Early fault detection method and device for complex power system and electronic equipment

Abstract

The invention provides an early fault detection method and device for a complex power system and electronic equipment. The method comprises the following steps: acquiring an early fault weak signal in the complex power system; performing signal processing on the early fault weak signal to obtain a plurality of IMF components and residual components corresponding to the early fault weak signal; inputting the plurality of IMF components and the residual component into a trained weak fault diagnosis model to obtain a detection result of the early fault weak signal in the complex power system; wherein the weak fault diagnosis model is constructed based on a convolutional neural network. By adopting the method, the accuracy of early fault diagnosis of a complex power system in the prior art is improved, and high-precision fault diagnosis is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and in particular to a method, device and electronic equipment for early fault detection of a complex power system. Background Art

[0002] With the progress of society, mechanical equipment is constantly developing in the direction of diversification, clustering, and intelligence. The probability of failure is also increasing, and the resulting failures are more complex, increasing the instability of mechanical equipment. Since the various components of mechanical equipment are closely connected, the failure of a component will affect the normal operation of other components, causing equipment damage and then causing serious accidents. Therefore, more and more researchers are committed to studying the health status monitoring and fault diagnosis of large equipment. The key to solving the huge economic losses, serious accidents and casualties caused by equipment failure is to carry out research on early fault diagnosis technology for mechanical equipment. Early failure is a gradual evolution process. With the normal operation and loss of equipment, it accumulates and evolves and gradually develops into a significant failure.

[0003] Early faults are usually highly destructive, hidden, random and complex. Compared with traditional fault diagnosis, early fault diagnosis is a more sophisticated diagnostic method that is more difficult and has higher requirements. Summary of the invention

[0004] The present invention provides a complex power system early fault detection method, device and electronic equipment, which are used to improve the accuracy of early fault diagnosis of complex power systems in the prior art and achieve high-precision fault diagnosis.

[0005] The present invention provides a complex power system early fault detection method, comprising the following steps: Obtain weak signals of early faults in complex power systems; Processing the early fault weak signal to obtain a plurality of IMF components and residual components corresponding to the early fault weak signal; The multiple IMF components and the residual components are input into the trained weak fault diagnosis model to obtain the detection result of the weak signal of the early fault in the complex power system; wherein the weak fault diagnosis model is constructed based on a convolutional neural network.

[0006] According to the complex power system early fault detection method provided by the present invention, the early fault weak signal is processed to obtain multiple IMF components and residual components corresponding to the early fault weak signal, including: Taking the early fault weak signal as the signal to be decomposed, and obtaining the extreme value points in the signal to be decomposed, so as to construct the upper envelope and the lower envelope of the early fault weak signal; Obtaining a first mean value of the early fault weak signal according to the upper envelope and the lower envelope; Obtaining a first IMF component and a corresponding first residual component according to a difference between the early fault weak signal and the first mean value; The first residual component is used as the signal to be decomposed again, and multiple iterations are performed to obtain multiple IMF components and residual components; When the IMF components obtained from two consecutive iterative processes meet the preset conditions, the iterative process is stopped.

[0007] According to the complex power system early fault detection method provided by the present invention, when the IMF components obtained in two adjacent iterative processes meet the preset conditions, the iterative process is stopped, comprising: When the standard deviation between two consecutive IMF components is less than the preset standard deviation, the iteration process is stopped.

[0008] According to the complex power system early fault detection method provided by the present invention, the extreme value points include maximum value points and minimum value points; the early fault weak signal is used as the signal to be decomposed, and the extreme value points in the signal to be decomposed are obtained to construct the upper envelope and the lower envelope of the early fault weak signal, including: The upper envelope corresponding to each maximum point and the lower envelope corresponding to each minimum point are obtained by cubic spline interpolation method.

[0009] The complex power system early fault detection method provided by the present invention also includes: obtaining a trained weak fault diagnosis model for the weak fault diagnosis model through historical early fault weak signals in the complex power system.

[0010] According to any of the complex power system early fault detection methods provided by the present invention, the weak fault diagnosis model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; The convolution layer and the pooling layer each have 5 layers, and a maximum pooling strategy is adopted; wherein the first layer of the convolution layer is a large convolution kernel; The loss function of the weak fault diagnosis model is a cross entropy loss function.

[0011] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for early fault detection of a complex power system as described in any one of the above is implemented.

[0012] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for early fault detection of a complex power system as described in any one of the above is implemented.

[0013] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for early fault detection of a complex power system as described in any one of the above is implemented.

[0014] The complex power system early fault detection method, device and electronic device provided by the present invention process the early fault weak signal in the complex power system to obtain multiple IMF components and residual components corresponding to the early fault weak signal; the characteristic information in the early fault signal can be effectively extracted, and the interference noise can be removed, thereby improving the clarity of the signal. Ensure the accuracy of subsequent fault diagnosis. Subsequently, the weak fault diagnosis model constructed by convolutional neural network is used to perform in-depth feature extraction and classification of the denoised early fault weak signal. Convolutional neural network has powerful signal processing capabilities, can recognize complex signal patterns, and achieve high-precision fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 This is one of the flow charts of the complex power system early fault detection method provided by the present invention.

[0017] Figure 2 This is the second flow chart of the complex power system early fault detection method provided by the present invention.

[0018] Figure 3 It is a result schematic diagram of the loss curve of the noisy signal model provided by the present invention.

[0019] Figure 4 This is a schematic diagram of the accuracy curve of the noisy signal model provided by the present invention.

[0020] Figure 5 This is a schematic diagram of the loss curve of the denoising signal model provided by the present invention.

[0021] Figure 6 This is a schematic diagram of the accuracy curve of the denoising signal model provided by the present invention.

[0022] Figure 7 This is a comparison chart of the Loss values ​​of the two signal ten experimental test sets provided by the present invention.

[0023] Figure 8 This is a comparison chart of the accuracy of the two signal ten experimental test sets provided by the present invention

[0024] Fig. 9 It is a structural schematic diagram of the complex power system early fault detection device provided by the present invention.

[0025] Fig.10 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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 are within the scope of protection of the present invention.

[0027] The terms "first", "second", etc. in the specification and claims of the present application 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 descriptions used in this way can be interchanged where appropriate, so that the embodiments can be implemented in a sequence other than that illustrated or described in the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of the steps that appear in the present application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved. The division of modules in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between units can be electrical or other similar forms, which are not limited in this application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical units, or may be distributed in multiple circuit units, and some or all of the units may be selected according to actual needs to achieve the purpose of the present application.

[0028] The present invention is applied to the problem of early fault detection in the scenario of complex power systems. Early fault detection is to use modern mechanical monitoring and diagnosis technology to quickly and accurately identify the location, fault mode and fault severity of the equipment when the equipment fault has just occurred or the fault severity is still mild, that is, when the fault symptoms are not obvious, and guide the equipment maintenance, repair and optimization design work accordingly, so as to effectively control the spread and development of the fault and ensure the safe, stable and reliable operation of the equipment.

[0029] In the present invention, early faults are a gradual evolutionary process. With the normal operation and wear of the equipment, after accumulation and evolution, they gradually develop into significant faults. Early faults are usually highly destructive, hidden, random and complex, which are mainly reflected in the following aspects: 1. The impact of the fault is slight and the impact on system performance is not obvious, but if not handled, it is easy to deteriorate rapidly or cause secondary faults to cause more serious losses; 2. The fault symptom signal is weak and easily masked by unknown disturbances and strong noise in the system; 3. Under the same working conditions, the changes in normal and faulty operating states are very small; 4. The fault characteristics are weakly separable, the fault characteristics are not significant and are coupled with each other, and fault diagnosis is very difficult. Compared with traditional fault diagnosis, early fault diagnosis is a more sophisticated diagnostic method, which requires increasing the signal-to-noise ratio between faults and noise, and removing the masking of minor faults by disturbances and noise. Therefore, the diagnosis is more difficult and has higher requirements.

[0030] Combine the following Figure 1-Figure 10 The specific contents of the present invention are described.

[0031] Figure 1 It is a flow chart of a complex power system early fault detection method provided by the present invention, including steps S101-S104.

[0032] Step S101, obtaining a weak signal of an early fault in a complex power system.

[0033] For example, take the electric vehicle power system as an example. The system has a complex structure, the power source is a high-voltage power supply, and the components are closely connected. When a component fails, a weak signal may be generated. If these signals are not captured and analyzed in time, they may gradually evolve into significant faults, causing equipment damage or even serious accidents. Therefore, a complex power system early fault detection method is needed to obtain these early fault weak signals.

[0034] Step S102: Process the early fault weak signal to obtain a plurality of IMF components and residual components corresponding to the early fault weak signal.

[0035] In the embodiments of the present invention, since early faults are usually highly destructive, hidden, random and complex, they are mainly manifested in the following aspects: the fault impact is slight, the fault symptom signal is weak, and it is easily masked by unknown disturbances and strong noise of the system, under the same working conditions, the changes in normal and faulty operating states are very small, the fault characteristics are weakly separable, and the fault characteristics are not significant and are mutually coupled. Compared with traditional fault diagnosis, early fault diagnosis is a more sophisticated diagnostic method, which requires increasing the signal-to-noise ratio between faults and noise and removing the masking of minor faults by disturbances and noise. Based on the above reasons, the present invention processes the weak signal of the early fault to obtain multiple IMF components and residual components corresponding to the weak signal of the early fault.

[0036] In the embodiment of the present invention, the IMF component is an IMF type signal, and IMF (Intrinsic Mode Function) is an intrinsic mode function.

[0037] Step S103, inputting multiple IMF components and residual components into the trained weak fault diagnosis model to obtain the detection results of weak signals of early faults in the complex power system; wherein the weak fault diagnosis model is constructed based on a convolutional neural network.

[0038] In a possible implementation, Figure 2 As shown, in step S102, the early fault weak signal is processed to obtain multiple IMF components and residual components corresponding to the early fault weak signal, including steps S201-S205.

[0039] Step S201: taking the early fault weak signal as the signal to be decomposed, and obtaining the extreme value points in the signal to be decomposed, so as to construct the upper envelope and the lower envelope of the early fault weak signal.

[0040] In a possible implementation, the extreme points include maximum points and minimum points; in step S201, the early fault weak signal is used as the signal to be decomposed, and the extreme points in the signal to be decomposed are obtained to construct the upper envelope and lower envelope of the early fault weak signal, including: obtaining the upper envelope corresponding to each maximum point and the lower envelope corresponding to each minimum point by cubic spline interpolation method.

[0041] Step S202: Obtain a first mean value of the early fault weak signal according to the upper envelope and the lower envelope.

[0042] Specifically, in steps S201-S202, for any signal to be decomposed , calculate all its extreme points. Use cubic spline interpolation to fit the upper envelope of each maximum value and the lower envelope of each minimum value, and calculate their mean and record them as .

[0043] Step S203: obtaining a first IMF component and a corresponding first residual component according to the difference between the early fault weak signal and the first mean value.

[0044] Specifically, the signal to be decomposed Subtract the mean get : .

[0045] Step S204: taking the first residual component as the signal to be decomposed again, performing multiple iterations to obtain multiple IMF components and residual components.

[0046] Specifically, ideally, The two conditions of the eigenmode function have been met, then that is However, when applied to actual signals, it is difficult to solve the theoretical envelope, and the envelope is fitted by the cubic spline interpolation method, so new extreme points may be generated, and the original extreme points may also be slightly shifted, which were not included in the previous screening process. Only by continuous screening and multiple iterations can all the small fluctuations in the signal to be decomposed be distinguished.

[0047] Exemplarily, the second screening is to As the original signal, let the mean of the envelope calculated by this screening be , then the screening can be expressed as:

[0048] After repeated screening k times, the true intrinsic mode function is obtained :

[0049] remember ,but for The first IMF component of:

[0050] In the embodiment of the present invention, the above iterative steps are repeated to obtain multiple IMF components and a residual component in sequence. , their frequencies are arranged from high to low, then the final original signal can be expressed as:

[0051] From the above formula we can see that is decomposed into n and a residual term representing the average trend or mean of the signal .

[0052] Step S205: When the IMF components obtained in two consecutive iterative processes meet a preset condition, the iterative process is stopped.

[0053] In a possible implementation, in step S205, when the IMF components obtained in two adjacent iterative processes meet a preset condition, the iterative process is stopped, including: when the standard deviation between the IMF components obtained in two adjacent iterative processes is less than a preset standard deviation, the iterative process is stopped.

[0054] In the embodiments of the present invention, the number of iterations also needs to be controlled to a certain extent. If the number is too large, the IMF component may become a constant, and if the number is too small, the IMF component may not have good local symmetry. Therefore, when decomposing the signal, the number of screening should be limited. In order to make the obtained IMF components have practical physical meaning, the present invention proposes to stop the iteration process when the IMF components obtained by two adjacent iterations meet the preset conditions. The preset conditions are determined by the two adjacent iterations. and The standard deviation SD is defined as:

[0055] Where T is When the above conditions are met, the iteration process stops and the required IMF can be obtained. The value of is exemplarily given as 0.2~0.3 in the embodiment of the present invention.

[0056] In a possible implementation, the complex power system early fault detection method further includes: obtaining a trained weak fault diagnosis model based on historical early fault weak signals in the complex power system.

[0057] In a possible implementation, the weak fault diagnosis model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0058] Among them, there are 5 convolutional layers and pooling layers respectively, and the maximum pooling strategy is adopted; among them, the first layer of the convolutional layer is a large convolution kernel; the loss function of the weak fault diagnosis model is the cross entropy loss function.

[0059] In the embodiment of the present invention, the first layer of the convolution layer is a large convolution kernel, the purpose of which is to extract short-time features, and its function is similar to that of the short-time Fourier transform. The difference is that the window function of the short-time Fourier transform is a sine function, and the large convolution kernel of the first layer of the convolution layer is obtained by training through an optimization algorithm, which has the advantage of automatically learning features for diagnosis and automatically removing features that are not helpful for diagnosis.

[0060] In a possible implementation, in the convolutional layers of the weak fault diagnosis model, except for the first layer, the convolutional kernel size of the convolutional layers is 3×1. Since the convolutional kernel has fewer parameters, this is conducive to deepening the network and suppressing overfitting. Batch normalization (BN) is performed after each convolution operation, and then 2×1 maximum pooling is performed.

[0061] Next, the specific structure of the weak fault diagnosis model for model training in the embodiment of the present invention is described.

[0062] Based on the convolution layer of the above-mentioned weak fault diagnosis model, the signal passes through the convolution layer, is activated by the ReLU function, and then downsamples the features extracted by the convolution layer through the maximum pooling layer. After repeating the above operation three times, the Dropout layer is introduced, and Dropout=0.3 is set. The fully connected layer converts the feature map into the vector required for classification to achieve end-to-end learning, and finally the fault features are classified by the Softmax classifier. At the same time, the BN batch normalization process is introduced between the convolution layer and the pooling layer to simplify the hyperparameter adjustment, eliminate overfitting, better generalize and retain more feature information.

[0063] The setting of convolutional neural network parameters has a significant impact on the diagnosis results. The main network parameters include the number of convolutional layers, loss function, learning rate, activation function and pooling layer configuration. The number of convolutional layers directly determines the depth and feature extraction ability of the model. Deeper networks can extract more complex features, but too deep networks may lead to overfitting. The choice of loss function and its parameter setting are crucial to the fault diagnosis results. Different types of loss functions (such as cross entropy and mean square error) are suitable for different tasks, and their optimization directly affects the learning effect of the model. By adjusting the weight coefficient in the loss function, the sensitivity of the model to early faults in the fault data set can be improved. The learning rate is another key parameter that controls the speed of weight update during model training. Too high a learning rate may lead to unstable model convergence, while too low a learning rate will lead to slow convergence, affecting training efficiency and final results. The choice of activation function will also affect the nonlinear expression ability of the model. For example, RelU (rectified linear unit) can effectively solve the gradient vanishing problem. The configuration of the pooling layer can reduce the feature dimension and improve computational efficiency while maintaining key information. Different pooling strategies (such as maximum pooling or average pooling) have different effects on feature selection, thus affecting the accuracy of diagnosis. Therefore, the weak fault diagnosis model will use 5 convolutional layers and pooling layers, adopt the maximum pooling strategy, select the cross entropy function as the loss function when training the weak fault diagnosis model, and optimize it with the Adam optimization algorithm. The learning rate value is set to 0.0001, and the epoch and batch sizes are 100 and 16 respectively.

[0064] In the embodiment of the present invention, the parameters of the weak fault diagnosis model used are shown in Table 1. The weak fault diagnosis model has a total of 5 convolution and pooling layers, the size of the first convolution kernel is 9×1, and except for the first layer, the sizes of the remaining convolution kernels are all 3×1. The number of neurons in the hidden layer is 128, and the Softmax layer has a total of 7 outputs, corresponding to 7 states.

[0065] Table 1 Model parameters for weak fault diagnosis

[0066] In an embodiment of the present invention, the model training process of the present invention is described.

[0067] Specifically, the denoised early fault signal data obtained by the above signal processing process and the early fault data containing Gaussian white noise are used to train the weak fault diagnosis model, and the accuracy and generalization of the weak fault diagnosis model are evaluated. 70% of the data is used to train the weak fault diagnosis model, 15% of the data is used for verification, and 15% of the data is used to test the trained weak fault diagnosis model. The noisy early fault signal and the denoised early fault signal of size 512*1 are respectively input into the constructed weak fault diagnosis model. In order to prevent the randomness of the experimental results, ten experiments are repeated on the two models to verify the effectiveness of the model. The loss rate and accuracy curves of the noisy early fault signal and the denoised early fault signal in an experiment on the weak fault diagnosis model are shown in the figure. Figure 3 , Figure 4 , Figure 5 , Figure 6 shown.

[0068] in, Figure 3 Schematic diagram of the loss curve of the noisy signal model.

[0069] Figure 4 This is a schematic diagram of the accuracy curve of the noisy signal model.

[0070] Figure 5 Schematic diagram of the loss curve of the denoising signal model.

[0071] Figure 6 Schematic diagram of the accuracy curve of the denoising signal model.

[0072] from Figures 3 to 6 It can be seen that the weak fault diagnosis model with denoised early fault signal as input has a higher recognition rate, a lower Loss value and a better effect on early fault identification than the weak fault diagnosis model with noisy early fault signal as input, although the Loss value curve and the accuracy curve are still in a fluctuating state at the end of the iteration, and the model is not stable.

[0073] Figure 7 This is a comparison chart of the loss values ​​of the test sets of ten experiments on two signals.

[0074] Figure 8 This is a comparison chart of the accuracy of the test sets of ten experiments on two signals.

[0075] Depend on Figure 7 and Figure 8 It can be seen that the accuracy of the weak fault diagnosis model run ten times on the denoised signal dataset is higher than that on the noisy signal dataset, and the average recognition rate can reach 90.95%, while the highest accuracy of the weak fault diagnosis model with the noisy signal dataset as input is less than 90%; at the same time, the loss of the weak fault diagnosis model with the noisy signal dataset as input is lower in comparison, which verifies the effectiveness of the above method in judging the early fault characteristics after denoising.

[0076] The present invention can achieve the following beneficial effects by adopting the above method: By processing the weak signals of early faults in complex power systems, multiple IMF components and residual components corresponding to the weak signals of early faults are obtained; the characteristic information in the early fault signals can be effectively extracted, and the interference noise can be removed, thereby improving the clarity of the signal. Ensure the accuracy of subsequent fault diagnosis. Subsequently, the weak fault diagnosis model constructed by convolutional neural network is used to perform in-depth feature extraction and classification of the denoised weak signals of early faults. Convolutional neural networks have powerful signal processing capabilities, can recognize complex signal patterns, and achieve high-precision fault diagnosis.

[0077] The complex power system early fault detection device provided by the present invention is described below. The complex power system early fault detection device described below and the complex power system early fault detection method described above can be referenced to each other.

[0078] Fig. 9 The structural schematic diagram of a complex power system early fault detection device provided by the present invention includes the following parts: The signal acquisition module 910 is used to acquire weak signals of early faults in complex power systems.

[0079] The signal processing module 920 is used to perform signal processing on the early fault weak signal to obtain multiple IMF components and residual components corresponding to the early fault weak signal.

[0080] The fault detection module 930 is used to input multiple IMF components and residual components into the trained weak fault diagnosis model to obtain the detection results of weak signals of early faults in complex power systems; wherein the weak fault diagnosis model is constructed based on a convolutional neural network.

[0081] In a possible implementation, the signal processing module 920 is specifically configured to: The early fault weak signal is used as the signal to be decomposed, and the extreme value points in the signal to be decomposed are obtained to construct the upper envelope and the lower envelope of the early fault weak signal; According to the upper envelope and the lower envelope, the first mean value of the weak signal of the early fault is obtained; According to the difference between the early fault weak signal and the first mean value, a first IMF component and a corresponding first residual component are obtained; The first residual component is used as the signal to be decomposed again, and multiple iterations are performed to obtain multiple IMF components and residual components; When the IMF components obtained from two consecutive iterative processes meet the preset conditions, the iterative process is stopped.

[0082] In a possible implementation manner, the signal processing module 920 is specifically configured to stop the iteration process when a standard deviation between two adjacent IMF components is less than a preset standard deviation.

[0083] In a possible implementation, the signal processing module 920 is specifically configured to obtain an upper envelope corresponding to each maximum value point and a lower envelope corresponding to each minimum value point by using a cubic spline interpolation method.

[0084] In a possible implementation, it also includes a model training module for training the weak fault diagnosis model through historical early fault weak signals in the complex power system to obtain a trained weak fault diagnosis model.

[0085] In a possible implementation, the weak fault diagnosis model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; the convolution layer and the pooling layer have 5 layers respectively, and a maximum pooling strategy is adopted; wherein the first layer of the convolution layer is a large convolution kernel; and the loss function of the weak fault diagnosis model is a cross entropy loss function.

[0086] Fig.10 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.10 As shown, the electronic device may include: a processor 1010, a communications interface 1020, a memory 1030 and a communication bus 1040, wherein the processor 1010, the communications interface 1020 and the memory 1030 communicate with each other through the communication bus 1040. The processor 1010 may call the logic instructions in the memory 1030 to execute a complex power system early fault detection method, the method comprising: Obtain weak signals of early faults in complex power systems; Processing the early fault weak signal to obtain multiple IMF components and residual components corresponding to the early fault weak signal; Multiple IMF components and residual components are input into the trained weak fault diagnosis model to obtain the detection results of weak signals of early faults in complex power systems; among them, the weak fault diagnosis model is constructed based on a convolutional neural network.

[0087] In addition, the logic instructions in the above-mentioned memory 1030 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the 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 several 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, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0088] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can execute a complex power system early fault detection method provided by the above methods, the method comprising: Obtain weak signals of early faults in complex power systems; Processing the early fault weak signal to obtain multiple IMF components and residual components corresponding to the early fault weak signal; Multiple IMF components and residual components are input into the trained weak fault diagnosis model to obtain the detection results of weak signals of early faults in complex power systems; among them, the weak fault diagnosis model is constructed based on a convolutional neural network.

[0089] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the computer program is implemented to execute a complex power system early fault detection method provided by the above methods, the method comprising: Obtain weak signals of early faults in complex power systems; Processing the early fault weak signal to obtain multiple IMF components and residual components corresponding to the early fault weak signal; Multiple IMF components and residual components are input into the trained weak fault diagnosis model to obtain the detection results of weak signals of early faults in complex power systems; among them, the weak fault diagnosis model is constructed based on a convolutional neural network.

[0090] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0091] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A complex power system early fault detection method, characterized in that: include: Obtain weak signals of early faults in complex power systems; Processing the early fault weak signal to obtain a plurality of IMF components and residual components corresponding to the early fault weak signal; The multiple IMF components and the residual components are input into the trained weak fault diagnosis model to obtain the detection result of the weak signal of the early fault in the complex power system; wherein the weak fault diagnosis model is constructed based on a convolutional neural network.

2. The complex power system early fault detection method according to claim 1, characterized in that: The signal processing of the early fault weak signal to obtain a plurality of IMF components and residual components corresponding to the early fault weak signal includes: Taking the early fault weak signal as the signal to be decomposed, and obtaining the extreme value points in the signal to be decomposed, so as to construct the upper envelope and the lower envelope of the early fault weak signal; Obtaining a first mean value of the early fault weak signal according to the upper envelope and the lower envelope; Obtaining a first IMF component and a corresponding first residual component according to a difference between the early fault weak signal and the first mean value; The first residual component is used as the signal to be decomposed again, and multiple iterations are performed to obtain multiple IMF components and residual components; When the IMF components obtained from two consecutive iterative processes meet the preset conditions, the iterative process is stopped.

3. The complex power system early fault detection method according to claim 2, characterized in that: When the IMF components obtained from two adjacent iterative processes meet a preset condition, stopping the iterative process includes: When the standard deviation between two consecutive IMF components is less than the preset standard deviation, the iteration process is stopped.

4. The complex power system early fault detection method according to claim 2, characterized in that: The extreme value points include maximum value points and minimum value points; the method of using the early fault weak signal as the signal to be decomposed and obtaining the extreme value points in the signal to be decomposed to construct the upper envelope and the lower envelope of the early fault weak signal includes: The upper envelope corresponding to each maximum point and the lower envelope corresponding to each minimum point are obtained by cubic spline interpolation method.

5. The complex power system early fault detection method according to claim 1, characterized in that: Also includes: Through the historical early fault weak signals in the complex power system, the weak fault diagnosis model is trained to obtain a weak fault diagnosis model.

6. The complex power system early fault detection method according to any one of claims 1 to 5, characterized in that: The weak fault diagnosis model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; The convolution layer and the pooling layer each have 5 layers, and adopt a maximum pooling strategy; wherein the first layer of the convolution layer is a large convolution kernel; The loss function of the weak fault diagnosis model is a cross entropy loss function.

7. A complex power system early fault detection device, characterized in that: include: Signal acquisition module, used to obtain weak signals of early faults in complex power systems; A signal processing module, used for performing signal processing on the early fault weak signal to obtain a plurality of IMF components and residual components corresponding to the early fault weak signal; A fault detection module is used to input multiple IMF components and the residual component into a trained weak fault diagnosis model to obtain the detection result of the weak signal of the early fault in the complex power system; wherein the weak fault diagnosis model is constructed based on a convolutional neural network.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for early fault detection of a complex power system as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for early fault detection of a complex power system as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for early fault detection of a complex power system as claimed in any one of claims 1 to 6 is implemented.

Images