CN113315100A - Micro-grid protection method and system based on convolutional neural network - Google Patents
Micro-grid protection method and system based on convolutional neural network Download PDFInfo
- Publication number
- CN113315100A CN113315100A CN202110556505.6A CN202110556505A CN113315100A CN 113315100 A CN113315100 A CN 113315100A CN 202110556505 A CN202110556505 A CN 202110556505A CN 113315100 A CN113315100 A CN 113315100A
- Authority
- CN
- China
- Prior art keywords
- signal
- current signal
- neural network
- convolutional neural
- current
- 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
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02H—EMERGENCY PROTECTIVE CIRCUIT ARRANGEMENTS
- H02H7/00—Emergency protective circuit arrangements specially adapted for specific types of electric machines or apparatus or for sectionalised protection of cable or line systems, and effecting automatic switching in the event of an undesired change from normal working conditions
- H02H7/26—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured
- H02H7/261—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations
- H02H7/262—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations involving transmissions of switching or blocking orders
-
- 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
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
-
- 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
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02H—EMERGENCY PROTECTIVE CIRCUIT ARRANGEMENTS
- H02H1/00—Details of emergency protective circuit arrangements
- H02H1/0007—Details of emergency protective circuit arrangements concerning the detecting means
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02H—EMERGENCY PROTECTIVE CIRCUIT ARRANGEMENTS
- H02H1/00—Details of emergency protective circuit arrangements
- H02H1/0061—Details of emergency protective circuit arrangements concerning transmission of signals
- H02H1/0084—Details of emergency protective circuit arrangements concerning transmission of signals by means of pilot wires or a telephone network; watching of these wires
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02H—EMERGENCY PROTECTIVE CIRCUIT ARRANGEMENTS
- H02H1/00—Details of emergency protective circuit arrangements
- H02H1/0092—Details of emergency protective circuit arrangements concerning the data processing means, e.g. expert systems, neural networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02H—EMERGENCY PROTECTIVE CIRCUIT ARRANGEMENTS
- H02H7/00—Emergency protective circuit arrangements specially adapted for specific types of electric machines or apparatus or for sectionalised protection of cable or line systems, and effecting automatic switching in the event of an undesired change from normal working conditions
- H02H7/26—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured
- H02H7/261—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations
- H02H7/263—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations involving transmissions of measured values
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Electromagnetism (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a microgrid protection method and system based on a convolutional neural network, and relates to the field of microgrids, wherein a three-phase current signal of a microgrid is obtained through current sampling; acquiring a negative sequence current signal and a zero sequence current signal under the symmetrical components of the three-phase current signals according to the three-phase current signals; acquiring a characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation, wherein the characteristic value set comprises an energy change value, a Shannon entropy and a standard difference value; and acquiring the fault position and the fault type of the microgrid by utilizing a convolutional neural network model according to the characteristic value set. According to the invention, the fault position and the fault type of the microgrid are intelligently judged through the convolutional neural network model, and the problem that the fault position and the fault type cannot be intelligently judged because the conventional microgrid protection scheme is based on a fixed structure or an operation mode is solved.
Description
Technical Field
The invention relates to the field of micro-grids, in particular to a micro-grid protection method and system based on a convolutional neural network.
Background
In recent years, with the increasing utilization rate of distributed power supplies, the micro-grid technology has been developed rapidly and becomes an important component of the power system. Renewable energy sources, combustion batteries, storage batteries, micro generators and the like directly supply power to loads through a micro grid, so that power transmission loss is reduced, energy utilization efficiency is improved, and reliability of power transmission is ensured. The micro-grid adopts a special communication, control and protection device, and when the grid generates voltage fluctuation, frequency deviation and other disturbances, grid-connected operation is switched to island operation to supply power for critical loads. However, different types of distributed power supplies, flexible modes of operation, and complex topologies present new challenges to the operation, control, and protection of the microgrid.
In particular, the access of the high-permeability distributed power supply makes the conventional relay protection device prone to malfunction or malfunction. The inertia of an inverter type distributed generation (IIDG) is lower than that of a synchronous generator. When a line fault cannot be quickly removed, the stability of a microgrid consisting of low-inertia power supplies is insufficient. The IIDG operates in different control modes depending on the mode of operation of the microgrid and the type of power source connected (intermittent or non-intermittent). Typically, IIDG has a maximum current of about twice the rated current and limited carrying capacity. Compared with a grid-connected mode, the fault allowable current value in an island operation mode is lower, and the traditional high fault current protection method is not applicable any more. Furthermore, power flow is dynamic and bi-directional, and the protection problem becomes more complex as the topology changes.
The protection scheme of the micro-grid ensures stability and quick-acting property, and simultaneously solves the problems of bidirectional tide and fault current in an island mode and a grid-connected mode. Aiming at the problems, scholars at home and abroad put forward different schemes to protect the micro-grid. For example, the equivalent impedance is calculated by adopting a three-impedance circular intersection point method, and then a self-adaptive mutation algorithm is added to optimize and perfect the traditional protection mode. For example, a concept of a composite compensation factor is provided, voltage characteristics and impedance characteristics during fault are analyzed, the compensation factor is added in inverse time limit overcurrent protection, the problem of action delay is solved, and high speed is improved. For example, wavelet transformation is utilized to extract zero sequence current characteristics of a bus, and characteristic directions and associated directions are constructed according to data of singular points of the highest frequency component, so that the fault position is determined, and the selectivity of a micro-grid protection threshold value is improved. For example, a differential protection technology suitable for a medium-voltage microgrid with a synchronous generator and an IIDG is provided. For example, a protection scheme only suitable for an IIDG-based low-voltage microgrid is proposed, and the scheme adopts a current and voltage scheme to protect the power grid. However, further analysis of these prior art techniques shows that the existing protection schemes are based on fixed structures or operation modes, and lack the ability to intelligently determine the location and type of the fault.
Therefore, the invention provides a microgrid protection method based on a convolutional neural network. And processing the current signal through wavelet transformation, extracting a characteristic value of the current signal, and inputting the characteristic value to the trained convolutional neural network model to realize intelligent identification of the relay protection on the fault.
Disclosure of Invention
In order to solve the problem that the existing microgrid protection scheme is based on a fixed structure or an operation mode and cannot intelligently judge the fault position and the fault type, the invention provides a microgrid protection method based on a convolutional neural network, which comprises the following steps:
s1: acquiring a three-phase current signal of the microgrid through current sampling;
s2: acquiring a negative sequence current signal and a zero sequence current signal under the symmetrical components of the three-phase current signals according to the three-phase current signals;
s3: acquiring a characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation, wherein the characteristic value set comprises an energy change value, a Shannon entropy and a standard difference value;
s4: and acquiring the fault position and the fault type of the microgrid by utilizing a convolutional neural network model according to the characteristic value set, wherein the convolutional neural network model is acquired after fault and non-fault analysis training and testing are carried out on a current data set.
Further, the current data set in the step S4 includes a current signal training set and a current signal testing set; the concrete steps of the convolutional neural network model obtained after performing fault and non-fault analysis training and testing through the current data set in the step S4 include:
s41: acquiring a current signal training set and a current signal testing set according to the preset database and the short-circuit fault conditions of each load level, each operation mode and each fault position simulated by virtual simulation;
s42: analyzing and training faults and non-faults of the preset model through a current signal training set;
s43: and obtaining a convolutional neural network model through the preset model after the current signal test set test training.
Further, the specific steps of obtaining the feature value set of the three-phase current signal, the negative-sequence current signal and the zero-sequence current signal by using wavelet transform in step S3 are as follows:
s31: acquiring a three-phase current frequency multiplication signal, a negative sequence current frequency multiplication signal and a zero sequence current frequency multiplication signal through wavelet transformation according to the three-phase current signal, the negative sequence current signal and the zero sequence current signal;
s32: and respectively obtaining an energy change value, a Shannon entropy and a standard difference value corresponding to each current frequency doubling signal by utilizing a preset signal energy formula, a preset Shannon entropy formula and a preset standard difference value formula according to the frequency doubling signal of each current.
Further, the preset signal energy formula in step S32 is as follows:
in the formula EiThe total signal energy of the frequency doubling signal under the time span i, j is the jth data point on the frequency doubling signal wave corresponding to the frequency doubling signal under the time span i, dijThe signal energy of the frequency doubling signal at the jth data point on the corresponding frequency doubling signal wave is shown, and n is the number of data points on the frequency doubling signal corresponding to the frequency doubling signal wave under the time span i.
Further, the obtaining formula of the preset energy variation value in step S32 is as follows:
ΔE=Ei-Ei-m;
wherein m is the number of time spans, Ei-mThe total signal energy of the frequency doubling signal under the time span i-m is shown, and the delta E is the energy change value of the total signal energy of the frequency doubling signal from the time span i to the time span i-m.
Further, the shannon entropy formula preset in step S32 is as follows:
in the formula, PjA preset energy probability distribution e corresponding to the jth data point on the frequency multiplication signal wave corresponding to the frequency multiplication signaliIs the shannon entropy of the multiplied signal at time span i.
Further, the standard deviation formula preset in step S32 is as follows:
wherein x isiThe data value corresponding to the ith data point on the frequency doubling signal wave corresponding to the frequency doubling signal is shown as mu, the average value of a data set formed by the data values corresponding to the data points is shown as mu, the total number of the data points is shown as N, and the standard deviation of the frequency doubling signal is shown as sigma.
The invention also provides a micro-grid protection system based on the convolutional neural network, which comprises the following components:
the sampling module acquires three-phase current signals of the microgrid through current sampling;
a symmetrical component module: acquiring a negative sequence current signal and a zero sequence current signal under the symmetrical components of the three-phase current signals according to the three-phase current signals;
the characteristic value module is used for acquiring a characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation, wherein the characteristic value set comprises an energy change value, a Shannon entropy and a standard difference value;
and the network model module is used for acquiring the fault position and the fault type of the microgrid by utilizing a convolutional neural network model according to the characteristic value set, and the convolutional neural network model is acquired after fault and non-fault analysis training and testing are carried out on a current data set.
Further, the current data set in the network model module comprises a current signal training set and a current signal testing set; the specific method for carrying out fault and non-fault analysis training and testing on the convolutional neural network model in the network model module through the current data set is as follows:
acquiring a current signal training set and a current signal testing set according to the preset database and the short-circuit fault conditions of each load level, each operation mode and each fault position simulated by virtual simulation; analyzing and training faults and non-faults of the preset model through a current signal training set; and obtaining a convolutional neural network model through the preset model after the current signal test set test training.
Further, the specific method for acquiring the characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing the wavelet transform in the characteristic value module comprises the following steps:
acquiring a three-phase current frequency multiplication signal, a negative sequence current frequency multiplication signal and a zero sequence current frequency multiplication signal through wavelet transformation according to the three-phase current signal, the negative sequence current signal and the zero sequence current signal; and respectively obtaining an energy change value, a Shannon entropy and a standard difference value corresponding to each current frequency doubling signal by utilizing a preset signal energy formula, a preset Shannon entropy formula and a preset standard difference value formula according to the frequency doubling signal of each current.
Compared with the prior art, the invention at least has the following beneficial effects:
(1) according to the method, the characteristic value set of each current frequency doubling signal is obtained through wavelet transformation, and the trained convolutional neural network is input to obtain the fault position and the fault type, so that the problem that the fault position and the fault type cannot be intelligently judged due to the fact that the existing microgrid protection scheme is based on a fixed structure or an operation mode is solved;
(2) the method intelligently judges the fault position and the fault type through the convolutional neural network model, and is different from the traditional protection mode, the scheme of the method is more intelligent, is not influenced by the operation mode, the load numerical value, the fault resistance and the like, and is high in stability;
(3) the wavelet transform and convolution neural network model utilized in the invention has the characteristics of high efficiency, small input data volume and high calculation speed, and improves the protection capability of the microgrid and the speed of identifying the fault position and the fault type;
(4) the microgrid protection method based on the convolutional neural network can quickly and accurately reflect the position and the fault type of the fault, and greatly improves the safety of the microgrid.
Drawings
FIG. 1 is a method flow diagram of a method and system for microgrid protection based on a convolutional neural network;
FIG. 2 is a convolutional neural network model acquisition flow diagram of a convolutional neural network-based microgrid protection method and system;
fig. 3 is a system block diagram of a method and system for microgrid protection based on a convolutional neural network.
Detailed Description
The following are specific embodiments of the present invention and are further described with reference to the drawings, but the present invention is not limited to these embodiments.
Example one
In order to solve the problem that the existing microgrid protection scheme is based on a fixed structure or an operation mode and cannot intelligently judge the fault position and the fault type, as shown in fig. 1, the invention provides a microgrid protection method based on a convolutional neural network, which comprises the following steps:
s1: acquiring a three-phase current signal of the microgrid through current sampling;
s2: acquiring a negative sequence current signal and a zero sequence current signal under the symmetrical components of the three-phase current signals according to the three-phase current signals;
s3: acquiring a characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation, wherein the characteristic value set comprises an energy change value, a Shannon entropy and a standard difference value;
the specific steps of obtaining the characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by using wavelet transform in the step S3 are as follows:
s31: acquiring a three-phase current frequency multiplication signal, a negative sequence current frequency multiplication signal and a zero sequence current frequency multiplication signal through wavelet transformation according to the three-phase current signal, the negative sequence current signal and the zero sequence current signal;
in this embodiment, the specific steps of obtaining the three-phase current frequency doubling signal, the negative sequence current frequency doubling signal and the zero sequence current frequency doubling signal through wavelet transformation are as follows:
s311: decomposing the three-phase current signals into six-time frequency three-phase current signals through wavelet transformation; decomposing the negative sequence current signal into a six-time frequency negative sequence current signal through wavelet transformation; decomposing the zero-sequence current signal into a six-time frequency zero-sequence current signal through wavelet transformation; the six-time frequency comprises a first frequency doubling, a second frequency doubling, a third frequency doubling, a fourth frequency doubling, a fifth frequency doubling and a sixth frequency doubling;
s312: acquiring a triple-frequency three-phase current signal, a triple-frequency negative sequence current signal and a triple-frequency zero sequence current signal; the frequency tripling three-phase current signal is a three-phase current frequency doubling signal; the triple frequency negative sequence current signal is a negative sequence current frequency doubling signal; and the frequency tripled zero-sequence current signal is a zero-sequence current frequency doubling signal.
It should be noted that the three-phase current frequency doubling signal includes an a-phase current frequency doubling signal, a B-phase current frequency doubling signal, and a C-phase current frequency doubling signal.
S32: and respectively obtaining an energy change value, a Shannon entropy and a standard difference value corresponding to each current frequency doubling signal by utilizing a preset signal energy formula, a preset Shannon entropy formula and a preset standard difference value formula according to the frequency doubling signal of each current.
In this embodiment, the energy variation value, shannon entropy, and standard deviation value corresponding to each current frequency doubling signal are respectively:
characteristic value 1: when a fault occurs, the energy change value of the A-phase current frequency doubling signal;
characteristic value 2: when a fault occurs, the Shannon entropy of the A-phase current frequency doubling signal;
characteristic value 3: when a fault occurs, the standard difference value of the A-phase current frequency doubling signal;
characteristic value 4: when a fault occurs, the energy change value of the B-phase current frequency doubling signal;
characteristic value 5: when a fault occurs, the Shannon entropy of a B-phase current frequency doubling signal;
characteristic value 6: when a fault occurs, the standard difference value of the phase B current frequency doubling signal;
characteristic value 7: when a fault occurs, the energy change value of the C-phase current frequency doubling signal;
characteristic value 8: when a fault occurs, the Shannon entropy of the C-phase current frequency doubling signal;
characteristic value 9: when a fault occurs, the standard difference value of the C-phase current frequency doubling signal;
characteristic value 10: when a fault occurs, the energy change value of the negative sequence current frequency multiplication signal;
characteristic value 11: when a fault occurs, the Shannon entropy of the negative sequence current frequency multiplication signal;
characteristic value 12: when a fault occurs, the standard difference value of the negative sequence current frequency multiplication signal;
characteristic value 13: when a fault occurs, the energy change value of the zero-sequence current frequency doubling signal;
characteristic value 14: when a fault occurs, the Shannon entropy of the zero sequence current frequency doubling signal;
characteristic value 15: when a fault occurs, the standard difference value of zero sequence current frequency doubling signals;
a total of 15 eigenvalues make up an eigenvalue set.
The preset signal energy formula in step S32 is:
in the formula EiThe total signal energy of the frequency doubling signal under the time span i, j is the jth data point on the frequency doubling signal wave corresponding to the frequency doubling signal under the time span i, dijThe signal energy of the frequency doubling signal at the jth data point on the corresponding frequency doubling signal wave is shown, and n is the number of data points on the frequency doubling signal corresponding to the frequency doubling signal wave under the time span i.
The formula for obtaining the preset energy variation value in step S32 is:
ΔE=Ei-Ei-m;
wherein m is the number of time spans, Ei-mThe total signal energy of the frequency doubling signal under the time span i-m is shown, and the delta E is the energy change value of the total signal energy of the frequency doubling signal from the time span i to the time span i-m.
The shannon entropy formula preset in step S32 is as follows:
in the formula, PjA preset energy probability distribution e corresponding to the jth data point on the frequency multiplication signal wave corresponding to the frequency multiplication signaliIs the shannon entropy of the multiplied signal at time span i.
The preset standard deviation formula in step S32 is as follows:
wherein x isiThe data value corresponding to the ith data point on the frequency doubling signal wave corresponding to the frequency doubling signal is shown as mu, the average value of a data set formed by the data values corresponding to the data points is shown as mu, the total number of the data points is shown as N, and the standard deviation of the frequency doubling signal is shown as sigma.
S4: and acquiring the fault position and the fault type of the microgrid by utilizing a convolutional neural network model according to the characteristic value set, wherein the convolutional neural network model is acquired after fault and non-fault analysis training and testing are carried out on a current data set.
In this embodiment, the fault position and the fault type are obtained by inputting the feature value set composed of the 15 feature values to the trained convolutional neural network model.
The current data set in the step S4 includes a current signal training set and a current signal testing set; the concrete steps of the convolutional neural network model obtained after performing fault and non-fault analysis training and testing through the current data set in the step S4 include:
in this embodiment, the obtaining flow of the convolutional neural network model is shown in fig. 2.
S41: acquiring a current signal training set and a current signal testing set according to the preset database and the short-circuit fault conditions of each load level, each operation mode and each fault position simulated by virtual simulation;
s42: analyzing and training faults and non-faults of the preset model through a current signal training set;
s43: and obtaining a convolutional neural network model through the preset model after the current signal test set test training.
Example two
In order to better understand the inventive concept of the present invention, this embodiment explains the present invention in the form of a system structure, as shown in fig. 3, a microgrid protection system based on a convolutional neural network includes:
the sampling module acquires three-phase current signals of the microgrid through current sampling;
a symmetrical component module: acquiring a negative sequence current signal and a zero sequence current signal under the symmetrical components of the three-phase current signals according to the three-phase current signals;
the characteristic value module is used for acquiring a characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation, wherein the characteristic value set comprises an energy change value, a Shannon entropy and a standard difference value;
the specific method for acquiring the characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation in the characteristic value module comprises the following steps:
acquiring a three-phase current frequency multiplication signal, a negative sequence current frequency multiplication signal and a zero sequence current frequency multiplication signal through wavelet transformation according to the three-phase current signal, the negative sequence current signal and the zero sequence current signal; and respectively obtaining an energy change value, a Shannon entropy and a standard difference value corresponding to each current frequency doubling signal by utilizing a preset signal energy formula, a preset Shannon entropy formula and a preset standard difference value formula according to the frequency doubling signal of each current.
And the network model module is used for acquiring the fault position and the fault type of the microgrid by utilizing a convolutional neural network model according to the characteristic value set, and the convolutional neural network model is acquired after fault and non-fault analysis training and testing are carried out on a current data set.
It should be noted that the invention acquires the characteristic value set of each current frequency multiplication signal through wavelet transformation, and inputs the trained convolutional neural network to acquire the fault position and fault type, thereby solving the problem that the existing microgrid protection scheme is based on a fixed structure or an operation mode and cannot intelligently judge the fault position and fault type; the method intelligently judges the fault position and the fault type through the convolutional neural network model, and is different from the traditional protection mode, the scheme of the method is more intelligent, is not influenced by the operation mode, the load numerical value, the fault resistance and the like, and is high in stability;
in addition, the wavelet transformation and convolution neural network model utilized in the invention has the characteristics of high efficiency, small input data volume and high calculation speed, and improves the protection capability of the microgrid and the speed of identifying the fault position and the fault type.
The current data set in the network model module comprises a current signal training set and a current signal testing set; the specific method for carrying out fault and non-fault analysis training and testing on the convolutional neural network model in the network model module through the current data set is as follows:
acquiring a current signal training set and a current signal testing set according to the preset database and the short-circuit fault conditions of each load level, each operation mode and each fault position simulated by virtual simulation; analyzing and training faults and non-faults of the preset model through a current signal training set; and obtaining a convolutional neural network model through the preset model after the current signal test set test training.
The microgrid protection method based on the convolutional neural network can quickly and accurately reflect the position and the fault type of the fault, and greatly improves the safety of the microgrid.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (10)
1. A micro-grid protection method based on a convolutional neural network is characterized by comprising the following steps:
s1: acquiring a three-phase current signal of the microgrid through current sampling;
s2: acquiring a negative sequence current signal and a zero sequence current signal under the symmetrical components of the three-phase current signals according to the three-phase current signals;
s3: acquiring a characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation, wherein the characteristic value set comprises an energy change value, a Shannon entropy and a standard difference value;
s4: and acquiring the fault position and the fault type of the microgrid by utilizing a convolutional neural network model according to the characteristic value set, wherein the convolutional neural network model is acquired after fault and non-fault analysis training and testing are carried out on a current data set.
2. The microgrid protection method based on a convolutional neural network of claim 1, wherein the current data set in the step S4 includes a current signal training set and a current signal testing set; the concrete steps of the convolutional neural network model obtained after performing fault and non-fault analysis training and testing through the current data set in the step S4 include:
s41: acquiring a current signal training set and a current signal testing set according to the preset database and the short-circuit fault conditions of each load level, each operation mode and each fault position simulated by virtual simulation;
s42: analyzing and training faults and non-faults of the preset model through a current signal training set;
s43: and obtaining a convolutional neural network model through the preset model after the current signal test set test training.
3. The microgrid protection method based on a convolutional neural network of claim 1, wherein the specific steps of obtaining the characteristic value set of the three-phase current signal, the negative-sequence current signal and the zero-sequence current signal by using wavelet transform in step S3 are as follows:
s31: acquiring a three-phase current frequency multiplication signal, a negative sequence current frequency multiplication signal and a zero sequence current frequency multiplication signal through wavelet transformation according to the three-phase current signal, the negative sequence current signal and the zero sequence current signal;
s32: and respectively obtaining an energy change value, a Shannon entropy and a standard difference value corresponding to each current frequency doubling signal by utilizing a preset signal energy formula, a preset Shannon entropy formula and a preset standard difference value formula according to the frequency doubling signal of each current.
4. The microgrid protection method based on a convolutional neural network of claim 3, wherein the preset signal energy formula in the step S32 is as follows:
in the formula EiThe total signal energy of the frequency doubling signal under the time span i, j is the jth data point on the frequency doubling signal wave corresponding to the frequency doubling signal under the time span i, dijThe signal energy of the frequency doubling signal at the jth data point on the corresponding frequency doubling signal wave is shown, and n is the number of data points on the frequency doubling signal corresponding to the frequency doubling signal wave under the time span i.
5. The microgrid protection method based on a convolutional neural network of claim 4, wherein the preset energy variation value in the step S32 is obtained by the following formula:
ΔE=Ei-Ei-m;
wherein m is the number of time spans, Ei-mThe total signal energy of the frequency doubling signal under the time span i-m is shown, and the delta E is the energy change value of the total signal energy of the frequency doubling signal from the time span i to the time span i-m.
6. The microgrid protection method based on a convolutional neural network of claim 5, wherein the shannon entropy formula preset in the step S32 is as follows:
in the formula, PjA preset energy probability distribution e corresponding to the jth data point on the frequency multiplication signal wave corresponding to the frequency multiplication signaliIs the shannon entropy of the multiplied signal at time span i.
7. The microgrid protection method based on a convolutional neural network of claim 6, wherein the standard deviation formula preset in the step S32 is as follows:
wherein x isiThe data value corresponding to the ith data point on the frequency doubling signal wave corresponding to the frequency doubling signal is shown as mu, the average value of a data set formed by the data values corresponding to the data points is shown as mu, the total number of the data points is shown as N, and the standard deviation of the frequency doubling signal is shown as sigma.
8. A microgrid protection system based on a convolutional neural network, comprising:
the sampling module acquires three-phase current signals of the microgrid through current sampling;
a symmetrical component module: acquiring a negative sequence current signal and a zero sequence current signal under the symmetrical components of the three-phase current signals according to the three-phase current signals;
the characteristic value module is used for acquiring a characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation, wherein the characteristic value set comprises an energy change value, a Shannon entropy and a standard difference value;
and the network model module is used for acquiring the fault position and the fault type of the microgrid by utilizing a convolutional neural network model according to the characteristic value set, and the convolutional neural network model is acquired after fault and non-fault analysis training and testing are carried out on a current data set.
9. The convolutional neural network-based microgrid protection system of claim 8, wherein the current data sets in the network model module comprise a current signal training set and a current signal testing set; the specific method for carrying out fault and non-fault analysis training and testing on the convolutional neural network model in the network model module through the current data set is as follows:
acquiring a current signal training set and a current signal testing set according to the preset database and the short-circuit fault conditions of each load level, each operation mode and each fault position simulated by virtual simulation; analyzing and training faults and non-faults of the preset model through a current signal training set; and obtaining a convolutional neural network model through the preset model after the current signal test set test training.
10. The microgrid protection system based on a convolutional neural network of claim 8, wherein the specific method for acquiring the characteristic value set of the three-phase current signal, the negative sequence current signal and the zero sequence current signal by utilizing wavelet transformation in the characteristic value module is as follows:
acquiring a three-phase current frequency multiplication signal, a negative sequence current frequency multiplication signal and a zero sequence current frequency multiplication signal through wavelet transformation according to the three-phase current signal, the negative sequence current signal and the zero sequence current signal; and respectively obtaining an energy change value, a Shannon entropy and a standard difference value corresponding to each current frequency doubling signal by utilizing a preset signal energy formula, a preset Shannon entropy formula and a preset standard difference value formula according to the frequency doubling signal of each current.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110556505.6A CN113315100A (en) | 2021-05-21 | 2021-05-21 | Micro-grid protection method and system based on convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110556505.6A CN113315100A (en) | 2021-05-21 | 2021-05-21 | Micro-grid protection method and system based on convolutional neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113315100A true CN113315100A (en) | 2021-08-27 |
Family
ID=77374092
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110556505.6A Pending CN113315100A (en) | 2021-05-21 | 2021-05-21 | Micro-grid protection method and system based on convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113315100A (en) |
-
2021
- 2021-05-21 CN CN202110556505.6A patent/CN113315100A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | Fault diagnosis of wind turbine power converter considering wavelet transform, feature analysis, judgment and BP neural network | |
Khalid et al. | Existing developments in adaptive smart grid protection: A review | |
Dash et al. | A hybrid time–frequency approach based fuzzy logic system for power island detection in grid connected distributed generation | |
CN102508076B (en) | Fault diagnosis device and method based on multi-agent system and wavelet analysis | |
Padhee et al. | A fast Gauss-Newton algorithm for islanding detection in distributed generation | |
CN108667005B (en) | Power grid static and dynamic combination vulnerability assessment method considering new energy influence | |
CN111103503B (en) | Micro-grid fault positioning method based on simulation | |
Jia et al. | Sparse voltage amplitude measurement based fault location in large-scale photovoltaic power plants | |
Assis et al. | Impact of multi-terminal HVDC grids on enhancing dynamic power transfer capability | |
CN108802570B (en) | Fault detection system and method for alternating current-direct current series-parallel micro-grid | |
Abdulwahid et al. | A new differential protection scheme for microgrid using Hilbert space based power setting and fuzzy decision processes | |
CN111190121A (en) | Distributed power supply containing power distribution network fault classification method and system based on sequence components | |
Zamani et al. | Islanding detection of synchronous machine-based distributed generators using signal trajectory pattern recognition | |
Chaitanya et al. | Communication assisted fuzzy based adaptive protective relaying scheme for microgrid | |
Uddin et al. | Hybrid machine learning-based intelligent distance protection and control schemes with fault and zonal classification capabilities for grid-connected wind farms | |
Tremblay et al. | Real-time simulation of a fully detailed type-IV wind turbine | |
Venkata et al. | Data mining model based differential microgrid fault classification using svm considering voltage and current distortions | |
CN116054224A (en) | Quick evaluation method for commutation failure of multi-feed direct-current transmission system | |
CN113437732B (en) | Pilot protection method and system for photovoltaic power generation grid-connected tie line | |
CN113315100A (en) | Micro-grid protection method and system based on convolutional neural network | |
Stanisavljević et al. | A method for real-time prediction of the probability of voltage sag duration based on harmonic footprint | |
Wang et al. | LPV Model-Based Fault Detection and Isolation in DC Microgrids Through Signature Recognition | |
Torres et al. | A comparison of new methods based on ANNs for detecting and locating faults in MTDC systems | |
Tao et al. | Transient stability analysis of AC/DC system considering electromagnetic transient model | |
CN107565547A (en) | A kind of power distribution network operation reliability evaluation and optimization 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 |