CN108535572A - Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence feature - Google Patents
Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence feature Download PDFInfo
- Publication number
- CN108535572A CN108535572A CN201810443022.3A CN201810443022A CN108535572A CN 108535572 A CN108535572 A CN 108535572A CN 201810443022 A CN201810443022 A CN 201810443022A CN 108535572 A CN108535572 A CN 108535572A
- Authority
- CN
- China
- Prior art keywords
- indicate
- vector
- sequence
- current
- zero
- 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.)
- Granted
Links
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
Abstract
The invention discloses a kind of metering system secondary circuit monitoring method and device based on fundamental wave zero sequence feature, the present invention obtains the three-phase voltage in high pressure measurement system voltage circuit, three-phase current and neutral current in current loop；Zero sequence voltage component and zero-sequence current component are obtained using vector superposed；Extract fundamental wave zero sequence component amplitude and phase；The amplitude of fundamental voltage and phase, the amplitude of fundamental wave zero sequence point electric current are established into the classification and identification realized based on depth belief network grader to failure as input variable, all kinds of failures with neutral current absolute value of the difference with phase, fundamental wave zero sequence electric current as output variable.The present invention can precisely extract high pressure measurement secondary system loop fault feature, and accurately all kinds of failures are classified and identified, effectively to cut off failure and to be monitored in real time to secondary circuit, to ensure electrical energy measurement it is safe and stable with it is accurate.
Description
Technical field
The present invention relates to high pressure measurement secondary system loop state monitoring and fault diagnosis technologies, and in particular to one kind is based on
The metering system secondary circuit monitoring method and device of fundamental wave zero sequence feature, for high pressure electrical metering system voltage circuit and electricity
Flow back to the condition monitoring and fault diagnosis on road.
Background technology
With China's expanding economy, electricity consumption sharply increases, and power supply and distribution stability requirement is higher and higher.High-tension electricity meter
Amount system, as one important link of electrical energy measurement, the safety and stablization of system are particularly significant, therefore monitor high pressure measurement in real time
Secondary system loop current, the state of voltage are of great significance.Traditional decompression defluidization warning device is from decompression, defluidization
Definition is set out, and realizes that the monitoring to voltage, current status brings many mistakes since the definition of decompression defluidization is changed
A problem that alarm.These unfavorable conditions lead to problems with：1) inaccuracy of electrical energy measurement is caused, it is public with power supply to user
Department brings property loss；2) malfunction for causing failure removal device interferes the normal work of metering secondary；3)
Failure, the serious stable operation for threatening metering secondary cannot timely be cut off.Therefore, it is lost for presently used decompression
The problems such as stream warning device is in the presence of that can not find failure and special, erroneous judgement and real time monitoring energy force difference, studies suitable method simultaneously
Relevant apparatus is designed, realizes the real-time monitoring to voltage, current status, improves the accuracy of decompression defluidization fault identification, to dimension
The safety and stablization in shield high pressure measurement secondary system circuit are of great significance.
Invention content
The technical problem to be solved in the present invention：For the above problem of the prior art, provide a kind of special based on fundamental wave zero sequence
The metering system secondary circuit monitoring method and device of sign, the present invention can precisely extract high pressure measurement secondary system loop fault
Feature is accurately classified and is identified to all kinds of failures, effectively to cut off failure and to be monitored in real time to secondary circuit, to protect
Demonstrate,prove electrical energy measurement it is safe and stable with it is accurate.
In order to solve the above-mentioned technical problem, the technical solution adopted by the present invention is：
A kind of metering system secondary circuit monitoring method based on fundamental wave zero sequence feature, implementation steps include：
1) it is directed to monitored high pressure measurement system, obtains the three-phase voltage U in its voltage circuita、Ub、Uc, obtain electric current
Three-phase current I in circuita、Ib、IcAnd neutral current In；
2) by three-phase voltage Ua、Ub、UcCarry out the vector superposed zero sequence voltage component obtained containing a large amount of harmonic waves and white noise
3U0, by three-phase current Ia、Ib、IcCarry out the vector superposed zero-sequence current component 3I obtained containing a large amount of harmonic waves and white noise0；
3) according to zero sequence voltage component 3U0/ zero-sequence current component 3I0Fundamental wave zero sequence component is extracted, zero-sequence fundamental voltage is obtained
Amplitude 3U01And phaseThe amplitude 3I of/fundamental wave zero sequence electric current01And phase
4) fundamental wave zero sequence electric current and neutral current I are calculatednBetween absolute value of the difference obtain fundamental wave zero sequence electric current in
Property line current absolute value of the difference Ie；
5) by the amplitude 3U of zero-sequence fundamental voltage01, zero-sequence fundamental voltage amplitudeThe amplitude of fundamental wave zero sequence electric current
3I01, fundamental wave zero sequence electric current phaseAnd fundamental wave zero sequence electric current and neutral current absolute value of the difference IeAs input vector
The machine learning classification model of training is completed in input, and the machine learning classification model is trained to comprising input vector and as defeated
Go out the mapping relations between the secondary circuit failure of result, finally obtains the monitoring in monitored high pressure measurement secondary system circuit
As a result.
Preferably, the detailed step of step 3) includes：
3.1) value of initialization iterations k is 0, inputs original zero sequence voltage component 3U0/ zero-sequence current component 3I0Amount
Equation and state equation are surveyed, original residual voltage/original zero-sequence current component is randomly generated into initial sample and is adopted at random
Sample；
3.2) original zero sequence voltage component 3U is calculated0/ zero-sequence current component 3I0The state value and particle of produced particle
Weight；
3.3) generalized regression nerve networks GRNN is established, by generalized regression nerve networks GRNN to original residual voltage point
Measure 3U0/ zero-sequence current component 3I0State value optimize and adjust；
3.4) number of effective particles is calculated；
3.5) judge whether number of effective particles is less than the threshold value of setting number of effective particles, if it is carry out resampling, so
After redirect and execute next step；Otherwise, it directly redirects and executes next step；
3.6) it carries out residual voltage/zero-sequence current state estimation and calculates likelihood probability density, obtain current optimum state
The corresponding zero sequence voltage component 3U of estimated value0/ zero-sequence current component 3I0；
3.7) the when maximum cumulative log-likelihood ratio of the log-likelihood that adds up is calculated；
3.8) judge maximum cumulative log-likelihood ratio is more than whether decision threshold is true, to current optimal shape if setting up
The corresponding zero sequence voltage component 3U of state estimated value0/ zero-sequence current component 3I0It is preserved, then redirects and execute next step；Otherwise,
It directly redirects and executes next step；
3.9) state updates, and iterations k is added 1；
3.10) judge whether iteration terminates, if not yet terminated, redirect and execute step 3.2)；Otherwise it redirects under execution
One step；
3.11) the amplitude 3U of zero-sequence fundamental voltage is exported01And phaseThe amplitude 3I of/fundamental wave zero sequence electric current01And phase
Preferably, the detailed step of step 3.3) includes：
3.3.1 generalized regression nerve networks GRNN) is established, the input vector of generalized regression nerve networks GRNN is defined asObject vector is defined as zk, generalized regression nerve networks GRNN is trained according to formula (1)；
In formula (1),Indicate initial prediction, YiThe dependent variable value of training sample, zkFor object vector,For input
Element in vector, σ indicate that the spread factor of Gaussian function is also known as smoothing factor, and n is expressed as sample size；
3.3.2 a n-dimensional vector) is constructedJ Δ ＜ L (j=1,2 ..., n/2), whereinIt is defeated
Element in incoming vector, j Δs indicate that input vector each element adjusted value, parameter L indicate that the adjusting range of definition, n indicate vector
Dimension；
3.3.3) according to formula (2) by n-dimensional vectorAs input vector, the predicted value that formula (1) training is obtained
As the dependent variable value of training sample, generalized regression nerve networks GRNN is further trained；
In formula (2),Indicate transformed predicted value,Indicate initial prediction, zkFor object vector,For
Element in input vector, σ indicate that the spread factor of Gaussian function is also known as smoothing factor, and n is expressed as sample size；
3.3.4) pass through the output vector z of generalized regression nerve networks GRNNkInstruction, by sampleBy optimum pointSubstitution,Wherein j Δs indicate input vector each element adjusted value,The member being expressed as in input vector
Element.
Preferably, the machine learning classification model in step 5) be based on depth belief network grader, it is described based on deep
Degree belief network grader training step include：
The zero-sequence fundamental voltage in high pressure measurement secondary system circuit when S1) choosing normal, decompression, defluidization, broken neutral line,
With phase as sample data and characteristic variable, the decompression includes that TV transformer polarity is reversed, voltage list for the amplitude of electric current
Mutually broken string, voltage transformer internal fault, the defluidization include current transformer reverse polarity connection, current loop single-phase wire break, survey
Coil short circuit is measured, is classified as training set after being standardized to sample data according to a certain percentage；
S2) state in high pressure measurement secondary system circuit is encoded；
S3 secondary circuit failure classification and identification model based on depth belief network grader) are established；
S4) parameter of initialization failure classification and identification model enables the smaller random number that it is one group of Gaussian distributed
Value；
S5 the unlabeled exemplars in training set) are chosen, by the way that sdpecific dispersion algorithm is classified and identified to secondary circuit failure
The model bottom of model is limited Boltzmann machine layer and carries out pre-training；
S6 it) uses the exemplar in training set to carry out tuning to whole network by BP algorithm, completes to being based on depth
The training of the secondary circuit failure classification and identification model of belief network grader.
Preferably, step S5) detailed step include：
S5.1) the original state v of initialization secondary circuit failure classification and the visible layer unit of identification model0=x0, initially
It is the random compared with numerical value of Gaussian distributed to change W, a, b, and each limited Boltzmann machine layer maximum of setting trains iterations；Wherein
v0Indicate the initial state vector of visible layer unit, x0Indicate that training sample, W indicate that connection weight matrix, a indicate visible layer
Bias vector, b indicate the bias vector of hidden layer；
S5.2) to all hidden unit calculating formulas (3) of secondary circuit failure classification and identification model, from condition distribution P (h0j
|v0) in extract h0~P (h0|v0), wherein h0jIndicate the initial state value of j-th of neuron of hidden layer, v0Indicate visible layer list
The initial state vector of member, h0Indicate the initial state vector of hidden layer；
In formula (3), h0jIndicate the initial state value of j-th of neuron of hidden layer, v0Indicate the initial shape of visible layer unit
State vector, bjIndicate the bias of j-th of neuron of hidden layer, v0iIndicate the original state of i-th of neuron of visible layer unit
Value, WijIndicate that the connection weight weight values between visible layer node i and hidden layer node j, n indicate that visible layer number of nodes, σ () are
Sigmoid functions；
S5.3) to all visible element calculating formulas (4) of secondary circuit failure classification and identification model, it is distributed P from condition
(v1i|h0) in extract v1~P (v1|h0), wherein v1iThe state of i-th of neuron of visible layer unit after 1 gibbs sampler of expression
Value, v1The state vector of visible layer unit, h after 1 gibbs sampler of expression0Indicate the initial state vector of hidden layer；
In formula (4), v1iThe state value of i-th of neuron of visible layer unit, h after 1 gibbs sampler of expression0Indicate implicit
The initial state vector of layer, aiIndicate the bias of i-th of neuron of visible layer, h0jIndicate the first of j-th of neuron of hidden layer
Beginning state value, WijIndicate that the connection weight weight values between visible layer node i and hidden layer node j, m indicate node in hidden layer, σ
() is sigmoid functions；
S5.4) to all hidden unit calculating formulas (5) of secondary circuit failure classification and identification model；
In formula (5), h1jThe state value of j-th of neuron of layer unit, v are implied after 1 gibbs sampler of expression0Indicate visible
The initial state vector of layer unit, bjIndicate the bias of j-th of neuron of hidden layer, v1iIt can after 1 gibbs sampler of expression
See the state value of i-th of neuron of layer unit, WijIndicate the connection weight between visible layer node i and hidden layer node j
Value, n indicate that visible layer number of nodes, σ () are sigmoid functions；
S5.5 the parameter of secondary circuit failure classification and identification model) is updated according to formula (6)；
In formula (6), W indicates that connection weight matrix, a indicate that the bias vector of visible layer, b indicate the biasing of hidden layer
Vector, ρ indicate learning rate, h0Indicate the initial state vector of hidden layer, v0Indicate the initial state vector of visible layer unit,Indicate the transposition of visible layer unit initial state vector, v1The state vector of visible layer unit after 1 gibbs sampler of expression,The transposition of visible layer location mode vector after 1 gibbs sampler of expression.
The metering system secondary circuit monitoring device based on fundamental wave zero sequence feature that the present invention also provides a kind of, including computer
Equipment, the computer equipment are programmed to perform the metering system secondary circuit monitoring method the present invention is based on fundamental wave zero sequence feature
The step of.
The present invention is based on the metering system secondary circuit monitoring method of fundamental wave zero sequence feature tools to have the advantage that：Base of the present invention
The three-phase in high pressure measurement system voltage circuit is obtained respectively in the metering system secondary circuit monitoring method of fundamental wave zero sequence feature
Three-phase current in voltage, current loop and neutral current are obtained using vector superposed containing a large amount of harmonic waves and white noise
Zero sequence voltage component and zero-sequence current component, utilize based on neural network significant samples adjustment particle filter algorithm extract base
Wave zero-sequence component amplitude and phase, by the amplitude of fundamental voltage and phase, the amplitude of fundamental wave zero sequence point electric current and phase, fundamental wave zero
Sequence electric current is established using all kinds of failures as output variable as input variable with neutral current absolute value of the difference and is based on depth
The high pressure measurement secondary system Circuit Searching mechanism of belief network, the recognition mechanism contain five input variables and as outputs
Mapping relations between all kinds of failures of variable, the final classification and identification realized to failure, compared to current existing decompression
Defluidization warning device, have accurately identify failure and can precise positioning fault type, effectively identification neutral conductor failure, will not occur
The advantages that false alarm.The present invention can by zero-sequence component harmonic wave and white noise effectively remove, realize to fundamental wave zero sequence component
Extraction；Common defects (None- identified broken neutral line failure, easily generation false alarm, the reality of existing warning device can be made up
When monitoring capacity it is poor), accurately alarm failure, the state of real-time monitoring and metering secondary circuit, to ensure high-pressure gauge
The safety and stablization of amount system.
It is that the present invention is based on fundamental wave zero sequences the present invention is based on the metering system secondary circuit monitoring device of fundamental wave zero sequence feature
The device that the completely corresponding program unit of the metering system secondary circuit monitoring method of feature is constituted, therefore equally also there is this hair
The aforementioned advantages of the bright metering system secondary circuit monitoring method based on fundamental wave zero sequence feature, therefore details are not described herein.
Description of the drawings
Fig. 1 is the basic procedure schematic diagram of present invention method.
Fig. 2 is the voltage circuit three-phase voltage waveform in the embodiment of the present invention.
Fig. 3 is the current loop three-phase current waveform in the embodiment of the present invention.
Fig. 4 is in the embodiment of the present invention based on the fundamental wave zero sequence component extraction flow chart for improving particle filter.
Fig. 5 is the voltage circuit residual voltage original waveform in the embodiment of the present invention.
Fig. 6 is the current loop zero-sequence current original waveform in the embodiment of the present invention.
Fig. 7 is the amplitude of the zero-sequence fundamental voltage of present invention method.
Fig. 8 is the phase of the zero-sequence fundamental voltage of present invention method.
Fig. 9 is the amplitude of the fundamental wave zero sequence electric current of present invention method.
Figure 10 is the phase of the fundamental wave zero sequence electric current of present invention method.
Figure 11 is mesohigh metering system secondary circuit failure diagnostic flow chart of the embodiment of the present invention.
Figure 12 is the high pressure measurement secondary system loop fault diagnostic model schematic diagram in the embodiment of the present invention.
Specific implementation mode
As shown in Figure 1, the implementation step of metering system secondary circuit monitoring method of the present embodiment based on fundamental wave zero sequence feature
Suddenly include：
1) it is directed to monitored high pressure measurement system, obtains the three-phase voltage U in its voltage circuita、Ub、Uc, obtain electric current
Three-phase current I in circuita、Ib、IcAnd neutral current In；
2) by three-phase voltage Ua、Ub、UcCarry out the vector superposed zero sequence voltage component obtained containing a large amount of harmonic waves and white noise
3U0, by three-phase current Ia、Ib、IcCarry out the vector superposed zero-sequence current component 3I obtained containing a large amount of harmonic waves and white noise0；
3) according to zero sequence voltage component 3U0/ zero-sequence current component 3I0Fundamental wave zero sequence component is extracted, zero-sequence fundamental voltage is obtained
Amplitude 3U01And phaseThe amplitude 3I of/fundamental wave zero sequence electric current01And phase
4) fundamental wave zero sequence electric current and neutral current I are calculatednBetween absolute value of the difference obtain fundamental wave zero sequence electric current in
Property line current absolute value of the difference Ie(it can be write as function expression Ie=| 3I01-In|)；
5) by the amplitude 3U of zero-sequence fundamental voltage01, zero-sequence fundamental voltage amplitudeThe amplitude of fundamental wave zero sequence electric current
3I01, fundamental wave zero sequence electric current phaseAnd fundamental wave zero sequence electric current and neutral current absolute value of the difference IeAs input vector
The machine learning classification model of training is completed in input, and the machine learning classification model is trained to comprising input vector and as defeated
Go out the mapping relations between the secondary circuit failure of result, finally obtains the monitoring in monitored high pressure measurement secondary system circuit
As a result.
In the present embodiment, step 1) is returned by DL-PT202H1 current mode voltage measuring transformer high pressure measurement system voltages
The three-phase voltage U on roada、Ub、Uc, the three of high pressure measurement system power circuit are measured by HBC-LSP closed-loop Hall current sensors
Phase current Ia、Ib、IcAnd neutral current In；For the signal collected, it is lifted by RC filtering, ratio amplifier, voltage
Equal circuits, the range that the voltage of acquisition, current control reprocessor can be worked normally, by the way that A/D conversions will treated
Voltage, current transformation are in analog input to central processing unit.Three-phase voltage waveform such as Fig. 2 institutes are collected in the present embodiment
Show, wherein three-phase voltage waveform when (a) expression voltage circuit normal operation, (b) indicates three when voltage circuit single-phase wire break
Phase voltage waveform, (c) indicates three-phase voltage waveform when voltage transformer internal fault, (d) indicates that TV transformer polarity connects
The three-phase voltage waveform of inverse time；The three-phase current waveform collected is as shown in Figure 3, wherein (a) indicates that current loop is normally transported
Three-phase current waveform when row (b) indicates three-phase current waveform when current loop single-phase wire break, (c) indicates current transformer
Three-phase current waveform when reverse polarity connection (d) indicates three-phase current waveform when current loop measuring coil short circuit.
As shown in figure 4, the detailed step of step 3) includes：
3.1) value of initialization iterations k is 0, inputs original zero sequence voltage component 3U0/ zero-sequence current component 3I0Amount
Equation and state equation are surveyed, original residual voltage/original zero-sequence current component is randomly generated into initial sample and is adopted at random
Sample；
In the present embodiment, measurement equation ZkIt is shown below：
[x1,x2,…,x2n-1,x2n,A0]T+υk
=Hk·xk+υk
In above formula, k indicates sampling period number, TsIndicate sampling period, ωnIndicate that the angular speed of n-th harmonic, τ expressions decline
Subtract time constant, AnIndicate the amplitude of n-th harmonic, φnIndicate the initial phase of n-th harmonic, A0Indicate attenuating dc component
Amplitude, HkIndicate calculation matrix, xkIndicate quantity of state, υkIndicate mean value be zero, variance RkMeasurement noise, n indicates contained humorous
Wave number.State equation is shown below：
In above formula, x2nIndicate that k moment quantity of states, k indicate sampling period number, n is overtone order, A0Indicate attenuating dc component
Amplitude, wk-1Indicate mean value be zero, variance Qk-1Process noise.As k=0, by priori probability density distribution function P (x0) random
Sample is generated, according to P (x0) profile samples obtainState is carried out to zero sequence voltage component and zero-sequence current component
Prediction, is sampled according to the state equation of zero sequence voltage component and zero-sequence current component：
WhereinIndicate that the state value of i-th of particle of k moment, q indicate importance density function, xkIndicate the quantity of state at k moment,It indicates
The state value of i-th of particle of initial time, k indicate sampling time, z0Indicate that initial time measuring value, N indicate population.
3.2) original zero sequence voltage component 3U is calculated0/ zero-sequence current component 3I0The state value and particle of produced particle
Weight；
The calculating function expression of particle weights is as follows in the present embodiment：
In above formula,Indicate the weighted value of i-th of particle of k moment,Indicate the weighted value of i-th of particle of k-1 moment, p
Indicate prior distribution density, zkIndicate k moment measuring values,Indicate the state value of i-th of particle of k-1 moment,When indicating k
The state value of i-th of particle is carved, N indicates population.It also needs to carry out normalizing using following formula for the particle weights being calculated
Change is handled：
In above formula,Indicate that the weighted value of i-th of particle of k moment, N indicate population.
3.3) generalized regression nerve networks GRNN is established, by generalized regression nerve networks GRNN to original residual voltage point
Measure 3U0/ zero-sequence current component 3I0State value optimize and adjust；
3.4) number of effective particles is calculated；
3.5) judge whether number of effective particles is less than the threshold value of setting number of effective particles, if it is carry out resampling, so
After redirect and execute next step；Otherwise, it directly redirects and executes next step；
3.6) it carries out residual voltage/zero-sequence current state estimation and calculates likelihood probability density, obtain current optimum state
The corresponding zero sequence voltage component 3U of estimated value0/ zero-sequence current component 3I0；
When carrying out residual voltage/zero-sequence current state estimation in the present embodiment, optimal state estimation is shown below；
In above formula,Indicate the estimated value of k moment particles,Indicate that the weighted value of i-th of particle of k moment, N indicate grain
Subnumber,Indicate the quantity of state of i-th of particle of k moment.
According to optimal state estimation, zero-sequence fundamental voltage/fundamental wave zero under optimal state estimation can be calculated
The amplitude of sequence electric current：
In above formula, A1Indicate the amplitude of zero-sequence fundamental voltage/fundamental wave zero sequence electric current under optimal state estimation, x1It indicates
The estimated value of 1st quantity of state, x2Indicate the estimated value of the 2nd quantity of state.
According to optimal state estimation, zero-sequence fundamental voltage/fundamental wave zero under optimal state estimation can be calculated
The phase of sequence electric current：
In above formula, φ1Indicate the phase of zero-sequence fundamental voltage/fundamental wave zero sequence electric current under optimal state estimation, x1It indicates
The estimated value of 1st quantity of state, x2Indicate the estimated value of the 2nd quantity of state.
3.7) the when maximum cumulative log-likelihood ratio of the log-likelihood that adds up is calculated；
3.8) judge maximum cumulative log-likelihood ratio is more than whether decision threshold is true, to current optimal shape if setting up
The corresponding zero sequence voltage component 3U of state estimated value0/ zero-sequence current component 3I0It is preserved, then redirects and execute next step；Otherwise,
It directly redirects and executes next step；
3.9) state updates, and iterations k is added 1；
3.10) judge whether iteration terminates, if not yet terminated, redirect and execute step 3.2)；Otherwise it redirects under execution
One step；
3.11) the amplitude 3U of zero-sequence fundamental voltage is exported01And phaseThe amplitude 3I of/fundamental wave zero sequence electric current01And phase
In the present embodiment, the detailed step of step 3.3) includes：
3.3.1 generalized regression nerve networks GRNN) is established, the input vector of generalized regression nerve networks GRNN is defined asObject vector is defined as zk, generalized regression nerve networks GRNN is trained according to formula (1)；
In formula (1),Indicate initial prediction, YiThe dependent variable value of training sample, zkFor object vector,For input
Element in vector, σ indicate that the spread factor of Gaussian function is also known as smoothing factor, and n is expressed as sample size；
3.3.2 a n-dimensional vector) is constructedJ Δ ＜ L (j=1,2 ..., n/2), whereinIt is defeated
Element in incoming vector, j Δs indicate that input vector each element adjusted value, parameter L indicate that the adjusting range of definition, n indicate vector
Dimension；
3.3.3) according to formula (2) by n-dimensional vectorAs input vector, the predicted value that formula (1) training is obtained
As the dependent variable value of training sample, generalized regression nerve networks GRNN is further trained；
In formula (2),Indicate transformed predicted value,Indicate initial prediction, zkFor object vector,For
Element in input vector, σ indicate that the spread factor of Gaussian function is also known as smoothing factor, and n is expressed as sample size；
3.3.4) pass through the output vector z of generalized regression nerve networks GRNNkInstruction, by sampleBy optimum pointSubstitution,Wherein j Δs indicate input vector each element adjusted value,The member being expressed as in input vector
Element.
By the vector superposed original waveform for obtaining residual voltage as shown in figure 5, original waveform such as Fig. 6 institutes of zero-sequence current
Show.Particle filter algorithm is adjusted by the significant samples based on neural network and extracts fundamental wave zero sequence component amplitude and phase, is obtained
The amplitude and phase of zero-sequence fundamental voltage are as Figure 7-8, and Fig. 7 is the amplitude of the zero-sequence fundamental voltage of extraction, and Fig. 8 is extraction
The phase of zero-sequence fundamental voltage.The amplitude and phase of fundamental wave zero sequence electric current are as shown in figs. 9-10, wherein Fig. 9 is the fundamental wave of extraction
The amplitude of zero-sequence current, Figure 10 are the phase of the fundamental wave zero sequence electric current of extraction.
In the present embodiment, the machine learning classification model in step 5) is based on depth belief network grader, such as Figure 11
Shown, the training step based on depth belief network grader includes：
The zero-sequence fundamental voltage in high pressure measurement secondary system circuit when S1) choosing normal, decompression, defluidization, broken neutral line,
With phase as sample data and characteristic variable, the decompression includes that TV transformer polarity is reversed, voltage list for the amplitude of electric current
Mutually broken string, voltage transformer internal fault, the defluidization include current transformer reverse polarity connection, current loop single-phase wire break, survey
Coil short circuit is measured, is classified as training set after being standardized to sample data according to a certain percentage；
S2) state in high pressure measurement secondary system circuit is encoded, as shown in table 1；
Table 1：Secondary circuit state encoding.
S3 secondary circuit failure classification and identification model based on depth belief network grader) are established；
As shown in figure 12, which includes 5 inputs, respectively：The amplitude of fundamental voltage and phase, fundamental wave zero sequence
The amplitude and phase, fundamental wave zero sequence electric current and neutral current absolute value of the difference of point electric current, including 8 outputs, respectively：Voltage
Mutual inductor polarity is reversed, voltage circuit single-phase wire break, voltage transformer internal fault, current transformer reverse polarity connection, electric current return
It is road single-phase wire break, measuring coil short circuit, broken neutral line failure, normal.
S4) parameter of initialization failure classification and identification model enables the smaller random number that it is one group of Gaussian distributed
Value；
S5) choose training set in unlabeled exemplars, by sdpecific dispersion algorithm (Contrastive Divergence,
CD algorithms) Boltzmann machine (Restricted is limited to secondary circuit failure classification and the model bottom of identification model
Boltzmann Machine, RBM) layer progress pre-training；
S6 it) uses the exemplar in training set to carry out tuning to whole network by BP algorithm, completes to being based on depth
The training of the secondary circuit failure classification and identification model of belief network grader.
In the present embodiment, step S5) detailed step include：
S5.1) the original state v of initialization secondary circuit failure classification and the visible layer unit of identification model0=x0, initially
It is the random compared with numerical value of Gaussian distributed to change W, a, b, and each limited Boltzmann machine layer maximum of setting trains iterations；Wherein
v0Indicate the initial state vector of visible layer unit, x0Indicate that training sample, W indicate that connection weight matrix, a indicate visible layer
Bias vector, b indicate the bias vector of hidden layer；
S5.2) to all hidden unit calculating formulas (3) of secondary circuit failure classification and identification model, from condition distribution P (h0j
|v0) in extract h0~P (h0|v0), wherein h0jIndicate the initial state value of j-th of neuron of hidden layer, v0Indicate visible layer list
The initial state vector of member, h0Indicate the initial state vector of hidden layer；
In formula (3), h0jIndicate the initial state value of j-th of neuron of hidden layer, v0Indicate the initial shape of visible layer unit
State vector, bjIndicate the bias of j-th of neuron of hidden layer, v0iIndicate the original state of i-th of neuron of visible layer unit
Value, WijIndicate that the connection weight weight values between visible layer node i and hidden layer node j, n indicate that visible layer number of nodes, σ () are
Sigmoid functions；
S5.3) to all visible element calculating formulas (4) of secondary circuit failure classification and identification model, it is distributed P from condition
(v1i|h0) in extract v1~P (v1|h0), wherein v1iThe state of i-th of neuron of visible layer unit after 1 gibbs sampler of expression
Value, v1The state vector of visible layer unit, h after 1 gibbs sampler of expression0Indicate the initial state vector of hidden layer；
In formula (4), v1iThe state value of i-th of neuron of visible layer unit, h after 1 gibbs sampler of expression0Indicate implicit
The initial state vector of layer, aiIndicate the bias of i-th of neuron of visible layer, h0jIndicate the first of j-th of neuron of hidden layer
Beginning state value, WijIndicate that the connection weight weight values between visible layer node i and hidden layer node j, m indicate node in hidden layer, σ
() is sigmoid functions；
S5.4) to all hidden unit calculating formulas (5) of secondary circuit failure classification and identification model；
In formula (5), h1jThe state value of j-th of neuron of layer unit, v are implied after 1 gibbs sampler of expression0Indicate visible
The initial state vector of layer unit, bjIndicate the bias of j-th of neuron of hidden layer, v1iIt can after 1 gibbs sampler of expression
See the state value of i-th of neuron of layer unit, WijIndicate the connection weight between visible layer node i and hidden layer node j
Value, n indicate that visible layer number of nodes, σ () are sigmoid functions；
S5.5 the parameter of secondary circuit failure classification and identification model) is updated according to formula (6)；
In formula (6), W indicates that connection weight matrix, a indicate that the bias vector of visible layer, b indicate the biasing of hidden layer
Vector, ρ indicate learning rate, h0Indicate the initial state vector of hidden layer, v0Indicate the initial state vector of visible layer unit,Indicate the transposition of visible layer unit initial state vector, v1The state vector of visible layer unit after 1 gibbs sampler of expression,The transposition of visible layer location mode vector after 1 gibbs sampler of expression.
The present embodiment based on the metering system secondary circuit monitoring method of fundamental wave zero sequence feature particular by MATLAB and
Computer program (c#) is realized, is adjusted particle filter algorithm by significant samples of the MATLAB realizations based on neural network and is based on
The high pressure measurement secondary system loop fault of depth belief network grader is classified and recognition mechanism.
When following data are the various operating statuses in 10kV high pressure measurement secondary systems circuit, the weight based on neural network is utilized
The amplitude and phase of sample adjustment particle filter algorithm extraction fundamental wave zero sequence component, input is wanted to be based on depth belief network grader
High pressure measurement secondary system loop fault classification with recognition mechanism in, obtained recognition result, as shown in table 2.Wherein parameter
It is set as：It is 500 to select 5 layer network structures, RBM Hidden unit numbers, and number of training 1000, test sample number is 500.
Table 2：10kV metering system secondary circuit failure recognition results based on DBNC.
As shown in Table 2, metering system secondary circuit monitoring method of the present embodiment based on fundamental wave zero sequence feature identifies high pressure
The effect of metering system secondary circuit failure is preferable, and the accuracy rate of fault identification substantially remains in 95% or more.
In conclusion metering system secondary circuit monitoring method of the present embodiment based on fundamental wave zero sequence feature passes through the acquisition
Three-phase voltage in high pressure measurement system voltage circuit, three-phase current and neutral current in current loop, utilize vector
Superposition obtains zero sequence voltage component and the zero-sequence current component containing a large amount of harmonic waves and white noise；Utilize the weight based on neural network
Want sample adjustment particle filter algorithm extraction fundamental wave zero sequence component amplitude and phase；By the amplitude of fundamental voltage and phase, fundamental wave
The amplitude and phase, fundamental wave zero sequence electric current and neutral current absolute value of the difference of zero sequence point electric current, will be all kinds of as input variable
Failure establishes the high pressure measurement secondary system Circuit Searching mechanism based on depth belief network grader as output variable, should
Recognition mechanism contains five input variables and as the mapping relations between the failure of output variable to realize to failure
Classification and identification, can by zero-sequence component harmonic wave and white noise effectively remove and realize extraction to fundamental wave zero sequence component, energy
All kinds of failures are accurately classified and are identified, so as to effective by enough precisely extraction high pressure measurement secondary system loop fault features
Excision failure simultaneously monitors secondary circuit in real time, to ensure safe and stable and accurate, the real-time monitoring and metering two of electrical energy measurement
The state of minor loop is to ensure the safety and stablization of high pressure measurement system.Compared to current existing decompression defluidization alarm method
For, have accurately identify failure and can precise positioning fault type, effectively identification neutral conductor failure, false alarm will not occur etc.
Advantage.
The above is only a preferred embodiment of the present invention, protection scope of the present invention is not limited merely to above-mentioned implementation
Example, all technical solutions belonged under thinking of the present invention all belong to the scope of protection of the present invention.It should be pointed out that for the art
Those of ordinary skill for, several improvements and modifications without departing from the principles of the present invention, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (6)
1. a kind of metering system secondary circuit monitoring method based on fundamental wave zero sequence feature, it is characterised in that implementation steps include：
1) it is directed to monitored high pressure measurement system, obtains the three-phase voltage U in its voltage circuita、Ub、Uc, obtain current loop
In three-phase current Ia、Ib、IcAnd neutral current In；
2) by three-phase voltage Ua、Ub、UcCarry out the vector superposed zero sequence voltage component 3U obtained containing a large amount of harmonic waves and white noise0,
By three-phase current Ia、Ib、IcCarry out the vector superposed zero-sequence current component 3I obtained containing a large amount of harmonic waves and white noise0；
3) according to zero sequence voltage component 3U0/ zero-sequence current component 3I0Fundamental wave zero sequence component is extracted, the width of zero-sequence fundamental voltage is obtained
Value 3U01And phaseThe amplitude 3I of fundamental wave zero sequence electric current01And phase
4) fundamental wave zero sequence electric current and neutral current I are calculatednBetween absolute value of the difference obtain fundamental wave zero sequence electric current and the neutral conductor electricity
Flow absolute value of the difference Ie；
5) by the amplitude 3U of zero-sequence fundamental voltage01, zero-sequence fundamental voltage amplitudeThe amplitude 3I of fundamental wave zero sequence electric current01, base
The phase of wave zero-sequence currentAnd fundamental wave zero sequence electric current and neutral current absolute value of the difference IeIt has been inputted as input vector
At trained machine learning classification model, the machine learning classification model is trained to comprising input vector and as output result
Secondary circuit failure between mapping relations, finally obtain the monitoring result in monitored high pressure measurement secondary system circuit.
2. the metering system secondary circuit monitoring method according to claim 1 based on fundamental wave zero sequence feature, feature exist
In the detailed step of step 3) includes：
3.1) value of initialization iterations k is 0, inputs original zero sequence voltage component 3U0/ zero-sequence current component 3I0Measurement side
Original residual voltage/original zero-sequence current component is randomly generated initial sample and carries out stochastical sampling by journey and state equation；
3.2) original zero sequence voltage component 3U is calculated0/ zero-sequence current component 3I0The state value and particle weights of produced particle；
3.3) generalized regression nerve networks GRNN is established, by generalized regression nerve networks GRNN to original zero sequence voltage component
3U0/ zero-sequence current component 3I0State value optimize and adjust；
3.4) number of effective particles is calculated；
3.5) judge whether number of effective particles is less than the threshold value of setting number of effective particles, if it is carry out resampling, then jump
Turn to execute next step；Otherwise, it directly redirects and executes next step；
3.6) it carries out residual voltage/zero-sequence current state estimation and calculates likelihood probability density, obtain current optimal State Estimation
It is worth corresponding zero sequence voltage component 3U0/ zero-sequence current component 3I0；
3.7) the when maximum cumulative log-likelihood ratio of the log-likelihood that adds up is calculated；
3.8) judge maximum cumulative log-likelihood ratio is more than whether decision threshold is true, is estimated to current optimum state if setting up
The corresponding zero sequence voltage component 3U of evaluation0/ zero-sequence current component 3I0It is preserved, then redirects and execute next step；Otherwise, directly
It redirects and executes next step；
3.9) state updates, and iterations k is added 1；
3.10) judge whether iteration terminates, if not yet terminated, redirect and execute step 3.2)；Otherwise it redirects and executes next step；
3.11) the amplitude 3U of zero-sequence fundamental voltage is exported01And phaseThe amplitude 3I of fundamental wave zero sequence electric current01And phase
3. the metering system secondary circuit monitoring method according to claim 2 based on fundamental wave zero sequence feature, feature exist
In the detailed step of step 3.3) includes：
3.3.1 generalized regression nerve networks GRNN) is established, the input vector of generalized regression nerve networks GRNN is defined asObject vector is defined as zk, generalized regression nerve networks GRNN is trained according to formula (1)；
In formula (1),Indicate initial prediction, YiThe dependent variable value of training sample, zkFor object vector,For input vector
In element, σ indicate Gaussian function spread factor be also known as smoothing factor, n is expressed as sample size；
3.3.2 a n-dimensional vector) is constructedWhereinFor input vector
In element, j Δs indicate that input vector each element adjusted value, parameter L indicate that the adjusting range of definition, n indicate vector dimension；
3.3.3) according to formula (2) by n-dimensional vectorAs input vector, the predicted value that formula (1) training is obtainedAs instruction
Practice the dependent variable value of sample, further training generalized regression nerve networks GRNN；
In formula (2),Indicate transformed predicted value,Indicate initial prediction, zkFor object vector,For input
Element in vector, σ indicate that the spread factor of Gaussian function is also known as smoothing factor, and n is expressed as sample size；
3.3.4) pass through the output vector z of generalized regression nerve networks GRNNkInstruction, by sampleBy optimum pointIt takes
Generation,Wherein j Δs indicate input vector each element adjusted value,The element being expressed as in input vector.
4. the metering system secondary circuit monitoring method according to claim 1 based on fundamental wave zero sequence feature, feature exist
In the machine learning classification model in step 5) is based on depth belief network grader, and the depth belief network that is based on divides
The training step of class device includes：
Zero-sequence fundamental voltage, the electric current in high pressure measurement secondary system circuit when S1) choosing normal, decompression, defluidization, broken neutral line
Amplitude and phase as sample data and characteristic variable, the decompression, which includes that TV transformer polarity is reversed, voltage is single-phase, breaks
Line, voltage transformer internal fault, the defluidization include current transformer reverse polarity connection, current loop single-phase wire break, measure line
Short circuit is enclosed, is classified as training set after being standardized to sample data according to a certain percentage；
S2) state in high pressure measurement secondary system circuit is encoded；
S3 secondary circuit failure classification and identification model based on depth belief network grader) are established；
S4) parameter of initialization failure classification and identification model enables the smaller random number that it is one group of Gaussian distributed；
S5) choose training set in unlabeled exemplars, by sdpecific dispersion algorithm to secondary circuit failure classify and identification model
Model bottom be limited Boltzmann machine layer and carry out pre-training；
S6 it) uses the exemplar in training set to carry out tuning to whole network by BP algorithm, completes to being based on depth conviction
The training of the secondary circuit failure classification and identification model of network classifier.
5. the metering system secondary circuit monitoring method according to claim 4 based on fundamental wave zero sequence feature, feature exist
In step S5) detailed step include：
S5.1) the original state v of initialization secondary circuit failure classification and the visible layer unit of identification model0=x0, initialization W,
A, b is the random compared with numerical value of Gaussian distributed, and each limited Boltzmann machine layer maximum of setting trains iterations；Wherein v0Table
Show the initial state vector of visible layer unit, x0Indicate that training sample, W indicate that connection weight matrix, a indicate the biasing of visible layer
Vector, b indicate the bias vector of hidden layer；
S5.2) to all hidden unit calculating formulas (3) of secondary circuit failure classification and identification model, from condition distribution P (h0j|v0)
Middle extraction h0~P (h0|v0), wherein h0jIndicate the initial state value of j-th of neuron of hidden layer, v0Indicate visible layer unit
Initial state vector, h0Indicate the initial state vector of hidden layer；
In formula (3), h0jIndicate the initial state value of j-th of neuron of hidden layer, v0Indicate the original state of visible layer unit to
Amount, bjIndicate the bias of j-th of neuron of hidden layer, v0iIndicate the initial state value of i-th of neuron of visible layer unit, Wij
Indicate that the connection weight weight values between visible layer node i and hidden layer node j, n indicate that visible layer number of nodes, σ () are sigmoid
Function；
S5.3) to all visible element calculating formulas (4) of secondary circuit failure classification and identification model, from condition distribution P (v1i|
h0) in extract v1~P (v1|h0), wherein v1iThe state value of i-th of neuron of visible layer unit after 1 gibbs sampler of expression,
v1The state vector of visible layer unit, h after 1 gibbs sampler of expression0Indicate the initial state vector of hidden layer；
In formula (4), v1iThe state value of i-th of neuron of visible layer unit, h after 1 gibbs sampler of expression0Indicate hidden layer
Initial state vector, aiIndicate the bias of i-th of neuron of visible layer, h0jIndicate the initial shape of j-th of neuron of hidden layer
State value, WijIndicate that the connection weight weight values between visible layer node i and hidden layer node j, m indicate that node in hidden layer, σ () are
Sigmoid functions；
S5.4) to all hidden unit calculating formulas (5) of secondary circuit failure classification and identification model；
In formula (5), h1jThe state value of j-th of neuron of layer unit, v are implied after 1 gibbs sampler of expression0Indicate visible layer list
The initial state vector of member, bjIndicate the bias of j-th of neuron of hidden layer, v1iVisible layer after 1 gibbs sampler of expression
The state value of i-th of neuron of unit, WijIndicate the connection weight weight values between visible layer node i and hidden layer node j, n
Indicate that visible layer number of nodes, σ () are sigmoid functions；
S5.5 the parameter of secondary circuit failure classification and identification model) is updated according to formula (6)；
In formula (6), W indicates that connection weight matrix, a indicate that the bias vector of visible layer, b indicate being biased towards for hidden layer
Amount, ρ indicate learning rate, h0Indicate the initial state vector of hidden layer, v0Indicate the initial state vector of visible layer unit,
Indicate the transposition of visible layer unit initial state vector, v1The state vector of visible layer unit after 1 gibbs sampler of expression,
The transposition of visible layer location mode vector after 1 gibbs sampler of expression.
6. a kind of metering system secondary circuit monitoring device based on fundamental wave zero sequence feature, including computer equipment, feature exist
In the computer equipment is programmed to perform the metering based on fundamental wave zero sequence feature described in any one of Claims 1 to 5
The step of secondary system circuit monitoring method.
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 true CN108535572A (en) | 2018-09-14 |
| CN108535572B 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) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109753705A (en) * | 2018-12-24 | 2019-05-14 | 北京华大九天软件有限公司 | IC initial value evaluation method in a kind of IC design |
| CN109799423A (en) * | 2019-01-09 | 2019-05-24 | 西安科技大学 | A kind of method of cable fault inline diagnosis |
| 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 |
| CN109980614A (en) * | 2019-03-29 | 2019-07-05 | 广东电网有限责任公司 | A kind of distribution line zero-sequenceprotection direction-adaptive method of discrimination |
| CN114002547A (en) * | 2021-10-25 | 2022-02-01 | 国网湖北省电力有限公司恩施供电公司 | Secondary multipoint earth fault judgment equipment and analysis algorithm for mutual inductor |
| CN117269843A (en) * | 2023-11-21 | 2023-12-22 | 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 |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107449979A (en) * | 2017-06-19 | 2017-12-08 | 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 |
| CN107609569A (en) * | 2017-07-31 | 2018-01-19 | 北京映翰通网络技术股份有限公司 | A kind of distribution net work earthing fault localization method based on multidimensional characteristic vectors |
| CN107817404A (en) * | 2017-11-18 | 2018-03-20 | 广西电网有限责任公司电力科学研究院 | A kind of Portable metering automatization terminal trouble-shooter and its diagnostic method |
-
2018
- 2018-05-10 CN CN201810443022.3A patent/CN108535572B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107449979A (en) * | 2017-06-19 | 2017-12-08 | 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 |
| CN107609569A (en) * | 2017-07-31 | 2018-01-19 | 北京映翰通网络技术股份有限公司 | A kind of distribution net work earthing fault localization method based on multidimensional characteristic vectors |
| CN107817404A (en) * | 2017-11-18 | 2018-03-20 | 广西电网有限责任公司电力科学研究院 | A kind of Portable metering automatization terminal trouble-shooter and its diagnostic method |
Non-Patent Citations (2)
| Title |
|---|
| 李兴凯: "基于神经网络的粒子滤波算法在GPS中的应用研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 石鑫 等: "基于深度信念网络的电力变压器故障分类建模", 《电力系统保护与控制》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109753705A (en) * | 2018-12-24 | 2019-05-14 | 北京华大九天软件有限公司 | IC initial value evaluation method in a kind of IC design |
| 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 |
| CN109799423A (en) * | 2019-01-09 | 2019-05-24 | 西安科技大学 | A kind of method of cable fault inline diagnosis |
| 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 |
| CN109980614A (en) * | 2019-03-29 | 2019-07-05 | 广东电网有限责任公司 | A kind of distribution line zero-sequenceprotection direction-adaptive method of discrimination |
| 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 |
| CN114002547A (en) * | 2021-10-25 | 2022-02-01 | 国网湖北省电力有限公司恩施供电公司 | Secondary multipoint earth fault judgment equipment and analysis algorithm for mutual inductor |
| 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 |
| CN117269843A (en) * | 2023-11-21 | 2023-12-22 | 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 |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108535572B (en) | 2020-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108535572A (en) | Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence feature | |
| Afrasiabi et al. | Integration of accelerated deep neural network into power transformer differential protection | |
| Wang et al. | ArcNet: Series AC arc fault detection based on raw current and convolutional neural network | |
| Chen et al. | Fault detection, classification and location for transmission lines and distribution systems: a review on the methods | |
| CN106443316B (en) | Multi-information detection method and device for deformation state of power transformer winding | |
| Teimourzadeh et al. | High impedance single-phase faults diagnosis in transmission lines via deep reinforcement learning of transfer functions | |
| Jiang et al. | Spatial-temporal synchrophasor data characterization and analytics in smart grid fault detection, identification, and impact causal analysis | |
| CN108414923A (en) | A kind of analog-circuit fault diagnosis method based on the extraction of depth confidence network characterization | |
| Wang | Extension neural network for power transformer incipient fault diagnosis | |
| Afrasiabi et al. | Fast GRNN-based method for distinguishing inrush currents in power transformers | |
| Deng et al. | A deep learning method based on long short term memory and sliding time window for type recognition and time location of power quality disturbance | |
| Moloi et al. | High impedance fault classification and localization method for power distribution network | |
| Jiang et al. | Machine learning approach to detect arc faults based on regular coupling features | |
| Huang et al. | BIT-based intermittent fault diagnosis of analog circuits by improved deep forest classifier | |
| Saikia et al. | Detection and classification of power quality disturbances using wavelet transform, fuzzy logic and neural network | |
| Dwivedi et al. | DynamoPMU: A Physics Informed Anomaly Detection, Clustering and Prediction Method using Non-linear Dynamics on μPMU Measurements | |
| Paydarnia et al. | Improved structure of PNN using PCA in transformer fault diagnostic | |
| Hong et al. | Locating High-Impedance Fault Section in Electric Power Systems Using Wavelet Transform,-Means, Genetic Algorithms, and Support Vector Machine | |
| Afrasiabi et al. | Differential protection of power transformers based on RSLVQ-gradient approach considering SFCL | |
| CN115809407A (en) | Breaker fault diagnosis method and system based on vibration characteristic analysis | |
| Lala et al. | Fault Detection and Localization using Continuous Wavelet Transform and Artificial Neural Network Based Approach in Distribution System. | |
| Chang | Partial discharge pattern recognition of cast resin current transformers using radial basis function neural network | |
| Qu et al. | An arc fault detection method based on multidictionary learning | |
| bin Sulaiman et al. | Using probabilistic neural network for classification high impedance faults on power distribution feeders | |
| Li et al. | Intelligent diagnosis and recognition method of GIS partial discharge data map based on deep learning |
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 |