression neural network GRNNkIndication of (2), the sample
Figure GDA00025946944900000913
Is optimized point
Figure GDA00025946944900000914
The substitution is carried out by the following steps,
Figure GDA00025946944900000915
where j Δ represents the adjustment value for each element of the input vector,
Figure GDA00025946944900000916
represented as elements in the input vector.
The original waveform of the zero-sequence voltage obtained by vector superposition is shown in fig. 5, and the original waveform of the zero-sequence current is shown in fig. 6. The amplitude and the phase of the fundamental zero-sequence component are extracted through the neural network-based important sample adjustment particle filter algorithm, and the amplitude and the phase of the obtained fundamental zero-sequence voltage are shown in fig. 7-8, where fig. 7 is the amplitude of the extracted fundamental zero-sequence voltage, and fig. 8 is the phase of the extracted fundamental zero-sequence voltage. The amplitude and phase of the fundamental zero-sequence current are shown in fig. 9-10, where fig. 9 is the amplitude of the extracted fundamental zero-sequence current, and fig. 10 is the phase of the extracted fundamental zero-sequence current.
In this embodiment, the machine learning classification model in step 5) is based on a deep belief network classifier, and as shown in fig. 11, the training step based on the deep belief network classifier includes:
s1) selecting fundamental wave zero sequence voltage, amplitude and phase of current of a secondary loop of a high-voltage metering system when normal, voltage loss, current loss and neutral line break as sample data and characteristic variables, wherein the voltage loss comprises reverse polarity connection of a voltage transformer, single-phase voltage break and internal faults of the voltage transformer, the current loss comprises reverse polarity connection of the current transformer, single-phase current break of the current loop and short circuit of a measuring coil, and the sample data is divided into a training set according to a certain proportion after standardized processing;
s2) encoding the state of the secondary circuit of the high-pressure metering system, as shown in table 1;
table 1: and (5) encoding the state of the secondary loop.
Figure GDA0002594694490000101
S3) establishing a secondary circuit fault classification and identification model based on the deep belief network classifier;
as shown in fig. 12, the recognition mechanism includes 5 inputs, respectively: the absolute value of the amplitude and the phase of the fundamental wave voltage, the amplitude and the phase of the fundamental wave zero sequence point current, and the absolute value of the fundamental wave zero sequence current and the neutral line current difference comprises 8 outputs which are respectively: the polarity of the voltage transformer is connected reversely, the voltage loop single-phase line is broken, the voltage transformer has internal faults, the polarity of the current transformer is connected reversely, the current loop single-phase line is broken, the measuring coil is in short circuit, and the neutral line is broken.
S4) initializing the parameters of the fault classification and identification model to make the parameters into a group of small random values which obey Gaussian distribution;
s5) selecting unlabeled samples in a training set, and pre-training a model bottom limited Boltzmann Machine (RBM) layer of a secondary circuit fault classification and identification model through a contrast Divergence algorithm (CD algorithm);
s6) optimizing the whole network by BP algorithm by using label samples in the training set, and finishing the training of the secondary loop fault classification and recognition model based on the deep belief network classifier.
In this embodiment, the detailed step of step S5) includes:
s5.1) initializing the initial stage of visible layer units of the secondary circuit fault classification and identification modelInitial state v0＝x0Initializing W, a and b as random comparative values obeying Gaussian distribution, and setting the maximum training iteration times of each limited Boltzmann machine layer; wherein v is0Representing the initial state vector, x, of the visible layer element0Representing training samples, W representing a connection weight matrix, a representing a bias vector of a visible layer, and b representing a bias vector of a hidden layer;
s5.2) calculating formula (3) for all hidden units of secondary circuit fault classification and identification model according to condition distribution P (h)0j|v0) Middle extraction h0～P(h0|v0) Wherein h is0jRepresenting the initial state value, v, of the jth neuron of the hidden layer0Represents the initial state vector of the visible layer element, h0An initial state vector representing the hidden layer;
Figure GDA0002594694490000111
in the formula (3), h0jRepresenting the initial state value, v, of the jth neuron of the hidden layer0Representing the initial state vector of the visible layer element, bjBias value, v, representing the jth neuron of the hidden layer0iRepresents the initial state value, W, of the ith neuron of the visible layer unitijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, n representing the number of visible layer nodes, and sigma () being a sigmoid function;
s5.3) calculating an expression (4) for all visible units of the secondary circuit fault classification and identification model according to the condition distribution P (v)1i|h0) Middle extract v1～P(v1|h0) Wherein v is1iRepresents the state value, v, of the ith neuron of the visible layer unit after 1 Gibbs sampling1Represents the state vector of the visible layer unit after 1 Gibbs sampling, h0An initial state vector representing the hidden layer;
Figure GDA0002594694490000112
in the formula (4), v1iIndicates 1 order jeanThe state value h of the ith neuron of the visible layer unit after sampling0Initial state vector representing the hidden layer, aiRepresents the bias value, h, of the ith neuron of the visible layer0jRepresents the initial state value, W, of the jth neuron in the hidden layerijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, wherein m represents the number of hidden layer nodes, and sigma () is a sigmoid function;
s5.4) calculating all hidden units of the secondary circuit fault classification and identification model according to the formula (5);
Figure GDA0002594694490000113
in the formula (5), h1jRepresents the state value, v, of the jth neuron of the hidden layer cell after 1 Gibbs sampling0Representing the initial state vector of the visible layer element, bjBias value, v, representing the jth neuron of the hidden layer1iRepresents the state value of the ith neuron of the visible layer unit after 1 Gibbs sampling, WijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, wherein n represents the number of visible layer nodes, and sigma () is a sigmoid function;
s5.5) updating parameters of the secondary circuit fault classification and identification model according to the formula (6);
Figure GDA0002594694490000114
in the formula (6), W represents a connection weight matrix, a represents a bias vector of a visible layer, b represents a bias vector of a hidden layer, ρ represents a learning rate, and h represents a learning rate0Initial state vector, v, representing the hidden layer0Represents the initial state vector of the visible layer element,
Figure GDA0002594694490000115
representing the transpose of the initial state vector of the visible layer unit, v1Representing the state vector of the visible layer cell after 1 gibbs sampling,
Figure GDA0002594694490000116
representing the transpose of the visible layer cell state vector after 1 gibbs sampling.
The metering system secondary circuit monitoring method based on the fundamental wave zero sequence characteristic is specifically realized by MATLAB and a computer program (c #), and an important sample adjustment particle filter algorithm based on a neural network and a high-voltage metering system secondary circuit fault classification and identification mechanism based on a deep belief network classifier are realized by the MATLAB.
The following data are identification results obtained by extracting the amplitude and the phase of the fundamental zero-sequence component by using an important sample adjustment particle filter algorithm based on a neural network when the secondary circuit of the 10kV high-voltage metering system is in various running states, and inputting the amplitude and the phase into a secondary circuit fault classification and identification mechanism of the high-voltage metering system based on a deep belief network classifier, as shown in Table 2. Wherein the parameters are set as: a5-layer network structure is selected, the number of RBM hidden layer units is 500, the number of training samples is 1000, and the number of testing samples is 500.
Table 2: and (4) identifying a secondary circuit fault of the 10kV metering system based on the DBNC.
Figure GDA0002594694490000121
As can be seen from table 2, the metering system secondary circuit monitoring method based on the fundamental zero sequence feature of the embodiment has a good effect of identifying the secondary circuit fault of the high-voltage metering system, and the accuracy of fault identification is basically maintained above 95%.
In summary, in the metering system secondary circuit monitoring method based on the fundamental zero-sequence feature of the embodiment, by obtaining the three-phase voltage in the voltage circuit, the three-phase current in the current circuit and the neutral current in the high-voltage metering system, the zero-sequence voltage component and the zero-sequence current component containing a large amount of harmonics and white noise are obtained by vector superposition; extracting fundamental zero-sequence component amplitude and phase by using an important sample adjustment particle filter algorithm based on a neural network; the amplitude and the phase of fundamental wave voltage, the amplitude and the phase of fundamental wave zero sequence point current and the absolute value of the difference between the fundamental wave zero sequence point current and neutral line current are used as input variables, various faults are used as output variables, a secondary loop identification mechanism of the high-voltage metering system based on the deep belief network classifier is established, the identification mechanism comprises the mapping relation between five input variables and faults as output variables so as to classify and identify the faults, can effectively remove harmonic waves and white noise in zero-sequence components to extract fundamental zero-sequence components, can accurately extract the fault characteristics of a secondary circuit of a high-voltage metering system, accurately classify and identify various faults so as to effectively remove the faults and monitor the secondary circuit in real time, therefore, the safety, stability and accuracy of electric energy metering are guaranteed, and the state of the metering secondary loop is monitored in real time, so that the safety and stability of a high-voltage metering system are guaranteed. Compared with the existing voltage and current loss alarm method, the method has the advantages of accurately identifying the fault, accurately positioning the fault type, effectively identifying the fault of the neutral line, avoiding false alarm and the like.
The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may occur to those skilled in the art without departing from the principle of the invention, and are considered to be within the scope of the invention.

Claims (5)

1. A metering system secondary circuit monitoring method based on fundamental wave zero sequence characteristics is characterized by comprising the following implementation steps:
1) aiming at a monitored high-voltage metering system, acquiring three-phase voltage U in a voltage loop of the high-voltage metering systema、Ub、UcObtaining three-phase current I in current loopa、Ib、IcAnd a neutral current In；
2) Will three-phase voltage Ua、Ub、UcVector superposition is carried out to obtain a zero sequence voltage component 3U containing a large amount of harmonic waves and white noise0Will make three-phase current Ia、Ib、IcVector superposition is carried out to obtain zero sequence current component containing a large amount of harmonic waves and white noiseAmount 3I0；
3) Respectively according to the zero sequence voltage component 3U0And zero sequence current component 3I0Extracting fundamental wave zero sequence component to obtain amplitude 3U of fundamental wave zero sequence voltage01And phase
Figure FDA0002615200440000011
Amplitude 3I of fundamental wave zero sequence current01And phase
Figure FDA0002615200440000012
4) Calculating fundamental wave zero sequence current and neutral current InThe absolute value of the difference between the zero sequence current and the neutral current is obtainede；
5) The amplitude of the fundamental wave zero sequence voltage is 3U01Phase of fundamental wave zero sequence voltage
Figure FDA0002615200440000013
Amplitude 3I of fundamental wave zero sequence current01Phase of fundamental wave zero sequence current
Figure FDA0002615200440000014
And the absolute value I of the difference between the fundamental zero-sequence current and the neutral currenteInputting a machine learning classification model which is used as an input vector and completes training, wherein the machine learning classification model is trained to comprise a mapping relation between the input vector and a secondary circuit fault which is used as an output result, and finally obtaining a monitoring result of a secondary circuit of the monitored high-voltage metering system;
the detailed steps for extracting the fundamental wave zero sequence component in the step 3) comprise:
3.1) the value of the initialization iteration number k is 0, and the original zero sequence voltage component 3U is input0Or zero sequence current component 3I0The original zero sequence voltage component is 3U0Or zero sequence current component 3I0Randomly generating an initial sample and randomly sampling;
3.2) calculating the original zero sequence voltage component 3U0Or zeroSequence current component 3I0A state value and a particle weight of the generated particle;
3.3) establishing a generalized regression neural network GRNN, and carrying out 3U (zero sequence) on the original zero-sequence voltage component through the generalized regression neural network GRNN0Or zero sequence current component 3I0Optimizing and adjusting the state value of the system;
3.4) calculating the effective particle number;
3.5) judging whether the number of the effective particles is less than a threshold value of the set number of the effective particles, if so, resampling, and then skipping to execute the next step; otherwise, directly skipping to execute the next step;
3.6) carrying out zero sequence voltage or zero sequence current state estimation and calculating likelihood probability density to obtain a zero sequence voltage component 3U corresponding to the current optimal state estimation value0Or zero sequence current component 3I0；
3.7) calculating the accumulated log-likelihood ratio and the maximum accumulated log-likelihood ratio;
3.8) judging whether the maximum accumulated log-likelihood ratio exceeds a judgment threshold value, if so, judging that the zero sequence voltage component 3U corresponding to the current optimal state estimation value is zero0Or zero sequence current component 3I0Saving, and then skipping to execute the next step; otherwise, directly skipping to execute the next step;
3.9) updating the state, and adding 1 to the iteration number k;
3.10) judging whether the iteration is finished or not, and if not, skipping to execute the step 3.2); otherwise, skipping to execute the next step;
3.11) amplitude of output fundamental wave zero sequence voltage 3U01And phase
Figure FDA0002615200440000021
Or amplitude of fundamental zero-sequence current 3I01And phase
Figure FDA0002615200440000022
2. The metering system secondary circuit monitoring method based on fundamental wave zero sequence characteristics as claimed in claim 1, wherein the detailed step of step 3.3) comprises:
3.3.1) establishing a generalized regression neural network GRNN, wherein an input vector of the generalized regression neural network GRNN is defined as
Figure FDA0002615200440000023
The target vector is defined as zkTraining a generalized regression neural network GRNN according to the formula (1);
Figure FDA0002615200440000024
in the formula (1), the reaction mixture is,
Figure FDA0002615200440000025
indicates the initial predicted value, YiDependent variable value of training sample, zkIn order to be the target vector,
Figure FDA0002615200440000026
as elements in the input vector, σ represents the width coefficient of the gaussian function, also called smoothing factor, and n represents the sample capacity;
3.3.2) constructing an n-dimensional vector
Figure FDA0002615200440000027
Wherein
Figure FDA0002615200440000028
J delta is an element in the input vector, represents an adjustment value of each element of the input vector, a parameter L represents a defined adjustment range, and n represents a vector dimension;
3.3.3) vector n dimensions according to equation (2)
Figure FDA0002615200440000029
As an input vector, a predicted value obtained by training equation (1)
Figure FDA00026152004400000210
As a function of training samplesMeasuring values, and further training a generalized regression neural network GRNN;
Figure FDA00026152004400000211
in the formula (2), the reaction mixture is,
Figure FDA00026152004400000212
indicating the predicted value after the conversion and,
Figure FDA00026152004400000213
denotes the initial predicted value, zkIn order to be the target vector,
Figure FDA00026152004400000214
as elements in the input vector, σ represents the width coefficient of the gaussian function, also called smoothing factor, and n represents the sample capacity;
3.3.4) output vector z through generalized regression neural network GRNNkIndication of (2), the sample
Figure FDA00026152004400000215
Is optimized point
Figure FDA00026152004400000216
The substitution is carried out by the following steps,
Figure FDA00026152004400000217
where j delta represents the adjustment value of each element of the input vector,
Figure FDA00026152004400000218
represented as elements in the input vector.
3. The fundamental zero-sequence feature-based metering system secondary loop monitoring method as claimed in claim 1, wherein the machine learning classification model in step 5) is a deep belief network classifier, and the deep belief network classifier-based training step comprises:
s1) selecting fundamental wave zero sequence voltage, amplitude and phase of current of a secondary loop of a high-voltage metering system when normal, voltage loss, current loss and neutral line break as sample data and characteristic variables, wherein the voltage loss comprises reverse polarity connection of a voltage transformer, single-phase voltage break and internal faults of the voltage transformer, the current loss comprises reverse polarity connection of the current transformer, single-phase current break of the current loop and short circuit of a measuring coil, and the sample data is divided into a training set according to a certain proportion after standardized processing;
s2) encoding the state of the secondary circuit of the high-pressure metering system;
s3) establishing a secondary circuit fault classification and identification model based on the deep belief network classifier;
s4) initializing the parameters of the fault classification and identification model to make the parameters into a group of small random values which obey Gaussian distribution;
s5) selecting unlabeled samples in the training set, and pre-training the limited Boltzmann machine layer at the bottom of the model of the secondary circuit fault classification and recognition model through a contrast divergence algorithm;
s6) optimizing the whole network by BP algorithm by using label samples in the training set, and finishing the training of the secondary loop fault classification and recognition model based on the deep belief network classifier.
4. The metering system secondary circuit monitoring method based on fundamental wave zero sequence characteristics as claimed in claim 3, wherein the detailed step of the step S5) comprises:
s5.1) initializing the initial state v of the visible layer unit of the secondary loop fault classification and identification model0＝x0Initializing W, a and b as random comparative values obeying Gaussian distribution, and setting the maximum training iteration times of each limited Boltzmann machine layer; wherein v is0Representing the initial state vector, x, of the visible layer element0Representing training samples, W representing a connection weight matrix, a representing a bias vector of a visible layer, and b representing a bias vector of a hidden layer;
s5.2) classifying and identifying all hidden lists of secondary circuit faultsThe meta-value is calculated by formula (3) from the conditional distribution P (h)0j|v0) Middle extraction h0～P(h0|v0) Wherein h is0jRepresenting the initial state value, v, of the jth neuron of the hidden layer0Represents the initial state vector of the visible layer element, h0An initial state vector representing the hidden layer;
Figure FDA0002615200440000031
in the formula (3), h0jRepresenting the initial state value, v, of the jth neuron of the hidden layer0Representing the initial state vector of the visible layer element, bjBias value, v, representing the jth neuron of the hidden layer0iRepresents the initial state value, W, of the ith neuron of the visible layer unitijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, n representing the number of visible layer nodes, and sigma () being a sigmoid function;
s5.3) calculating an expression (4) for all visible units of the secondary circuit fault classification and identification model according to the condition distribution P (v)1i|h0) Middle extract v1～P(v1|h0) Wherein v is1iRepresents the state value, v, of the ith neuron of the visible layer unit after 1 Gibbs sampling1Represents the state vector of the visible layer unit after 1 Gibbs sampling, h0An initial state vector representing the hidden layer;
Figure FDA0002615200440000032
in the formula (4), v1iRepresents the state value h of the ith neuron of the visible layer unit after 1 Gibbs sampling0Initial state vector representing the hidden layer, aiRepresents the bias value, h, of the ith neuron of the visible layer0jRepresents the initial state value, W, of the jth neuron in the hidden layerijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, wherein m represents the number of hidden layer nodes, and sigma () is a sigmoid function;
s5.4) calculating all hidden units of the secondary circuit fault classification and identification model according to the formula (5);
Figure FDA0002615200440000041
in the formula (5), h1jRepresents the state value, v, of the jth neuron of the hidden layer cell after 1 Gibbs sampling0Representing the initial state vector of the visible layer element, bjBias value, v, representing the jth neuron of the hidden layer1iRepresents the state value of the ith neuron of the visible layer unit after 1 Gibbs sampling, WijRepresenting the connection weight value between a visible layer node i and a hidden layer node j, wherein n represents the number of visible layer nodes, and sigma () is a sigmoid function;
s5.5) updating parameters of the secondary circuit fault classification and identification model according to the formula (6);
Figure FDA0002615200440000042
in the formula (6), W represents a connection weight matrix, a represents a bias vector of a visible layer, b represents a bias vector of a hidden layer, ρ represents a learning rate, and h represents a learning rate0Initial state vector, v, representing the hidden layer0Represents the initial state vector of the visible layer element,
Figure FDA0002615200440000043
representing the transpose of the initial state vector of the visible layer unit, v1Representing the state vector of the visible layer cell after 1 gibbs sampling,
Figure FDA0002615200440000044
representing the transpose of the visible layer cell state vector after 1 gibbs sampling.
5. A metering system secondary circuit monitoring device based on fundamental wave zero sequence characteristics, comprising a computer device, wherein the computer device is programmed to execute the steps of the metering system secondary circuit monitoring method based on fundamental wave zero sequence characteristics according to any one of claims 1-4.
CN201810443022.3A 2018-05-10 2018-05-10 Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics Active CN108535572B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810443022.3A CN108535572B (en) 2018-05-10 2018-05-10 Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810443022.3A CN108535572B (en) 2018-05-10 2018-05-10 Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics

Publications (2)

Publication Number Publication Date
CN108535572A CN108535572A (en) 2018-09-14
CN108535572B true CN108535572B (en) 2020-10-23

Family

ID=63476792

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810443022.3A Active CN108535572B (en) 2018-05-10 2018-05-10 Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics

Country Status (1)

Country Link
CN (1) CN108535572B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109753705B (en) * 2018-12-24 2020-04-03 北京华大九天软件有限公司 IC initial value estimation method in integrated circuit design
CN109799423B (en) * 2019-01-09 2021-05-07 西安科技大学 Cable fault on-line diagnosis method
CN109932675A (en) * 2019-03-27 2019-06-25 国家电网有限公司 The method for detecting abnormality of current transformer loop
CN109980614B (en) * 2019-03-29 2020-12-29 广东电网有限责任公司 Self-adaptive discrimination method for zero sequence protection direction of distribution line
CN114002547B (en) * 2021-10-25 2024-02-02 国网湖北省电力有限公司恩施供电公司 Equipment for judging secondary multipoint ground fault of transformer and analysis algorithm
CN117269843B (en) * 2023-11-21 2024-04-19 云南电网有限责任公司 On-line monitoring method and system for neutral line running state of secondary current loop

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107449979B (en) * 2017-06-19 2019-11-29 国网湖南省电力公司计量中心 A kind of Current Voltage state monitoring method and device based on fundamental wave zero sequence feature
CN107609569B (en) * 2017-07-31 2020-11-03 北京映翰通网络技术股份有限公司 Power distribution network ground fault positioning method based on multi-dimensional feature vectors
CN107817404B (en) * 2017-11-18 2023-06-20 广西电网有限责任公司电力科学研究院 Portable metering automation terminal fault diagnosis device and diagnosis method thereof

Also Published As

Publication number Publication date
CN108535572A (en) 2018-09-14

Similar Documents

Publication Publication Date Title
CN108535572B (en) Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics
CN109635928B (en) Voltage sag reason identification method based on deep learning model fusion
CN105548862B (en) A kind of analog-circuit fault diagnosis method based on broad sense multi-kernel support vector machine
CN109145516B (en) Analog circuit fault identification method based on improved extreme learning machine
CN110414412B (en) Wide-area power grid multiple disturbance accurate identification method and device based on big data analysis
CN112051481A (en) Alternating current-direct current hybrid power grid fault area diagnosis method and system based on LSTM
CN111160241B (en) Power distribution network fault classification method, system and medium based on deep learning
Moloi et al. High impedance fault classification and localization method for power distribution network
CN113191192A (en) Breaker fault detection method based on wavelet analysis and fuzzy neural network algorithm
Yan et al. A simplified current feature extraction and deployment method for DC series arc fault detection
CN107689015A (en) A kind of improved power system bad data recognition method
CN109975634A (en) A kind of fault diagnostic method for transformer winding based on atom sparse decomposition
CN113112039A (en) Active power distribution system initial fault identification method based on time-frequency memory recurrent neural network
CN113159088A (en) Fault monitoring and diagnosis method based on multi-feature fusion and width learning
CN116566061A (en) Grid-connected inverter system stability on-line monitoring method and system
CN114384319A (en) Grid-connected inverter island detection method, system, terminal and medium
CN109506936B (en) Bearing fault degree identification method based on flow chart and non-naive Bayes inference
CN110135021B (en) ATRU system fault grading diagnosis method based on multi-source signals and RBF neural network
Wilson et al. Uncertainty Quantification of Capacitor Switching Transient Location Using Machine Learning
CN115130550A (en) Distribution transformer fault identification method based on gradient lifting decision tree
Xu et al. Partial discharge detection based on long short-term memory neural network classifier with efficient feature extraction methods
Adhikari et al. Real-Time Short-Term Voltage Stability Assessment using Temporal Convolutional Neural Network
Wang et al. Turn-to-turn short circuit of motor stator fault diagnosis using dropout rate improved deep sparse autoencoder
Laosiritaworn et al. Classification techniques for control chart pattern recognition: A case of metal frame for actuator production
CN111506045A (en) Fault diagnosis method based on single-value intelligent set correlation coefficient

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