Voltage waveform abnormity identification method in DC / DC power supply inspection process
Through the automated inspection platform and abnormal waveform recognition model, the problems of low efficiency and poor accuracy of voltage waveform abnormal recognition in DC/DC power supply equipment inspection are solved, and intelligent, fast and accurate voltage waveform abnormal recognition is achieved.
Patent Information
- Application Number
- CN202411100556.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-12-06
AI Technical Summary
During the inspection process of DC/DC power supply equipment, it takes a long time to manually identify abnormal voltage waveforms and is easy to be misidentified, which affects the inspection efficiency and accuracy.
Design an automated inspection platform, adopting high-precision waveform acquisition card, programmable DC power supply and programmable DC load, and construct a voltage abnormal waveform recognition model through time domain feature extraction network, frequency domain feature conversion module and abnormal waveform recognition network to realize automated recognition.
It improves the efficiency and accuracy of DC/DC power supply inspection, reduces the time and misidentification rate of manual identification, and realizes intelligent, fast and accurate identification of voltage waveform sudden changes.
Smart Images

Figure CN119089345A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of DC / DC technology, and in particular to a method for identifying voltage waveform anomalies during a DC / DC power supply inspection process. Background Art
[0002] As a key equipment in the power system, the performance stability of the DC / DC power supply equipment in the substation is directly related to the overall operation quality of the power system. In order to ensure the stable operation of the DC / DC power supply equipment in the substation, whether before the bidding for the purchase of equipment, after winning the bid or during use, it is necessary to strictly inspect the DC / DC power supply equipment according to national standards to ensure that it can meet the technical parameters of the substation's operating equipment.
[0003] According to industry standards, during the inspection of DC / DC power supply equipment, there are multiple tests that require the identification of sudden abnormal changes in the DC / DC power supply voltage waveform, including the overshoot amplitude test of the on / off machine, the dynamic voltage transient range and transient response recovery time test, the inverter power synchronization accuracy test, the voltage waveform distortion test, and the soft start time test of the charging device. In order to meet the inspection requirements of the above-mentioned inspection items, the voltage waveform timing and amplitude data output by the DC / DC power supply need to be collected using high-speed waveform acquisition technology during the inspection process, that is, the DC voltage is tracked, measured and recorded in real time during the test process through high-speed, high-precision sensors and real-time digital signal processors. Because of the use of high-speed sampling and storage technology, a large amount of waveform data will be generated in a very short time, and then the waveform data needs to be traced back and analyzed and abnormal waveform data needs to be identified. This requirement has become a key constraint affecting the efficiency and accuracy of the inspection. If the manual method is used, it will take a lot of time and energy for the quality inspectors of the substation network equipment, which is not only inefficient but also prone to misidentification and missed identification. Summary of the invention
[0004] In order to solve the above technical problem or at least partially solve the above technical problem, the present invention provides a method for identifying voltage waveform anomalies during a DC / DC power supply inspection process.
[0005] In a first aspect, the present invention provides a method for identifying abnormal voltage waveform during a DC / DC power supply inspection process, comprising:
[0006] Adopt high-precision waveform acquisition card, programmable DC power supply and programmable DC load to design and build an automated inspection platform for DC / DC power supply;
[0007] The voltage waveform data is acquired by using an automated inspection platform and the voltage waveform data is preprocessed to obtain training data; a voltage abnormal waveform recognition model is constructed by using the training data to train the voltage abnormal waveform recognition model so that the voltage abnormal waveform recognition model can locate the voltage abnormal waveform in the voltage waveform, wherein the voltage abnormal waveform recognition model comprises: a time domain feature extraction network for voltage waveform data, a frequency domain feature conversion module and an abnormal waveform data detection network; the time domain features extracted by the time domain feature extraction network are fused with the frequency domain features obtained by the frequency domain feature conversion module and converted into two-dimensional feature information which is input into the abnormal waveform recognition network
[0008] Deploy the pre-trained abnormal voltage waveform recognition model to the substation inspection system to identify and locate the abnormal voltage waveform of the DC / DC power supply.
[0009] Furthermore, an automated inspection platform for DC / DC power supplies is designed and constructed using a high-precision waveform acquisition card, a programmable DC power supply, and a programmable DC load. The high-precision waveform acquisition card acquires voltage waveform data of the front-stage input and the back-stage output of the DC / DC power supply. The programmable DC power supply is used to adjust and control the input voltage of the DC / DC power supply being inspected. The programmable DC load is used to adjust and control the output load of the DC / DC power supply being inspected. An inspection process rule base is designed for each inspection item of the DC / DC power supply. The control center controls the programmable DC power supply and the programmable DC load to operate according to the inspection process rules in the inspection process rule base, and simulates the operation of the DC / DC power supply in different scenarios. The inspection process rule base contains normal rules and abnormal rules. The DC / DC power supply is controlled to be in normal operating condition by the normal rules, and the DC / DC power supply is controlled to trigger abnormal mutation conditions by the abnormal rules.
[0010] Furthermore, when obtaining training data, preprocessing of the voltage waveform data includes: cutting into sub-segment data of uniform length; screening and removing irrelevant data and low-quality data; and performing data expansion on the sub-segment data.
[0011] Furthermore, the time domain feature extraction network is designed using a one-dimensional convolutional neural network, which includes: a one-dimensional convolutional layer, a maximum pooling layer and an average pooling layer, an activation function and a fully connected layer; the one-dimensional convolutional layer extracts local features of the voltage waveform from the voltage waveform data sequence; a combination of maximum pooling and average pooling technology is used to reduce the dimension and number of local features of the DC / DC power supply voltage abnormal waveform, thereby extracting the most important and critical features of the voltage abnormal waveform that appears in the entire inspection process; nonlinear mapping is introduced through the activation function so that the voltage waveform data time domain feature extraction network can learn a more complex feature representation of the DC / DC power supply voltage abnormal waveform; the fully connected layer is used to integrate the features of the DC / DC power supply voltage abnormal waveform extracted by the convolutional layer, so as to learn a higher-order feature representation of the DC / DC power supply voltage abnormal waveform.
[0012] Furthermore, the frequency domain feature conversion module uses Fourier transform to convert the time domain features of the DC / DC power supply voltage waveform into frequency domain features.
[0013] Furthermore, the number of channels, filter configuration and step size strategy of the one-dimensional convolutional layer of the time domain feature extraction network are designed to match the number of Fourier transform filter groups, preset window length and frame shift scale during frequency domain feature extraction, so as to ensure that the time domain feature dimension extracted from the voltage waveform signal is consistent with the frequency domain feature dimension of the Fourier transform. By superimposing the channel dimension, the deep fusion of time domain and frequency domain features is achieved.
[0014] Furthermore, the abnormal waveform recognition network includes: a stacked bottleneck layer, a global depth-by-depth convolution layer is connected after the bottleneck layer, and a linear transformation layer is arranged after the global depth-by-depth convolution layer.
[0015] Furthermore, the bottleneck layer includes: an expansion layer for expanding the number of feature channels implemented by a 1×1 convolution layer and a ReLU activation function; the expansion layer is connected to several deep convolution layers, the deep convolution layer is connected to a linear bottleneck layer as a linear activation function, and the linear bottleneck layer is connected to a projection layer implemented by a 1×1 convolution layer to restore the number of channels; the deep convolution layer step size of the bottleneck layer includes 1 and 2. When the step size is 2, the bottleneck layer is used to implement downsampling, and when the step size is 1, the input and output of the bottleneck layer are connected through a residual connection.
[0016] In a second aspect, the present invention provides a device for identifying anomalies in voltage waveforms during a DC / DC power supply inspection process, comprising: at least one processing unit, the processing unit, a collection unit, and a storage unit are interconnected via a bus unit, the storage unit stores a computer program, and when the computer program is executed by the processing unit, the method for identifying anomalies in voltage waveforms during a DC / DC power supply inspection process is implemented.
[0017] In a third aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for identifying voltage waveform anomalies during the DC / DC power supply inspection process is implemented.
[0018] The above technical solution provided by the embodiment of the present invention has the following advantages compared with the prior art:
[0019] The present invention proposes a method for identifying voltage waveform anomalies during the inspection of a DC / DC power supply in a substation. An automated inspection platform for a DC / DC power supply is constructed by designing a precision waveform acquisition card, a programmable DC power supply, and a programmable DC load, and the acquisition, clipping, and expansion operations of the voltage waveform data are performed to construct the training set and test set required for the voltage abnormal waveform recognition model. The time-frequency domain features of the voltage waveform data are extracted and fused by a time-domain feature extraction network and a frequency-domain feature conversion module. After the trained voltage abnormal waveform recognition model is deployed, the quality inspectors of the substation network-connected equipment can use the voltage abnormal waveform recognition model to automatically and intelligently identify the voltage waveform during the DC / DC power supply inspection process, thereby realizing intelligent, rapid, and accurate identification of voltage waveform mutations, and improving the inspection efficiency and accuracy of various complex DC / DC power supplies during the quality inspection of substation network-connected equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0022] Figure 1 A schematic diagram of an automated inspection platform provided by an embodiment of the present invention;
[0023] Figure 2 A schematic diagram of a voltage abnormal waveform recognition model provided by an embodiment of the present invention;
[0024] Figure 3 A schematic diagram of an abnormal waveform data detection network provided by an embodiment of the present invention;
[0025] Figure 4 A schematic diagram of a device for identifying abnormal voltage waveform during a DC / DC power supply inspection process provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.
[0028] Example 1
[0029] See also Figure 1 As shown, an embodiment of the present invention provides a method for identifying abnormal voltage waveform during a DC / DC power supply inspection process, comprising the following steps:
[0030] The automatic inspection platform of DC / DC power supply is designed by using high-precision waveform acquisition card, programmable DC power supply and programmable DC load. Figure 1 As shown in the figure, the high-precision waveform acquisition card is responsible for the fast and efficient acquisition of the voltage waveform data of the front-stage input and the rear-stage output of the DC / DC power supply, and converts the analog signal of the voltage waveform into a digital signal through an analog-to-digital converter; the programmable DC power supply is used to accurately adjust and control the input voltage of the DC / DC power supply under test; the programmable DC load is used to adjust and control the output load of the DC / DC power supply under test. Design the inspection process rule base for each inspection item of the DC / DC power supply, such as: according to the performance index requirements of the DC / DC power supply system for substations in the industry standard, the dynamic voltage transient range inspection of the DC / DC power supply is carried out: under the conditions of rated input voltage and rated output voltage, the load current (resistive load) suddenly changes from 25%→50%→25% of the rated value and from 50%→75%→50% of the rated value, measure the waveform of the DC / DC power supply DC output voltage, and record the sudden change value of the output voltage. The ratio of the output voltage sudden change value to the output voltage rated value should not exceed ±5% of the output nominal voltage.
[0031] The control center controls the programmable DC power supply and programmable DC load according to the inspection process rules in the inspection process rule base, and simulates the operation of the DC / DC power supply in different scenarios. In the specific implementation process, the inspection process rule base contains normal rules and abnormal rules; the DC / DC power supply is controlled to be in normal operation through normal rules, and the DC / DC power supply is controlled to trigger abnormal mutation conditions through abnormal rules.
[0032] The voltage waveform data is acquired using an automated inspection platform and preprocessed to obtain training data.
[0033] In various scenarios simulated by the automated inspection platform, high-precision waveform acquisition cards are used to obtain normal waveform signal data and abnormal waveform signal data. Normal waveform data is acquired when the DC / DC power supply is in normal operating conditions controlled by normal rules; abnormal waveform data is acquired when the DC / DC power supply triggers abnormal mutation conditions controlled by abnormal rules; the acquired normal waveform data and abnormal waveform data are classified and stored.
[0034] After the DC / DC power supply voltage waveform data is collected, the waveform data is clipped and preprocessed to decompose the voltage waveform data into sub-segment data of uniform length, such as a sub-segment duration of 1 segment / second. Since the collected data dimensions are huge and cannot be directly used for the training of the voltage abnormal waveform recognition model, it is clipped into sub-segment data of uniform length, and then the voltage waveform data is screened and processed to remove irrelevant data and low-quality data to ensure the validity of the data.
[0035] The sub-segment data cropped from the voltage waveform is expanded and categorized. Specifically, an abnormal ID that marks the abnormal type is constructed, and the corresponding abnormal voltage sub-segment data is named and labeled using the abnormal ID, and the training set and test set are divided according to the data set classification rules.
[0036] Model training requires a large amount of abnormal waveform data support. Therefore, in addition to collecting abnormal waveform data, the method of randomly generating one-dimensional waveform data is also used to expand the abnormal waveform data. The expanded abnormal waveform data is efficiently and automatically generated by writing Python language scripts, including: setting the threshold of normal waveform signals, randomly selecting points within the 1-second interval data to simulate the abnormal jump scenario of DC / DC power supply voltage, and generating signals exceeding the normal waveform threshold to expand the abnormal waveform data.
[0037] A voltage abnormal waveform recognition model is constructed, the voltage abnormal waveform recognition model is trained using a training set and a test set, and whether the voltage abnormal waveform recognition model is overfitting is verified using the test set.
[0038] like Figure 2 As shown in the figure, the design of the voltage abnormal waveform recognition model includes three parts: the first part is the voltage waveform data time domain feature extraction network, the second part is the frequency domain feature conversion module, and the third part is the abnormal waveform data detection network.
[0039] The time domain feature extraction network of voltage waveform data is designed using a one-dimensional convolutional neural network to effectively extract the time domain features of abnormal voltage waveform data.
[0040] The one-dimensional convolutional neural network involved in this scheme is designed as follows:
[0041] One-dimensional convolution layer, the one-dimensional convolution layer extracts the local features of the voltage waveform from the voltage waveform data sequence; for each sliding window of the one-dimensional convolution, the convolution operation multiplies the sequence fragment in the window by the convolution kernel element by element, and sums them up to obtain a characteristic value of the DC / DC power supply voltage waveform. According to the type of abnormal waveform data collected, the size and number of convolution kernels are changed to better extract the local features of abnormal DC / DC power supply voltage waveforms of different scales and numbers.
[0042] Pooling operation: After the local features of the DC / DC power supply voltage abnormal waveform are extracted, the pooling operation is performed. This solution design adopts a combination of maximum pooling and average pooling technology to reduce the dimension and number of local features of the DC / DC power supply voltage abnormal waveform, and then extract the most important and critical features of the voltage abnormal waveform that appears in the entire inspection process.
[0043] Activation function: Nonlinear mapping is introduced through the activation function so that the voltage waveform data time domain feature extraction network can learn a more complex feature representation of the DC / DC power supply voltage abnormal waveform.
[0044] Fully connected layer: The fully connected layer is used to integrate the features of the abnormal DC / DC power supply voltage waveform extracted by the convolutional layer. Each neuron in the fully connected layer is connected to all neurons in the previous layer to learn a higher-order feature representation of the abnormal DC / DC power supply voltage waveform.
[0045] Through the above-mentioned convolutional layer, pooling layer, activation function and fully connected layer, important time domain features of the abnormal waveform sequence data of the DC / DC power supply voltage can be effectively extracted during the inspection process.
[0046] The frequency domain feature conversion module uses Fourier transform to convert the time domain features of the DC / DC power supply voltage waveform into frequency domain features. The DC / DC power supply voltage waveform signal is designed to be x(t), and its Fourier transform is defined as follows:
[0047]
[0048] Wherein, X(ω) is the complex number representation of the frequency domain signal corresponding to the DC / DC power supply voltage waveform, ω is the frequency, e is the base of the natural logarithm, and j is the imaginary unit.
[0049] After the time domain features and frequency domain features of the DC / DC power supply voltage waveform are extracted through the above steps, the final abnormal waveform feature fusion processing is performed. In order to ensure that the time and frequency domain features can be fused, the structure of the one-dimensional convolutional Bible network for extracting time domain features is adapted and designed so that the number of channels, filter configuration and step size parameters of the one-dimensional convolution match the number of Fourier transform filter groups, the preset window length and the scale of the frame shift when extracting frequency domain features, so as to ensure that the time domain feature dimension extracted from the voltage waveform signal is consistent with the frequency domain feature dimension of the subsequent Fourier transform. Subsequently, the deep fusion of time domain and frequency domain features is finally achieved by superimposing the feature channel dimensions.
[0050] The fusion features of the one-dimensional DC / DC power supply voltage waveform are converted into two-dimensional feature information through convolution operation, and the two-dimensional feature information is input as a parameter into the abnormal waveform recognition network. Figure 3 As shown, the structure design of the abnormal waveform recognition network is as follows, including:
[0051] A stacked bottleneck layer is connected to a global depth-wise convolutional layer, and a linear transformation layer is arranged after the global depth-wise convolutional layer.
[0052] The bottleneck layer includes: an expansion layer that expands the number of feature channels implemented by a 1×1 convolution layer and a ReLU activation function; the expansion layer is connected to several deep convolution layers, the deep convolution layer is connected to a linear bottleneck layer as a linear activation function, and the linear bottleneck layer is connected to a projection layer implemented by a 1×1 convolution layer to restore the number of channels. The deep convolution layer step size of the bottleneck layer includes 1 and 2. When the step size is 2, the bottleneck layer is used to implement downsampling; when the step size is 1, the input and output of the bottleneck layer are connected through a residual connection. The deep convolution layer applies a single convolution filter to each input channel to perform filtering, and the 1×1 convolution layer is responsible for constructing new features by calculating the linear combination of the input channels.
[0053] In an exemplary abnormal waveform recognition network, the first bottleneck layer includes an expansion layer with an expansion ratio of 1 and a deep convolution layer with a step size of 1, the second bottleneck layer includes an expansion layer with an expansion ratio of 6 and two deep convolution layers with a step size of 2, the third bottleneck layer includes an expansion layer with an expansion ratio of 6 and three deep convolution layers with a step size of 2, the fourth bottleneck layer includes an expansion layer with an expansion ratio of 6 and four deep convolution layers with a step size of 2, the fifth bottleneck layer includes an expansion layer with an expansion ratio of 6 and three deep convolution layers with a step size of 1, the sixth bottleneck layer includes an expansion layer with an expansion ratio of 6 and three deep convolution layers with a step size of 2, the seventh bottleneck layer includes an expansion layer with an expansion ratio of 6 and a deep convolution layer with a step size of 1, and a bottleneck layer with a downsampling function is used at the beginning of the network for fast downsampling to reduce the amount of calculation of subsequent layers.
[0054] A global depth-wise convolution layer is added at the end of the bottleneck layer to help improve the network's feature extraction capabilities while keeping the amount of computation low. After the global depth-wise convolution layer, a linear transformation layer is added to output the final feature vector.
[0055] The characteristic distribution of abnormal voltage waveform data is learned through training. The loss function used in training adopts ArcFace function. ArcFace function is used as the loss function, which greatly improves the normalization and additive angle interval of the characteristic vector of abnormal DC / DC power supply voltage waveform, improves the inter-class separability, and strengthens the intra-class compactness and inter-class difference, which can better identify the abnormal DC / DC power supply voltage waveform.
[0056] The trained DC / DC power supply voltage abnormal waveform recognition model is deployed to the inspection system used by the quality inspectors of the substation network equipment. During the performance index inspection of the DC / DC power supply equipment, the DC / DC power supply voltage abnormal waveform can be intelligently, quickly and accurately identified, thereby improving the inspection efficiency and accuracy of the substation network equipment quality inspectors for various complex DC / DC power supplies.
[0057] Example 2
[0058] See also Figure 4As shown, an embodiment of the present invention provides a device for identifying abnormal voltage waveforms during a DC / DC power supply inspection process, comprising: at least one processing unit, wherein the processing unit, the acquisition unit, and the storage unit are interconnected via a bus unit, and the storage unit, as a computer-readable storage medium, can be used to store software programs, computer executable programs, and modules, such as the software programs, computer executable programs, and modules corresponding to a method for identifying abnormal voltage waveforms during a DC / DC power supply inspection process in an embodiment of the present invention. The processing unit implements the above-mentioned method for identifying abnormal voltage waveforms during a DC / DC power supply inspection process by running the software programs, computer executable programs, and modules stored in the storage unit, comprising:
[0059] The voltage waveform data is acquired by using an automated inspection platform and the voltage waveform data is preprocessed to obtain training data; a voltage abnormal waveform recognition model is constructed by using the training data to train the voltage abnormal waveform recognition model so that the voltage abnormal waveform recognition model can locate the voltage abnormal waveform in the voltage waveform, wherein the voltage abnormal waveform recognition model comprises: a time domain feature extraction network for voltage waveform data, a frequency domain feature conversion module and an abnormal waveform data detection network; the time domain features extracted by the time domain feature extraction network are fused with the frequency domain features obtained by the frequency domain feature conversion module and converted into two-dimensional feature information which is input into the abnormal waveform recognition network
[0060] Deploy the pre-trained abnormal voltage waveform recognition model to the substation inspection system to identify and locate the abnormal voltage waveform of the DC / DC power supply.
[0061] Of course, the computer program stored in the storage unit of the device for implementing the method for identifying anomaly of voltage waveform during the inspection process of a DC / DC power supply provided by an embodiment of the present invention is not limited to the method operation as described above, and can also execute the related operations of the method for identifying anomaly of voltage waveform during the inspection process of a DC / DC power supply provided by any embodiment of the present invention.
[0062] Example 3
[0063] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed, the method for identifying abnormal voltage waveform during a DC / DC power supply inspection is implemented, including:
[0064] Adopt high-precision waveform acquisition card, programmable DC power supply and programmable DC load to design and build an automated inspection platform for DC / DC power supply;
[0065] The voltage waveform data is acquired by using an automated inspection platform and the voltage waveform data is preprocessed to obtain training data; a voltage abnormal waveform recognition model is constructed by using the training data to train the voltage abnormal waveform recognition model so that the voltage abnormal waveform recognition model can locate the voltage abnormal waveform in the voltage waveform, wherein the voltage abnormal waveform recognition model comprises: a time domain feature extraction network for voltage waveform data, a frequency domain feature conversion module and an abnormal waveform data detection network; the time domain features extracted by the time domain feature extraction network are fused with the frequency domain features obtained by the frequency domain feature conversion module and converted into two-dimensional feature information which is input into the abnormal waveform recognition network
[0066] Deploy the pre-trained abnormal voltage waveform recognition model to the substation inspection system to identify and locate the abnormal voltage waveform of the DC / DC power supply.
[0067] A computer-readable storage medium provided in an embodiment of the present invention stores a computer program which is not limited to the method operations described above, but can also execute related operations in a method for identifying voltage waveform anomalies during a DC / DC power supply inspection process provided in any embodiment of the present invention.
[0068] In the embodiments provided by the present invention, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the structural embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as 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 an indirect coupling or communication connection through some interfaces, structures or units, which can be electrical, mechanical or other forms.
[0069] 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 network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0070] 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.
[0071] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.
Claims
1. A method for identifying abnormal voltage waveform during DC / DC power supply inspection, characterized in that: include: Adopt high-precision waveform acquisition card, programmable DC power supply and programmable DC load to design and build an automated inspection platform for DC / DC power supply; Using an automated inspection platform to obtain voltage waveform data and preprocessing the voltage waveform data to obtain training data; The voltage abnormal waveform recognition model constructed by training the training data enables the voltage abnormal waveform recognition model to locate the voltage abnormal waveform in the voltage waveform, wherein the voltage abnormal waveform recognition model comprises: a time domain feature extraction network for voltage waveform data, a frequency domain feature conversion module and an abnormal waveform data detection network; the time domain features extracted by the time domain feature extraction network are fused with the frequency domain features obtained by the frequency domain feature conversion module and converted into two-dimensional feature information which is input into the abnormal waveform recognition network Deploy the pre-trained abnormal voltage waveform recognition model to the substation inspection system to identify and locate the abnormal voltage waveform of the DC / DC power supply.
2. The method for identifying voltage waveform anomalies during DC / DC power supply inspection according to claim 1, characterized in that: An automated inspection platform for DC / DC power supplies is designed using a high-precision waveform acquisition card, a programmable DC power supply, and a programmable DC load. The high-precision waveform acquisition card collects voltage waveform data of the front-stage input and the back-stage output of the DC / DC power supply. The programmable DC power supply is used to adjust and control the input voltage of the DC / DC power supply under inspection; the programmable DC load is used to adjust and control the output load of the DC / DC power supply under inspection; an inspection process rule base for each inspection item of the DC / DC power supply is designed, and the control center controls the operation of the programmable DC power supply and the programmable DC load according to the inspection process rules in the inspection process rule base, simulating the operation of the DC / DC power supply in different scenarios; wherein the inspection process rule base contains normal rules and abnormal rules; the DC / DC power supply is controlled to be in normal operating condition by the normal rules, and the DC / DC power supply is controlled to trigger abnormal mutation conditions by the abnormal rules.
3. The method for identifying voltage waveform anomalies during DC / DC power supply inspection according to claim 1, characterized in that: When obtaining training data, the preprocessing of the voltage waveform data includes: cutting into sub-segment data of uniform length; screening and removing irrelevant data and low-quality data; and performing data expansion on the sub-segment data.
4. The method for identifying voltage waveform anomalies during DC / DC power supply inspection according to claim 1, characterized in that: The time domain feature extraction network is designed using a one-dimensional convolutional neural network, which includes: a one-dimensional convolutional layer, a maximum pooling layer and an average pooling layer, an activation function and a fully connected layer; the one-dimensional convolutional layer extracts local features of the voltage waveform from the voltage waveform data sequence; a combination of maximum pooling and average pooling technology is used to reduce the dimension and number of local features of the DC / DC power supply voltage abnormal waveform, thereby extracting the most important and critical features of the voltage abnormal waveform that appears in the entire inspection process; nonlinear mapping is introduced through the activation function, so that the voltage waveform data time domain feature extraction network can learn a more complex feature representation of the DC / DC power supply voltage abnormal waveform; the fully connected layer is used to integrate the features of the DC / DC power supply voltage abnormal waveform extracted by the convolutional layer, so as to learn a higher-order feature representation of the DC / DC power supply voltage abnormal waveform.
5. The method for identifying voltage waveform anomaly during DC / DC power supply inspection according to claim 1, characterized in that: The frequency domain feature conversion module uses Fourier transform to convert the time domain features of the DC / DC power supply voltage waveform into frequency domain features.
6. The method for identifying abnormal voltage waveform during the DC / DC power supply inspection process according to claim 1, characterized in that: The number of channels, filter configuration and step size strategy of the one-dimensional convolutional layer of the time domain feature extraction network are designed to match the number of Fourier transform filter groups, preset window length and frame shift scale during frequency domain feature extraction to ensure that the time domain feature dimension extracted from the voltage waveform signal is consistent with the frequency domain feature dimension of the Fourier transform. By superimposing the channel dimension, the deep fusion of time domain and frequency domain features is achieved.
7. The method for identifying voltage waveform anomalies during DC / DC power supply inspection according to claim 1, characterized in that: The abnormal waveform recognition network includes: a stacked bottleneck layer, a global depth-by-depth convolution layer is connected after the bottleneck layer, and a linear transformation layer is arranged after the global depth-by-depth convolution layer.
8. The method for identifying voltage waveform anomalies during the DC / DC power supply inspection process according to claim 7, characterized in that: The bottleneck layer includes: an expansion layer for expanding the number of feature channels implemented by a 1×1 convolution layer and a ReLU activation function; the expansion layer is connected to several deep convolution layers, the deep convolution layer is connected to a linear bottleneck layer as a linear activation function, and the linear bottleneck layer is connected to a projection layer implemented by a 1×1 convolution layer to restore the number of channels; the deep convolution layer step size of the bottleneck layer includes 1 and 2. When the step size is 2, the bottleneck layer is used to implement downsampling, and when the step size is 1, the input and output of the bottleneck layer are connected through a residual connection.
9. A device for identifying abnormal voltage waveform during DC / DC power supply inspection, characterized in that: include: At least one processing unit, the processing unit, the acquisition unit and the storage unit are interconnected through a bus unit, the storage unit stores a computer program, and when the computer program is executed by the processing unit, the method for identifying abnormal voltage waveform during the DC / DC power supply inspection process as described in any one of claims 1-8 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for identifying abnormal voltage waveform during a DC / DC power supply inspection process according to any one of claims 1 to 8 is implemented.
