Fault monitoring method and device of power grid loop, electronic equipment and storage medium

Abstract

The invention discloses a fault monitoring method and device for a power grid loop, electronic equipment and a storage medium. The method comprises the steps that real-time operation data of a power grid loop to be subjected to fault diagnosis is acquired, and the power grid loop at least comprises a power grid relay protection secondary loop; a pre-training fault diagnosis model is used for monitoring the power grid loop based on the real-time operation data to obtain a fault monitoring result, and the pre-training fault diagnosis model is used for representing a power grid fault diagnosis model which is pre-trained based on historical operation data of the power grid loop. The pre-trained fault diagnosis model at least comprises an asymmetric convolution module and a full connection layer, the asymmetric convolution module comprises an asymmetric convolution kernel, and the fault monitoring result at least comprises whether a power grid loop has a fault and a fault type under the fault condition. According to the invention, the technical problem of low efficiency of fault monitoring of the power grid loop in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of power grid fault monitoring, and in particular to a method, device, electronic equipment and storage medium for monitoring faults in a power grid loop. Background Art

[0002] Fault monitoring is performed on the power grid circuit, especially the secondary circuit of the power grid relay protection, that is, by monitoring the secondary circuit in the power grid relay protection system, to detect whether there is a fault or abnormality in it. This monitoring can help operators to find and locate faults in time, ensure the safe and stable operation of the power grid, and improve the reliability and safety of the power grid.

[0003] At present, the related technologies usually adopt the methods of regular inspection, fault recording analysis, etc. to monitor the fault of power grid circuits, and rely more on the experience of related technicians or experts, which is greatly affected by human factors, resulting in low efficiency of the related technologies in monitoring the fault of power grid circuits.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method, device, electronic device and storage medium for fault monitoring of a power grid loop, so as to at least solve the technical problem of low efficiency in fault monitoring of a power grid loop in the related art.

[0006] According to one aspect of an embodiment of the present invention, a fault monitoring method for a power grid loop is provided, comprising: obtaining real-time operating data of a power grid loop to be fault diagnosed, wherein the power grid loop includes at least a power grid relay protection secondary circuit; using a pre-trained fault diagnosis model to monitor the power grid loop based on the real-time operating data to obtain a fault monitoring result, wherein the pre-trained fault diagnosis model is used to represent a power grid fault diagnosis model pre-trained based on historical operating data of the power grid loop, the pre-trained fault diagnosis model includes at least an asymmetric convolution module and a fully connected layer, the output end of the asymmetric convolution module is connected to the input end of the fully connected layer, the asymmetric convolution module includes an asymmetric convolution kernel, and the fault monitoring result includes at least whether a fault occurs in the power grid loop and the fault type when a fault occurs.

[0007] Optionally, a pre-trained fault diagnosis model is used to monitor the power grid circuit based on real-time operation data to obtain a fault monitoring result, including: using an asymmetric convolution module in the pre-trained fault diagnosis model to extract features of the real-time operation data to obtain feature information; using a fully connected layer to integrate the feature information to obtain a fault monitoring result.

[0008] Optionally, the asymmetric convolution module includes at least one asymmetric convolution layer, an activation function layer and a splicing layer, the output end of at least one asymmetric convolution layer is connected to the input end of the activation function layer, and the output end of the activation function layer is connected to the input end of the splicing layer; using the asymmetric convolution module in the pre-trained fault diagnosis model, feature extraction is performed on the real-time operation data to obtain feature information, including: using at least one asymmetric convolution layer to extract features from the real-time operation data to obtain initial feature information; using the activation function layer to perform nonlinear transformation on the initial feature information to obtain transformed feature information; using the splicing layer to splice the transformed feature information to obtain feature information.

[0009] Optionally, the asymmetric convolution kernel is constructed based on the series connection of the first convolution kernel and the second convolution kernel, the first convolution kernel and the second convolution kernel are obtained by splitting the preset symmetric convolution kernel, and the receptive field of the asymmetric convolution kernel is the same as the receptive field of the preset symmetric convolution kernel.

[0010] Optionally, the fully connected layer includes at least a capsule fully connected layer, a capsule pooling layer and a global average pooling layer, the output end of the capsule fully connected layer is connected to the input end of the capsule pooling layer, and the output end of the capsule pooling layer is connected to the input end of the global average pooling layer; the feature information is integrated using the fully connected layer to obtain a fault monitoring result, including: using the capsule fully connected layer to perform linear transformation on the feature information to obtain a capsule vector; using the capsule pooling layer to aggregate the capsule vector to obtain an aggregated capsule vector; using the global average pooling layer to average pool the aggregated capsule vector to obtain a fault monitoring result.

[0011] Optionally, the historical operation data includes at least historical status operation information and historical warning information of the power grid loop, the historical status operation information includes at least software and hardware self-detection information and sampling information of the power grid loop, and the historical warning information includes at least maintenance information and disconnection information of the power grid loop.

[0012] According to another aspect of an embodiment of the present invention, a fault monitoring device for a power grid loop is also provided, including: an acquisition module, used to acquire real-time operation data of a power grid loop to be fault diagnosed, wherein the power grid loop includes at least a power grid relay protection secondary circuit; a monitoring module, used to monitor the power grid loop based on the real-time operation data using a pre-trained fault diagnosis model to obtain a fault monitoring result, wherein the pre-trained fault diagnosis model is used to represent a power grid fault diagnosis model pre-trained based on historical operation data of the power grid loop, the pre-trained fault diagnosis model includes at least an asymmetric convolution module and a fully connected layer, the output end of the asymmetric convolution module is connected to the input end of the fully connected layer, the asymmetric convolution module includes an asymmetric convolution kernel, and the fault monitoring result includes at least whether a fault occurs in the power grid loop and the fault type when a fault occurs.

[0013] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0014] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0015] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0016] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0017] According to another aspect of the embodiments of the present invention, a computer program is further provided. When the computer program is executed by a processor, the methods in the embodiments of the present invention are implemented.

[0018] In an embodiment of the present invention, a method for fault monitoring of a power grid loop is provided, comprising: obtaining real-time operating data of a power grid loop to be fault diagnosed, wherein the power grid loop includes at least a power grid relay protection secondary circuit; using a pre-trained fault diagnosis model to monitor the power grid loop based on the real-time operating data to obtain a fault monitoring result, wherein the pre-trained fault diagnosis model is used to represent a power grid fault diagnosis model pre-trained based on historical operating data of the power grid loop, the pre-trained fault diagnosis model includes at least an asymmetric convolution module and a fully connected layer, the output end of the asymmetric convolution module is connected to the input end of the fully connected layer, the asymmetric convolution module includes an asymmetric convolution kernel, and the fault monitoring result includes at least whether a fault occurs in the power grid loop and the type of fault when a fault occurs. It is easy to notice that

[0019] The present application analyzes the real-time data, and the monitoring system can timely diagnose whether there is a fault in the power grid circuit. In particular, for the secondary circuit of the power grid relay protection, the monitoring system can detect the fault conditions in the secondary circuit, including short circuit, open circuit and other problems. The pre-trained model used includes an asymmetric convolution module and a fully connected layer. The use of the asymmetric convolution module can better capture the asymmetric features in the power grid circuit, improve the sensitivity and accuracy of the model to faults, and the asymmetric convolution module can better identify and distinguish these fault modes. The use of the fully connected layer can integrate and classify the features extracted by the asymmetric convolution module to obtain more detailed and accurate fault monitoring results. The fully connected layer can learn the correlation between different features, improve the model's ability to judge the power grid circuit fault, and this structure can better adapt to the characteristics and fault types of different power grid circuits. The model has learned various possible fault modes during the training process, so it can adapt to the monitoring needs in different situations, thereby solving the technical problem of low efficiency in fault monitoring of power grid circuits in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] 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:

[0021] Figure 1 is a flow chart of a fault monitoring method for a power grid loop according to an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of an optional fault monitoring process of a power grid loop according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of an optional accuracy and loss value of a pre-trained fault diagnosis model for testing according to an embodiment of the present invention;

[0024] Figure 4 is a schematic diagram of a fault monitoring device for a power grid loop 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 can be explained 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 clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] According to an embodiment of the present invention, an embodiment of a fault monitoring method for a power grid loop is provided. It can be explained 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.

[0028] Figure 1 is a flow chart of a fault monitoring method for a power grid loop according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0029] Step S102, obtaining real-time operation data of the power grid loop to be diagnosed for fault.

[0030] Wherein, the power grid circuit at least includes a power grid relay protection secondary circuit.

[0031] In an optional embodiment, the real-time operation data of the power grid loop to be diagnosed for faults can be obtained. First, data acquisition equipment can be installed on the power grid loop. These equipment can be sensors, smart meters, monitoring devices, etc., which can collect parameter data such as current, voltage, frequency, etc. of the power grid loop in real time. The collected real-time data can be transmitted to the monitoring system through the communication network for processing, and wired or wireless communication can be used for data transmission to ensure that the data can reach the monitoring system in a timely and accurate manner. After receiving the real-time data, the monitoring system can process and analyze the data. Data mining, machine learning and other technologies can be used to monitor and diagnose the real-time data of the power grid loop to find abnormal conditions of the power grid loop. Through the analysis of real-time data, the monitoring system can timely diagnose whether there is a fault in the power grid loop. In particular, for the secondary circuit of the power grid relay protection, the monitoring system can detect faults in the secondary circuit, including short circuits, open circuits and other problems. Through real-time monitoring and diagnosis, the occurrence of faults in the power grid loop can be prevented in advance, maintenance and repair can be carried out in time, power outage time can be reduced, and the operation efficiency and reliability of the power grid can be improved.

[0032] Step S104, using the pre-trained fault diagnosis model, monitor the power grid loop based on the real-time operation data to obtain a fault monitoring result.

[0033] Among them, the pre-trained fault diagnosis model is used to represent the power grid fault diagnosis model that is pre-trained based on the historical operation data of the power grid loop. The pre-trained fault diagnosis model includes at least an asymmetric convolution module and a fully connected layer. The output end of the asymmetric convolution module is connected to the input end of the fully connected layer. The asymmetric convolution module includes an asymmetric convolution kernel. The fault monitoring result includes at least whether a fault occurs in the power grid loop and the fault type when a fault occurs.

[0034] In an optional embodiment, the pre-trained fault diagnosis model can be trained in advance using the sorted historical operation data. During the training process, a deep learning algorithm, such as a convolutional neural network, can be used to learn the characteristics and fault modes of the power grid loop. In the training of the pre-trained fault diagnosis model, the complex characteristics of the power grid loop can be better captured by combining the asymmetric convolution module and the fully connected layer. When the power grid loop is in actual operation, the real-time monitoring data is input into the pre-trained fault diagnosis model, and the fault monitoring result is obtained by model calculation. According to the monitoring results, it can be determined whether the power grid loop has a fault and the specific type of the fault. In the above steps, the use of the pre-trained fault diagnosis model for monitoring can effectively improve the accuracy and efficiency of fault monitoring. The pre-trained model has learned the characteristics and fault modes of the power grid loop through historical data, so it can make accurate judgments more quickly in real-time monitoring. The pre-trained model includes an asymmetric convolution module and a fully connected layer. The use of the asymmetric convolution module can better capture the asymmetric characteristics in the power grid loop, improve the sensitivity and accuracy of the model to faults, and the asymmetric convolution module can better identify and distinguish these fault modes. The use of the fully connected layer can integrate and classify the features extracted by the asymmetric convolution module to obtain more detailed and accurate fault monitoring results. The fully connected layer can learn the correlation between different features, which improves the model's ability to judge power circuit faults. This structure can better adapt to the characteristics and fault types of different power circuits. The model has learned various possible fault modes during the training process, so it can adapt to monitoring needs in different situations.

[0035] The present application analyzes the real-time data, and the monitoring system can timely diagnose whether there is a fault in the power grid circuit. In particular, for the secondary circuit of the power grid relay protection, the monitoring system can detect the fault conditions in the secondary circuit, including short circuit, open circuit and other problems. The pre-trained model used includes an asymmetric convolution module and a fully connected layer. The use of the asymmetric convolution module can better capture the asymmetric features in the power grid circuit, improve the sensitivity and accuracy of the model to faults, and the asymmetric convolution module can better identify and distinguish these fault modes. The use of the fully connected layer can integrate and classify the features extracted by the asymmetric convolution module to obtain more detailed and accurate fault monitoring results. The fully connected layer can learn the correlation between different features, improve the model's ability to judge the power grid circuit fault, and this structure can better adapt to the characteristics and fault types of different power grid circuits. The model has learned various possible fault modes during the training process, so it can adapt to the monitoring needs in different situations, thereby solving the technical problem of low efficiency in fault monitoring of power grid circuits in related technologies.

[0036] Optionally, a pre-trained fault diagnosis model is used to monitor the power grid circuit based on real-time operation data to obtain a fault monitoring result, including: using an asymmetric convolution module in the pre-trained fault diagnosis model to extract features of the real-time operation data to obtain feature information; using a fully connected layer to integrate the feature information to obtain a fault monitoring result.

[0037] In an optional embodiment, in the fault monitoring process of the power grid loop, especially for the secondary circuit of the power grid relay protection, the real-time operation data is monitored using a pre-trained fault diagnosis model to improve the accuracy and efficiency of fault diagnosis, reduce the need for manual intervention, and improve the operation stability and reliability of the power grid relay protection system. First, the asymmetric convolution module in the pre-trained fault diagnosis model is used to extract features of the real-time operation data. The asymmetric convolution module can capture asymmetric features in the data and improve the accuracy of feature extraction. By performing a convolution operation on the real-time operation data, a feature map with rich information can be obtained, and these features can reflect the fault conditions existing in the power grid loop. Secondly, the feature information is integrated using a fully connected layer to obtain a fault monitoring result. The fully connected layer can integrate the feature map obtained by the convolution operation and extract a more advanced feature representation. Through the processing of the fully connected layer, the features of different levels can be fused to obtain a more comprehensive and accurate fault monitoring result. The accuracy and efficiency of fault diagnosis can be improved, the need for manual intervention can be reduced, and the operation stability and reliability of the power grid relay protection system can be improved.

[0038] Optionally, the asymmetric convolution module includes at least one asymmetric convolution layer, an activation function layer and a splicing layer, the output end of at least one asymmetric convolution layer is connected to the input end of the activation function layer, and the output end of the activation function layer is connected to the input end of the splicing layer; using the asymmetric convolution module in the pre-trained fault diagnosis model, feature extraction is performed on the real-time operation data to obtain feature information, including: using at least one asymmetric convolution layer to extract features from the real-time operation data to obtain initial feature information; using the activation function layer to perform nonlinear transformation on the initial feature information to obtain transformed feature information; using the splicing layer to splice the transformed feature information to obtain feature information.

[0039] In an optional embodiment, the asymmetric convolution module includes at least one asymmetric convolution layer, an activation function layer and a splicing layer. The asymmetric convolution layer can capture feature information of different directions and scales, the activation function layer can introduce nonlinear transformation, and the splicing layer can integrate feature information of different levels, thereby improving the representation ability of the feature. In the actual implementation process, the asymmetric convolution module in the pre-trained fault diagnosis model can first be used to extract features of the real-time operation data. Specifically, at least one asymmetric convolution layer can be used to extract features of the real-time operation data to obtain initial feature information. The asymmetric convolution layer can capture feature information of different directions and scales, thereby improving the diversity and representation ability of the features. The activation function layer can be used to perform nonlinear transformation on the initial feature information to obtain transformed feature information. Nonlinear transformation can introduce nonlinear factors to improve the distinguishability and representation ability of the features. The splicing layer can be used to splice the transformed feature information to obtain the final feature information. The splicing layer can integrate feature information of different levels to improve the comprehensive representation ability of the features. Through the above steps, the feature information of the real-time operation data can be obtained for subsequent fault monitoring and diagnosis. The advantage of using an asymmetric convolution module for feature extraction is that it can make full use of the diversity and nonlinear information of the data, improve the characterization ability and discrimination of the features, and thus improve the accuracy and efficiency of fault monitoring.

[0040] Optionally, the asymmetric convolution kernel is constructed based on the series connection of the first convolution kernel and the second convolution kernel, the first convolution kernel and the second convolution kernel are obtained by splitting the preset symmetric convolution kernel, and the receptive field of the asymmetric convolution kernel is the same as the receptive field of the preset symmetric convolution kernel.

[0041] In an optional embodiment, the application of an asymmetric convolution kernel can improve the generalization ability and accuracy of the model. The asymmetric convolution kernel is constructed based on the first convolution kernel and the second convolution kernel connected in series, and the first convolution kernel and the second convolution kernel are obtained by splitting the preset symmetric convolution kernel. The network can better capture the characteristics of the input data, and the number of parameters of the model can be reduced, thereby improving the computational efficiency. Specifically, a preset symmetric convolution kernel can be defined first, and then split into two different convolution kernels, namely the first convolution kernel and the second convolution kernel. Then, the two convolution kernels are connected in series to form an asymmetric convolution kernel. During the model training process, the parameters of the convolution kernel are updated by the back propagation algorithm, so that the network can learn more complex feature representations. The asymmetric convolution kernel finally obtained can be applied to the convolution layer of the network to extract the characteristics of the input data. The receptive field of the asymmetric convolution kernel is the same as the receptive field of the preset symmetric convolution kernel, that is, the network will not lose the ability to perceive the information of the input data in the process of learning the features. The design of asymmetric convolution kernel can increase the flexibility of the network, enable the network to better adapt to different types of input data, and improve the generalization ability of the model. In addition, by reducing the number of model parameters, asymmetric convolution kernel can also reduce the complexity of the model, reduce the computational cost, and improve the training and reasoning speed of the model. It can bring better results in fault monitoring of secondary circuits of power grid relay protection. By extracting features more effectively, the accuracy and generalization ability of the model are improved, while the computational cost is reduced and the efficiency of the model is improved.

[0042] Optionally, the fully connected layer includes at least a capsule fully connected layer, a capsule pooling layer and a global average pooling layer, the output end of the capsule fully connected layer is connected to the input end of the capsule pooling layer, and the output end of the capsule pooling layer is connected to the input end of the global average pooling layer; the feature information is integrated using the fully connected layer to obtain a fault monitoring result, including: using the capsule fully connected layer to perform linear transformation on the feature information to obtain a capsule vector; using the capsule pooling layer to aggregate the capsule vector to obtain an aggregated capsule vector; using the global average pooling layer to average pool the aggregated capsule vector to obtain a fault monitoring result.

[0043] In an optional embodiment, a fully connected layer structure can be used for feature integration and fault monitoring. The fully connected layer realizes information transmission and integration between different layers by linearly transforming the input feature information. In the fault monitoring of the secondary circuit of the power grid relay protection, the design of the fully connected layer includes a capsule fully connected layer, a capsule pooling layer, and a global average pooling layer. First, the capsule fully connected layer linearly transforms the input feature information to obtain a capsule vector. The capsule vector contains feature information at different levels and can better describe the characteristics of the input data. Next, the capsule pooling layer aggregates the capsule vector, integrates and extracts feature information at different levels, and obtains an aggregated capsule vector. Through the aggregation operation of the capsule pooling layer, the dimension of the feature information can be effectively reduced, the calculation efficiency can be improved, and the risk of overfitting can be reduced. Finally, the global average pooling layer performs an average pooling operation on the aggregated capsule vector to obtain the final fault monitoring result. The global average pooling operation can average the weights of different features to obtain a more stable and reliable monitoring result. Through the design and operation process of the above fully connected layer, fault monitoring of the secondary circuit of the power grid relay protection can be achieved. The operations of integrating capsule vectors and aggregating capsule vectors can better extract and integrate feature information at different levels, improving the accuracy and robustness of monitoring. At the same time, the design of the fully connected layer can also reduce the dimension and complexity of feature information, improve computational efficiency and reduce the risk of overfitting.

[0044] Optionally, the historical operation data includes at least historical status operation information and historical warning information of the power grid loop, the historical status operation information includes at least software and hardware self-detection information and sampling information of the power grid loop, and the historical warning information includes at least maintenance information and disconnection information of the power grid loop.

[0045] In an optional embodiment, the accuracy and stability of the model can be improved by pre-training the power grid fault diagnosis model using historical operation data. In actual operation, the historical operation data of the power grid loop can be collected and sorted first, which includes the historical state operation information and historical warning information of the power grid loop. The historical state operation information may include the software and hardware self-detection information and sampling information of the power grid loop, which can reflect important parameters such as the operation status and equipment health status of the power grid loop. The historical warning information includes the maintenance information and disconnection information of the power grid loop, which can guide the model to predict and diagnose possible faults. It is convenient to train the historical operation data by using machine learning or deep learning algorithms to build a pre-trained fault diagnosis model. Through the learning and training of historical operation data, the model can better understand the characteristics and laws of the power grid loop, thereby improving the accuracy and precision of diagnosis.

[0046] The technical solution proposed in this application is described below in combination with an optional embodiment. This application proposes a method and system for diagnosing secondary circuit faults of power grid relay protection based on capsule network. With the continuous increase in the scale of power grid construction, it is necessary to build an optimized control model of the secondary circuit of power grid relay protection, and combine fault fusion and feature detection technology to control the power grid relay protection. In the secondary circuit of relay protection, the output stability of the secondary circuit of power grid relay protection is poor and prone to failure due to the influence of environmental working condition information. It is necessary to build an optimized secondary circuit fault monitoring system of power grid relay protection, and control and monitor the secondary circuit of power grid relay protection through fault state information fusion and feature optimization extraction technology, extract state parameters under fault conditions, and combine big data analysis technology to realize fault state parameter fusion. The research on the fault monitoring system of secondary circuit of power grid relay protection is of great significance in improving the output stability and reliability of substations. The purpose of this application is to provide a method and system for diagnosing secondary circuit faults of power grid relay protection based on capsule network, so as to solve the diagnosis of fault signals under strong noise and variable working conditions by capsule network, and realize intelligent monitoring of secondary circuit faults. The technical solution adopted is as follows: train the historical operation data of the relay protection secondary circuit to obtain a power grid relay protection fault diagnosis model based on capsule network; collect the operation data of the relay protection secondary circuit in real time; use the trained relay protection fault diagnosis model to monitor the secondary circuit operation data in real time to identify whether there is a fault; once the system detects fault information, use the relay protection fault diagnosis model to extract and classify fault features, and output the identification results.

[0047] In this embodiment, the operation history data of the relay protection secondary circuit may include "status operation information" and "warning information". The status operation information includes software and hardware self-detection information, sampling information, etc. Take the software and hardware self-detection information as an example: when the relay protection secondary circuit is running, the software and hardware run continuously, which will inevitably generate a certain amount of heat. If the operating temperature is too high, the risk of failure will be significantly increased. The significance of collecting hardware and software self-detection information is not only to discover the above problems, but also to prevent failures in time. The system needs to monitor early warning information, mainly maintenance information and line break information. In order to effectively distinguish various information, the system divides the operation status of the secondary circuit into three types: steady state, transient state and fault state.

[0048] The present application uses an improved multi-scale asymmetric convolution module and an improved fully connected capsule layer, and proposes a new capsule pooling to extract features from the input data, including: the improved multi-scale asymmetric convolution module is based on the criss-cross network (Inception) structure. Compared with the convolution of the ordinary symmetric k×k convolution kernel, the asymmetric convolution is to decompose a k×k convolution into a k×1 convolution and then connect a 1×k convolution in series. The receptive fields of the two are the same, but the asymmetric convolution can effectively reduce the number of parameters and the amount of calculation, because multiple size-compatible two-dimensional kernels operate on the same input with the same step size, output the same resolution, and calculate the convolution of the corresponding positions of the output and the asymmetric convolution. The kernels can be added to obtain equivalent kernels with the same output. In the first layer, 1×1, 3×1 and 1×3 are connected in series, and 5×1 and 1×5 are connected in series, and parallel convolution layers extract features of different scales for the input fault data, and the number of channels is set to 16, 8, and 8; in the second layer, the 5×5 and 3×3 convolutions are equivalently decomposed into 5×1 and 1×5, 3×1 and 1×3 asymmetric convolutions in series in the first layer, and the number of channels is 16; the third layer uses two 1×1 convolution layers with 32 channels. In order to increase the nonlinear expression ability of the model, batch normalization and ReLU activation function are used after each convolution layer, and then the feature dimensions of different branches are stacked and spliced ​​through the splicing layer.

[0049] The capsule fully connected layer is composed of sub-capsules, which retain the feature information extracted by the parent capsule in the upper layer. When the model is compressed and output at the end, the capsules in the lower layer are flattened into a capsule list and sent to the fully connected capsule layer. Each sub-capsule in the capsule layer is multiplied by the transformation matrix W. d , routed by agreement to produce the final capsule v for each class j and its probability a j , because the protocol routing uses a function (Softmax) to generate the logarithmic prior probability b between capsules ij , so this paper uses the activation function (ReLU) to perform a recompression inside the fully connected layer, for each final capsule v j Decoding is performed and the function (Softmax) inside the improved capsule fully connected layer is overused, resulting in a numerical overflow problem. Therefore, an activation function (ReLU) is used inside the capsule to perform unilateral suppression. While enhancing the computational efficiency, it does not lose too much feature information like the traditional front-connected layer. Because each capsule retains the feature vector as much as possible, when the feature information extracted by each digital capsule layer is connected, a high amount of information is still maintained. Finally, the activation function (ReLU) is used to suppress overfitting, which can improve the classification accuracy of the model.

[0050] Capsule pooling can reduce the number of capsules. The capsule with the largest weight value is selected at the corresponding position of each feature map in the capsule attention layer, redundant capsules are removed, and important capsules related to classification are found, which effectively reduces the computational overhead. The sub-capsule layer contains H×W×C capsules, where K is the number of capsule types; H is the height of the capsule layer U; W is the width of the capsule layer U. There is a capsule at each pixel point (x, y) in the capsule layer. There are K categories of capsules in total. The set of capsules is as follows:

[0051] U(x,y)={u 1|(x,y) ,u 2|(x,y) ,…,u K|(x,y)};

[0052] In the formula, b represents the training batch size; O = H × W; D = D in represents the primary capsule dimension; capsule pooling is performed on (x, y) at K category capsules, and the capsule with the largest weight value among the K is selected. The final capsule layer after pooling is There are a total of O capsules, as follows:

[0053]

[0054] The number of capsules after the capsule pooling layer is reduced by 1 / K compared to before pooling. It is the most important capsule at the corresponding position in the K-group capsule layer;

[0055] The output features of the capsule network are processed using global average pooling, and the data features are as follows:

[0056] f pooling =[z1,z2,…,z m ] T ;

[0057] In the formula, f pooling represents the output value of the global average pooling layer, z m represents the mean value of each feature data, and m is the number of fault types. The classification function divides the output value of the global average pooling layer into fault types, converts the output value into a probability distribution in the range [0,1] with a sum of 1, and takes the category corresponding to the maximum probability as the output result. The calculation process is as follows:

[0058]

[0059] In the formula, p s represents the classification function and e represents the exponential function.

[0060] The present application also proposes a capsule network-based power grid relay protection secondary circuit fault diagnosis system, which supports the above-mentioned capsule network-based power grid relay protection secondary circuit fault diagnosis method, and the system includes: a model training unit, which is used to train the relay protection secondary circuit operation history data to obtain a capsule network-based power grid relay protection fault diagnosis model; a data acquisition unit, which is used to collect the relay protection secondary circuit operation data in real time; a processing unit, which is used to use the capsule network-based power grid relay protection secondary circuit fault diagnosis model obtained in the model training unit to monitor the secondary circuit operation data in real time; an identification unit, which is used to identify whether the secondary circuit operation data monitored in real time in the processing unit has a fault, and if there is a fault, the relay protection fault diagnosis model is used to extract and classify the fault features; an output and display unit, which outputs the secondary circuit fault type output by the identification unit and displays it.

[0061] The technical solution proposed in this application can effectively extract features from noisy secondary circuit operation data. Compared with other classic deep learning models, the power grid relay protection secondary circuit fault diagnosis model based on capsule network enables the capsule to avoid the loss of feature information in space when outputting vector feature information. On the one hand, it is conducive to the centralized processing of faults, and on the other hand, it can effectively reduce the scope of power distribution interruption and the frequency of power outages for users during faults, thereby effectively improving the social benefits of the use of power systems and having high application value.

[0062] Figure 2 is a schematic diagram of an optional fault monitoring process of a power grid loop according to an embodiment of the present invention, such as Figure 2 As shown, it includes starting, acquiring operation data, performing feature extraction, feature processing and processing based on the capsule network power grid relay protection fault diagnosis model, judging whether a fault occurs, and when it is judged that no fault occurs, continuing to acquire operation data for monitoring, and when it is judged that a fault occurs, performing fault classification, and ending.

[0063] Figure 3 is a schematic diagram of an optional accuracy and loss value of a pre-trained fault diagnosis model for testing according to an embodiment of the present invention, such as Figure 3 As shown in the figure, the horizontal axis represents the number of iterations, the vertical axis on the left represents the accuracy, and the vertical axis on the right represents the loss value. The curve at the top of the figure represents the accuracy curve, and the curve at the bottom of the figure represents the loss value curve. The figure records the development trend of the loss value and the test accuracy of the power grid relay protection secondary circuit fault diagnosis model based on capsule network during the training and testing process. It can be seen that the loss value gradually decreases and the test accuracy gradually increases.

[0064] According to another aspect of an embodiment of the present application, a fault monitoring device for a power grid loop is also provided, which can execute the fault monitoring method for a power grid loop of the above embodiment. The specific implementation method and preferred application scenario are the same as those of the above embodiment and will not be repeated here.

[0065] Figure 4 is a schematic diagram of a fault monitoring device for a power grid loop according to an embodiment of the present application, such as Figure 4 As shown, the device includes the following: an acquisition module 402 and a monitoring module 404 .

[0066] Among them, the acquisition module is used to obtain real-time operation data of the power grid circuit to be fault diagnosed, wherein the power grid circuit at least includes a power grid relay protection secondary circuit; the monitoring module is used to use a pre-trained fault diagnosis model to monitor the power grid circuit based on the real-time operation data to obtain a fault monitoring result, wherein the pre-trained fault diagnosis model is used to represent a power grid fault diagnosis model pre-trained based on the historical operation data of the power grid circuit, the pre-trained fault diagnosis model includes at least an asymmetric convolution module and a fully connected layer, the output end of the asymmetric convolution module is connected to the input end of the fully connected layer, the asymmetric convolution module includes an asymmetric convolution kernel, and the fault monitoring result includes at least whether a fault occurs in the power grid circuit and the fault type when a fault occurs.

[0067] Among them, the monitoring module is also used to use the asymmetric convolution module in the pre-trained fault diagnosis model to extract features of real-time operation data to obtain feature information; and use the fully connected layer to integrate the feature information to obtain fault monitoring results.

[0068] Among them, the asymmetric convolution module includes at least one asymmetric convolution layer, an activation function layer and a splicing layer, the output end of at least one asymmetric convolution layer is connected to the input end of the activation function layer, and the output end of the activation function layer is connected to the input end of the splicing layer; the monitoring module is also used to use at least one asymmetric convolution layer to extract features of real-time operation data to obtain initial feature information; use the activation function layer to perform nonlinear transformation on the initial feature information to obtain transformed feature information; use the splicing layer to splice the transformed feature information to obtain feature information.

[0069] Among them, the asymmetric convolution kernel is constructed based on the series connection of the first convolution kernel and the second convolution kernel. The first convolution kernel and the second convolution kernel are obtained by splitting the preset symmetric convolution kernel. The receptive field of the asymmetric convolution kernel is the same as the receptive field of the preset symmetric convolution kernel.

[0070] Among them, the fully connected layer at least includes a capsule fully connected layer, a capsule pooling layer and a global average pooling layer, the output end of the capsule fully connected layer is connected to the input end of the capsule pooling layer, and the output end of the capsule pooling layer is connected to the input end of the global average pooling layer; wherein the monitoring module is also used to use the capsule fully connected layer to perform linear transformation on the feature information to obtain a capsule vector; use the capsule pooling layer to aggregate the capsule vector to obtain an aggregated capsule vector; use the global average pooling layer to average pool the aggregated capsule vector to obtain a fault monitoring result.

[0071] Among them, the historical operation data at least includes the historical status operation information and historical warning information of the power grid loop, the historical status operation information at least includes the software and hardware self-detection information and sampling information of the power grid loop, and the historical warning information at least includes the maintenance information and disconnection information of the power grid loop.

[0072] An embodiment of the present application further provides an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of the present invention when running.

[0073] The above-mentioned memory may refer to a device inside a computer for storing data and programs, and may include memory, hard disk, etc., wherein the memory may be used to temporarily store running programs and data, the hard disk may be used to store programs and data for a long time, and the memory may be used to enable the computer to read and write data, and execute programs; the above-mentioned processor may be responsible for executing instructions in computer programs and performing data processing, and may be responsible for controlling and executing various operations, including arithmetic operations, logical operations, data transmission, etc.

[0074] An embodiment of the present application further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0075] The above-mentioned computer storage medium may refer to a medium in a computer memory used to store certain discontinuous physical quantities. Computer storage media mainly include semiconductors, magnetic cores, magnetic drums, magnetic tapes, laser disks, etc.; the stored program included in the computer-readable storage medium may be a set of instructions that can be recognized and executed by a computer, running on an electronic computer, and is an information tool that meets certain needs of people.

[0076] An embodiment of the present application further provides a computer program product, including a computer program, which implements the methods in various embodiments of the present invention when executed by a processor.

[0077] The above-mentioned computer program product may refer to a software program that has been written, tested and released and can be run on a computer or other device. The computer program product may include an application, an operating system, tool software, etc., which is used to implement specific functions or solve specific problems.

[0078] An embodiment of the present application further provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0079] The above-mentioned non-volatile computer-readable storage medium may refer to a medium for storing data. The non-volatile computer-readable storage medium can keep the data from being lost when the power is off, and can be used to store long-term data, such as operating systems, applications, and user files. The non-volatile storage medium may include hard disk drives, solid-state drives, optical disks, and flash memory storage devices, etc.

[0080] The embodiments of the present application further provide a computer program, which implements the methods in the above-mentioned embodiments of the present invention when executed by a processor.

[0081] The above-mentioned computer program may refer to a collection of instructions used to tell a computer to perform a specific task or operation. A computer program may be written by a programmer using a specific programming language and may include algorithms, data structures, logic, and control flows. A computer program may be used for a variety of purposes, including application software, operating systems, and the like.

[0082] 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.

[0083] 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 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.

[0084] 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.

[0085] 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.

[0086] 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 storage medium. Based on this understanding, the technical solution of the present invention, in essence, 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 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 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.

[0087] 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 monitoring faults in a power grid loop, characterized in that: include: Acquiring real-time operating data of a power grid circuit to be diagnosed for fault, wherein the power grid circuit at least includes a power grid relay protection secondary circuit; The pre-trained fault diagnosis model is used to monitor the power grid loop based on the real-time operation data to obtain a fault monitoring result, wherein the pre-trained fault diagnosis model is used to represent a power grid fault diagnosis model pre-trained based on the historical operation data of the power grid loop, the pre-trained fault diagnosis model includes at least an asymmetric convolution module and a fully connected layer, the output end of the asymmetric convolution module is connected to the input end of the fully connected layer, the asymmetric convolution module includes an asymmetric convolution kernel, and the fault monitoring result includes at least whether the power grid loop fails and the fault type when a fault occurs.

2. The method for monitoring faults in a power grid loop according to claim 1, characterized in that: The pre-trained fault diagnosis model is used to monitor the power grid loop based on the real-time operation data to obtain fault monitoring results, including: Using the asymmetric convolution module in the pre-trained fault diagnosis model, feature extraction is performed on the real-time operation data to obtain feature information; The feature information is integrated using the fully connected layer to obtain the fault monitoring result.

3. The fault monitoring method of the power grid loop according to claim 2, characterized in that: The asymmetric convolution module includes at least one asymmetric convolution layer, an activation function layer and a splicing layer, the output end of the at least one asymmetric convolution layer is connected to the input end of the activation function layer, and the output end of the activation function layer is connected to the input end of the splicing layer; The asymmetric convolution module in the pre-trained fault diagnosis model is used to extract features from the real-time operation data to obtain feature information, including: Using the at least one asymmetric convolutional layer to extract features from the real-time operation data to obtain initial feature information; Using the activation function layer to perform nonlinear transformation on the initial feature information to obtain transformed feature information; The transformed feature information is spliced ​​using the splicing layer to obtain the feature information.

4. The fault monitoring method of the power grid loop according to claim 1, characterized in that: The asymmetric convolution kernel is constructed based on the series connection of the first convolution kernel and the second convolution kernel, the first convolution kernel and the second convolution kernel are obtained by splitting the preset symmetric convolution kernel, and the receptive field of the asymmetric convolution kernel is the same as the receptive field of the preset symmetric convolution kernel.

5. The method for monitoring faults in a power grid circuit according to claim 2, characterized in that: The fully connected layer at least includes a capsule fully connected layer, a capsule pooling layer and a global average pooling layer, the output end of the capsule fully connected layer is connected to the input end of the capsule pooling layer, and the output end of the capsule pooling layer is connected to the input end of the global average pooling layer; The feature information is integrated by using the fully connected layer to obtain the fault monitoring result, including: Using the capsule fully connected layer to perform linear transformation on the feature information to obtain a capsule vector; Aggregating the capsule vectors using the capsule pooling layer to obtain an aggregated capsule vector; The aggregated capsule vector is average pooled using the global average pooling layer to obtain the fault monitoring result.

6. The method for monitoring faults in a power grid circuit according to claim 1, characterized in that: The historical operation data includes at least the historical status operation information and historical warning information of the power grid loop, the historical status operation information includes at least the software and hardware self-detection information and sampling information of the power grid loop, and the historical warning information includes at least the maintenance information and disconnection information of the power grid loop.

7. A fault monitoring device for a power grid circuit, characterized in that: include: An acquisition module, used for acquiring real-time operation data of a power grid loop to be diagnosed for fault, wherein the power grid loop at least includes a power grid relay protection secondary loop; A monitoring module is used to use a pre-trained fault diagnosis model to monitor the power grid loop based on the real-time operation data to obtain a fault monitoring result, wherein the pre-trained fault diagnosis model is used to represent a power grid fault diagnosis model pre-trained based on the historical operation data of the power grid loop, the pre-trained fault diagnosis model at least includes an asymmetric convolution module and a fully connected layer, the output end of the asymmetric convolution module is connected to the input end of the fully connected layer, the asymmetric convolution module includes an asymmetric convolution kernel, and the fault monitoring result at least includes whether the power grid loop has a fault and the fault type when a fault occurs.

8. An electronic device, characterized in that: include: A memory storing an executable program; A processor is used to run the program, wherein the program, when running, executes the fault monitoring method for the power grid loop as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the storage medium is located is controlled to execute the fault monitoring method for the power grid loop according to any one of claims 1 to 6.

10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the fault monitoring method for a power grid loop according to any one of claims 1 to 6.

Images