CN110570114A - Power quality disturbance identification method, system and medium based on variable cycle neural network - Google Patents
Power quality disturbance identification method, system and medium based on variable cycle neural network Download PDFInfo
- Publication number
- CN110570114A CN110570114A CN201910827961.2A CN201910827961A CN110570114A CN 110570114 A CN110570114 A CN 110570114A CN 201910827961 A CN201910827961 A CN 201910827961A CN 110570114 A CN110570114 A CN 110570114A
- Authority
- CN
- China
- Prior art keywords
- neural network
- disturbance
- variable
- power quality
- quality disturbance
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims description 38
- 230000004913 activation Effects 0.000 claims description 25
- 230000000306 recurrent effect Effects 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 20
- 239000013598 vector Substances 0.000 claims description 19
- 230000014509 gene expression Effects 0.000 claims description 14
- 210000002569 neuron Anatomy 0.000 claims description 14
- 125000004122 cyclic group Chemical group 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 7
- 239000013589 supplement Substances 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 230000003213 activating effect Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 3
- 238000004880 explosion Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- G—PHYSICS
- 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
- G06N3/045—Combinations of 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Educational Administration (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a method, a system and a medium for recognizing power quality disturbance based on a variable cycle neural network. The electric energy quality disturbance identification method based on the variable cycle neural network can realize the identification and classification of the electric energy quality disturbance types at any time when the electric energy quality is monitored in a power grid, can accurately identify the start-stop time of the disturbance, and can realize multi-node, quick and accurate electric energy quality disturbance identification after a common household electric meter introduces the algorithm.
Description
Technical Field
The invention relates to the technical field of power quality detection, in particular to a power quality disturbance identification method, a system and a medium based on a variable cycle neural network.
Background
The existing power quality monitoring method generally utilizes signal processing (power quality disturbance detection algorithms such as short-time windowed Fourier, wavelet and variational modal decomposition) to perform disturbance detection and extraction features, and then utilizes a series of algorithms such as Bayes criterion, k nearest neighbor, fisher criterion and artificial neural network to perform recognition. However, these methods can only perform disturbance recognition by time period since there is no pertinence to the processing of the sequence data. The neural network structure can make full use of a large amount of existing power quality data, after new data is added in model training, the identification accuracy rate can reach a new peak value after continuous iteration, and the neural network algorithm cannot iterate. The whole network structure of the common recurrent neural network is completely determined, the front and back logics are related, the probabilities are mutually influenced, the randomness is lost, and the problems of gradient explosion and the like are easily generated.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: aiming at the problems in the prior art, the invention provides a method, a system and a medium for identifying the power quality disturbance based on the variable circulation neural network.
In order to solve the technical problems, the invention adopts the technical scheme that:
a power quality disturbance identification method based on a variable cycle neural network comprises the following implementation steps:
1) Inputting power quality disturbance waveform data;
2) Performing feature extraction on the power quality disturbance waveform data to obtain a multi-dimensional feature vector;
3) Inputting a multi-dimensional feature vector into a pre-trained variable cycle neural network, wherein the variable cycle neural network comprises a random layer used for introducing potential random variables into an output layer, and the pre-trained variable cycle neural network establishes a mapping relation between the multi-dimensional feature vector and output probabilities of N disturbances corresponding to the multi-dimensional feature vector;
4) Generating an output probability curve corresponding to N disturbances by the output of the variable cyclic neural network through an activation function;
5) And determining the disturbance type at a certain moment and/or the starting and stopping moments of a certain disturbance according to the output probability curve.
Optionally, the power quality disturbance waveform data in step 1) is voltage or current waveform data.
Optionally, the feature extraction in step 2) specifically refers to using one of fourier transform, short-time windowed fourier transform, wavelet transform, variational modal decomposition, hilbert-yellow transform, and S transform.
optionally, the variable recurrent neural network is a four-layer structure formed by adding random layers to three-layer structures of an input layer, a hidden layer and an output layer of a common recurrent neural network.
Optionally, the functional expression of the random layer is as follows:
(μt,σt)＝g(ht-1)
yt＝f2(Wozzt+mt)
In the above formula, μ (t) and σ (t) are the mean and standard deviation of the normal distribution, ht-1T-1 neuron of hidden layer, g is activation function, g (h)t-1) T-1 st neuron h representing hidden layert-1upon reaching the activation condition, the random variable z is activatedt，Representing a random variable ztSatisfy normal distribution, ytIs the output of the t-th neuron, f2For activating functions, mtBeing intermediate variables of the network, WozIs a random variable ztThe weight coefficient of (2).
Optionally, when the output of the variable recurrent neural network is used to generate an output probability curve corresponding to the N kinds of disturbances through the activation function in step 4), the activation function is selected to be the softmax activation function.
Optionally, step 3) is preceded by a step of training the variable-cycle neural network, and the detailed steps include:
S1) establishing a disturbance function aiming at various power quality disturbances, generating power quality disturbance sample library data by adopting a Monte Carlo algorithm based on the disturbance function, and randomly dividing according to a specified proportion to obtain a training set, a testing set and a verifying set;
S2) establishing a variable cyclic neural network and initializing;
S3) training the variable-cycle neural network based on the training set, the testing set and the verifying set until a preset end condition is met, ending the training and exiting.
Alternatively, the functional expression of the disturbance function established in step S1) is as follows:
In the above formula, v (t) is a disturbance function, α is a voltage amplitude range coefficient, Δ ffIs the amount of frequency deviation, t is time, betaiIs the coefficient of the i-th disturbance, c is the harmonic amplitude coefficient, tstart,iRepresenting the start time, t, of the i-th disturbance on the disturbance time axisend,iRepresenting the end time of the i-th disturbance on the disturbance time axis, fhIs the harmonic voltage frequency, tstartFor perturbation start time, μ (t) satisfies the normalThe distributed voltage amplitude supplement amount, gamma (t), is the direct current voltage amplitude supplement amount.
in addition, the invention also provides a power quality disturbance identification system based on the variable cycle neural network, which comprises a computer device, wherein the computer device is programmed or configured to execute the steps of the power quality disturbance identification method based on the variable cycle neural network, or a storage medium of the computer device is stored with a computer program which is programmed or configured to execute the power quality disturbance identification method based on the variable cycle neural network.
Furthermore, the present invention also provides a computer-readable storage medium having stored thereon a computer program programmed or configured to execute the variable-cycle neural network-based power quality disturbance identification method.
compared with the prior art, the invention has the following advantages:
1. the invention introduces the variable circulation neural network structure into the traditional electric energy quality disturbance identification, can accurately identify the start and stop and the duration of the disturbance, can judge the disturbance type at any time, and greatly improves the disturbance identification speed and the disturbance accuracy.
2. the method can be carried on the intelligent ammeter, and then the application and popularization of the intelligent ammeter are combined, so that the cost can be reduced, the application scenes of the electric energy quality disturbance identification algorithm can be enriched, the electric energy quality disturbance identification can be realized rapidly and accurately at multiple nodes at any time, and the method is very helpful for monitoring the electric energy quality of the whole power grid.
drawings
FIG. 1 is a schematic diagram of the basic principle of the method according to the embodiment of the present invention.
Fig. 2 is a diagram of a structure of a general recurrent neural network for comparison.
Fig. 3 is a structural diagram of a variable recurrent neural network according to an embodiment of the present invention.
FIG. 4 is a graph of the output probability of a single perturbation obtained according to an embodiment of the present invention.
Fig. 5 shows 10s power quality disturbance data in a sample database according to an embodiment of the present invention.
Fig. 6 shows 0.3s power quality disturbance data in a sample database according to an embodiment of the present invention.
Fig. 7 is a mathematical model diagram of power quality disturbance data according to an embodiment of the invention.
Fig. 8 is a table of disturbance parameters corresponding to a power quality disturbance function in an embodiment of the present invention.
FIG. 9 is a table of disturbance parameters in an embodiment of the present invention.
Detailed Description
As shown in fig. 1, the implementation steps of the power quality disturbance identification method based on the variable cycle neural network of the embodiment include:
1) Inputting power quality disturbance waveform data;
2) Performing feature extraction on the power quality disturbance waveform data to obtain a multi-dimensional feature vector;
3) inputting a multi-dimensional feature vector into a pre-trained variable cycle neural network, wherein the variable cycle neural network comprises a random layer used for introducing potential random variables into an output layer, and the pre-trained variable cycle neural network establishes a mapping relation between the multi-dimensional feature vector and output probabilities of N disturbances corresponding to the multi-dimensional feature vector;
4) generating an output probability curve corresponding to N disturbances by the output of the variable cyclic neural network through an activation function;
5) And determining the disturbance type at a certain moment and/or the starting and stopping moments of a certain disturbance according to the output probability curve.
the power quality disturbance is referred to as disturbance for short, and mainly refers to a series of waveforms of the abnormal voltage quality state of the power grid, and in this embodiment, the N disturbances are specifically 14 disturbances related to the power quality disturbance specified by IEEE, as shown in table 1.
Table 1: IEEE power quality disturbance definition.
The number and the type of the disturbances can be adjusted according to needs, and the scheme of the embodiment can also realize the power quality disturbance identification by correspondingly adjusting in the variable cycle neural network training stage.
In this embodiment, the power quality disturbance waveform data in step 1) is voltage, and current waveform data may also be adopted as needed.
In this embodiment, the performing of feature extraction in step 2) specifically refers to performing fourier transform, performing feature extraction to obtain a multi-dimensional feature vector, where the multi-dimensional feature vector is a vector formed by values of fundamental waves and harmonic voltages at this time, and then screening all the obtained multi-dimensional feature vectors as needed, and removing part of feature vectors with poor correlation through screening to improve the accuracy of identification. In addition, short-time windowed fourier transform, wavelet transform, variational modal decomposition, hilbert-yellow transform, S transform, and the like may be used as necessary.
The general recurrent neural network is a neural network structure dedicated to sequence data, as shown in fig. 3, the basic network structure of the general recurrent neural network is a three-layer structure of an input layer, a hidden layer, and an output layer, and functional expressions are shown in the following formulas (1) to (4):
ht-1＝f1(Whxxt-1+bt-1) (1)
ht＝f1(Whhht-1+Whxxt+bt) (2)
yt-1＝f2(Wohht-1+ct-1)＝f(mt-1) (3)
in formulae (1) to (4), htthe t-th neuron of the hidden layer, ht-1T-1 neuron as a hidden layer, f1 f2Are all activation functions, f1Using Relu as the activation function, f2adopting softmax as an activation function; w in all expressionsoh,Woz,Whh,WhxAre all weight coefficients, have no unitThe back propagation automatically updates the weights. x is the number oft,ytInput and output of the t-th neuron, respectively, bt-1B, m, analogous to y ═ ax + b, for a constant offset with inputtFor intermediate variables of the network, setting the expression (4) as an expression of a softmax activation function and g as a sigmoid activation function for simplifying the expression, and activating a random variable z when the activation condition is mett。
in this embodiment, the variable recurrent neural network is a four-layer structure formed by adding random layers to three-layer structures of an input layer, a hidden layer, and an output layer of a common recurrent neural network. The improvement of the variable recurrent neural network on the common recurrent neural network introduces a random layer to increase the uncertainty of the network, and is more helpful to identify the actual random sequence.
In this embodiment, the functional expression of the random layer is shown as follows:
(μt,σt)＝g(ht-1) (5)
yt＝f2(Wozzt+mt) (7)
In the formulae (5) to (7), μ (t) and σ (t) are the mean and standard deviation of the normal distribution, respectively, and ht-1T-1 neuron of hidden layer, g is activation function, g (h)t-1) T-1 st neuron h representing hidden layert-1upon reaching the activation condition, the random variable z is activatedt，representing a random variable ztSatisfy normal distribution, ytIs the output of the t-th neuron, f2For activating functions, mtBeing intermediate variables of the network, WozIs a random variable ztThe weight coefficient of (2). In this example introducemesh of normal distributionThat is, z is obtained to increase the randomness of the networktIs a random variable, the output of each neuron is affected by the random variable, and the output ytis a 14-dimensional perturbation probability vector.
Referring to fig. 3, compared with the common recurrent neural network, the variable recurrent neural network has a random layer z in addition to the hidden layer h, the value of the random layer z satisfies the normal distribution, and the basic parameters are determined by the equations (5) to (7). All the weights in fig. 3 are automatically updated during the backward propagation, and are not described in detail herein. The last layer of fig. 3 is the network output, the output y after the softmax activation function corresponds to a 14-dimensional probability vector, and the probability of each disturbance at each moment must have a maximum value, which is the most probable disturbance. The variable recurrent neural network is still a non-linear mapping in nature, and the new recurrent neural network is four layers, input layer xtH of the hidden layertZ of random layertAnd y of the output layertCompared with the common cyclic neural network, the method increases the uncertainty of the network and can better identify the transient power quality disturbance. Fig. 4 shows one of the 14 outputs of the variable-cycle neural network in this embodiment, the outputs of the variable-cycle neural network have 14 different probability curves, one of the probability curves is selected as an exhibition, and it is desirable to know the disturbance type at a certain time, and the maximum value that is parallel to the probability axis and intersects with the probability curve at the required time is the disturbance type with the maximum probability, and it is desirable to know the start-stop time of a certain disturbance, and it is only necessary to require that the disturbance probability curve maintain the abscissa start-stop point corresponding to the maximum value.
In this embodiment, when the output of the variable recurrent neural network is used to generate an output probability curve corresponding to N types of disturbances through an activation function in step 4), the activation function is selected as the softmax activation function.
in this embodiment, step 3) is preceded by a step of training a variable-cycle neural network, and the detailed steps include:
S1) establishing a disturbance function aiming at various power quality disturbances, generating power quality disturbance sample library data by adopting a Monte Carlo algorithm based on the disturbance function, and randomly dividing according to a specified proportion to obtain a training set, a testing set and a verifying set;
S2) establishing a variable cyclic neural network and initializing;
S3) training the variable-cycle neural network based on the training set, the testing set and the verifying set until a preset end condition is met, ending the training and exiting.
When the sample library data of the power quality disturbance is generated in step S1), based on the special condition of the power quality disturbance, it is difficult to actually obtain the sample library data of the power quality disturbance meeting the requirement, so in this embodiment, a monte carlo algorithm is used to generate data to replace actual data as truly as possible, and the trained model can still continue to adjust parameters to improve the accuracy of recognition after receiving the recognition of the actual data. Fig. 5 and 6 are both from the sample library data, fig. 5 is the intercepted power quality disturbance data for 10 seconds, fig. 6 is the waveform from 9.7 seconds to 10 seconds, and four disturbance types of voltage drop, voltage rise, noise and normal voltage can be seen.
In the present embodiment, the mathematical model is established according to the IEEE Power quality disturbance definition shown in Table 1, as shown in FIG. 7, wherein v isnormal(t) is a normal voltage type, v (t) refers to a voltage waveform, alpha is a voltage amplitude range coefficient, and a range is selected according to a disturbance type. f is the voltage frequency, 50Hz at home and 60Hz generally at abroad. Δ ffIs the frequency deviation, t is time, and in the actual calculation is the abscissa of the discrete sampling point, tstart,irepresenting the ith point, beta, on the disturbance time axisiAlso coefficients, the index i is the number of points, an integer ranging from 1 to 14 (14 perturbations), c is the harmonic amplitude coefficient, given value. t is tstartAnd tendFor disturbance start and end times, fhFor harmonic voltage frequency, mu (t) and gamma (t) are voltage amplitude supplementary quantities, and mu (t) meets normal distribution, represents small random fluctuation which cannot be avoided by voltage per se under actual working conditions, can enable disturbance waveforms to be closer to actual conditions, and all disturbance types are available. Gamma (t) is a direct current component, and only a direct current bias is provided, so that a voltage amplitude is generatedThe values are shifted by an amount overall. On the basis, in the present embodiment, the functional expression of the perturbation function established in step S1) is as follows:
In the formula (8), v (t) is a disturbance function, α is a voltage amplitude range coefficient, and Δ ffIs the amount of frequency deviation, t is time, betaiIs the coefficient of the i-th disturbance, c is the harmonic amplitude coefficient, tstart,iRepresenting the start time, t, of the i-th disturbance on the disturbance time axisend,iRepresenting the end time of the i-th disturbance on the disturbance time axis, fhIs the harmonic voltage frequency, tstartTo disturb the start time, μ (t) satisfies the voltage amplitude supplement amount of the normal distribution. And gamma (t) is the supplement quantity of the DC voltage amplitude. For different disturbances, the corresponding parameters of fig. 8 and 9 are substituted into the expression (8) in sequence to obtain the expression. Substituting the value of fig. 8 into expression (8) can obtain fig. 7, where the three parameters of fig. 9 correspond to three disturbances respectively, and the parameters of the remaining disturbances are all 0. The 14 disturbance probabilities are supplemented as shown in table 2:
Table 2: the disturbance probability supplements the table.
Different expression parameters represent different disturbance types, and different start times and end times determine different disturbance samples. The data of the sample library is further divided after being generated to start training, and the generated sample library is randomly divided into a training set, a testing set and a verification set according to the ratio of 6:2: 2.
In step S2), before inputting data into the neural network, initial parameters of the variable recurrent neural network are set, and these parameters include the number of hidden neurons, step size, learning rate, and so on. All these input samples then correspond to x in the expressiontAnd the first layer of figure 3.
The power quality disturbance identification method based on the variable cycle neural network solves the three defects of the existing power quality monitoring algorithm: firstly, the current algorithm utilizes signal processing (short-time windowed Fourier, wavelet, electric energy quality disturbance detection algorithm such as variational modal decomposition and the like) to carry out disturbance detection and extract characteristics, and then utilizes a series of algorithms such as Bayes criterion, k nearest neighbor, fisher criterion, artificial neural network and the like to carry out recognition, the algorithms have no pertinence to the processing of sequence data, disturbance recognition can be carried out only according to time periods, the input and output of the cyclic neural network can be high-dimensional sequences, and the electric energy quality disturbance type at any moment can be accurately recognized. Then, the neural network structure can fully utilize a large amount of existing power quality data, after new data is added in model training, the identification accuracy rate can reach a new peak value after continuous iteration, and the neural network algorithm cannot iterate. And finally, the whole network structure of the common recurrent neural network is completely determined, the front and the back are logically related, the probabilities are mutually influenced, the randomness is lost, and the problems of gradient explosion and the like are easily generated. And by using the characteristics that the output of potential random variables introduced by the Bayesian network has certain probability randomness besides the position certainty before and after the output is ensured, the actual situation of the power quality disturbance is relatively met, and the improved VRNN structure is more suitable for processing the power quality disturbance identification problem. In summary, when the power quality is monitored in the power grid by the power quality disturbance identification method based on the variable-cycle neural network, the identification and classification of the power quality disturbance types at any time can be realized, the start-stop time of the disturbance can be accurately identified, and after the algorithm is introduced into a common household electric meter, the multi-node, rapid and accurate power quality disturbance identification can be realized.
In addition, the invention also provides a power quality disturbance identification system based on the variable cycle neural network, which comprises a computer device, wherein the computer device is programmed or configured to execute the steps of the power quality disturbance identification method based on the variable cycle neural network, or a storage medium of the computer device is stored with a computer program which is programmed or configured to execute the power quality disturbance identification method based on the variable cycle neural network.
Furthermore, the present invention also provides a computer readable storage medium, which stores thereon a computer program programmed or configured to execute the variable cycle neural network-based power quality disturbance identification method of the present embodiment.
the above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may occur to those skilled in the art without departing from the principle of the invention, and are considered to be within the scope of the invention.
Claims (10)
1. a power quality disturbance identification method based on a variable cycle neural network is characterized by comprising the following implementation steps:
1) Inputting power quality disturbance waveform data;
2) Performing feature extraction on the power quality disturbance waveform data to obtain a multi-dimensional feature vector;
3) inputting a multi-dimensional feature vector into a pre-trained variable cycle neural network, wherein the variable cycle neural network comprises a random layer used for introducing potential random variables into an output layer, and the pre-trained variable cycle neural network establishes a mapping relation between the multi-dimensional feature vector and output probabilities of N disturbances corresponding to the multi-dimensional feature vector;
4) generating an output probability curve corresponding to N disturbances by the output of the variable cyclic neural network through an activation function;
5) And determining the disturbance type at a certain moment and/or the starting and stopping moments of a certain disturbance according to the output probability curve.
2. The method for identifying the power quality disturbance based on the variable cycle neural network as claimed in claim 1, wherein the power quality disturbance waveform data in the step 1) is voltage or current waveform data.
3. the method for identifying the power quality disturbance based on the variable cycle neural network as claimed in claim 1, wherein the step 2) of extracting the features specifically comprises one of Fourier transform, short-time windowed Fourier transform, wavelet transform, decomposition of variation modes, Hilbert-Huang transform and S transform.
4. the method for identifying the power quality disturbance based on the variable cyclic neural network as claimed in claim 1, wherein the variable cyclic neural network is a four-layer structure formed by adding random layers to a three-layer structure of an input layer, a hidden layer and an output layer of a common cyclic neural network.
5. The method for identifying the power quality disturbance based on the variable cycle neural network as claimed in claim 1, wherein the functional expression of the random layer is as follows:
(μt,σt)＝g(ht-1)
yt＝f2(Wozzt+mt)
In the above formula, μ (t) and σ (t) are the mean and standard deviation of the normal distribution, ht-1T-1 neuron of hidden layer, g is activation function, g (h)t-1) T-1 st neuron h representing hidden layert-1Upon reaching the activation condition, the random variable z is activatedt，Representing a random variable ztSatisfy normal distribution, ytIs the output of the t-th neuron, f2for activating functions, mtBeing intermediate variables of the network, WozIs a random variable ztThe weight coefficient of (2).
6. The method for identifying the power quality disturbance based on the variable recurrent neural network as claimed in claim 1, wherein when the output of the variable recurrent neural network is used to generate the output probability curve corresponding to the N disturbances through the activation function in step 4), the activation function is selected to be a softmax activation function.
7. the method for identifying the power quality disturbance based on the variable-cycle neural network as claimed in claim 1, wherein step 3) is preceded by a step of training the variable-cycle neural network, and the detailed steps comprise:
S1) establishing a disturbance function aiming at various power quality disturbances, generating power quality disturbance sample library data by adopting a Monte Carlo algorithm based on the disturbance function, and randomly dividing according to a specified proportion to obtain a training set, a testing set and a verifying set;
S2) establishing a variable cyclic neural network and initializing;
S3) training the variable-cycle neural network based on the training set, the testing set and the verifying set until a preset end condition is met, ending the training and exiting.
8. The method for identifying the power quality disturbance based on the variable-cycle neural network as claimed in claim 7, wherein the function expression of the disturbance function established in the step S1) is as follows:
In the above formula, v (t) is a disturbance function, α is a voltage amplitude range coefficient, Δ ffIs the amount of frequency deviation, t is time, betaiIs the coefficient of the i-th disturbance, c is the harmonic amplitude coefficient, tstart,iRepresenting the start time, t, of the i-th disturbance on the disturbance time axisend,iRepresenting the end time of the i-th disturbance on the disturbance time axis, fhIs the harmonic voltage frequency, tstartFor the disturbance start time, μ (t) satisfies the normally distributed voltage amplitude supplement amount, and γ (t) is the direct current voltage amplitude supplement amount.
9. A variable-cycle neural network-based power quality disturbance identification system, comprising a computer device, wherein the computer device is programmed or configured to execute the steps of the variable-cycle neural network-based power quality disturbance identification method according to any one of claims 1 to 8, or a storage medium of the computer device has stored thereon a computer program programmed or configured to execute the variable-cycle neural network-based power quality disturbance identification method according to any one of claims 1 to 8.
10. a computer-readable storage medium, wherein the computer-readable storage medium stores thereon a computer program programmed or configured to execute the method for identifying a disturbance in power quality based on a variable-cycle neural network according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910827961.2A CN110570114A (en) | 2019-09-03 | 2019-09-03 | Power quality disturbance identification method, system and medium based on variable cycle neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910827961.2A CN110570114A (en) | 2019-09-03 | 2019-09-03 | Power quality disturbance identification method, system and medium based on variable cycle neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110570114A true CN110570114A (en) | 2019-12-13 |
Family
ID=68777575
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910827961.2A Pending CN110570114A (en) | 2019-09-03 | 2019-09-03 | Power quality disturbance identification method, system and medium based on variable cycle neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110570114A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111325485A (en) * | 2020-03-22 | 2020-06-23 | 东北电力大学 | Light-weight gradient elevator power quality disturbance identification method considering internet-of-things bandwidth constraint |
| CN114487894A (en) * | 2021-12-24 | 2022-05-13 | 中铁二院工程集团有限责任公司 | System for carrying out real-time quality monitoring on vehicle-mounted power supply equipment |
| CN114660531A (en) * | 2022-05-24 | 2022-06-24 | 江西西平计量检测有限公司 | Detection method, system and device based on ammeter measurement error compensation |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101034389A (en) * | 2007-03-19 | 2007-09-12 | 江西省电力科学研究院 | Electrical energy power quality disturbance automatic identification method and system based on information fusion |
| CN108921285A (en) * | 2018-06-22 | 2018-11-30 | 西安理工大学 | Single-element classification method in sequence based on bidirectional valve controlled Recognition with Recurrent Neural Network |
| CN109766853A (en) * | 2019-01-16 | 2019-05-17 | 华北电力大学 | Voltage Sag Disturbance classification method based on LSTM |
-
2019
- 2019-09-03 CN CN201910827961.2A patent/CN110570114A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101034389A (en) * | 2007-03-19 | 2007-09-12 | 江西省电力科学研究院 | Electrical energy power quality disturbance automatic identification method and system based on information fusion |
| CN108921285A (en) * | 2018-06-22 | 2018-11-30 | 西安理工大学 | Single-element classification method in sequence based on bidirectional valve controlled Recognition with Recurrent Neural Network |
| CN109766853A (en) * | 2019-01-16 | 2019-05-17 | 华北电力大学 | Voltage Sag Disturbance classification method based on LSTM |
Non-Patent Citations (1)
| Title |
|---|
| JUNYOUNG CHUNG: "A Recurrent Latent Variable Model for Sequential Data", 《ARXIV:1506.02216V6 [CS.LG]》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111325485A (en) * | 2020-03-22 | 2020-06-23 | 东北电力大学 | Light-weight gradient elevator power quality disturbance identification method considering internet-of-things bandwidth constraint |
| CN114487894A (en) * | 2021-12-24 | 2022-05-13 | 中铁二院工程集团有限责任公司 | System for carrying out real-time quality monitoring on vehicle-mounted power supply equipment |
| CN114660531A (en) * | 2022-05-24 | 2022-06-24 | 江西西平计量检测有限公司 | Detection method, system and device based on ammeter measurement error compensation |
| CN114660531B (en) * | 2022-05-24 | 2022-08-26 | 江西西平计量检测有限公司 | Detection method, system and device based on ammeter measurement error compensation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110175386B (en) | Method for predicting temperature of electrical equipment of transformer substation | |
| CN110570114A (en) | Power quality disturbance identification method, system and medium based on variable cycle neural network | |
| Zheng et al. | State of health estimation for lithium battery random charging process based on CNN-GRU method | |
| CN112560079B (en) | Hidden false data injection attack method based on deep belief network and migration learning | |
| CN112200038B (en) | CNN-based quick identification method for oscillation type of power system | |
| CN110084148A (en) | A kind of Mechanical Failure of HV Circuit Breaker diagnostic method | |
| CN110543921A (en) | cable early fault identification method based on waveform learning | |
| CN110796120A (en) | Time domain feature-based circuit breaker mechanical fault XGboost diagnosis method | |
| CN114897144A (en) | Complex value time sequence signal prediction method based on complex value neural network | |
| Omar et al. | Fault classification on transmission line using LSTM network | |
| CN110610226A (en) | Generator fault prediction method and device | |
| CN113447759A (en) | Multi-classification RVM power grid fault discrimination method and system | |
| CN117421571A (en) | Topology real-time identification method and system based on power distribution network | |
| CN110929835B (en) | Novel silicon carbide-based aviation power converter fault diagnosis method and system | |
| CN112329370A (en) | SiC power tube online fault diagnosis method based on extreme learning machine | |
| CN116432006A (en) | Photovoltaic grid-connected inverter fault diagnosis method based on CEEMDAN-SE-IHHO-LSTM model | |
| Ma et al. | Long short-term memory autoencoder neural networks based dc pulsed load monitoring using short-time fourier transform feature extraction | |
| CN117669656A (en) | TCN-Semi PN-based direct-current micro-grid stability real-time monitoring method and device | |
| CN104834816A (en) | Short-term wind speed prediction method | |
| CN117726478A (en) | Intelligent decision-making method for dispatching of power system unit, terminal equipment and storage medium | |
| CN116643177A (en) | Online battery health degree prediction method, device, equipment and medium | |
| CN114565051B (en) | Method for testing product classification model based on influence degree of neurons | |
| CN116566061A (en) | Grid-connected inverter system stability on-line monitoring method and system | |
| CN113128130B (en) | Real-time monitoring method and device for judging stability of direct-current power distribution system | |
| CN112329535B (en) | CNN-based quick identification method for low-frequency oscillation modal characteristics of power system |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191213 |