Classification method, device, equipment and storage medium for distribution network fault causes

Abstract

The present invention discloses a classification method, device, equipment and storage medium for distribution network fault causes. The method includes: obtaining zero-sequence current waveform data of distribution network faults of different fault types; determining a training set and a test set based on target waveform data selected from the zero-sequence current waveform data; decomposing the training set and the test set to obtain a first eigenvector and a second eigenvector respectively; normalizing the first eigenvector and inputting it into an initial fault cause classification model for training to obtain a trained fault cause classification model; normalizing the second eigenvector and inputting it into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault. The present invention solves the technical problem of large labor cost and low recognition accuracy caused by the use of manual line patrol in the prior art to classify and identify the causes of distribution network faults.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, device, equipment and storage medium for classifying causes of distribution network faults. Background Art

[0002] As an important public infrastructure for national economic and social development, the distribution network is directly connected to the majority of electricity users and is an important link in the electric energy transmission chain. Failures in the distribution network will affect the safe and stable operation of the power system and the reliability of power supply.

[0003] Due to the dispersed geographical location and complex operating environment of the distribution network, it is very easy to be affected by various factors, leading to the occurrence of faults, such as tree-line conflict faults, insulation breakdown faults, etc. Most of my country's distribution networks use small current grounding, that is, the neutral point is not grounded or grounded through an arc suppression coil. More than 80% of the faults in the small current grounding system are single-phase grounding faults. Therefore, for the single-phase grounding fault of the distribution network, accurate and rapid identification of the fault cause is crucial for fault clearance and power supply restoration.

[0004] At present, the method for identifying and classifying the causes of single-phase grounding faults in distribution networks mainly relies on "manual line patrol", in which operators patrol along the lines to find the causes of the faults. This not only consumes a lot of manpower and material resources and prolongs the power outage time, but is also prone to misjudgment and has a low identification accuracy rate.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiments of the present invention provide a method, device, equipment and storage medium for classifying the causes of distribution network faults, so as to at least solve the technical problems of high manpower cost and low identification accuracy caused by the manual line patrol method used in the prior art to classify and identify the causes of distribution network faults.

[0007] According to one aspect of an embodiment of the present invention, a method for classifying the causes of distribution network faults is provided, comprising: obtaining zero-sequence current waveform data of distribution network faults of different fault types; determining a training set and a test set based on target waveform data selected from the zero-sequence current waveform data, wherein the target waveform data comprises: zero-sequence current waveform data within a first predetermined time period before the occurrence of the distribution network fault, and zero-sequence current waveform data within a second predetermined time period after the occurrence of the distribution network fault; decomposing the training set and the test set to obtain intrinsic mode function components with local signal characteristics, and obtaining first eigenvectors and second eigenvectors correspondingly; normalizing the first eigenvectors and inputting them into an initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network; normalizing the second eigenvectors and inputting them into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault.

[0008] Optionally, based on the target waveform data selected from the above-mentioned zero-sequence current waveform data, a training set and a test set are determined, including: selecting from the above-mentioned zero-sequence current waveform data the zero-sequence current waveform data within a first predetermined time period before the above-mentioned distribution network fault occurs, and the zero-sequence current waveform data within a second predetermined time period after the above-mentioned distribution network fault occurs; performing data enhancement processing on the zero-sequence current waveform data within the above-mentioned first predetermined time period and the zero-sequence current waveform data within the above-mentioned second predetermined time period to obtain the above-mentioned training set and the above-mentioned test set.

[0009] Optionally, the training set and the test set are decomposed to obtain multiple intrinsic mode function components with local signal characteristics, corresponding to a first eigenvector and a second eigenvector, including: using an empirical mode decomposition algorithm to decompose each zero-sequence current waveform data in the training set according to different time scales to obtain multiple intrinsic mode function components; using an empirical mode decomposition algorithm to decompose each zero-sequence current waveform data in the test set according to different time scales to obtain multiple intrinsic mode function components; combining the multiple intrinsic mode function components corresponding to the training set to obtain a first eigenvector, and combining the multiple intrinsic mode function components corresponding to the test set to obtain a second eigenvector.

[0010] Optionally, the first feature vector is normalized and then input into the initial fault cause classification model for training to obtain the trained fault cause classification model, including: normalizing the first feature vector and then inputting into the initial fault cause classification model for training to obtain the first output feature vector of the distribution network fault; inputting the first output feature vector into a classifier, and the classifier maps the first output feature vector into a first probability distribution value of each fault cause; using the first probability distribution value of each fault cause and a predetermined label distribution to construct a loss function to train the initial fault cause classification model until a predetermined stop condition is met to obtain the trained fault cause classification model, wherein the predetermined stop condition includes at least one of the following: the initial fault cause classification model converges and a predetermined number of training times is met.

[0011] Optionally, the second eigenvector is normalized and input into the trained fault cause classification model to obtain the fault cause classification result of the distribution network fault, including: the second eigenvector is normalized and input into the trained fault cause classification model to obtain the second output eigenvector of the distribution network fault; the second output eigenvector is input into the classifier, and the classifier maps the second output eigenvector to the second probability distribution value of each fault cause; determines the maximum probability distribution value among the second probability distribution values ​​of each fault cause; and determines the corresponding fault cause classification results based on the maximum probability distribution value of each fault cause.

[0012] According to another aspect of an embodiment of the present invention, there is also provided a distribution network fault cause classification device, comprising: an acquisition module, used to acquire zero-sequence current waveform data of distribution network faults of different fault types; a determination module, used to determine a training set and a test set based on target waveform data selected from the zero-sequence current waveform data, wherein the target waveform data comprises: zero-sequence current waveform data within a first predetermined time period before the occurrence of the distribution network fault, and zero-sequence current waveform data within a second predetermined time period after the occurrence of the distribution network fault; a first processing module, used to decompose the training set and the test set respectively to obtain intrinsic mode function components with local signal characteristics, and obtain first eigenvectors and second eigenvectors correspondingly; a training module, used to normalize the first eigenvectors and input them into an initial fault cause classification model for training, to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network; a second processing module, used to normalize the second eigenvectors and input them into the trained fault cause classification model, to obtain a fault cause classification result of the distribution network fault.

[0013] Optionally, the above-mentioned determination module includes: a selection unit, used to select zero-sequence current waveform data within a first predetermined time period before the above-mentioned distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the above-mentioned distribution network fault occurs from the above-mentioned zero-sequence current waveform data; a first processing unit, used to perform data enhancement processing on the zero-sequence current waveform data within the above-mentioned first predetermined time period and the zero-sequence current waveform data within the above-mentioned second predetermined time period to obtain the above-mentioned training set and the above-mentioned test set.

[0014] Optionally, the above-mentioned first processing module includes: a second processing unit, used to use an empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned training set according to different time scales, so as to obtain multiple above-mentioned intrinsic mode function components; a third processing unit, used to use an empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned test set according to different time scales, so as to obtain multiple above-mentioned intrinsic mode function components; a combining unit, used to combine the above-mentioned multiple intrinsic mode function components corresponding to the above-mentioned training set to obtain a first eigenvector, and combine the above-mentioned multiple intrinsic mode function components corresponding to the above-mentioned test set to obtain a second eigenvector.

[0015] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing any one of the above-mentioned methods for classifying causes of distribution network faults.

[0016] According to another aspect of an embodiment of the present invention, there is also provided an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any one of the above-mentioned methods for classifying causes of distribution network faults.

[0017] In an embodiment of the present invention, a classification method of the causes of distribution network faults is adopted, and zero-sequence current waveform data of distribution network faults of different fault types are obtained; a training set and a test set are determined based on target waveform data selected from the zero-sequence current waveform data, wherein the target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the occurrence of the distribution network fault, and zero-sequence current waveform data within a second predetermined time period after the occurrence of the distribution network fault; the intrinsic mode function components with local signal characteristics obtained by decomposing the training set and the test set are respectively obtained, and the first eigenvector and the second eigenvector are correspondingly obtained; the first eigenvector is normalized The first feature vector is normalized and then input into the initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network; the second feature vector is normalized and then input into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault, thereby achieving the purpose of automatically classifying the causes of distribution network faults, thereby achieving the technical effect of improving the efficiency and accuracy of classification and identification of distribution network fault causes and reducing labor costs, thereby solving the technical problem of large labor cost investment and low identification accuracy caused by the use of manual line patrol method to classify and identify distribution network fault causes in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0019] Figure 1 is a flow chart of a method for classifying causes of distribution network faults according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of an optional LSTM network-based initial fault cause classification model construction principle according to an embodiment of the present invention;

[0021] Figure 3 is a schematic diagram of an optional output result of an intrinsic mode function component and a residual component according to an embodiment of the present invention;

[0022] Figure 4 is a flow chart of an optional method for classifying causes of distribution network faults according to an embodiment of the present invention;

[0023] Figure 5 is a two-dimensional visualization schematic diagram of an optional fault cause classification result according to an embodiment of the present invention;

[0024] Figure 64 is a schematic diagram of the structure of a device for classifying causes of distribution network faults according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] Example 1

[0028] According to an embodiment of the present invention, a method embodiment for classifying causes of distribution network faults is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0029] Figure 1 is a flow chart of a method for classifying causes of distribution network faults according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0030] Step S102, obtaining zero-sequence current waveform data of distribution network faults of different fault types;

[0031] Step S104, determining a training set and a test set based on target waveform data selected from the zero-sequence current waveform data;

[0032] Step S106, respectively decomposing the training set and the test set to obtain intrinsic mode function components with local signal characteristics, and correspondingly obtaining a first eigenvector and a second eigenvector;

[0033] Step S108, normalizing the first feature vector and inputting it into an initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network;

[0034] Step S110, normalizing the second eigenvector and inputting it into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault.

[0035] Optionally, the above fault types include at least one of the following: burning of the transformer flexible cable, insulation breakdown of the transformer elbow head, and tree-line conflict.

[0036] Optionally, the above-mentioned target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the above-mentioned distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the above-mentioned distribution network fault occurs. For example, the zero-sequence current waveform data N1 seconds before the above-mentioned distribution network fault occurs and N2 seconds after the fault occurs are intercepted as the above-mentioned target waveform data.

[0037] Optionally, the training set and the test set are decomposed by using an empirical mode decomposition (EMD) algorithm to obtain intrinsic mode function components with local signal characteristics, and the first eigenvector and the second eigenvector are correspondingly obtained.

[0038] Optionally, the first eigenvector is used to train the initial fault cause classification model to obtain a trained fault cause classification model; the second eigenvector is input as test data into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault.

[0039] In an embodiment of the present invention, a classification method of the causes of distribution network faults is adopted, and zero-sequence current waveform data of distribution network faults of different fault types are obtained; a training set and a test set are determined based on target waveform data selected from the zero-sequence current waveform data, wherein the target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the occurrence of the distribution network fault, and zero-sequence current waveform data within a second predetermined time period after the occurrence of the distribution network fault; the intrinsic mode function components with local signal characteristics obtained by decomposing the training set and the test set are respectively obtained, and the first eigenvector and the second eigenvector are correspondingly obtained; the first eigenvector is normalized The first feature vector is normalized and then input into the initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network; the second feature vector is normalized and then input into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault, thereby achieving the purpose of automatically classifying the causes of distribution network faults, thereby achieving the technical effect of improving the efficiency and accuracy of classification and identification of distribution network fault causes and reducing labor costs, thereby solving the technical problem of large labor cost investment and low identification accuracy caused by the use of manual line patrol method to classify and identify distribution network fault causes in the prior art.

[0040] Optionally, the initial fault cause classification model in the embodiment of the present invention is constructed based on a long short-term memory network (i.e., an LSTM network). The initial fault cause classification model constructed based on the LSTM network includes: a plurality of unit structures connected in sequence, such as Figure 2 As shown, each unit structure includes: forget gate f t , input gate i t and output gate o t , add or delete information to the cell state. The forget gate determines the cell state c at the previous moment t-1 How many are retained to the current state c t , the input gate determines the input IMF of the network at the current moment t 'How much is saved to the cell state c t , the output gate controls the control unit state c t How many outputs are there to the LSTM's current output value h t The calculation process is expressed as:

[0041] f t =σ(W f ·[h t-1 ,IMF t ']+b f )

[0042] i t=σ(W i ·[h t-1 ,IMF t ']+b i )

[0043] o t =σ(W o ·[h t-1 ,IMF t ']+b o )

[0044] Through the forget gate t and input gate i t For the unit state c at the previous moment t-1 Update and get the new unit state c t , using the new cell state c t :Calculate the output feature vector h of the fault at the current moment t , expressed as:

[0045]

[0046]

[0047] h t =o t *tanh(c t )

[0048] in, is the current input unit state, W f , W i , W o and W c f t 、i t , o t as well as The weight matrix, b i , b f , b o and b c is the corresponding bias; σ and tanh are the Sigmoid activation function and the hyperbolic tangent activation function respectively; * represents matrix multiplication; t and t-1 represent the current moment and the previous moment respectively.

[0049] It should be noted that, in the embodiment of the present invention, a fault cause classification model is constructed based on the LSTM network, and the model is trained using feature vectors extracted from waveform data containing rich fault information. Compared with traditional methods, it can quickly and accurately classify the fault causes of the actual distribution network and realize real-time monitoring of the dynamic behavior of the power system.

[0050] In an optional embodiment, based on the target waveform data selected from the above zero-sequence current waveform data, determining the training set and the test set includes:

[0051] Step S202, selecting from the zero-sequence current waveform data the zero-sequence current waveform data within a first predetermined time period before the distribution network fault occurs, and the zero-sequence current waveform data within a second predetermined time period after the distribution network fault occurs;

[0052] Step S204, performing data enhancement processing on the zero-sequence current waveform data within the first predetermined time period and the zero-sequence current waveform data within the second predetermined time period to obtain the training set and the test set.

[0053] Optionally, the zero-sequence current waveform data within a first predetermined time period before the above-mentioned distribution network fault occurs, and the zero-sequence current waveform data within a second predetermined time period after the above-mentioned distribution network fault occurs, for example, the zero-sequence current waveform data N1 seconds before the above-mentioned distribution network fault occurs and N2 seconds after the fault occurs are taken as the above-mentioned target waveform data, and data enhancement processing is performed on the above-mentioned target waveform data to obtain the above-mentioned training set and the above-mentioned test set.

[0054] Optionally, a fault recording device is used to record the zero-sequence current waveform data of the distribution network fault. For example, the first predetermined time period and the first predetermined time period are respectively set to N1=0.1 and N2=0.3, then a total of 0.04 seconds of zero-sequence current waveform data is intercepted as the target waveform data, and the waveform segment contains rich fault information. In addition, for the zero-sequence current waveform data before and after the fault occurs, the sampling frequency can be set to 4096Hz and 12800Hz respectively.

[0055] In an optional embodiment, the training set and the test set are decomposed to obtain a plurality of intrinsic mode function components having local signal characteristics, and the first eigenvector and the second eigenvector are obtained correspondingly, including:

[0056] Step S302, using an empirical mode decomposition algorithm to decompose each zero-sequence current waveform data in the training set according to different time scales to obtain a plurality of intrinsic mode function components;

[0057] Step S304, using an empirical mode decomposition algorithm to decompose each zero-sequence current waveform data in the test set according to different time scales to obtain a plurality of intrinsic mode function components;

[0058] Step S306, combining the multiple intrinsic mode function components corresponding to the training set to obtain a first eigenvector, and combining the multiple intrinsic mode function components corresponding to the test set to obtain a second eigenvector.

[0059] It can be understood that the above-mentioned method of using the empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned training set according to different time scales to obtain the above-mentioned multiple intrinsic mode function components is the same as the above-mentioned algorithm process of using the empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned test set according to different time scales to obtain the above-mentioned multiple intrinsic mode function components. Taking the use of the empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned training set as an example, the method includes the following steps:

[0060] Step S311, determining all local maximum points and local minimum points in each zero-sequence current waveform data x(t) in the training set;

[0061] Step S312, performing cubic spline function interpolation processing on all the local maximum points and the local minimum points to obtain an upper envelope and a lower envelope;

[0062] Step S313, calculating the average value m(t) between the upper envelope and the lower envelope, and the difference c(t)=x(t)-m(t) between the zero-sequence current waveform data x(t) and the average value m(t);

[0063] Step S314, determining whether the difference c(t) satisfies the conditions for the establishment of the intrinsic mode function, wherein the conditions include: a first condition and a second condition, the first condition being that the difference between the number of poles and the number of zeros in the difference c(t) is less than or equal to a first preset value (such as setting the first preset value to 1), and the second condition being that the average values ​​of the upper envelope and the lower envelope are 0;

[0064] Step S315, if the difference c(t) satisfies both the first condition and the second condition, the difference c(t) is used as the first intrinsic mode function component, and a residual function r(t)=x(t)-c(t) is obtained;

[0065] Step S316, if the difference c(t) does not satisfy both the first condition and the second condition, the difference c(t) is used as the original signal, and the process returns to step S311;

[0066] Step S317, repeat the above steps S311 to S316 until the number of extreme points of the above residual function r(t) (i.e., the total number of the maximum and minimum values ​​of r(t)) is less than or equal to the second preset value (for example, the above second preset value is set to 2), then the decomposition is completed, and the final residual function r(t) is recorded as the residual component res. The output results of the intrinsic mode function component and the residual component are finally obtained as follows: Figure 3 shown.

[0067] Optionally, after each zero-sequence current waveform data in the training set is subjected to eigenmode decomposition processing, the multiple eigenmode function components obtained can be expressed as:

[0068]

[0069] Among them, IMF j is the jth intrinsic mode function component after empirical mode decomposition; res is the residual component; n is the total number of decomposed intrinsic mode function components.

[0070] In an optional embodiment, the first feature vector is normalized and then input into an initial fault cause classification model for training to obtain a trained fault cause classification model, including:

[0071] Step S402, normalizing the first feature vector and inputting the normalized feature vector into an initial fault cause classification model for training to obtain a first output feature vector of the distribution network fault;

[0072] Step S404, inputting the first output feature vector into a classifier, and the classifier maps the first output feature vector into a first probability distribution value of each fault cause;

[0073] Step S406, using the above-mentioned first probability distribution value of each of the above-mentioned fault causes and the predetermined label distribution to construct a loss function to train the above-mentioned initial fault cause classification model until the predetermined stopping condition is met, and the above-mentioned trained fault cause classification model is obtained, wherein the above-mentioned predetermined stopping condition includes at least one of the following: the initial fault cause classification model converges and meets the predetermined number of training times.

[0074] Optionally, the above classifier may be, but is not limited to, a Softmax classifier.

[0075] It should be noted that the above-mentioned first eigenvector includes multiple intrinsic mode function components. Since the amplitudes of the intrinsic mode function components fluctuate greatly, affecting the classification performance of the fault cause classification model, each intrinsic mode function component in the above-mentioned first eigenvector is normalized and the amplitude is transformed to [0, 1] to improve the classification effect of the fault cause classification model.

[0076] Optionally, the normalization formula may be: Among them, IMF j ' represents the normalized intrinsic mode function component (IMF component), IMF jmax 、IMF jmin are the maximum and minimum values ​​of the jth intrinsic mode function component in all eigenvectors. After EMD decomposition and normalization, the eigenvector is obtained, denoted as IMF'=[IMF1',IMF1',...,IMF n ']. Where n is the number of IMF components obtained after EMD decomposition.

[0077] Optionally, the first feature vector corresponding to the training set is normalized to obtain the normalized first feature vector IMF'; the first feature vector after normalization is input into the trained fault cause classification model to obtain the first output feature vector h of the distribution network fault. t ; The first output feature vector h t Input to the Softmax classifier, which converts the first output feature vector h t Mapped to the first probability distribution value p of each fault cause, the formula is expressed as:

[0078] p=Softmax(Wh t +b)

[0079] Where p is the first probability distribution value of each fault cause, W and b are the weight and bias of the Softmax classifier respectively.

[0080] Optionally, during the model training stage, the first probability distribution value p of each fault cause and a preset loss function are used to train the initial fault cause classification model until a predetermined stopping condition is met, wherein the preset loss function is a loss function constructed based on a given label distribution, and the predetermined stopping condition includes at least one of the following: the initial fault cause classification model converges and a predetermined number of training times is met.

[0081] In an alternative embodiment, Figure 4 is a flow chart of an optional method for classifying causes of distribution network faults according to an embodiment of the present invention, such as Figure 4 As shown, the second eigenvector is normalized and then input into the trained fault cause classification model to obtain the fault cause classification result of the distribution network fault, including:

[0082] Step S502, normalizing the second feature vector and inputting it into the trained fault cause classification model to obtain a second output feature vector of the distribution network fault;

[0083] Step S504, inputting the second output feature vector into a classifier, and the classifier maps the second output feature vector into a second probability distribution value of each fault cause;

[0084] Step S506, determining the maximum probability distribution value among the second probability distribution values ​​of each of the fault causes;

[0085] Step S508: determining the corresponding fault cause classification result based on the maximum probability distribution value of each fault cause.

[0086] Optionally, the above classifier may be, but is not limited to, a Softmax classifier.

[0087] It should be noted that the above-mentioned second eigenvector includes multiple intrinsic mode function components. Since the amplitudes of the intrinsic mode function components fluctuate greatly, affecting the classification performance of the fault cause classification model, each intrinsic mode function component in the above-mentioned second eigenvector is normalized and the amplitude is transformed to [0, 1] to improve the classification effect of the fault cause classification model.

[0088] Optionally, the second feature vector corresponding to the test set is normalized to obtain the normalized second feature vector; the normalized second feature vector is input into the trained fault cause classification model to obtain the second output feature vector of the distribution network fault; the second output feature vector is input into a Softmax classifier, and the Softmax classifier maps the second output feature vector to a second probability distribution value of each fault cause; the fault type corresponding to the maximum probability distribution value is selected as the fault cause classification result, thereby achieving accurate classification of the fault cause. Figure 5 is a two-dimensional visualization schematic diagram of an optional fault cause classification result according to an embodiment of the present invention. It can be seen that the fault cause classification method has a good classification effect.

[0089] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0090] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the above methods of each embodiment of the present invention.

[0091] Example 2

[0092] In this embodiment, a device for classifying the causes of distribution network faults is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the terms "module" and "device" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0093] According to an embodiment of the present invention, there is also provided an embodiment of a device for implementing the above-mentioned method for classifying causes of distribution network faults. Figure 6 is a schematic diagram of the structure of a device for classifying causes of distribution network faults according to an embodiment of the present invention. Figure 6 As shown, the above-mentioned distribution network fault cause classification device includes: an acquisition module 40, a determination module 42, a first processing module 44, a training module 46, and a second processing module 48, wherein:

[0094] An acquisition module 40 is used to acquire zero-sequence current waveform data of distribution network faults of different fault types;

[0095] A determination module 42 is used to determine a training set and a test set based on target waveform data selected from the zero-sequence current waveform data, wherein the target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the distribution network fault occurs;

[0096] A first processing module 44 is used to decompose the training set and the test set to obtain the intrinsic mode function components with local signal characteristics, and obtain the first eigenvector and the second eigenvector accordingly;

[0097] A training module 46 is used to normalize the first feature vector and input it into an initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network;

[0098] The second processing module 48 is used to normalize the second feature vector and input it into the trained fault cause classification model to obtain the fault cause classification result of the distribution network fault.

[0099] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0100] It should be noted that the acquisition module 40, the determination module 42, the first processing module 44, the training module 46, and the second processing module 48 correspond to steps S102 to S110 in Example 1, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the modules as part of the device can be run in a computer terminal.

[0101] Optionally, the above-mentioned determination module includes: a selection unit, used to select zero-sequence current waveform data within a first predetermined time period before the above-mentioned distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the above-mentioned distribution network fault occurs from the above-mentioned zero-sequence current waveform data; a first processing unit, used to perform data enhancement processing on the zero-sequence current waveform data within the above-mentioned first predetermined time period and the zero-sequence current waveform data within the above-mentioned second predetermined time period to obtain the above-mentioned training set and the above-mentioned test set.

[0102] Optionally, the above-mentioned first processing module includes: a second processing unit, used to use an empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned training set according to different time scales, so as to obtain multiple above-mentioned intrinsic mode function components; a third processing unit, used to use an empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned test set according to different time scales, so as to obtain multiple above-mentioned intrinsic mode function components; a combining unit, used to combine the above-mentioned multiple intrinsic mode function components corresponding to the above-mentioned training set to obtain a first eigenvector, and combine the above-mentioned multiple intrinsic mode function components corresponding to the above-mentioned test set to obtain a second eigenvector.

[0103] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in Example 1, which will not be repeated here.

[0104] The above-mentioned distribution network fault cause classification device may also include a processor and a memory. The above-mentioned acquisition module 40, determination module 42, first processing module 44, training module 46, second processing module 48, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0105] The processor includes a kernel, which retrieves the corresponding program unit from the memory. The kernel may be one or more. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one memory chip.

[0106] According to an embodiment of the present application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute any of the above-mentioned methods for classifying causes of distribution network failures.

[0107] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group, and the non-volatile storage medium includes a stored program.

[0108] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: obtain zero-sequence current waveform data of distribution network faults of different fault types; determine a training set and a test set based on target waveform data selected from the above zero-sequence current waveform data, wherein the above target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the occurrence of the above distribution network fault, and zero-sequence current waveform data within a second predetermined time period after the occurrence of the above distribution network fault; respectively decompose the above training set and the above test set to obtain intrinsic mode function components with local signal characteristics, and obtain first eigenvectors and second eigenvectors correspondingly; normalize the above first eigenvectors and input them into an initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the above initial fault cause classification model is constructed based on a long short-term memory network; normalize the above second eigenvectors and input them into the above trained fault cause classification model to obtain a fault cause classification result of the above distribution network fault.

[0109] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: select zero-sequence current waveform data within a first predetermined time period before the above-mentioned distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the above-mentioned distribution network fault occurs from the above-mentioned zero-sequence current waveform data; perform data enhancement processing on the zero-sequence current waveform data within the above-mentioned first predetermined time period and the zero-sequence current waveform data within the above-mentioned second predetermined time period to obtain the above-mentioned training set and the above-mentioned test set.

[0110] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: using an empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned training set according to different time scales to obtain multiple intrinsic mode function components; using an empirical mode decomposition algorithm to decompose and process each zero-sequence current waveform data in the above-mentioned test set according to different time scales to obtain multiple intrinsic mode function components; combining the multiple intrinsic mode function components corresponding to the above-mentioned training set to obtain a first eigenvector, and combining the multiple intrinsic mode function components corresponding to the above-mentioned test set to obtain a second eigenvector.

[0111] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: normalize the above-mentioned first feature vector and input it into the initial fault cause classification model for training to obtain the first output feature vector of the above-mentioned distribution network fault; input the above-mentioned first output feature vector into the classifier, and the above-mentioned classifier maps the above-mentioned first output feature vector into the first probability distribution value of each fault cause; use the above-mentioned first probability distribution value of each of the above-mentioned fault causes and the predetermined label distribution to construct a loss function to train the above-mentioned initial fault cause classification model until a predetermined stopping condition is met to obtain the above-mentioned trained fault cause classification model, wherein the above-mentioned predetermined stopping condition includes at least one of the following: the initial fault cause classification model converges and meets the predetermined number of training times.

[0112] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: normalize the above-mentioned second feature vector and input it into the above-mentioned trained fault cause classification model to obtain the second output feature vector of the above-mentioned distribution network fault; input the above-mentioned second output feature vector into a classifier, and the above-mentioned classifier maps the above-mentioned second output feature vector into a second probability distribution value of each fault cause; determine the maximum probability distribution value among the above-mentioned second probability distribution values ​​of each of the above-mentioned fault causes; and determine the corresponding fault cause classification results based on the above-mentioned maximum probability distribution value of each of the above-mentioned fault causes.

[0113] According to an embodiment of the present application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein when the program is run, any of the above-mentioned methods for classifying causes of power distribution network faults is executed.

[0114] According to an embodiment of the present application, an embodiment of a computer program product is also provided. When executed on a data processing device, the program is suitable for executing the steps of initializing the classification method of the causes of distribution network faults of any one of the above-mentioned methods.

[0115] Optionally, the above-mentioned computer program product, when executed on a data processing device, is suitable for executing a program initialized with the following method steps: obtaining zero-sequence current waveform data of distribution network faults of different fault types; determining a training set and a test set based on target waveform data selected from the above-mentioned zero-sequence current waveform data, wherein the above-mentioned target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the occurrence of the above-mentioned distribution network fault, and zero-sequence current waveform data within a second predetermined time period after the occurrence of the above-mentioned distribution network fault; decomposing the above-mentioned training set and the above-mentioned test set to obtain intrinsic mode function components with local signal characteristics, and correspondingly obtaining a first eigenvector and a second eigenvector; normalizing the above-mentioned first eigenvector and inputting it into an initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the above-mentioned initial fault cause classification model is constructed based on a long short-term memory network; normalizing the above-mentioned second eigenvector and inputting it into the above-mentioned trained fault cause classification model to obtain a fault cause classification result of the above-mentioned distribution network fault.

[0116] According to an embodiment of the present application, an embodiment of an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any one of the above-mentioned methods for classifying causes of distribution network faults.

[0117] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0118] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0119] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the above-mentioned units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0120] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0121] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0122] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a non-volatile storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0123] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for classifying causes of distribution network failures, characterized in that: include: Obtain zero-sequence current waveform data of distribution network faults of different fault types; Determine a training set and a test set based on target waveform data selected from the zero-sequence current waveform data, wherein the target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the distribution network fault occurs; Decomposing the training set and the test set to obtain intrinsic mode function components with local signal characteristics, respectively, to obtain a first eigenvector and a second eigenvector; Normalizing the first feature vector and inputting it into an initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network; Normalizing the second feature vector and inputting it into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault; Decomposing the training set and the test set to obtain a plurality of intrinsic mode function components with local signal characteristics, and obtaining a first eigenvector and a second eigenvector correspondingly, includes: Using an empirical mode decomposition algorithm, each zero-sequence current waveform data in the training set is decomposed according to different time scales to obtain multiple intrinsic mode function components; using an empirical mode decomposition algorithm, each zero-sequence current waveform data in the test set is decomposed according to different time scales to obtain multiple intrinsic mode function components; combining the multiple intrinsic mode function components corresponding to the training set to obtain a first eigenvector, and combining the multiple intrinsic mode function components corresponding to the test set to obtain a second eigenvector; wherein the intrinsic mode function component is expressed as: ; Among them, IMF j is the jth intrinsic mode function component after empirical mode decomposition; res is the residual component; n is the total number of decomposed intrinsic mode function components; The first feature vector is normalized and then input into the initial fault cause classification model for training to obtain a trained fault cause classification model, including: The first feature vector is normalized and then input into the initial fault cause classification model for training to obtain the first output feature vector of the distribution network fault; the first output feature vector is input into the classifier, and the classifier maps the first output feature vector into the first probability distribution value of each fault cause; the initial fault cause classification model is trained by using the first probability distribution value of each fault cause and a predetermined label distribution to construct a loss function until a predetermined stop condition is met to obtain the trained fault cause classification model, wherein the predetermined stop condition includes at least one of the following: the initial fault cause classification model converges and a predetermined number of training times is met; The second feature vector is normalized and then input into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault, including: The second feature vector is normalized and input into the trained fault cause classification model to obtain a second output feature vector of the distribution network fault; the second output feature vector is input into a classifier, and the classifier maps the second output feature vector into a second probability distribution value of each fault cause; the maximum probability distribution value among the second probability distribution values ​​of each fault cause is determined; and the corresponding fault cause classification results are determined based on the maximum probability distribution value of each fault cause.

2. The classification method according to claim 1, characterized in that: Based on target waveform data selected from the zero-sequence current waveform data, determining a training set and a test set includes: Selecting from the zero-sequence current waveform data the zero-sequence current waveform data within a first predetermined time period before the distribution network fault occurs, and the zero-sequence current waveform data within a second predetermined time period after the distribution network fault occurs; The zero-sequence current waveform data within the first predetermined time period and the zero-sequence current waveform data within the second predetermined time period are subjected to data enhancement processing to obtain the training set and the test set.

3. A device for classifying causes of distribution network faults, characterized in that: include: An acquisition module, used to acquire zero-sequence current waveform data of distribution network faults of different fault types; A determination module, configured to determine a training set and a test set based on target waveform data selected from the zero-sequence current waveform data, wherein the target waveform data includes: zero-sequence current waveform data within a first predetermined time period before the distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the distribution network fault occurs; A first processing module, configured to respectively decompose the training set and the test set to obtain intrinsic mode function components having local signal characteristics, and correspondingly obtain a first eigenvector and a second eigenvector; A training module, used for normalizing the first feature vector and inputting the first feature vector into an initial fault cause classification model for training to obtain a trained fault cause classification model, wherein the initial fault cause classification model is constructed based on a long short-term memory network; A second processing module, used for normalizing the second feature vector and inputting it into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault; The first processing module comprises: The second processing unit is used to use the empirical mode decomposition algorithm to decompose each zero-sequence current waveform data in the training set according to different time scales to obtain multiple intrinsic mode function components; the third processing unit is used to use the empirical mode decomposition algorithm to decompose each zero-sequence current waveform data in the test set according to different time scales to obtain multiple intrinsic mode function components; the combination unit is used to combine the multiple intrinsic mode function components corresponding to the training set to obtain a first eigenvector, and combine the multiple intrinsic mode function components corresponding to the test set to obtain a second eigenvector; wherein the intrinsic mode function component is expressed as: ; Among them, IMF j is the jth intrinsic mode function component after empirical mode decomposition; res is the residual component; n is the total number of decomposed intrinsic mode function components; The first feature vector is normalized and then input into the initial fault cause classification model for training to obtain the trained fault cause classification model, including: The first feature vector is normalized and then input into the initial fault cause classification model for training to obtain the first output feature vector of the distribution network fault; the first output feature vector is input into the classifier, and the classifier maps the first output feature vector into the first probability distribution value of each fault cause; the initial fault cause classification model is trained by using the first probability distribution value of each fault cause and a predetermined label distribution to construct a loss function until a predetermined stop condition is met to obtain the trained fault cause classification model, wherein the predetermined stop condition includes at least one of the following: the initial fault cause classification model converges and a predetermined number of training times is met; The second feature vector is normalized and then input into the trained fault cause classification model to obtain a fault cause classification result of the distribution network fault, including: The second feature vector is normalized and input into the trained fault cause classification model to obtain a second output feature vector of the distribution network fault; the second output feature vector is input into a classifier, and the classifier maps the second output feature vector into a second probability distribution value of each fault cause; the maximum probability distribution value among the second probability distribution values ​​of each fault cause is determined; and the corresponding fault cause classification results are determined based on the maximum probability distribution value of each fault cause.

4. The classification device according to claim 3, characterized in that: The determining module comprises: A selection unit, configured to select, from the zero-sequence current waveform data, zero-sequence current waveform data within a first predetermined time period before the distribution network fault occurs, and zero-sequence current waveform data within a second predetermined time period after the distribution network fault occurs; The first processing unit is used to perform data enhancement processing on the zero-sequence current waveform data within the first predetermined time period and the zero-sequence current waveform data within the second predetermined time period to obtain the training set and the test set.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method for classifying causes of distribution network faults as claimed in claim 1 or 2.

6. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to execute the method for classifying causes of distribution network faults as claimed in claim 1 or 2.

Images