CN115879044A - CNN network-based GIS switching-on/off state current detection method and device - Google Patents
CNN network-based GIS switching-on/off state current detection method and device Download PDFInfo
- Publication number
- CN115879044A CN115879044A CN202211481173.0A CN202211481173A CN115879044A CN 115879044 A CN115879044 A CN 115879044A CN 202211481173 A CN202211481173 A CN 202211481173A CN 115879044 A CN115879044 A CN 115879044A
- Authority
- CN
- China
- Prior art keywords
- cnn network
- optimal
- gis
- network
- switching
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 14
- 239000002245 particle Substances 0.000 claims abstract description 83
- 238000005457 optimization Methods 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 22
- 230000008878 coupling Effects 0.000 claims abstract description 17
- 238000010168 coupling process Methods 0.000 claims abstract description 17
- 238000005859 coupling reaction Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 44
- 230000008569 process Effects 0.000 claims description 16
- 238000011176 pooling Methods 0.000 claims description 14
- 238000010586 diagram Methods 0.000 claims description 12
- 230000009466 transformation Effects 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 8
- 238000012423 maintenance Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 5
- 210000002364 input neuron Anatomy 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 230000001052 transient effect Effects 0.000 claims description 5
- 239000000203 mixture Substances 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 4
- 238000013527 convolutional neural network Methods 0.000 description 97
- 230000009471 action Effects 0.000 description 9
- 230000007246 mechanism Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 239000003990 capacitor Substances 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007789 sealing Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Images
Landscapes
- Measurement Of Resistance Or Impedance (AREA)
Abstract
The invention discloses a GIS switching-on and switching-off state current detection method and device based on a CNN network, wherein the method comprises the following steps: collecting the opening and closing states of a disconnecting link and a grounding switch and corresponding coupling capacitance current data sets in all states of GIS equipment; preprocessing the acquired current data, and constructing a training set of the network; constructing a CNN network; optimizing the CNN network parameters by using a particle swarm optimization algorithm to obtain optimal CNN network parameters; acquiring fault current data of a disconnecting link and a grounding switch of GIS equipment in real time, preprocessing the fault current data, inputting the preprocessed data into an optimal CNN network, and outputting a GIS switching-on and switching-off state judgment result by using the optimal CNN network; the invention has the advantages that: the execution speed is fast, the real-time performance is high, and after the model is trained, the output result is more accurate, and the reliability is strong.
Description
Technical Field
The invention relates to the field of detection of a switch-on/off state of a knife switch in GIS equipment, in particular to a method and a device for detecting a current of a switch-on/off state of a GIS based on a CNN network.
Background
The Gas-Insulated metal-enclosed Switchgear (GIS) is a metal enclosed Switchgear which totally or partially uses Gas instead of air under atmospheric pressure as an insulating medium, and compared with the traditional open type power distribution unit, the GIS has the advantages of small floor area, no environmental interference on the whole sealing of elements, convenient operation, small maintenance workload, convenient equipment installation, short construction period and the like.
A500 kV GIS of a certain power station is produced by Mitsubishi corporation and Western-style switch factories in Japan, leaves a factory in 1990, and is put into operation in 1992; the model is 500-GNS-MS50, the main connection is one and a half times connection, 4 strings of 12 intervals are provided, the switch, the disconnecting link (isolating disconnecting link) and the grounding switch (grounding disconnecting link) are all pneumatic operating mechanisms, a centralized air supply mode is adopted, the rated operating air pressure is 1.5MPa, and the disconnecting link 33 (99 phases) and the grounding switch 43 (129 phases) are adopted. During 2015-2017, the 500kV GIS of the power station exposes the defects of unsuccessful disconnecting link and grounding knife switching caused by the failure of internal mechanisms of disconnecting links and grounding knives due to 4 causes, and the safe and stable operation of the system is seriously influenced. In order to avoid equipment accidents, the conditions of disconnecting link and grounding switch opening and closing need to be researched.
Chinese patent publication No. CN113484741A discloses a device and method for monitoring the operating state of a GIS disconnecting switch, which utilizes an open-type current transformer to obtain the current value of a GIS disconnecting switch operating mechanism driving motor changing with time during the action process, and further indirectly judges the action state of a disconnecting switch disconnecting knife according to the current value of the operating mechanism driving motor changing with time during the action process; the angular displacement sensor is used for obtaining a rotational displacement value of a driving motor of the GIS disconnecting switch operating mechanism along with time change, and the position state of a GIS disconnecting switch is judged; the opening and closing action times of the GIS isolating switch are accumulated, the mechanical life of the GIS isolating switch is predicted, and the monitoring accuracy of the working state of the GIS isolating switch is effectively improved. However, the current value and the sensor data in the action process of the operating mechanism driving motor are relied on to judge the action and the position of the knife switch, the real-time performance is not high, the action and the position of the knife switch obtained at present may actually be the historical action and the position of the knife switch calculated by the current at the historical moment and the sensor data, so that the actual judgment result is not accurate, and the reliability is not high.
Disclosure of Invention
The invention aims to solve the technical problems that a GIS isolating switch working state judging method in the prior art is low in real-time performance, not accurate in judging result and low in reliability.
The invention solves the technical problems through the following technical means: the GIS switching-on and switching-off state current detection method based on the CNN network comprises the following steps:
the method comprises the following steps: collecting the opening and closing states of a disconnecting link and a grounding switch and corresponding coupling capacitance current data sets in all states of GIS equipment;
step two: preprocessing the acquired current data, and constructing a training set of the network;
step three: constructing a CNN network, wherein the input of the CNN network is a training set, and the output of the CNN network is the opening and closing state of a disconnecting link and a grounding switch;
step four: optimizing the CNN network parameters by using a particle swarm optimization algorithm to obtain optimal CNN network parameters;
step five: and setting the CNN network according to the optimal CNN network parameters to obtain the optimal CNN network, acquiring fault current data of a disconnecting link and a grounding switch of the GIS equipment in real time, preprocessing the fault current data, inputting the preprocessed data into the optimal CNN network, and outputting a GIS switching-closing state judgment result by using the optimal CNN network.
Has the beneficial effects that: according to the method, the CNN network is constructed, the particle swarm optimization algorithm is utilized to optimize CNN network parameters to obtain optimal CNN network parameters, so that the optimal CNN network is obtained, finally, the optimal CNN network is utilized to judge fault current data of the disconnecting link and the grounding switch collected in real time, and the GIS switching-on and switching-off state is output.
Further, the first step comprises:
when the GIS equipment is in a hot standby state, the opening and closing conditions of each group of disconnecting link and earthing knife are disconnecting link closing and earthing knife opening, and the opening and closing state Y of the earthing knife is recorded H ＝{y 1 H ,y 2 H …y k H …y K H Simultaneously acquiring a current data set I H ＝{I 1 H ,I 2 H …I k H …I K H Therein of
Representing the kth earth-knife current in a hot standby state;
when the GIS equipment is in a cold standby state, the opening and closing conditions of each group of disconnecting link and earthing switch are disconnecting link and earthing switch, and the opening and closing states Y of the earthing switch and the disconnecting link are recorded C ＝{y 1 C ,y 2 C …y k C …y K C ，y 1 C' ,y 2 C' …y m C' …y M C' Simultaneously collecting a ground knife current data set I and a knife switch current data set I C ＝{I 1 C ,I 2 C …I k C …I K C ,I 1 C' ,I 2 C' …I m C' …I M C' Therein of
Indicates the kth ground switch current in the cold standby state, < > is present>
The mth knife switch current in the cold standby state is shown;
when the GIS equipment is in an overhaul state, the opening and closing conditions of each group of disconnecting link and grounding switch are disconnecting link and grounding switch closing, and the opening and closing states of the disconnecting link and Y are recorded R ＝{y 1 R ,y 2 R …y m R …y M R Simultaneously acquiring a disconnecting link current data set I R ＝{I 1 R ,I 2 R …I m R …I M R Therein of
Indicating the mth boundary knife-switch current in the maintenance state.
Further, the second step comprises:
s21, collecting fault current data of the disconnecting link and the earthing switch, and performing wavelet transformation on a fault current signal to obtain a wavelet coefficient W i (a, b) and a candidate frequency f ins ；
S22, passing formula
Obtain a zero matrix, where, let (Δ a) k ＝a k -a k-1 ，Δf＝f k -f k-1 ，a k Is the scale factor of the k-th transient component, f k Is the center frequency of the kth instantaneous component, taking k e [0, n a ]，f s Is the signal sampling frequency; from f in S21 ins According to the formula f ins ＝2 kΔf ·f s /n a Calculating a k value, wherein &>
Judging k belongs to [0, n ] a ]Whether it is satisfied, if so, then
Repeating the steps for a preset number of times, so that the signals become thin and clearly appear on a time-frequency diagram,
passing through type
Reconstruct the signal and its respective component, wherein>
R e The real part is taken.
Further, the S21 includes:
selecting a mother wavelet function psi (t) such that its Fourier transform psi (omega) satisfies the admissibility condition of the following wavelet function
Selecting a scaling factor a and a shifting factor b, and performing scaling and shifting operations on the wavelet mother function psi (t) to obtain a wavelet basis function:
wavelet transformation is carried out on the current signal according to the following formula to obtain a wavelet coefficient W i (a，b)
Wherein psi * (t) is the complex conjugate of ψ (t).
Further, the CNN network comprises an input layer, a convolution layer, a pooling layer, a full-connection layer and an output layer which are sequentially connected, wherein the convolution layer comprises 3 layers, the convolution kernel extracts the characteristics of a current time-frequency spectrogram, the number of the pooling layers is also 3, and the output data of the 3 convolution layers are subjected to dimensionality reduction respectively; the output layer is composed of a Softmax module and a classification module.
Further, the convolutional layer is represented by the formula
Obtaining the current neuron>
In the formula, M j For the total number of vectors input to neuron j>
Characteristic diagram of the input neuron of the previous layer; ki j l Weight of the convolution kernel; b is a mixture of j Is the bias of the jth profile; σ is the ReLU activation function;
softmax module output
In the formula, x j Is the jth output of the full link layer, k is SoftmThe number of input of ax module, k =2, P1 and P2 are
P1, P2 represent probabilities of outputs of y =1 and y =0, respectively; inputting the probability values P1, P2 into a classification module, and calculating a cross-loss entropy function value L in the classification module, wherein
y i Is an event tag value, y 1 ＝1，y 2 ＝0。
Further, the CNN network further includes an inclusion-v 3 structure that uses filters of sizes 1 × n and n × 1 instead of convolution of n × n in the CNN network.
Still further, the fourth step includes: setting a loss function L loss When loss value L of CNN network<L loss The network training is reliable, and the label value with a larger value in P1 and P2 is taken as the network output; otherwise, the network weight is improperly distributed, and the process is repeated by redistributing the weight by adopting the particle swarm optimization algorithm until the network output condition is met to obtain the optimal CNN network parameter.
Further, the weight redistribution by the particle swarm optimization algorithm comprises:
s41, after the error between the expected value and the actual value is calculated by the CNN network, the batch processing sample number batchsize, the training data discarding rate dropout and the convolution kernel number N of the CNN network are calculated by each particle c Size of convolution kernel M c And network initial bias b j The parameters are used as hyper-parameters to be optimized, and the hyper-parameters to be optimized are used as particle dimensions;
s42, calculating the fitness value of each particle by taking the error between the expected value and the actual value of the CNN as a fitness function, and taking the optimal point of each particle in n-time iterative updating as an individual optimal solution P best Storing, comparing with the whole population particles, and selecting a global optimum value G best Storing;
s43, comparing the fitness value of the particle with the optimal fitness value of the individual, and replacing the fitness value of the particle with the optimal fitness value of the individual if the fitness value of the particle is larger;
s44, comparing the individual optimal fitness values of all the particles with the group global optimal fitness value, and if the individual optimal fitness value is larger, replacing the individual optimal fitness value with the group global optimal fitness value;
s45, updating the speed and the position of the particles, and returning to execute the step S42;
s46, comparing the current iteration times with the maximum iteration times, and when N is less than N, continuing the iteration; and when N is larger than N, finishing iteration or finishing iteration when the minimum error is reached, returning to the step S42 if the two iteration finishing conditions are not met, otherwise, exiting the loop to obtain a global optimal fitness value, wherein each parameter of the particles corresponding to the global optimal fitness value is the optimal CNN network parameter obtained by particle swarm optimization.
The invention also provides a GIS switching-on and switching-off state current detection device based on the CNN network, which comprises:
the data acquisition module is used for acquiring the opening and closing states of a disconnecting link and a grounding switch and corresponding coupling capacitance current data sets in all states of the GIS equipment;
the training set construction module is used for preprocessing the acquired current data and constructing a training set of the network;
the network construction module is used for constructing a CNN network, the input of the CNN network is a training set, and the output of the CNN network is the opening and closing state of a disconnecting link and a grounding switch;
the parameter optimization module is used for optimizing the CNN network parameters by using a particle swarm optimization algorithm to obtain optimal CNN network parameters;
and the result output module is used for setting the CNN network according to the optimal CNN network parameters to obtain the optimal CNN network, acquiring fault current data of the disconnecting link and the grounding switch of the GIS equipment in real time, preprocessing the fault current data, inputting the preprocessed data into the optimal CNN network, and outputting a GIS switching-on/off state judgment result by using the optimal CNN network.
Further, the data acquisition module is further configured to:
the GIS equipment is in heatIn the standby state, the opening and closing conditions of each group of disconnecting link and earthing switch are disconnecting link closing and earthing switch opening, and the earthing switch opening and closing state Y is recorded H ＝{y 1 H ,y 2 H …y k H …y K H Simultaneously acquiring a current data set I H ＝{I 1 H ,I 2 H …I k H …I K H Therein of
Representing the kth earth-knife current in a hot standby state;
when the GIS equipment is in a cold standby state, the opening and closing conditions of each group of disconnecting link and grounding switch are disconnecting link and grounding switch, and the opening and closing states Y of the grounding switch and the disconnecting link are recorded C ＝{y 1 C ,y 2 C …y k C …y K C ，y 1 C' ,y 2 C' …y m C' …y M C' And (5) simultaneously acquiring a ground knife current data set I and a knife switch current data set I C ＝{I 1 C ,I 2 C …I k C …I K C ,I 1 C' ,I 2 C' …I m C' …I M C' Therein of
Indicates the kth ground switch current in the cold standby state, < > is present>
Indicating the m-th knife switch current in a cold standby state;
when the GIS equipment is in an overhaul state, the opening and closing conditions of each group of disconnecting link and grounding switch are disconnecting link and grounding switch closing, and the opening and closing states of the disconnecting link and Y are recorded R ＝{y 1 R ,y 2 R …y m R …y M R Simultaneously acquiring a disconnecting link current data set I R ＝{I 1 R ,I 2 R …I m R …I M R Therein of
Indicating the mth boundary knife-switch current in the maintenance state.
Further, the training set construction module is further configured to:
s21, collecting fault current data of the disconnecting link and the grounding switch, and performing wavelet transformation on a fault current signal to obtain a wavelet coefficient W i (a, b) and a candidate frequency f ins ；
S22, passing formula
A zero matrix is obtained, wherein (Δ a) k ＝a k -a k-1 ，Δf＝f k -f k-1 ，a k Is the scale factor of the k-th transient component, f k Is the center frequency of the kth instantaneous component, taking k e [0, n a ]，f s A signal sampling frequency; from f in S21 ins According to the formula f ins ＝2 kΔf ·f s /n a Calculating a k value, wherein &>
Judging k is belonged to [0, n' a ]Whether or not, if so, has
Repeating the steps for a preset number of times, so that the signals become thin and clearly appear on a time-frequency diagram,
passing through type
Reconstructing a signal and its individual components, in which>
R e The real part is taken.
Further, the S21 includes:
a mother wavelet function psi (t) is selected such that its Fourier transform psi (w) satisfies the admissibility conditions of the wavelet functions
Selecting a scaling factor a and a shifting factor b, and performing scaling and shifting operations on the wavelet mother function psi (t) to obtain a wavelet base function:
wavelet transformation is carried out on the current signal according to the following formula to obtain a wavelet coefficient W i (a，b)
Wherein psi * (t) is the complex conjugate of ψ (t).
Further, the CNN network comprises an input layer, a convolution layer, a pooling layer, a full-connection layer and an output layer which are sequentially connected, wherein the convolution layer comprises 3 layers, the convolution kernel extracts the characteristics of a current time-frequency spectrogram, the number of the pooling layers is also 3, and the output data of the 3 convolution layers are subjected to dimensionality reduction respectively; the output layer is composed of a Softmax module and a classification module.
Further, the convolutional layer is expressed by the formula
Obtaining the current neuron>
In the formula, M j For the total number of vectors input to neuron j, <' >>
Is frontA layer of input neuron feature map; k is a radical of ij l Weight of the convolution kernel; b j Is the bias of the jth feature map; σ is the ReLU activation function;
softmax module output
In the formula, x j J is the j output of the full connection layer, k is the input number of the Softmax module, k =2, and P1 and P2 are
P1, P2 represent probabilities of outputs of y =1 and y =0, respectively; inputting the probability values P1, P2 into a classification module, and calculating a cross-loss entropy function value L in the classification module, wherein
y i Is an event tag value, y 1 ＝1，y 2 ＝0。
Further, the CNN network further includes an inclusion-v 3 structure that uses filters of sizes 1 × n and n × 1 instead of convolution of n × n in the CNN network.
Still further, the parameter optimization module is further configured to: setting a loss function L loss When loss value L of CNN network<L loss The network training is reliable, and the label value with a larger value in P1 and P2 is taken as the network output; otherwise, indicating that the network weight is not properly distributed, and redistributing the weight by adopting a particle swarm optimization algorithm to repeat the process until the network output condition is met to obtain the optimal CNN network parameter.
Further, the re-assigning weights by the particle swarm optimization algorithm comprises:
s41, after the error between the expected value and the actual value is calculated by the CNN network, the batch processing sample number batchsize, the training data discarding rate dropout and the convolution of the CNN network are calculated by each particleNumber of cores N c Convolution kernel size M c And network initial bias b j The parameters are used as the hyper-parameters to be optimized, and the hyper-parameters to be optimized are used as particle dimensions;
s42, calculating the fitness value of each particle by taking the error between the expected value and the actual value of the CNN as a fitness function, and taking the optimal point of each particle in the n times of iterative updating as the individual optimal solution P best Storing, comparing with the whole population particles, and selecting the global optimum value G best Storing;
s43, comparing the fitness value of the particle with the optimal fitness value of the individual, and replacing the fitness value of the particle with the optimal fitness value of the individual if the fitness value of the particle is larger;
s44, comparing the individual optimal fitness values of all the particles with the group global optimal fitness value, and replacing the individual optimal fitness value with the global optimal fitness value if the individual optimal fitness value is larger;
s45, updating the speed and the position of the particles, and returning to execute the step S42;
s46, comparing the current iteration times with the maximum iteration times, and when N is less than N, continuing the iteration; and when N is larger than N, finishing iteration or finishing iteration when the minimum error is reached, returning to the step S42 if the two iteration finishing conditions are not met, otherwise, exiting the loop to obtain a global optimal fitness value, wherein each parameter of the particles corresponding to the global optimal fitness value is the optimal CNN network parameter obtained by particle swarm optimization.
The invention has the advantages that: according to the method, the CNN network is constructed, the particle swarm optimization algorithm is utilized to optimize CNN network parameters to obtain optimal CNN network parameters, so that the optimal CNN network is obtained, finally, the optimal CNN network is utilized to judge fault current data of the disconnecting link and the grounding switch collected in real time, and the GIS switching-on and switching-off state is output.
Drawings
FIG. 1 is a schematic diagram of three situations of a prior art disconnecting link and a grounding switch;
fig. 2 is a CNN network architecture diagram in the method for detecting a GIS opening/closing state current based on a CNN network disclosed in embodiment 1 of the present invention;
fig. 3 is a schematic structural diagram of an initiation module in the CNN network-based GIS opening/closing state current detection method disclosed in embodiment 1 of the present invention;
fig. 4 is a flowchart of a particle group optimization algorithm in the CNN network-based GIS opening/closing state current detection method disclosed in embodiment 1 of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
Example 1
The inventor researches the opening and closing conditions of the disconnecting link and the grounding switch, carries out statistical analysis on the faults of the disconnecting link and the grounding switch, finds out the common problem of the GIS equipment that the disconnecting link and the grounding switch do not complete the action according to the preset command due to the fault of the internal transmission mechanism, and does not have a half-opening or half-closing state. The adjacent parallel lines are in a strong electric field, can generate induced voltage, and generate electric coupling capacitance current through the grounding switch grounding plate in the grounding switch closing state, and do not have circuit coupling capacitance current in the grounding switch grounding plate in the opening state. The voltage-sharing covers on two sides of the disconnecting link fracture can be regarded as two polar plates of a capacitor, the fracture can be equivalent to a fixed capacitor, one end of the fracture is provided with high voltage, the other end of the fracture is grounded, and coupling capacitor current inevitably passes through the fracture and flows into the grounding grid through the grounding plate of the grounding switch. As shown in fig. 1, the knife switch and the ground knife have the following three conditions:
1) The knife break is connected in series with the earth break (see fig. 1 (a)), and then the calculation formula of the coupling capacitance current is as follows:
wherein j is an imaginary unit.
2) When the knife switch is in an open state and the ground switch is in a closed state (see fig. 1 (b)), the calculation formula of the coupling capacitance current is as follows:
I c ＝j2πfC g U
3) When the grounding knife is in an off state and the knife switch is in an on state (see fig. 1 (c)), the coupling capacitance current calculation formula is as follows:
I c ＝j2πfC j U
for fractures of GIS disconnecting links and grounding knife, the resistive current is 0 and the total leakage current is the coupling capacitance current because the fractures are of a pure capacitive structure; therefore, by detecting the characteristic change of the coupling capacitance current in the operation process of the disconnecting link and the grounding switch, the switching-on and switching-off conditions of the disconnecting link and the grounding switch can be judged.
Based on the analysis, the invention provides a GIS switching-on and switching-off state coupling capacitance current detection method based on a CNN network, and the judgment of the GIS switching-on and switching-off is realized by collecting the characteristic change of the coupling capacitance current in the operation process of disconnecting link and earthing switch switching-off. The process of the method of the present invention is described in detail below.
S1, collecting the opening and closing states of a disconnecting link and a grounding switch and a coupling capacitor current data set in each state of GIS equipment; the specific process of the step is as follows:
collecting current states of all equipment when the GIS equipment is powered off and powered back; when being in the hot standby state, the condition that every group switch, earthing switch open and shut is: switching on the disconnecting link and switching off the grounding link; recording ground knife on-off state Y H ＝{y 1 H ,y 2 H …y k H …y K H Simultaneously acquiring a current data set I H ＝{I 1 H ,I 2 H …I k H …I K H Therein of
Indicating the kth earth-knife current in hot standby state(ii) a When being in cold standby state, every group switch, the earthing switch condition of opening and shutting do: disconnecting link and grounding switch; recording the on-off state Y of the ground knife and the disconnecting link C ＝{y 1 C ,y 2 C …y k C …y K C ，y 1 C' ,y 2 C' …y m C' …y M C' Simultaneously collecting a ground knife current data set I and a knife switch current data set I C ＝{I 1 C ,I 2 C …I k C …I K C ,I 1 C' ,I 2 C' …I m C' …I M C' Get out of>
Indicates the kth ground fault current in the cold standby state, medium>
The mth knife switch current in the cold standby state is shown; when being in the maintenance state, every group switch, the earthing switch condition of opening and shutting do: disconnecting the knife switch and closing the ground knife; recording the on-off state of the knife switch, Y R ＝{y 1 R ,y 2 R …y m R …y M R Simultaneously acquiring a disconnecting link current data set I R ＝{I 1 R ,I 2 R …I m R …I M R Is in which>
Indicating the mth boundary knife switch current in the maintenance state.
S2, preprocessing the acquired current signals, acquiring frequency band information of the signals by adopting a wavelet decomposition algorithm, removing clutter signals, finishing noise reduction processing on the signals, and constructing a training set of a network; the specific process is as follows:
s21, collecting fault current data I (t) of the disconnecting link and the grounding switch, and performing wavelet transformation on a fault current signal to obtain a wavelet coefficient W i (a, b) and a candidate frequency f ins A wavelet mother function psi (t) is selected such that its Fourier transform psi (ω) satisfies the admissibility conditions of the wavelet function as follows
Selecting a scale factor a and a shift factor b, and performing scaling and shifting operations on the wavelet mother function ψ (t) to obtain a wavelet basis function:
wavelet transform is performed on the current signal I (t) according to the following formula to obtain W i (a，b)
Wherein psi * (t) is the complex conjugate of ψ (t).
S22, reconstructing each component of the multi-component signal through SWT inverse transformation of synchronous compression wavelet transformation, and converting the time-scale planes (a, b) into time-frequency planes (f) ins B) above, obtaining S by the following formula i (f k ，b)；
Wherein (Delta a) k ＝a k -a k-1 ，Δf＝f k -f k-1 ，a k Scale factor of the kth instantaneous component, f k Is the center frequency of the kth instantaneous component; instantaneous frequency is compressed at (f) k -0.5(Δf) k ，f k -0.5(Δf) k ) Sharpening the time-frequency graph; taking k as element [0, n ] a ]，S i (f k B) is a zero matrix and the signal sampling frequency is f s Therefore, the signal frequency takes the value of [ f s /n a ，f s /2](ii) a From S2F in 1 ins Calculating a k value according to the following formula;
f ins ＝2 kΔf ·f s /n a
wherein
Judging k belongs to [0, n ] a ]Whether or not, and if so, whether or not>
Repeating the above steps to make the signal fine and clearly present on the time-frequency diagram, and reconstructing the signal and each component thereof by the following formula of SWT inverse transformation
Wherein
R e Taking a real part;
and S3, constructing a CNN network model, wherein the network model is mainly divided into an input layer, a CNN layer and an output layer. Inputting a model by a training set through an input layer, performing feature extraction through a CNN layer, generating a feature vector, and outputting a model result through an output layer; the specific process is as follows:
s31, constructing an initial CNN network, wherein the CNN structure comprises an input layer, a convolutional layer, a pooling layer, a full-connection layer and an output layer, and the convolutional layer, the pooling layer and the full-connection layer form the CNN layer. And the input layer receives the current time-frequency spectrogram after wavelet decomposition processing, and each pixel point in the time-frequency spectrogram is used as input. The convolution layer acts on the current time-frequency spectrogram through different convolution kernel numbers and sizes and extracts the characteristics of the current time-frequency spectrogram. In order to avoid the situation that the same characteristic of time-frequency spectrograms of current is repeatedly extracted by different convolution cores, a network structure of 3 layers of convolution is designed, the number of the convolution cores of each convolution layer is respectively the optimal network parameter obtained by particle swarm optimization, and the convolutionThe characteristics of current time-frequency spectrogram are extracted by the product kernel, and the current neuron can be obtained by adding the bias and the activation function
And different feature graphs are formed according to the characteristic graph, the ReLU function is selected by the activation function, and the calculation formula is as follows:
in the formula, M j For the total number of vectors input to neuron j,
is a characteristic diagram of the previous layer of input neurons; k is a radical of formula ij l Weight of the convolution kernel; b j Is the bias of the jth feature map; σ is the ReLU activation function;
3 pooling layer structures are designed, maximum pooling is adopted, and the purpose is to reduce the dimension of the convolutional layer output data so as to reduce the complexity of the data. And ensuring that the full connection layer is connected with all the previous neurons, reducing the matrix data dimension of the pooling layer, extracting characteristic samples, and adopting a Softmax loss function as a full connection layer conversion function. The structure of the CNN model is shown in FIG. 2.
S32, adding an inclusion module, and directly replacing an n × n convolution layer in the neural network with the inclusion module, so that the n × n convolution in the particle swarm optimization process is decomposed into filters of 1 × n and n × 1, and the complexity of a CNN network structure is reduced. Generally, when n is relatively large, the inclusion module is used to replace the convolution layer of n × n, and in practical application, the value of n is set as required. The CNN network structure is formed by stacking convolutional layers, more layers are needed to achieve a better effect, the network structure is deepened, the inclusion is added to combine different convolutional layers in a parallel mode, the width of the network is stretched, the depth of the network structure is reduced, filters with multiple sizes are operated on the same level of network, and the complexity of the CNN is reduced. An inclusion-v 3 architecture is used that uses filters of size 1 xn and nx 1 instead of the n x n convolution, which performs the convolution and pooling operations on the previous layer. By adopting the structure, the filter is smaller, the number of parameters is reduced, the calculation speed is higher, and the training efficiency is improved; the structure of the initiation module is shown in FIG. 3.
S33, the output layer is composed of a Softmax module and a classification module, x1 and x2 are full connection layer outputs, and P1 and P2 are output of the output layer Softmax module
Wherein k is the input number of the Softmax module, and k =2, then P1 and P2 are
P1, P2 represent probabilities of outputs of y =1 and y =0, respectively; inputting the probability values P1 and P2 into a classification module, and calculating a cross loss entropy function value L in the classification module;
wherein
y i Is an event tag value, y 1 ＝1，y 2 =0; setting a loss function L loss ＝1.2×10 -3 . When L is<L loss The network training is reliable, and the label value with larger value in P1 and P2 is taken as the network output; otherwise, the network weight is not distributed properly, and the process is repeated until the weight is distributed properly. As shown in fig. 4, the weight assignment adopts a particle swarm optimization algorithm, and the algorithm process refers to the following steps.
S4, selecting and optimizing the established CNN network structure and initial parameters by using a particle swarm optimization algorithm;
s41, after the error between the expected value and the actual value is calculated by the CNN network, the batch processing sample number batchsize, the training data discarding rate dropout and the convolution kernel number N of the CNN network are calculated by each particle c Size of convolution kernel M c And network initial bias b j As the hyper-parameter to be optimized, as the particle dimension;
and S42, calculating the fitness value of each particle by taking the error between the expected value and the actual value of the CNN network as a fitness function. Taking the optimal point of each particle in the n iterative updates as the individual optimal solution P best Storing, comparing with the whole population particles, and selecting the global optimum value G best Storing;
s43, comparing the fitness value of the particle with the individual optimal fitness value, and replacing the fitness value of the particle with the individual optimal fitness value if the fitness value of the particle is larger;
and S44, comparing the individual optimal fitness values of all the particles with the group global optimal fitness value, and replacing the individual optimal fitness value with the global optimal fitness value if the individual optimal fitness value is larger.
S45, updating the speed and the position of the particles according to the following formula, and returning to execute the step S42;
w n is a coefficient of the inertial weight, and,
representing the speed of the i-th particle at the nth iteration>
For a global learning factor, <' >>
As a local learning factor, R 1 And R 2 Is a random number between (0, 1)>
For individual extrema after the nth iteration>
Is the population extremum after the nth iteration->
Is the position after the nth iteration;
while weighting the inertia weight coefficient w therein by n Carrying out nonlinear degressive processing to enhance the local and global searching capability of multiple targets;
w o initial inertial weight coefficient, w f The value is the inertia weight value after the iteration is finished, and N is the maximum iteration number;
for learning factors at the same time
Updating according to the following formula;
By matching learning factors
And an inertia weight coefficient w n The dynamic adjustment of the particle swarm optimization algorithm is realized, the local and global searching capabilities are balanced, and the global searching level and the convergence speed of the particle swarm optimization algorithm are improved.
S46, comparing the current iteration times with the maximum iteration times, and when N is less than N, continuing the iteration; and when N is larger than N, finishing iteration or finishing iteration when the minimum error is reached, if the two iteration finishing conditions are not met, returning to the step S42, otherwise, exiting the loop to obtain a global optimal fitness value, wherein each parameter of the particles corresponding to the global optimal fitness value is the optimal CNN network parameter obtained by particle swarm optimization.
S5, setting the CNN network according to the optimal CNN network parameters to obtain the optimal CNN network, acquiring fault current data of a disconnecting link and a grounding switch of the GIS equipment in real time, preprocessing the fault current data, inputting the fault current data into the optimal CNN network, and outputting a GIS switching-on and switching-off state judgment result by using the optimal CNN network.
Example 2
Based on embodiment 1, embodiment 2 of the present invention further provides a device for detecting a current in a GIS switching-on/off state based on a CNN network, where the device includes:
the data acquisition module is used for acquiring the opening and closing states of a disconnecting link and a grounding switch and corresponding coupling capacitance current data sets in all states of the GIS equipment;
the training set construction module is used for preprocessing the acquired current data and constructing a training set of the network;
the network construction module is used for constructing a CNN network, the input of the CNN network is a training set, and the output of the CNN network is the opening and closing states of a disconnecting link and a grounding switch;
the parameter optimization module is used for optimizing the CNN network parameters by using a particle swarm optimization algorithm to obtain optimal CNN network parameters;
and the result output module is used for setting the CNN network according to the optimal CNN network parameters to obtain the optimal CNN network, acquiring fault current data of a disconnecting link and a grounding switch of the GIS equipment in real time, preprocessing the fault current data, inputting the preprocessed data into the optimal CNN network, and outputting a GIS switching-on/off state judgment result by using the optimal CNN network.
Specifically, the data acquisition module is further configured to:
when the GIS equipment is in a hot standby state, the opening and closing conditions of each group of disconnecting link and earthing knife are disconnecting link closing and earthing knife opening, and the opening and closing state Y of the earthing knife is recorded H ＝{y 1 H ,y 2 H …y k H …y K H }, simultaneously acquiring a current data set I H ＝{I 1 H ,I 2 H …I k H …I K H Therein of
Representing the kth earth-knife current in a hot standby state;
when the GIS equipment is in a cold standby state, the opening and closing conditions of each group of disconnecting link and grounding switch are disconnecting link and grounding switch, and the opening and closing states Y of the grounding switch and the disconnecting link are recorded C ＝{y 1 C ,y 2 C …y k C …y K C ，y 1 C' ,y 2 C' …y m C' …y M C' And (5) simultaneously acquiring a ground knife current data set I and a knife switch current data set I C ＝{I 1 C ,I 2 C …I k C …I K C ,I 1 C' ,I 2 C' …I m C' …I M C' Therein of
Indicating a kth ground knife current in a cold standby state>
The mth knife switch current in the cold standby state is shown;
when the GIS equipment is in an overhaul state, the opening and closing conditions of each group of disconnecting link and grounding switch are disconnecting link and grounding switch closing, and the opening and closing states of the disconnecting link and Y are recorded R ＝{y 1 R ,y 2 R …y m R …y M R At the same time, collectingKnife switch current data set I R ＝{I 1 R ,I 2 R …I m R …I M R Therein of
Indicating the mth boundary knife-switch current in the maintenance state.
Specifically, the training set constructing module is further configured to:
s21, collecting fault current data of the disconnecting link and the earthing switch, and performing wavelet transformation on a fault current signal to obtain a wavelet coefficient W i (a, b) and a candidate frequency f ins ；
S22, passing formula
A zero matrix is obtained, where (Δ a) k ＝a k -a k-1 ，Δf＝f k -f k-1 ，a k Is the scale factor of the k-th transient component, f k Is the center frequency of the kth instantaneous component, taking k e [0, n a ]，f s Is the signal sampling frequency; from f in S21 ins According to the formula f ins ＝2 kΔf ·f s /n a Calculating a k value, wherein &>
Judging k belongs to [0, n ] a ]Whether it is satisfied, if so, then
Repeating the steps for a preset number of times, so that the signals become thin and clearly appear on a time-frequency diagram,
passing through type
Reconstruct the signal and its respective component, wherein>
R e The real part is taken.
More specifically, the S21 includes:
selecting a mother wavelet function psi (t) such that its Fourier transform psi (omega) satisfies the admissibility condition of the following wavelet function
Selecting a scaling factor a and a shifting factor b, and performing scaling and shifting operations on the wavelet mother function psi (t) to obtain a wavelet base function:
wavelet transformation is carried out on the current signal according to the following formula to obtain a wavelet coefficient W i (a，b)
Wherein psi * (t) is the complex conjugate of ψ (t).
Specifically, the CNN network includes an input layer, a convolution layer, a pooling layer, a full-link layer, and an output layer, which are sequentially connected, where the convolution layer includes 3 layers, the convolution kernel extracts the characteristics of a current time-frequency spectrogram, the pooling layer also includes 3 layers, and the 3 convolution layers output data are respectively reduced in dimension; the output layer is composed of a Softmax module and a classification module.
More specifically, the convolutional layer is represented by the formula
Obtaining the current neuron>
In the formula, M j For the total number of vectors input to neuron j>
Characteristic diagram of the input neuron of the previous layer; k is a radical of ij l Weight of the convolution kernel; b is a mixture of j Is the bias of the jth feature map; σ is a ReLU activation function;
softmax module output
In the formula, x j J is the j output of the full connection layer, k is the input number of the Softmax module, k =2, and P1 and P2 are
P1, P2 represent probabilities of outputs of y =1 and y =0, respectively; inputting the probability values P1, P2 into a classification module, and calculating a cross-loss entropy function value L in the classification module, wherein
y i Is an event tag value, y 1 ＝1，y 2 ＝0。
More specifically, the CNN network further includes an inclusion-v 3 structure that uses filters of sizes 1 × n and n × 1 instead of the convolution of n × n in the CNN network.
More specifically, the parameter optimization module is further configured to: setting a loss function L loss When loss value L of CNN network<L loss The network training is reliable, and the label value with a larger value in P1 and P2 is taken as the network output; otherwise, the network weight is improperly distributed, and the process is repeated by redistributing the weight by adopting the particle swarm optimization algorithm until the network output condition is met to obtain the optimal CNN network parameter.
More specifically, the weight redistribution by the particle swarm optimization algorithm comprises:
s41, after error between the expected value and the actual value is calculated in the CNN networkThe batch processing sample number batchsize, the training data discarding rate dropout and the convolution kernel number N of the CNN network are calculated for each particle c Size of convolution kernel M c And network initial bias b j The parameters are used as hyper-parameters to be optimized, and the hyper-parameters to be optimized are used as particle dimensions;
s42, calculating the fitness value of each particle by taking the error between the expected value and the actual value of the CNN as a fitness function, and taking the optimal point of each particle in the n times of iterative updating as the individual optimal solution P best Storing, comparing with the whole population particles, and selecting the global optimum value G best Storing;
s43, comparing the fitness value of the particle with the optimal fitness value of the individual, and replacing the fitness value of the particle with the optimal fitness value of the individual if the fitness value of the particle is larger;
s44, comparing the individual optimal fitness values of all the particles with the group global optimal fitness value, and if the individual optimal fitness value is larger, replacing the individual optimal fitness value with the group global optimal fitness value;
s45, updating the speed and the position of the particles, and returning to execute the step S42;
s46, comparing the current iteration times with the maximum iteration times, and when N is less than N, continuing the iteration; and when N is larger than N, finishing iteration or finishing iteration when the minimum error is reached, returning to the step S42 if the two iteration finishing conditions are not met, otherwise, exiting the loop to obtain a global optimal fitness value, wherein each parameter of the particles corresponding to the global optimal fitness value is the optimal CNN network parameter obtained by particle swarm optimization.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. GIS switching-on and switching-off state current detection method based on CNN network, characterized in that the method comprises:
the method comprises the following steps: collecting the opening and closing states of a disconnecting link and a grounding switch and corresponding coupling capacitance current data sets in all states of GIS equipment;
step two: preprocessing the acquired current data, and constructing a training set of the network;
step three: constructing a CNN network, wherein the input of the CNN network is a training set, and the output of the CNN network is the opening and closing state of a disconnecting link and a grounding switch;
step four: optimizing CNN network parameters by using a particle swarm optimization algorithm to obtain optimal CNN network parameters;
step five: and setting the CNN network according to the optimal CNN network parameters to obtain the optimal CNN network, acquiring fault current data of a disconnecting link and a grounding switch of the GIS equipment in real time, preprocessing the fault current data, inputting the fault current data into the optimal CNN network, and outputting a GIS switching-on and switching-off state judgment result by using the optimal CNN network.
2. The method for detecting GIS opening and closing state current based on CNN network according to claim 1, wherein the step one includes:
when the GIS equipment is in a hot standby state, the switching conditions of each group of disconnecting link and grounding switch are disconnecting link switching-on and grounding switch switching-off, and the switching state Y of the grounding switch is recorded H ＝{y 1 H ,y 2 H …y k H …y K H }, simultaneously acquiring a current data set I H ＝{I 1 H ,I 2 H …I k H …I K H Therein of
Representing the kth earth-knife current in a hot standby state;
when the GIS equipment is in a cold standby state, the opening and closing conditions of each group of disconnecting link and earthing switch are disconnecting link and earthing switch, and the opening and closing states Y of the earthing switch and the disconnecting link are recorded C ＝{y 1 C ,y 2 C …y k C …y K C ，y 1 C' ,y 2 C' …y m C' …y M C' Simultaneously collecting a ground knife current data set I and a knife switch current data set I C ＝{I 1 C ,I 2 C …I k C …I K C ,I 1 C' ,I 2 C' …I m C' …I M C' Therein of
Indicates the kth ground switch current in the cold standby state, < > is present>
Indicating the m-th knife switch current in a cold standby state;
when the GIS equipment is in the maintenance state, the switching conditions of each group of disconnecting link and grounding switch are disconnecting link and grounding switch switching, and the switching states of the disconnecting links and the grounding switch are recorded, namely Y R ＝{y 1 R ，y 2 R …y m R …y M R Simultaneously acquiring a disconnecting link current data set I R ＝{I 1 R ，I 2 R …I m R …I M R Therein of
Indicating the mth boundary knife switch current in the maintenance state.
3. The method for detecting GIS opening and closing state current based on CNN network according to claim 1, wherein the second step comprises:
s21, collecting fault current data of the disconnecting link and the grounding switch, and performing wavelet transformation on a fault current signal to obtain a wavelet coefficient W i (a, b) and a candidate frequency f ins ；
S22, passing formula
A zero matrix is obtained, where (Δ a) k ＝a k -a k-1 ，Δf＝f k -f k-1 ，a k Is the scale factor of the k-th transient component, f k Is the center frequency of the kth transient component, taking k e [0, n a ]，f s A signal sampling frequency; from f in S21 ins According to the formula f ins ＝2 kΔf ·f s /n a Calculating a k value, wherein &>
Judging k belongs to [0, n ] a ]Whether it is satisfied, if so, then
Repeating the steps for preset times, making the signal thin and clearly showing on the time-frequency diagram,
passing through type
Reconstruct the signal and its respective component, wherein>
R e The real part is taken. />
4. The method for detecting GIS opening/closing state current based on CNN network according to claim 3, wherein S21 includes:
a mother wavelet function psi (t) is selected such that its Fourier transform psi (omega) satisfies the admissibility conditions of the following wavelet functions
Selecting a scaling factor a and a shifting factor b, and performing scaling and shifting operations on the wavelet mother function psi (t) to obtain a wavelet basis function:
wavelet transformation is carried out on the current signal according to the following formula to obtain a wavelet coefficient W i (a，b)
Wherein psi * (t) is the complex conjugate of ψ (t).
5. The method for detecting GIS switching-on/off state current based on CNN network according to claim 1, wherein CNN network comprises sequentially connected input layer, convolution layer, pooling layer, full-connection layer and output layer, the convolution layer is 3 layers, convolution kernel extracts current time-frequency spectrogram feature, the pooling layer is also 3, 3 convolution layers output data are respectively reduced in dimension; the output layer is composed of a Softmax module and a classification module.
6. The GIS switching-on/off state current detection method based on CNN network of claim 5, characterized in that the convolution layer passes through a formula
Obtain the current neuron>
In the formula, M j For the total number of vectors input to neuron j, <' >>
Is a characteristic diagram of the previous layer of input neurons; k is a radical of formula ij 1 Weight of the convolution kernel; b is a mixture of j Is the bias of the jth profile; σ is a ReLU activation function;
softmax module output
In the formula, x j J is the j output of the full connection layer, k is the input number of the Softmax module, and k =2, then P1 and P2 are
P1, P2 represent probabilities of outputs of y =1 and y =0, respectively; inputting the probability values P1, P2 into a classification module, and calculating a cross-loss entropy function value L in the classification module, wherein
y i Is an event tag value, y 1 ＝1，y 2 ＝0。
7. The method of claim 6, wherein the CNN network further comprises an inclusion-v 3 structure that uses filters of 1 × n and n × 1 sizes to replace the convolution of n × n in the CNN network.
8. The GIS switching-on/off state current detection method based on CNN network according to claim 6, characterized by that, the fourth step includes: setting a loss function L loss When the loss value L of the CNN network is less than L loss The network training is reliable, and the label value with a larger value in P1 and P2 is taken as the network output; otherwise, indicating that the network weight is not properly distributed, and redistributing the weight by adopting a particle swarm optimization algorithm to repeat the process until the network output condition is met to obtain the optimal CNN network parameter.
9. The GIS switching-on/off state current detection method based on CNN network of claim 8, wherein the weight redistribution by the particle swarm optimization comprises:
s41, after the error between the expected value and the actual value is calculated by the CNN network, the batch processing sample number batchsize, the training data discarding rate dropout and the convolution kernel number N of the CNN network are calculated by each particle c Size of convolution kernel M c And network initial bias b j The parameters are used as hyper-parameters to be optimized, and the hyper-parameters to be optimized are used as particle dimensions;
s42, calculating the fitness value of each particle by taking the error between the expected value and the actual value of the CNN as a fitness function, and taking the optimal point of each particle in the n times of iterative updating as the individual optimal solution P best Storing, comparing with the whole population particles, and selecting the global optimum value G best Storing;
s43, comparing the fitness value of the particle with the individual optimal fitness value, and replacing the fitness value of the particle with the individual optimal fitness value if the fitness value of the particle is larger;
s44, comparing the individual optimal fitness values of all the particles with the group global optimal fitness value, and replacing the individual optimal fitness value with the global optimal fitness value if the individual optimal fitness value is larger;
s45, updating the speed and the position of the particles, and returning to execute the step S42;
s46, comparing the current iteration times with the maximum iteration times, and when N is less than N, continuing the iteration; and when N is larger than N, finishing iteration or finishing iteration when the minimum error is reached, returning to the step S42 if the two iteration finishing conditions are not met, otherwise, exiting the loop to obtain a global optimal fitness value, wherein each parameter of the particles corresponding to the global optimal fitness value is the optimal CNN network parameter obtained by particle swarm optimization.
10. GIS divide-shut brake state current detection device based on CNN network, its characterized in that, the device includes:
the data acquisition module is used for acquiring the opening and closing states of a disconnecting link and a grounding switch and corresponding coupling capacitance current data sets in all states of the GIS equipment;
the training set construction module is used for preprocessing the acquired current data and constructing a training set of the network;
the network construction module is used for constructing a CNN network, the input of the CNN network is a training set, and the output of the CNN network is the opening and closing states of a disconnecting link and a grounding switch;
the parameter optimization module is used for optimizing the CNN network parameters by using a particle swarm optimization algorithm to obtain optimal CNN network parameters;
and the result output module is used for setting the CNN network according to the optimal CNN network parameters to obtain the optimal CNN network, acquiring fault current data of a disconnecting link and a grounding switch of the GIS equipment in real time, preprocessing the fault current data, inputting the preprocessed data into the optimal CNN network, and outputting a GIS switching-on/off state judgment result by using the optimal CNN network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211481173.0A CN115879044A (en) | 2022-11-24 | 2022-11-24 | CNN network-based GIS switching-on/off state current detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211481173.0A CN115879044A (en) | 2022-11-24 | 2022-11-24 | CNN network-based GIS switching-on/off state current detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115879044A true CN115879044A (en) | 2023-03-31 |
Family
ID=85763771
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211481173.0A Pending CN115879044A (en) | 2022-11-24 | 2022-11-24 | CNN network-based GIS switching-on/off state current detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115879044A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117216673A (en) * | 2023-11-08 | 2023-12-12 | 国网江西省电力有限公司电力科学研究院 | Current transformer monitoring evaluation overhauls platform |
| CN117805607A (en) * | 2024-02-29 | 2024-04-02 | 山西漳电科学技术研究院(有限公司) | DC level difference matching test method for power plant DC system |
-
2022
- 2022-11-24 CN CN202211481173.0A patent/CN115879044A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117216673A (en) * | 2023-11-08 | 2023-12-12 | 国网江西省电力有限公司电力科学研究院 | Current transformer monitoring evaluation overhauls platform |
| CN117216673B (en) * | 2023-11-08 | 2024-03-12 | 国网江西省电力有限公司电力科学研究院 | Current transformer monitoring evaluation overhauls platform |
| CN117805607A (en) * | 2024-02-29 | 2024-04-02 | 山西漳电科学技术研究院(有限公司) | DC level difference matching test method for power plant DC system |
| CN117805607B (en) * | 2024-02-29 | 2024-05-07 | 山西漳电科学技术研究院(有限公司) | DC level difference matching test method for power plant DC system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115879044A (en) | CNN network-based GIS switching-on/off state current detection method and device | |
| CN109948597B (en) | High-voltage circuit breaker mechanical fault diagnosis method | |
| CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
| CN110926778B (en) | Mechanical fault diagnosis method for gas insulated switchgear assembly based on abnormal vibration | |
| Yang et al. | Condition evaluation for opening damper of spring operated high-voltage circuit breaker using vibration time-frequency image | |
| CN106482937B (en) | Method for monitoring mechanical state of high-voltage circuit breaker | |
| CN105891707A (en) | Opening-closing fault diagnosis method for air circuit breaker based on vibration signals | |
| CN103575525A (en) | Intelligent diagnosis method for mechanical fault of circuit breaker | |
| Wang et al. | Few-shot transfer learning with attention mechanism for high-voltage circuit breaker fault diagnosis | |
| CN103616635A (en) | Method and device for diagnosing mechanical characteristic failures of high-voltage circuit-breaker | |
| CN113219328B (en) | Intelligent fault diagnosis method for circuit breaker operating mechanism based on information fusion | |
| CN116524200A (en) | High-voltage circuit breaker fault diagnosis method based on image recognition | |
| CN116186520A (en) | Multi-task combined high-voltage isolating switch fault diagnosis method | |
| CN115659551A (en) | Water turbine set monitoring data anomaly detection method based on graph neural network | |
| CN117607672A (en) | Intelligent monitoring method and system for GIS circuit breaker | |
| CN108445382A (en) | A kind of rapid failure detection method of high-voltage circuitbreaker | |
| CN111079647A (en) | Circuit breaker defect identification method | |
| Wang et al. | CNN based mechanical fault diagnosis of high voltage circuit breaker using sound and current signal | |
| CN112883639A (en) | GIS equipment service life prediction device and method based on machine learning | |
| Wang et al. | Synthetic data in DC microgrids: Label creation for ensemble learning for fault isolation | |
| de Aguiar et al. | Features extraction and selection with the scattering transform for electrical load classification | |
| CN204065345U (en) | A kind of mechanical characteristic of high-voltage circuit breaker trouble-shooter | |
| Li et al. | Fault Diagnosis Technology of Industrial Circuit Breaker Optimized by Genetic Algorithm in LVQ Network | |
| Ma et al. | Mechanical fault diagnosis of GIS disconnector based on synchrosqueezing wavelet transform and stacked autoencoder | |
| Khokhar et al. | Automated recognition of single & hybrid power quality disturbances using wavelet transform based support vector machine |
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 |