CN113987906A - Electromagnetic voltage transformer voltage back calculation method based on deep learning - Google Patents
Electromagnetic voltage transformer voltage back calculation method based on deep learning Download PDFInfo
- Publication number
- CN113987906A CN113987906A CN202110903339.2A CN202110903339A CN113987906A CN 113987906 A CN113987906 A CN 113987906A CN 202110903339 A CN202110903339 A CN 202110903339A CN 113987906 A CN113987906 A CN 113987906A
- Authority
- CN
- China
- Prior art keywords
- voltage
- voltage transformer
- gate
- input
- data
- 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
- 238000004364 calculation method Methods 0.000 title claims abstract description 61
- 238000013135 deep learning Methods 0.000 title claims abstract description 23
- 230000004913 activation Effects 0.000 claims description 53
- 239000013598 vector Substances 0.000 claims description 52
- 230000006870 function Effects 0.000 claims description 41
- 238000012549 training Methods 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 14
- 230000005350 ferromagnetic resonance Effects 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 6
- 238000012886 linear function Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 5
- 238000010200 validation analysis Methods 0.000 claims 1
- 238000012544 monitoring process Methods 0.000 abstract description 4
- 238000000034 method Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000020411 cell activation Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Protection Of Transformers (AREA)
Abstract
The invention discloses a deep learning-based voltage back calculation method for an electromagnetic voltage transformer, which comprises the following steps of: 1) collecting secondary side voltage data of a voltage transformer to be tested; 2) establishing an LSTM electromagnetic voltage transformer voltage back calculation model; 3) and inputting the secondary side voltage data of the voltage transformer to be tested into an LSTM electromagnetic voltage transformer voltage back calculation model, and calculating to obtain the primary side voltage value of the voltage transformer to be tested. The invention is based on the LSTM model, so that the primary side voltage of the voltage transformer can be accurately reduced only by measuring the secondary side voltage of the voltage transformer, and more accurate data is provided for the safety monitoring of a power grid.
Description
Technical Field
The invention relates to the field of power system technology and big data processing, in particular to a voltage back-calculation method of an electromagnetic voltage transformer based on deep learning.
Background
The artificial neural network deep learning is a brand new concept proposed in recent years and is a brand new research direction in the field of machine learning. The artificial neural network deep learning is to reveal the intrinsic rules and the expression levels of sample data by continuously and deeply learning the sample data, and the final aim is to enable a machine to have the analysis and learning capability like a human. The concept of deep learning is derived from the research of artificial neural networks, and a multi-layer perceptron comprising a plurality of hidden layers is a deep learning structure. Deep learning is achieved by continuously grouping elementary units into complex, abstract properties or features. And selecting proper input and output layers, and continuously and deeply learning based on sample data to further construct a more accurate input and output function relation.
The voltage transformer is a main voltage sensing terminal for monitoring the voltage in the power grid, and is used for stably monitoring the voltage operation state of the power grid for a long time so as to ensure the safe operation of the power grid and provide real-time data support for risk early warning, accident source analysis, harmonic analysis and the like. However, the existing voltage transformer with large-scale grid distribution has low-frequency deep saturation distortion and medium-high frequency distortion, so that the frequency band range or amplitude range which can be effectively measured is extremely narrow, and the digital construction of a power grid is severely restricted. In order to solve the problem that the voltage transformer is difficult to accurately measure the voltage full wave, a wideband nonlinear model capable of accurately representing the voltage transformer must be established.
Disclosure of Invention
The invention aims to provide a voltage back calculation method of an electromagnetic voltage transformer based on deep learning, which comprises the following steps:
1) and collecting secondary side voltage data of the voltage transformer to be measured.
2) And establishing an LSTM electromagnetic voltage transformer voltage back calculation model.
The LSTM electromagnetic voltage transformer voltage back calculation model comprises an input layer, a hidden layer and an output layer. The activation functions of the input gate, the output gate and the forgetting gate are sigmoid functions. Each gate has a sigmoid function and a bitwise multiplication operation.
The input layer is used for receiving secondary side voltage data of the voltage transformer to be tested.
The hidden layer comprises a plurality of memory modules, and each memory module comprises an input unit, an output unit, an input gate, an output gate, a forgetting gate and a Ceil. The activation function of the input unit and the output unit is a tanh function.
The activation function of the output layer is a linear function, and the output layer is used for outputting the primary side voltage of the voltage transformer to be tested.
LSTM electromagnetic voltage transformer voltage inverse calculation model input door itForgetting door ftAnd an output gate otUnit activation vector ctA hidden layer unit htRespectively as follows:
it＝σ(WxiXt+Whiht-1+Wcict-1+bi)
ft＝σ(WxfXt+Whfht-1+Wcfct-1+bf)
ct＝ftct-1+it tanh(WxcXt+Whcht-1+bc)
ot＝σ(WxoXt+Whoht-1+Wcoct-1+bo) (1)
ht＝ot tanh(ct)
in the formula, σ represents a logical sigmoid function. Wxi、Whi、WciRespectively representing the weight matrixes between the input gate and the input feature vector, between the input gate and the hidden layer unit, and between the input gate and the unit activation vector. Wxf、Whf、WcfRespectively representing weight matrixes between the forgetting gate and the input characteristic vector, between the forgetting gate and the hidden layer unit, and between the forgetting gate and the unit activation vector, Wxo、Who、WcoAnd respectively representing the weight matrixes between the output gate and the input feature vector, between the output gate and the hidden layer unit, and between the output gate and the unit activation vector. Wxc、WhcRespectively representing the weight matrixes between the unit activation vector and the input feature vector and between the unit activation vector and the hidden layer unit. t denotes the sampling instant. tanh is the activation function. bi、bf、bc、boAnd respectively representing the deviation values of the input gate, the forgetting gate, the unit activation vector and the output gate. XtRepresenting the input.
The LSTM electromagnetic voltage transformer voltage back calculation model is obtained by training a voltage transformer port voltage training data set. And the voltage transformer port voltage training data set comprises voltage transformer port voltage data after intervention processing.
The step of collecting voltage data of the voltage transformer port comprises the following steps:
a) and connecting a resistance-capacitance voltage divider in parallel at the primary side of the voltage transformer, and carrying out high-frequency acquisition on a secondary side signal of the resistance-capacitance voltage divider to obtain a secondary side signal of the resistance-capacitance voltage divider.
And multiplying the secondary side signal of the resistance-capacitance voltage divider by the transformation ratio of the resistance-capacitance voltage divider to obtain a primary side voltage signal of the voltage transformer.
b) And carrying out high-frequency acquisition on the secondary side signal of the voltage transformer to obtain secondary side voltage data of the voltage transformer.
c) A typical overvoltage is applied to the primary side of the transformer and the primary side voltage and the secondary side voltage of the voltage transformer are measured. The typical overvoltage comprises standard lightning overvoltage, non-standard lightning overvoltage, power frequency overvoltage, low frequency overvoltage, frequency division ferromagnetic resonance overvoltage, high frequency ferromagnetic resonance overvoltage, fundamental frequency ferromagnetic resonance overvoltage, arc grounding overvoltage and standard operation overvoltage.
d) Changing waveform parameters and amplitude of the overvoltage, and returning to the step 3) to obtain a plurality of groups of voltage transformer port voltage data.
The step of acquiring the voltage data of the preprocessed voltage transformer port comprises the following steps:
I) and carrying out abnormal data detection on the voltage data of the voltage transformer port by using an isolated forest algorithm, and removing abnormal data according to a detection result.
II) carrying out waveform normalization processing on the data, and changing the named value into a per unit value. The normalized calculation formula is:
in the formula, m' is a per unit value, m is a named value, and REF is a reference value.
II) establishing an input parameter matrix, namely:
in the formula, XnAt a time tnThe secondary side voltage of (c). Y isnAt a time tnThe primary side voltage of (1). And n is a time step.
The LSTM electromagnetic voltage transformer voltage back calculation model is verified by a voltage transformer port voltage verification data set. The voltage transformer port voltage verification data set comprises voltage transformer port voltage data after intervention processing, and the data of the voltage transformer port voltage verification data set is different from the data of the voltage transformer port voltage training data set.
3) And inputting the secondary side voltage data of the voltage transformer to be tested into an LSTM electromagnetic voltage transformer voltage back calculation model, and calculating to obtain the primary side voltage value of the voltage transformer to be tested.
The invention has the advantages that the invention is based on the LSTM model, so that the primary side voltage can be accurately reduced only by measuring the secondary side voltage of the voltage transformer, and more accurate data is provided for the safety monitoring of the power grid. Meanwhile, the LSTM mode adopted by the invention has the characteristics of strong learning ability, strong universality and the like.
Drawings
Fig. 1 is a main flow chart of a voltage transformer back calculation method based on an LSTM model.
FIG. 2 is a schematic diagram of the LSTM model.
Fig. 3 is a graph showing the results of a comparative experiment using a voltage transformer back-calculation method based on the LSTM model.
Detailed Description
The present invention is further illustrated by the following examples, but it should not be construed that the scope of the above-described subject matter is limited to the following examples. Various substitutions and alterations can be made without departing from the technical idea of the invention and the scope of the invention is covered by the present invention according to the common technical knowledge and the conventional means in the field.
Example 1:
referring to fig. 1 to 2, a method for back-computing a voltage of an electromagnetic voltage transformer based on deep learning includes the following steps:
1) and collecting secondary side voltage data of the voltage transformer to be measured.
2) And establishing an LSTM electromagnetic voltage transformer voltage back calculation model.
Referring to fig. 2, the voltage back calculation model of the LSTM electromagnetic voltage transformer includes an input layer, a hidden layer, and an output layer. The activation functions of the input gate, the output gate and the forgetting gate are sigmoid functions. Each gate has a sigmoid function and a bitwise multiplication operation. The structure of a in fig. 2 is the same as that of the middle box in fig. 2.
The input layer is used for receiving secondary side voltage data of the voltage transformer to be tested.
The hidden layer comprises a plurality of memory modules, and each memory module comprises an input unit, an output unit, an input gate, an output gate, a forgetting gate and a Ceil. The activation function of the input unit and the output unit is a tanh function.
The activation function of the output layer is a linear function, and the output layer is used for outputting the primary side voltage of the voltage transformer to be tested.
LSTM electromagnetic voltage transformer voltage inverse calculation model input door itForgetting door ftAnd an output gate otUnit activation vector ctA hidden layer unit htRespectively as follows:
it＝σ(WxiXt+Whiht-1+Wcict-1+bi)
ft＝σ(WxfXt+Whfht-1+Wcfct-1+bf)
ct＝ftct-1+it tanh(WxcXt+Whcht-1+bc)
ot＝σ(WxoXt+Whoht-1+Wcoct-1+bo) (1)
ht＝ot tanh(ct)
in the formula, σ represents a logical sigmoid function. Wxi、Whi、WciRespectively representing the weight matrixes between the input gate and the input feature vector, between the input gate and the hidden layer unit, and between the input gate and the unit activation vector. Wxf、Whf、WcfRespectively representing weight matrixes between the forgetting gate and the input characteristic vector, between the forgetting gate and the hidden layer unit, and between the forgetting gate and the unit activation vector, Wxo、Who、WcoAnd respectively representing the weight matrixes between the output gate and the input feature vector, between the output gate and the hidden layer unit, and between the output gate and the unit activation vector. Wxc、WhcRespectively representing the weight matrixes between the unit activation vector and the input feature vector and between the unit activation vector and the hidden layer unit. t denotes the sampling instant. tanh is the activation function. bi、bf、bc、boRespectively representing deviation values of an input gate, a forgetting gate, a unit activation vector and an output gate。XtRepresenting the input to the model.
The LSTM electromagnetic voltage transformer voltage back calculation model is obtained by training a voltage transformer port voltage training data set. And the voltage transformer port voltage training data set comprises voltage transformer port voltage data after intervention processing.
The step of collecting voltage data of the voltage transformer port comprises the following steps:
a) and connecting a resistance-capacitance voltage divider in parallel at the primary side of the voltage transformer, and carrying out high-frequency acquisition on a secondary side signal of the resistance-capacitance voltage divider to obtain a secondary side signal of the resistance-capacitance voltage divider.
And multiplying the secondary side signal of the resistance-capacitance voltage divider by the transformation ratio of the resistance-capacitance voltage divider to obtain a primary side voltage signal of the voltage transformer.
b) And carrying out high-frequency acquisition on the secondary side signal of the voltage transformer to obtain secondary side voltage data of the voltage transformer.
c) A typical overvoltage is applied to the primary side of the transformer and the primary side voltage and the secondary side voltage of the voltage transformer are measured. The typical overvoltage comprises standard lightning overvoltage, non-standard lightning overvoltage, power frequency overvoltage, low frequency overvoltage, frequency division ferromagnetic resonance overvoltage, high frequency ferromagnetic resonance overvoltage, fundamental frequency ferromagnetic resonance overvoltage, arc grounding overvoltage and standard operation overvoltage.
d) Changing waveform parameters and amplitude of the overvoltage, and returning to the step 3) to obtain a plurality of groups of voltage transformer port voltage data.
The step of acquiring the voltage data of the preprocessed voltage transformer port comprises the following steps:
I) and carrying out abnormal data detection on the voltage data of the voltage transformer port by using an isolated forest algorithm, and removing abnormal data according to a detection result.
II) carrying out waveform normalization processing on the data, and changing the named value into a per unit value. The normalized calculation formula is:
in the formula, m' is a per unit value, m is a named value, and REF is a reference value.
II) establishing an input parameter matrix, namely:
in the formula, XnAt a time tnThe secondary side voltage of (c). Y isnAt a time tnThe primary side voltage of (1). And n is a time step.
The LSTM electromagnetic voltage transformer voltage back calculation model is verified by a voltage transformer port voltage verification data set. The voltage transformer port voltage verification data set comprises voltage transformer port voltage data after intervention processing, and the data of the voltage transformer port voltage verification data set is different from the data of the voltage transformer port voltage training data set.
3) And inputting the secondary side voltage data of the voltage transformer to be tested into an LSTM electromagnetic voltage transformer voltage back calculation model, and calculating to obtain the primary side voltage value of the voltage transformer to be tested.
Example 2:
a voltage back calculation method of an electromagnetic voltage transformer based on deep learning comprises the following steps:
1) collecting secondary side voltage data of a voltage transformer to be tested;
2) establishing an LSTM electromagnetic voltage transformer voltage back calculation model;
3) and inputting the secondary side voltage data of the voltage transformer to be tested into an LSTM electromagnetic voltage transformer voltage back calculation model, and calculating to obtain the primary side voltage value of the voltage transformer to be tested.
Example 3:
the electromagnetic voltage transformer voltage back calculation method based on deep learning mainly comprises the following steps of embodiment 2, wherein an LSTM electromagnetic voltage transformer voltage back calculation model comprises an input layer, a hidden layer and an output layer; the activation functions of the input gate, the output gate and the forgetting gate are sigmoid functions; each gate has a sigmoid function and a bitwise multiplication operation.
The input layer is used for receiving secondary side voltage data of the voltage transformer to be tested;
the hidden layer comprises a plurality of memory modules, and each memory module comprises an input unit, an output unit, an input gate, an output gate, a forgetting gate and a Ceil; the activation functions of the input unit and the output unit are tanh functions;
the activation function of the output layer is a linear function, and the output layer is used for outputting the primary side voltage of the voltage transformer to be tested.
Example 4:
the method for reversely calculating the voltage of the electromagnetic voltage transformer based on deep learning mainly comprises the steps of embodiment 2, wherein an input door i of a voltage reverse calculation model of the LSTM electromagnetic voltage transformertForgetting door ftAnd an output gate otUnit activation vector ctA hidden layer unit htRespectively as follows:
it＝σ(Wxixt+Whiht-1+Wcict-1+bi)
ft＝σ(Wxfxt+Whfht-1+Wcfct-1+bf)
ct＝ftct-1+it tanh(WxcXt+Whcht-1+bc)
ot＝σ(Wxoxt+Whoht-1+Wcoct-1+bo) (1)
ht＝ot tanh(ct)
in the formula, sigma represents a logic sigmoid function; wxi、Whi、WciRespectively representing weight matrixes between the input gate and the input characteristic vector, between the input gate and the hidden layer unit, and between the input gate and the unit activation vector; wxf、Whf、WcfRespectively representing forgetting gate and input feature vector, forgetting gate and hidden layer unit, forgetting gateAnd a weight matrix between the cell activation vectors, Wxo、Who、WcoRespectively representing weight matrixes between the output gate and the input feature vector, between the output gate and the hidden layer unit, and between the output gate and the unit activation vector; wxc、WhcRespectively representing weight matrixes between the unit activation vector and the input characteristic vector and between the unit activation vector and the hidden layer unit; t represents a sampling instant; tan h is an activation function; bi、bf、bc、boAnd respectively representing the deviation values of the input gate, the forgetting gate, the unit activation vector and the output gate.
Example 5:
the electromagnetic voltage transformer voltage back calculation method based on deep learning mainly comprises the following steps of embodiment 2, wherein an LSTM electromagnetic voltage transformer voltage back calculation model is obtained by training a voltage transformer port voltage training data set; and the voltage transformer port voltage training data set comprises voltage transformer port voltage data after intervention processing.
Example 6:
the electromagnetic voltage transformer voltage back calculation method based on deep learning mainly comprises the following steps of embodiment 5, wherein the step of collecting voltage data of a voltage transformer port comprises the following steps:
1) connecting a resistance-capacitance voltage divider in parallel at the primary side of the voltage transformer, and performing high-frequency acquisition on a secondary side signal of the resistance-capacitance voltage divider to obtain a secondary side signal of the resistance-capacitance voltage divider;
multiplying the secondary side signal of the resistance-capacitance voltage divider by the transformation ratio of the resistance-capacitance voltage divider to obtain a primary side voltage signal of the voltage transformer;
2) performing high-frequency acquisition on a secondary side signal of the voltage transformer to obtain secondary side voltage data of the voltage transformer;
3) applying typical overvoltage on a primary side of the transformer, and measuring primary side voltage and secondary side voltage of the voltage transformer;
4) changing waveform parameters and amplitude of the overvoltage, and returning to the step 3) to obtain a plurality of groups of voltage transformer port voltage data.
Example 7:
the voltage back calculation method of the electromagnetic voltage transformer based on deep learning mainly comprises the following steps of embodiment 6, wherein typical overvoltage comprises standard lightning overvoltage, non-standard lightning overvoltage, power frequency overvoltage, low frequency overvoltage, frequency division ferromagnetic resonance overvoltage, high frequency ferromagnetic resonance overvoltage, fundamental frequency ferromagnetic resonance overvoltage, arc grounding overvoltage and standard operation overvoltage.
Example 8:
the electromagnetic voltage transformer voltage back calculation method based on deep learning mainly comprises the following steps of embodiment 5, wherein the step of acquiring voltage data of a voltage transformer port after preprocessing comprises the following steps:
1) carrying out abnormal data detection on voltage data of a voltage transformer port by using an isolated forest algorithm, and removing abnormal data according to a detection result;
2) carrying out waveform normalization processing on the data, and changing the named value into a per unit value; the normalized calculation formula is:
in the formula, m' is a per unit value, m is a named value, and REF is a reference value.
3) Establishing an input parameter matrix, namely:
in the formula, XnAt a time tnThe secondary side voltage of (d); y isnAt a time tnA primary side voltage of (d); and n is a time step.
Example 9:
the electromagnetic voltage transformer voltage back calculation method based on deep learning mainly comprises the following steps of embodiment 2, wherein an LSTM electromagnetic voltage transformer voltage back calculation model is verified by a voltage transformer port voltage verification data set; the voltage transformer port voltage verification data set comprises voltage transformer port voltage data after intervention processing, and the numerical value of the voltage transformer port voltage verification data set is different from the numerical value of the voltage transformer port voltage training data set.
Example 10:
a voltage back calculation method of an electromagnetic voltage transformer based on deep learning comprises the following steps:
1) voltage data of a voltage transformer port are collected; connecting a resistance-capacitance voltage divider in parallel at the primary side of the transformer to be tested, carrying out high-frequency acquisition on signals at the secondary side of the resistance-capacitance voltage divider, directly multiplying the obtained data by the transformation ratio of the resistance-capacitance voltage divider to obtain voltage signals at the primary side of the voltage transformer, and storing the acquired data at the primary side in an analysis instrument;
performing high-frequency acquisition on a secondary side signal of the voltage transformer to obtain secondary side voltage data of the voltage transformer;
applying a typical overvoltage of different magnitudes on the primary side of the transformer includes: the primary side voltage and the secondary side voltage of the mutual inductor are obtained through measurement, wherein the primary side voltage and the secondary side voltage of the mutual inductor are obtained through measurement; and forming a primary voltage data set and a secondary voltage data set which are sufficient for a deep learning algorithm by changing the waveform parameters and the amplitude of the overvoltage.
2) Preprocessing voltage data of a voltage transformer port; randomly disordering and segmenting the obtained primary and secondary voltage data sets of the mutual inductor, dividing the data sets into a 75% training set and a 25% testing set, wherein the training set is used for deep learning training of the LSTM model, and the testing set is used for testing the effect of the trained model.
Detecting abnormal voltage data of the detected port based on an isolated forest algorithm, and removing the abnormal data according to a detection result;
the data is subjected to waveform normalization processing, the named value is changed into a per unit value, and the normalization calculation formula is as follows:
in the formula, m' is a per unit value, m is a named value, and REF is a reference value.
Recording the primary side voltage of the voltage transformer as Y and the secondary side voltage of the voltage transformer as X;
the normalized data set is further collated to form an input parameter matrix as shown below
The first row in the input parameter matrix represents X, Y state values for the quantities respectively;
[X0,Y0]to [ X ]n,Yn]Respectively at time t0To time tnTime series based variable states of (1); and n is the time step, and is 1000.
3) Designing an LSTM back calculation model; the LSTM model comprises an input layer, a hidden layer and an output layer, and characteristic rules of sequence data are extracted by forgetting and memorizing historical information controlled by an input gate, a forgetting gate and an output gate, and a prediction result is finally output.
An input layer: receiving the input of the secondary side voltage data of the processed voltage transformer;
hiding the layer: the system comprises 8 memory modules, wherein each memory module comprises an input unit, an output unit, an input gate, an output gate, a forgetting gate and a Ceil; the dimensions of the input unit, the output unit, the input gate, the output gate, the forgetting gate and the Ceil are all 16; the length of the sequence is 8, 8 memory modules are arranged after the sequence is expanded, and the output of the module at the previous moment is fed back to the module at the next moment; the activation functions of the input gate, the output gate and the forgetting gate are sigmoid functions, and the activation functions of the input unit and the output unit are tanh functions; each gate has a sigmoid function and bitwise multiplication operation, so that the hidden unit only memorizes useful information as far as possible and discards useless information, thereby solving the problem of long-term dependence;
an output layer: the activation function is a linear function and outputs the primary side voltage of the voltage transformer;
the calculation process at the moment t is as follows:
it＝σ(Wxixt+Whiht-1+Wcict-1+bi)
ft＝σ(Wxfxt+Whfht-1+Wcfct-1+bf)
ct＝ftct-1+it tanh(WxcXt+Whcht-1+bc)
ot＝σ(Wxoxt+Whoht-1+Wcoct-1+bo) (3)
ht＝ot tanh(ct)
where σ denotes a logical sigmoid function, itDenotes an input gate, ftIndicating forgetting to leave door otRepresents an output gate, ctRepresents the unit activation vector, htFor hiding layer units, Wxi、Whi、WciRespectively representing the weight matrix between the input gate and the input eigenvector, hidden layer element, and element activation vector, Wxf、Whf、WcfRespectively representing weight matrixes among the forgetting gate, the input characteristic vector, the hidden layer unit and the unit activation vector, Wxo、Who、WcoRespectively representing the weight matrix between the output gate and the input eigenvector, hidden layer element, and element activation vector, Wxc、WhcRespectively representing weight matrixes among the unit activation vector, the input feature vector and the hidden layer unit; t denotes the sampling instant, tanh is the activation function, bi、bf、bc、boAnd respectively representing the deviation values of the input gate, the forgetting gate, the unit activation vector and the output gate.
4) Training an LSTM back calculation model by using voltage transformer port voltage training set data; and feeding the input data [ X, Y ] in the training set into the network, obtaining the error of the corresponding output current primary side voltage calculation and the standard value through the forward propagation process of the LSTM model, and repeatedly training to improve the accuracy of the network model to obtain the LSTM model.
5) Inputting the secondary side voltage of the voltage transformer obtained in the test set into a voltage transformer port voltage back calculation method of the LSTM model to obtain the corresponding primary side voltage; this is compared with the primary side real voltage.
The result chart of the comparison experiment performed by using the voltage transformer inverse calculation method based on the LSTM model is shown in FIG. 3, and the method can accurately restore the primary side voltage.
6) And acquiring secondary side voltage data of the voltage transformer to be tested, inputting the secondary side voltage data into an LSTM electromagnetic voltage transformer voltage back calculation model, and calculating to obtain a primary side voltage value of the voltage transformer to be tested.
Claims (8)
1. A voltage back calculation method of an electromagnetic voltage transformer based on deep learning is characterized by comprising the following steps:
1) and collecting secondary side voltage data of the voltage transformer to be measured.
2) Establishing a voltage back calculation model of the LSTM electromagnetic voltage transformer;
3) and inputting the secondary side voltage data of the voltage transformer to be tested into an LSTM electromagnetic voltage transformer voltage back calculation model, and calculating to obtain the primary side voltage value of the voltage transformer to be tested.
2. The LSTM model-based voltage transformer port voltage back-calculation method of claim 1, wherein the LSTM electromagnetic voltage transformer voltage back-calculation model comprises an input layer, a hidden layer and an output layer; the activation functions of the input gate, the output gate and the forgetting gate are sigmoid functions; each gate has a sigmoid function and a bitwise multiplication operation;
the input layer is used for receiving secondary side voltage data of the voltage transformer to be tested;
the hidden layer comprises a plurality of memory modules, and each memory module comprises an input unit, an output unit, an input gate, an output gate, a forgetting gate and a Ceil; the activation functions of the input unit and the output unit are tanh functions;
the activation function of the output layer is a linear function, and the output layer is used for outputting the primary side voltage of the voltage transformer to be tested.
3. The LSTM model-based voltage transformer port voltage back-calculation method of claim 1, wherein the LSTM electromagnetic voltage transformer voltage back-calculation model input gate itForgetting door ftAnd an output gate otUnit activation vector ctA hidden layer unit htRespectively as follows:
in the formula, sigma represents a logic sigmoid function; wxi、Whi、WciRespectively representing weight matrixes between the input gate and the input characteristic vector, between the input gate and the hidden layer unit, and between the input gate and the unit activation vector; wxf、Whf、WcfRespectively representing weight matrixes between the forgetting gate and the input characteristic vector, between the forgetting gate and the hidden layer unit, and between the forgetting gate and the unit activation vector, Wxo、Who、WcoRespectively representing weight matrixes between the output gate and the input feature vector, between the output gate and the hidden layer unit, and between the output gate and the unit activation vector; wxc、WhcRespectively representing weight matrixes between the unit activation vector and the input characteristic vector and between the unit activation vector and the hidden layer unit; t represents a sampling instant; tan h is an activation function; bi、bf、bc、boRespectively representing the deviation values of an input gate, a forgetting gate, a unit activation vector and an output gate; xtRepresenting the input.
4. The LSTM model-based voltage transformer port voltage back-calculation method of claim 1, wherein the LSTM electromagnetic voltage transformer voltage back-calculation model is trained from a voltage transformer port voltage training dataset; and the voltage transformer port voltage training data set comprises voltage transformer port voltage data after intervention processing.
5. The LSTM model-based voltage transformer port voltage back-calculation method of claim 4, wherein the step of collecting voltage transformer port voltage data comprises:
1) connecting a resistance-capacitance voltage divider in parallel at the primary side of the voltage transformer, and performing high-frequency acquisition on a secondary side signal of the resistance-capacitance voltage divider to obtain a secondary side signal of the resistance-capacitance voltage divider;
multiplying the secondary side signal of the resistance-capacitance voltage divider by the transformation ratio of the resistance-capacitance voltage divider to obtain a primary side voltage signal of the voltage transformer;
2) performing high-frequency acquisition on a secondary side signal of the voltage transformer to obtain secondary side voltage data of the voltage transformer;
3) applying typical overvoltage on a primary side of the transformer, and measuring primary side voltage and secondary side voltage of the voltage transformer;
4) changing waveform parameters and amplitude of the overvoltage, and returning to the step 3) to obtain a plurality of groups of voltage transformer port voltage data.
6. The LSTM model-based voltage transformer port voltage back-calculation method of claim 5, wherein the typical overvoltages include standard lightning overvoltages, non-standard lightning overvoltages, power frequency overvoltages, low frequency overvoltages, frequency division ferromagnetic resonance overvoltages, high frequency ferromagnetic resonance overvoltages, fundamental frequency ferromagnetic resonance overvoltages, arc grounding overvoltages, standard operation overvoltages.
7. The LSTM model-based voltage transformer port voltage back-calculation method of claim 5, wherein the step of obtaining the preprocessed voltage transformer port voltage data comprises:
1) carrying out abnormal data detection on voltage data of a voltage transformer port by using an isolated forest algorithm, and removing abnormal data according to a detection result;
2) carrying out waveform normalization processing on the data, and changing the named value into a per unit value; the normalized calculation formula is:
wherein m' is a per unit value, m is a named value, and REF is a reference value;
3) establishing an input parameter matrix, namely:
in the formula, XnAt a time tnThe secondary side voltage of (d); y isnAt a time tnA primary side voltage of (d); and n is a time step.
8. The LSTM model-based voltage transformer port voltage back-calculation method of claim 1, wherein the LSTM electromagnetic voltage transformer voltage back-calculation model is further validated by a voltage transformer port voltage validation dataset; the voltage transformer port voltage verification data set comprises voltage transformer port voltage data after intervention processing, and the numerical value of the voltage transformer port voltage verification data set is different from the numerical value of the voltage transformer port voltage training data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110903339.2A CN113987906A (en) | 2021-08-06 | 2021-08-06 | Electromagnetic voltage transformer voltage back calculation method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110903339.2A CN113987906A (en) | 2021-08-06 | 2021-08-06 | Electromagnetic voltage transformer voltage back calculation method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113987906A true CN113987906A (en) | 2022-01-28 |
Family
ID=79735106
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110903339.2A Pending CN113987906A (en) | 2021-08-06 | 2021-08-06 | Electromagnetic voltage transformer voltage back calculation method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113987906A (en) |
-
2021
- 2021-08-06 CN CN202110903339.2A patent/CN113987906A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109612708B (en) | Power transformer on-line detection system and method based on improved convolutional neural network | |
| CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
| CN110705456A (en) | Micro motor abnormity detection method based on transfer learning | |
| GB2431726A (en) | Identification of partial discharge using a neural network | |
| CN110672905A (en) | CNN-based self-supervision voltage sag source identification method | |
| CN113695713A (en) | Online monitoring method and device for welding quality of inner container of water heater | |
| CN111239549A (en) | Power distribution fault rapid positioning method based on discrete wavelet transform | |
| CN115580445B (en) | Unknown attack intrusion detection method, unknown attack intrusion detection device and computer readable storage medium | |
| 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 | |
| CN116610998A (en) | Switch cabinet fault diagnosis method and system based on multi-mode data fusion | |
| CN112596016A (en) | Transformer fault diagnosis method based on integration of multiple one-dimensional convolutional neural networks | |
| CN110020637A (en) | A kind of analog circuit intermittent fault diagnostic method based on more granularities cascade forest | |
| CN116432112A (en) | Arc fault detection method based on wavelet packet transformation and residual convolution neural network | |
| CN112486137A (en) | Method and system for constructing fault feature library of active power distribution network and fault diagnosis method | |
| CN114386452B (en) | Nuclear power circulating water pump sun gear fault detection method | |
| CN111291481A (en) | Bayesian model-based structure early warning analysis method | |
| CN116400244B (en) | Abnormality detection method and device for energy storage battery | |
| CN113987906A (en) | Electromagnetic voltage transformer voltage back calculation method based on deep learning | |
| CN116952586A (en) | Axle fatigue crack identification method based on 1D-DCNN | |
| CN116628620A (en) | Non-invasive load identification calculation method | |
| CN116758922A (en) | Voiceprint monitoring and diagnosing method for transformer | |
| Zhu et al. | Aiming to Complex Power Quality Disturbances: A Novel Decomposition and Detection Framework | |
| Mei et al. | Classification of the type of harmonic source based on image-matrix transformation and deep convolutional neural network | |
| CN114492543A (en) | Partial discharge classification and identification method based on time-frequency domain characteristics | |
| CN114689991A (en) | Method and system for determining fault time of high-voltage cable and related components |
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 |