Secondary screen cabinet wiring terminal alternating current and direct current voltage identification circuit and separation and classification method

Abstract

The invention discloses a secondary screen cabinet wiring terminal alternating current and direct current voltage identification circuit and a separation and classification method, and relates to the technical field of power electronics and signal identification. According to the invention, firstly, a pre-amplification and AC / DC separation circuit mainly composed of a pre-amplification module, a high-pass filtering module and a low-pass filtering module is used for completing amplification and AC / DC separation after voltage signals of a secondary screen cabinet wiring terminal are acquired; converting the time sequence of the AC voltage signal into image data through a Grubby angle difference field; and then optimizing by using the image data through establishing and training a model based on a convolutional neural network, and continuously iterating to identify and classify the optimal parameter of the AC voltage signal of the secondary screen cabinet wiring terminal. Finally, after model training is completed, the test set is input into the model to determine the accuracy of the model.

Description

Technical Field

[0001] The invention relates to the field of power electronics and signal recognition, and in particular to an AC / DC voltage recognition circuit for secondary panel cabinet connection terminals, and a separation and classification method. Background Art

[0002] The secondary equipment of the substation is mainly used to protect and control the primary equipment. It reflects the location of the fault point by collecting the current, voltage and status information of the primary equipment, and acts accurately and quickly at the fault point, ensuring the safety of the primary equipment and the reliable power supply of the power system. It also affects the safety and stability of the power grid to a certain extent. Therefore, it is of great significance to identify and classify the status of the secondary equipment and ensure the safe and stable operation of the secondary equipment.

[0003] During the equipment replacement and maintenance of the secondary cabinet of the substation, after the DC power supply is connected, all the relevant circuits should be DC. However, in actual operation, AC will appear due to electromagnetic induction and other reasons, resulting in the phenomenon of AC and DC mixing.

[0004] In the prior art, two people usually work together, one person switches the DC power air switch of each circuit one by one, and the other uses a multimeter to measure the voltage status of each circuit power point on the terminal block. The staff will face the situation of using a multimeter to repeatedly test all the power points of the terminal block every time the air switch is pulled and closed. If the cooperation and records are not in place, it will lead to mismeasurement and omission of circuit power points, resulting in risks such as false start of the device and grounding of the DC power supply during the subsequent use of the device, threatening the safe and stable operation of the power system. At the same time, in the existing detection methods, the problem of mixed AC and DC voltage signals at the secondary cabinet terminals will make it impossible for the operation and maintenance personnel to determine the current power status of the secondary cabinet terminals, and it is difficult to determine the current status of the secondary cabinet equipment.

[0005] Therefore, there is an urgent need for a circuit that can effectively separate the AC and DC voltage signals of the secondary cabinet terminals, accurately distinguish and separate the AC and DC voltage signals of the secondary cabinet terminals, and a method that can better identify and classify the energized conditions of the secondary cabinet terminals, so as to reduce the equipment replacement and maintenance risks of the secondary cabinets in substations and ensure the personal safety of operation and maintenance personnel.

[0006] In the prior art, for example, the publication number CN110251101A discloses a pulse diagnosis instrument based on a limb clamp, which includes an AC / DC separation circuit, which separates the pressure sensing signal into a DC signal and an AC signal, wherein Out1 is a DC signal output, which is used to analyze the light, medium and heavy information during pulse diagnosis, and the DC signal can be adjusted and filtered by adjusting the parameters of R5 and C2; Out2 is an AC signal output, which is used to collect the pressure change value when the pulse beats, and the signal filtering parameters are adjusted by C1 and R4 to obtain the AC signal in the pulse frequency domain. This separation circuit adds a path in the hardware, but separates the pressure signal and the pulse signal of the pulse diagnosis, provides a signal basis for the subsequent amplification circuit, and ensures the high-precision acquisition of the AC signal. However, the specific parameter ratio of the resistor and the capacitor cannot be applied to the separation of the AC / DC voltage signal of the secondary panel cabinet terminal, and cannot achieve a steeper frequency response, and cannot accurately separate the AC / DC voltage of the secondary panel cabinet terminal to avoid the interference caused by the aliasing of the AC / DC voltage.

[0007] In the prior art, for example, the publication number CN114428235A discloses a spatial micro-motion target recognition method based on decision-level fusion, which uses a convolutional neural network to extract the physical characteristics of the target radar cross-sectional area and time-frequency diagram from narrowband echoes, and uses a convolutional neural network to extract data domain features from high-resolution one-dimensional range images and range-instantaneous Doppler images, and then uses particle swarm optimization for decision-level fusion, which not only effectively improves the performance of target recognition, but also has good generalization performance and noise resistance. However, its application scenarios and specific algorithms are not suitable for separating AC and DC voltage signals of secondary cabinet terminals. Summary of the invention

[0008] 1. Technical issues to be resolved

[0009] In view of the deficiencies in the prior art, the present invention provides an AC / DC voltage identification circuit for secondary panel cabinet terminals, and a separation and classification method, which solves the technical problems mentioned in the background technology.

[0010] (II) Technical solution

[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for identifying and classifying AC and DC voltage signals of secondary cabinet terminals, characterized in that it includes the following steps:

[0012] S1: The AC voltage signal time series is converted into image data through the Gram angle difference field. Based on the principle of the Gram angle difference field matrix, the Gram angle difference field is an image representing the time series in the polar coordinate system. Therefore, each element of the Gram angle difference field matrix represents the relationship between each time point.

[0013] S2: For a typical AC voltage signal with n observations, the time series is represented as Z = {z1, z2, ..., z n}, standardize the time series Z to between -1 and 1, and the calculation formula is as follows:

[0014]

[0015] Among them, z i Represents the typical AC voltage signal of the i-th observation value, Represents the value of the typical AC voltage signal of the i-th observation value after being standardized to between -1 and 1;

[0016] S3: Convert the one-dimensional time series Z = {z1,z2,......,z n}Convert to polar coordinate system, the conversion formula is as follows:

[0017]

[0018] The angle φ is obtained by taking the standardized observed value The arc cosine is determined by the time label OK, i is the timestamp, Q is a constant factor used to regularize the span of polar coordinates;

[0019] S4: The change of the original amplitude sequence value will cause the angle change in the polar coordinate system over time. The time correlation between different time intervals is calculated by the trigonometric difference between each sampling point. The calculated polar coordinates are used to evaluate the GADF matrix as follows:

[0020]

[0021] S5: The AC voltage signal time series is converted into a GADF image through a GADF matrix, and its features are extracted through a convolutional neural network. The polar coordinates are extracted from the time signal using the formula in S3. The GADF matrix is ​​calculated using the formula in S4 using the obtained polar coordinates. The obtained matrix is ​​visualized as an image, which is consistent near the main diagonal. Different AC voltage signals are represented by different images, thereby generating different features, thereby improving classification accuracy.

[0022] S6: Use convolutional neural networks to assign weights and biases to each feature in the image to distinguish different features; convolutional neural networks include multiple layers of filtering stages to extract features from the input image and a classification stage;

[0023] The convolution layer uses a two-dimensional convolution method to transform lower-level features into relatively higher-level features;

[0024] The convolutional layer consists of multiple filters, each of which outputs a feature map;

[0025] The kernel function generates a feature map based on the provided image;

[0026] The convolution operation uses the following formula:

[0027]

[0028] Among them, W k It is c k The weight matrix of the feature, the input data is represented as y, the symbol * represents a two-dimensional convolution operation, f(.) is a nonlinear activation function, B is a bias term, and k is the number of kernels; this convolution layer extracts all hidden features;

[0029] S7: In order to improve the convergence speed and reduce overfitting, batch normalization is used after each convolution layer. Then, the convolution layer and batch normalization layer are followed by activation operations. The purpose is to add nonlinear characteristics to the CNN framework to improve the learning effect. Due to the general short training time and the existence of gradient reduction problems, the present invention selects the rectified linear unit (ReLU) activation function to implement the nonlinear activation operation. The ReLU function is as follows:

[0030]

[0031] Where N is the total set of AC voltage categories, H is the output of the fully connected layer (FC layer), I is the total number of AC voltage categories, and the exp(x) function calculates the exponential value e of the floating point parameter x. x , where e is equal to 2.17128128…);

[0032] The correct category is determined by the maximum probability value, and the category with the largest probability value is the final judgment result of the category.

[0033] Preferably, the classification steps of S5 and S6 are as follows:

[0034] S51: convert the AC voltage signal time series data into a polar coordinate system, and then apply the S4 formula to calculate the corresponding GADF matrix;

[0035] S52: Set random values ​​for CNN parameters such as learning rate, batch size, and number of cores per layer;

[0036] S53: All original images are divided into two categories: training set and test set. Through the training images, the model performs forward propagation to evaluate the performance probability of each type;

[0037] In order to determine the output layer, the cross entropy function is used as the loss function. The specific formula is as follows:

[0038]

[0039] in, represents the true distribution, b i Represents the network output distribution, n represents the total number of categories; S54: Estimates the error gradient of all parameters and weights in the model. In order to minimize the output error, the gradient descent method is used to adjust all convolution kernel parameters;

[0040] S55: for each image in the training set, repeat steps S52 to S54 until the error is within a predetermined range;

[0041] S56: After the model training is completed, the test set is input into the model to determine its accuracy.

[0042] Preferably, the framework using convolutional neural network includes three convolutional layers, three maximum pooling layers, one fully connected layer, and one output layer. In practical applications, the learning rate, the number of layers in the network, and the size of the convolution kernel have a significant impact on the accuracy of the framework using convolutional neural network. Appropriate parameters are selected by heuristic methods. The convolutional layers include 0, 1, 2, 3, 4, 5, 6, 7, and 8 layers. The fully connected layers include 9, 10, and 11 layers. The maximum pooling layers include 3, 4, 5, 6, 7, 8, 9, 10, and 11 layers.

[0043] The types of the 0th layer, the 1st layer, and the 2nd layer are batch normalization of the input image convolution, and the activation function is ReLU. The convolution kernel size and stride of the 1st layer are 6x6x6 and 2 respectively. The output size of the 0th layer is 400x400, the output size of the 1st layer is 400x400x32, and the output size of the 2nd layer is 400x400x32.

[0044] The types of the 3rd, 4th and 5th layers are maximum pooling convolution batch normalization, the activation function is ReLU, the convolution kernel size and stride of the 3rd layer are 4x4 and 4 respectively, the convolution kernel size and stride of the 4th layer are 6x6x32 and 2 respectively, the output size of the 3rd layer is 200x200x32, the output size of the 4th layer is 200x200x64, and the output size of the 5th layer is 200x200x64;

[0045] The types of the 6th, 7th and 8th layers are maximum pooling convolution batch normalization, the activation function is ReLU, the convolution kernel size and stride of the 6th layer are 4x4 and 4 respectively, the convolution kernel size and stride of the 7th layer are 6x6x64 and 2 respectively, the output size of the 6th layer is 100x100x64, the output size of the 7th layer is 100x100x128, and the output size of the 8th layer is 100x100x128;

[0046] The types of the 9th, 10th and 11th layers are maximum pooling fully connected normalization, the convolution kernel size and stride of the 9th layer are 4x4 and 4 respectively, and the output size of the 9th layer is 50x50x128; the output size of the 10th layer is 2x2x24, and the output size of the 11th layer is 2x2x24.

[0047] A secondary panel cabinet terminal AC / DC voltage signal separation circuit, comprising a preamplifier and AC / DC separation module, the preamplifier and AC / DC separation module comprising a preamplifier module, a high-pass filter module and a low-pass filter module, the preamplifier module and the AC / DC separation module are used to preliminarily amplify and filter the input signal obtained from the secondary panel cabinet terminal, and separate the AC / DC signal;

[0048] The preamplifier module includes an operational amplifier U1, a resistor R1, a resistor R2, and a resistor R3. The positive input terminal of the operational amplifier U1 is connected to the input signal INPUT. The resistor R1 is arranged in series between the positive input terminal of the operational amplifier U1 and the input signal INPUT. The negative input terminal of the operational amplifier U1 is grounded. The resistor R2 is arranged in series between the negative input terminal of the operational amplifier U1 and the ground. The resistor R3 is connected in parallel between the output terminal of the operational amplifier U1 and the negative input terminal of the operational amplifier U1. The output terminal of the operational amplifier U1 is connected to the high-pass filter module and the low-pass filter module.

[0049] The high-pass filter module includes an operational amplifier U2, a capacitor C1, a capacitor C2, a capacitor C3, a resistor R4, and a resistor R5. The positive input terminal of the operational amplifier U2 is connected to the output terminal of the operational amplifier U1. The capacitors C1, C2, and C3 are sequentially connected in series between the positive input terminal of the operational amplifier U2 and the output terminal of the operational amplifier U1. The resistor R4 is a resistor grounded between the capacitor C1 and the capacitor C2. The resistor R5 is a resistor grounded between the capacitor C2 and the capacitor C3. The negative input terminal of the operational amplifier U2 is connected to the output terminal of the operational amplifier U2 and then outputs an AC voltage after the AC and DC voltages of the secondary cabinet wiring terminal are separated.

[0050] The low-pass filter module is mainly composed of an operational amplifier U3, a resistor R6, a resistor R7, a resistor R8, a capacitor C4, and a capacitor C5. The positive input terminal of the operational amplifier U3 is connected to the output terminal of the operational amplifier U1. The resistors R6, R7, and R8 are connected in series between the positive input terminal of the operational amplifier U3 and the output terminal of the operational amplifier U1. The capacitor C4 is a resistor grounded between the resistor R6 and the resistor R7. The capacitor C5 is a resistor grounded between the resistor R7 and the resistor R8. The negative input terminal of the operational amplifier U3 is connected to the output terminal of the operational amplifier U3. The negative input terminal of the operational amplifier U3 is connected to the output terminal of the operational amplifier U3, and the DC voltage after the AC and DC voltages of the secondary panel cabinet terminal are separated is output after the connection point between the negative input terminal of the operational amplifier U3 and the output terminal of the operational amplifier U3.

[0051] Preferably, the amplification factor of the input voltage signal by the preamplifier module is the ratio of the resistor R3 to the resistor R2 plus one.

[0052] Preferably, the ratio of the capacitances of the capacitors C1, C2 and C3 of the high-pass filter module is 1:2:1, and the resistances of the resistors R4 and R5 are equal.

[0053] Preferably, the resistance ratio of the resistor R6, the resistor R7, and the resistor R8 of the low-pass filter module is 1:2:1, and the capacitance of the capacitor C4 is equal to that of the capacitor C5.

[0054] A separation method for an AC / DC voltage signal separation circuit of a secondary panel cabinet terminal, the separation method comprising the following steps:

[0055] S1A: The preamplifier module is used to amplify the input signal to increase the signal strength and ensure the signal stability and accuracy in the subsequent processing links;

[0056] S2A: The operational amplifier U1 is a core component, which is used to amplify the input signal. The gain of the amplifier is set by the resistor network R1, R2 and R3 to ensure that the signal strength meets the requirements of the subsequent processing module; the resistor R1 connects the input signal INPUT and the positive input terminal of U1, which is used to guide the signal to the amplifier and appropriately attenuate or adjust the signal voltage; one end of the resistor R2 is grounded, and the other end is connected to the negative input terminal of U1, which is used to form the feedback circuit of the amplifier; the resistor R3 connects the output terminal and the negative input terminal of U1 as part of the feedback path to ensure the linearity and stability of the signal after amplification; through this pre-amplification circuit, the input signal is sufficiently amplified without distortion or noise introduction, so that the subsequent filtering process can more effectively separate the AC and DC signals;

[0057] S3A: The high-pass filter module is used to extract the AC component from the amplified signal and remove the DC part. This module is based on the operational amplifier U2 and combines the filter network composed of capacitors and resistors to separate the AC signal;

[0058] S4A: Capacitors C1, C2 and C3 form a three-stage series capacitor network, which, as the core component of the high-pass filter, can block the passage of DC signals and allow AC components with higher frequencies to pass smoothly; resistor R4 connects capacitors C1 and C2 and is grounded to provide a signal discharge path to prevent DC signal accumulation; resistor R5 is similar to resistor R4, connecting capacitors C2 and C3 and is grounded to further enhance the filtering effect; the positive input terminal of the operational amplifier U2 is connected to the capacitor network to extract the filtered AC signal and perform appropriate amplification;

[0059] The negative input terminal of U2 is directly connected to its output terminal to form a unit gain configuration to ensure that the output signal is not distorted; finally, after passing through the high-pass filter module, the AC signal in the secondary cabinet wiring terminal is separated and output for subsequent identification and classification;

[0060] S5A: The low-pass filter module is specifically used to extract the DC component in the amplified signal. The low-pass filter module is based on the operational amplifier U3 and a filter network composed of resistors and capacitors, which can effectively remove AC noise and extract a stable DC voltage.

[0061] S6A: Resistors R6, R7 and R8 form a resistor network for DC signal filtering. R6 is connected to the output end of the operational amplifier U1. Resistors R7 and R8 are connected in series to form a filtering path, which adjusts the time constant of the circuit to determine the passing range of the low-frequency signal. Capacitors C4 and C5 are connected in parallel with resistors R6 and R7 to remove high-frequency AC components so that low-frequency DC signals can pass through the resistor network smoothly. The positive input end of the operational amplifier U3 is connected to the filtered signal to amplify and buffer the DC signal. The negative input end of U3 is directly connected to the output end to maintain unit gain. Finally, the output of the low-pass filter module is a DC voltage signal extracted from the input signal, which can be used for subsequent detection and analysis.

[0062] (III) Beneficial effects

[0063] The present invention provides a secondary panel cabinet terminal AC and DC voltage identification circuit, separation and classification method. It has the following beneficial effects:

[0064] (1) The secondary cabinet terminal AC and DC voltage identification circuit adopts a multi-stage filter, which can achieve a steeper frequency response, thereby accurately separating the secondary cabinet terminal AC and DC voltages and avoiding interference caused by AC and DC voltage aliasing.

[0065] (2) The preamplifier module of the secondary cabinet terminal AC / DC voltage identification circuit and separation method provides clear and accurate voltage signal amplification. By reasonably selecting the resistance value, adjustable gain is achieved to meet the voltage signal amplification requirements of equipment replacement and maintenance scenarios of secondary cabinets in different substations.

[0066] (3) In the classification method of AC and DC voltage at the secondary cabinet terminals, GADF-CNN involves fewer model parameters and layers, takes the shortest training time, and has a smaller standard deviation than other methods. In terms of recognition and classification of AC and DC voltage signals at the secondary cabinet terminals, GADF-CNN has good results in terms of accuracy and noise resistance. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a flow chart of AC and DC voltage signal separation of the present invention;

[0068] Figure 2 It is a schematic diagram of the structure of the AC / DC voltage signal separation circuit of the present invention;

[0069] Figure 3 The present invention is a flow chart of the AC voltage signal recognition and classification method. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.

[0071] The specific operation of the preamplifier and AC / DC separation circuit is as follows:

[0072] like Figure 2As shown, the present invention provides a secondary cabinet wiring terminal AC / DC voltage signal separation circuit, which is used to preliminarily amplify and filter the input signal obtained from the secondary cabinet wiring terminal, and separate the AC / DC signal. It is mainly composed of three submodules: a preamplifier module, a high-pass filter module and a low-pass filter module, each of which performs a specific task to ensure that the input signal can be effectively processed and classified. It includes an operational amplifier U1, an operational amplifier U2, an operational amplifier U3, a resistor R1, a resistor R2, a resistor R3, a resistor R4, a resistor R5, a resistor R6, a resistor R7, a resistor R8, a capacitor C1, a capacitor C2, a capacitor C3, a capacitor C4, and a capacitor C5. Among them, the resistor R1 is 50Ω, the resistor R2 is 40Ω, the resistor R3 is 200Ω, the resistor R4 is 200Ω, the resistor R5 is 200Ω, the resistor R6 is 100Ω, the resistor R7 is 200Ω, and the resistor R8 is 100Ω. Meanwhile, the capacitor C1 is 10 μF, the capacitor C2 is 20 μF, the capacitor C3 is 10 μF, the capacitor C4 is 10 μF, and the capacitor C5 is 10 μF. In implementation, the operational amplifier U1, the operational amplifier U2, and the operational amplifier U3 are all implemented using the OPA2134UA operational amplifier.

[0073] The preamplifier module is used to amplify the input signal to improve the signal strength and ensure the signal stability and accuracy in the subsequent processing links. The design of this module is based on the operational amplifier U1, and the gain is adjusted through an external resistor network. The preamplifier module is mainly composed of the operational amplifier U1, a resistor R1 with one end connected to the input signal INPUT and the other end connected to the positive input terminal of the operational amplifier U1, a resistor R2 with one end connected to the ground and the other end connected to the negative input terminal of the operational amplifier U1, and a resistor R3 with one end connected to the negative input terminal of the operational amplifier U1 and the other end connected to the output terminal of the operational amplifier U1;

[0074] The operational amplifier U1 is the core component, which is used to amplify the input signal. The gain of the amplifier is set by the resistor network R1, R2 and R3 to ensure that the signal strength meets the requirements of the subsequent processing module; the resistor R1 connects the input signal (INPUT) and the positive input terminal of U1, which is used to guide the signal to the amplifier and appropriately attenuate or adjust the signal voltage; one end of the resistor R2 is grounded, and the other end is connected to the negative input terminal of U1 to form the feedback circuit of the amplifier; the resistor R3 connects the output terminal and the negative input terminal of U1 as part of the feedback path to ensure the linearity and stability of the signal after amplification; through this pre-amplification circuit, the input signal is sufficiently amplified without distortion or noise introduction, so that the subsequent filtering process can more effectively separate the AC and DC signals;

[0075] The high-pass filter module is used to extract the AC component from the amplified signal and remove the DC part. The module is based on the operational amplifier U2, and is combined with a filter network composed of capacitors and resistors to separate the AC signal. The high-pass filter module mainly consists of the operational amplifier U2, a capacitor C1 with one end connected to the output end of the operational amplifier U1 and the other end connected to the capacitor C2, a capacitor C2 with one end connected to the capacitor C1 and the other end connected to the capacitor C3, a capacitor C3 with one end connected to the capacitor C2 and the other end connected to the positive input end of the operational amplifier U2, a resistor R4 with one end connected to the capacitor C1 and the capacitor C2 and the other end grounded, a resistor R5 with one end connected to the capacitor C2 and the capacitor C3 and the other end grounded, the negative input end of the operational amplifier U2 is connected to the output end of the operational amplifier U2, and the negative input end of the operational amplifier U2 is connected to the output end of the operational amplifier U2, and the output point of the secondary panel cabinet connection terminal outputs the AC voltage after the AC and DC voltages are separated;

[0076] Capacitors C1, C2, and C3 form a three-stage series capacitor network. As the core component of the high-pass filter, it can block the passage of DC signals and allow higher-frequency AC components to pass smoothly. Resistor R4 connects capacitors C1 and C2 and is grounded to provide a signal discharge path to prevent DC signal accumulation. Resistor R5 is similar to R4, connecting capacitors C2 and C3 and grounding to further enhance the filtering effect. The positive input terminal of operational amplifier U2 is connected to the capacitor network to extract the filtered AC signal and amplify it appropriately. The negative input terminal of U2 is directly connected to its output terminal to form a unit gain configuration to ensure that the output signal is not distorted. Finally, after passing through the high-pass filter module, the AC signal in the secondary cabinet terminal is separated and output for subsequent identification and classification.

[0077] The low-pass filter module is specifically used to extract the DC component in the amplified signal. This module is based on the operational amplifier U3, combined with a filter network composed of resistors and capacitors, which can effectively remove AC noise and extract a stable DC voltage. The low-pass filter module mainly consists of the operational amplifier U3, a resistor R6 with one end connected to the output end of the operational amplifier U1 and the other end connected to the resistor R7, a resistor R7 with one end connected to the resistor R6 and the other end connected to the resistor R8, a resistor R8 with one end connected to the resistor R7 and the other end connected to the positive input end of the operational amplifier U3, a capacitor C4 with one end connected to the resistor R6 and the resistor R7 and the other end grounded, a capacitor C5 with one end connected to the resistor R7 and the resistor R8 and the other end grounded, the negative input end of the operational amplifier U3 is connected to the output end of the operational amplifier U3, and the negative input end of the operational amplifier U3 is connected to the output end of the operational amplifier U3. The DC voltage after the AC and DC voltages of the secondary cabinet wiring terminal are separated after the connection point between the negative input end of the operational amplifier U3 and the output end of the operational amplifier U3;

[0078] Resistors R6, R7, and R8 form a resistor network for DC signal filtering. R6 is connected to the output end of the operational amplifier U1, and R7 and R8 are connected in series to form a filtering path to adjust the time constant of the circuit to determine the passing range of the low-frequency signal. Capacitors C4 and C5 are connected in parallel with resistors R6 and R7 to remove high-frequency AC components so that low-frequency DC signals can pass through the resistor network smoothly. The positive input end of the operational amplifier U3 is connected to the filtered signal to amplify and buffer the DC signal, and the negative input end of U3 is directly connected to the output end to maintain unit gain. Finally, the output of the low-pass filter module is a DC voltage signal extracted from the input signal, which can be used for subsequent detection and analysis.

[0079] When connected, one end of the resistor R1 is connected to the input voltage signal INPUT, and the other end is connected to the positive input terminal of the operational amplifier U1, one end of the resistor R2 is grounded, and the other end is connected to the negative input terminal of the operational amplifier U1, one end of the resistor R3 is connected to the negative input terminal of the operational amplifier U1, and the other end is connected to the output terminal of the operational amplifier U1, one end of the capacitor C1 is connected to the output terminal of the operational amplifier U1, and the other end is connected to the capacitor C2, one end of the capacitor C2 is connected to the capacitor C1, and the other end is connected to the capacitor C3, one end of the capacitor C3 is connected to the capacitor C2, and the other end is connected to the positive input terminal of the operational amplifier U2, one end of the resistor R4 is connected to the capacitor C1 and the capacitor C2, and the other end is grounded, one end of the resistor R5 is connected to the capacitor C2 and the capacitor C3, and the other end is grounded, the negative input terminal of the operational amplifier U2 is connected to the operational amplifier The output end of the operational amplifier U2, the negative input end of the operational amplifier U2 is connected to the output end of the operational amplifier U2, and then the AC voltage of the secondary panel cabinet connection terminal after the AC and DC voltages are separated is output; one end of the resistor R6 is connected to the output end of the operational amplifier U1, and the other end is connected to the resistor R7; one end of the resistor R7 is connected to the resistor R6, and the other end is connected to the resistor R8; one end of the resistor R8 is connected to the resistor R7, and the other end is connected to the positive input end of the operational amplifier U3; one end of the capacitor C4 is connected to the resistor R6 and the resistor R7, and the other end is grounded; one end of the capacitor C5 is connected to the resistor R7 and the resistor R8, and the other end is grounded; the negative input end of the operational amplifier U3 is connected to the output end of the operational amplifier U3; the negative input end of the operational amplifier U3 is connected to the output end of the operational amplifier U3, and then the DC voltage of the secondary panel cabinet connection terminal after the AC and DC voltages are separated is output;

[0080] The preamplification and AC / DC separation circuit of the present invention has a simple structure, low cost, stable performance and high separation efficiency. When in use, the input voltage signal is connected to the positive input terminal of the operational amplifier U1 through the resistor R1, and the input voltage signal INPUT and the operational amplifier U1, the resistor R1, the resistor R2 and the resistor R3 constitute an active preamplification module to amplify the input voltage signal, and the amplification factor is the ratio of the resistor R3 to the resistor R2 plus one, so that it is suitable for subsequent filtering and processing. The amplified voltage signal and the operational amplifier U2, the resistor R4, the resistor R5, the capacitor C1, the capacitor C2 and the capacitor C3 constitute an active high-pass filter module to filter out the low-frequency DC voltage and retain the high-frequency AC voltage, and output the AC voltage after the AC / DC voltage separation of the secondary panel cabinet terminal. The amplified voltage signal and the operational amplifier U3, the resistor R6, the resistor R7, the resistor R8, the capacitor C4, the capacitor C5 constitute an active low-pass filter module to filter out the high-frequency AC voltage and retain the low-frequency DC voltage, and output the DC voltage after the AC / DC voltage separation of the secondary panel cabinet terminal.

[0081] The specific operations of AC voltage signal recognition and classification are as follows:

[0082] The AC voltage signal time series is converted into image data through the Gram angle difference field. Based on the principle of the Gram angle difference field matrix, the Gram angle difference field is an image representing the time series in the polar coordinate system. Therefore, each element of the Gram angle difference field matrix represents the relationship between each time point.

[0083] For a typical AC voltage signal time series with n observations, Z = {z1, z2, ..., z n}. Normalize the time series Z to between -1 and 1, and the calculation formula is as follows:

[0084]

[0085] Where zi represents the typical AC voltage signal of the i-th observation value, Represents the value of the typical AC voltage signal of the i-th observation value after being standardized to between -1 and 1.

[0086] Furthermore, the one-dimensional time series Z = {z1, z2, ..., z n}Convert to polar coordinate system, the conversion formula is as follows:

[0087]

[0088] The angle φ is obtained by taking the standardized observed value The arc cosine is determined by the time label OK, iis the timestamp and Q is a constant factor used to regularize the span of the polar coordinates.

[0089] The change of the original amplitude sequence value will cause the angle change in the polar coordinate system over time. The time correlation between different time intervals is calculated by the trigonometric difference between each sampling point. The calculated polar coordinates are used to evaluate the GADF matrix as follows:

[0090]

[0091] The AC voltage signal time series is converted into a GADF image through a GADF matrix, and its features are extracted through a convolutional neural network. The polar coordinates are extracted from the time signal using the formula in S3, and the GADF matrix is ​​calculated using the obtained polar coordinates according to the formula in S4. The resulting matrix is ​​visualized as an image, which is consistent near the main diagonal. Different AC voltage signals are represented by different images, thereby generating different features, thereby improving classification accuracy.

[0092] Convolutional neural networks are a type of neural network that takes an image as input and distinguishes between different features by assigning weights and biases to each feature in the image. Convolutional neural networks consist of multiple layers of filtering stages to extract features from the input image and a classification stage. The convolutional layer uses a two-dimensional convolution method to convert lower-level features into relatively higher-level features. The convolutional layer consists of multiple filters, each of which outputs a feature map. The kernel function generates a feature map based on the provided image. The following formula is used for the convolution operation:

[0093]

[0094] Among them, W k It is c k The weight matrix of the feature, the input data is represented as y, the symbol * represents a two-dimensional convolution operation, f(.) is a nonlinear activation function, B is a bias term, and k is the number of kernels. This convolution layer extracts all hidden features.

[0095] In order to improve the convergence speed and reduce overfitting, batch normalization is used after each convolution layer. Then, the convolution layer and batch normalization layer are followed by activation operations, the purpose of which is to add nonlinear characteristics to the CNN framework to improve the learning effect. Due to the general short training time and the existence of gradient fading problems, the present invention selects the rectified linear unit (ReLU) activation function to implement the nonlinear activation operation. The ReLU function is shown below:

[0096]

[0097] Where N is the total set of AC voltage categories, H is the output of the fully connected layer (FC layer), I is the total number of AC voltage categories, and the exp(x) function calculates the exponential value (e) of the floating-point parameter x. x , where e is equal to 2.17128128...). The correct category is determined by the maximum probability value, and the category with the largest probability value is the final judgment result of the category.

[0098] The CNN framework designed by the present invention consists of three convolutional layers, three maximum pooling layers, a fully connected layer (FC layer) and an output layer. In practical applications, key parameters such as learning rate, number of layers in the network and convolution kernel size have a significant impact on the accuracy of the CNN architecture. The CNN size and parameters of each frame are shown in Table 1, where all frames share the same weights.

[0099] Table 1 Detailed parameters of the CNN framework designed by the present invention

[0100]

[0101]

[0102] The GADF-CNN classification steps are as follows:

[0103] Step 1: Convert the AC voltage signal time series data into a polar coordinate system, and then apply the S24 formula to calculate the corresponding GADF matrix.

[0104] Step 2: Set random values ​​for CNN parameters such as learning rate, batch size, and number of cores per layer.

[0105] Step 3: All original images are divided into two categories: training set and test set. Through the training images, the model performs forward propagation to evaluate the performance probability of each type. In order to determine the output layer, the cross entropy function is used as the loss function, and the specific formula is as follows:

[0106]

[0107] in, represents the true distribution, b i Represents the network output distribution, and n represents the total number of categories.

[0108] Step 4: Estimate the error gradients of all parameters and weights in the model. In order to minimize the output error, the gradient descent method is used to adjust all convolution kernel parameters.

[0109] Step 5: For each image in the training set, repeat steps 2 to 4 until the error is within the predetermined range.

[0110] Step 6: After the model is trained, the test set is fed into the model to determine its accuracy.

[0111] Based on Matlab R2018a and Python3.7, according to IEEE-1159, 6 single AC voltage signals and 3 composite AC voltage signals were generated to verify the effectiveness of the GADF-CNN method in AC voltage signal recognition and classification. The amplitude of the synthetic signal was standardized to 1p.u., and the base frequency was set to 50Hz. Actual signals are usually affected by noise. In order to enhance the robustness of the model, Gaussian white noise was randomly added to the synthesized AC voltage signal data. In this specific example, three signal-to-noise ratios (SNRs) were used for the synthetic AC voltage signal: 15dB, 30dB, and 45dB. GADF was used to encode the AC voltage time series into a two-dimensional image. The input of the CNN framework is an image of 400*400 pixels. A maximum of 315 images were used for training (35 images per class), and 85 images per class were used for testing. On all test sets, the AC voltage signal recognition and classification method of the secondary cabinet terminal of the present invention was compared with the existing methods.

[0112] After the training process, the proposed method showed good performance. Table 2 shows the classification results of the test set. According to the research results, the overall classification accuracy of the AC voltage signal is 99.5%, which shows the accuracy of the GADF-CNN framework proposed in the present invention in the recognition and classification of AC voltage signals of secondary cabinet terminals.

[0113] Table 2 GADF-CNN classification results

[0114]

[0115]

[0116] In order to further evaluate the effectiveness of GADF-CNN in the present invention in identifying and classifying AC voltage signals at the secondary cabinet terminals, two other CNN technologies, namely CNN-LSTM and KBN-CNN, were used in this embodiment for comparative study.

[0117] In this embodiment, the CNN-LSTM includes two convolutional layers and a long short-term memory (LSTM) layer with 100 memory nodes. The depth of the CNN is set to 128 and 256. The downsampling factor and convolution kernel size are set to 6 and 4 respectively.

[0118] In the KBN-CNN of this embodiment, KBN is used to decompose the original signal to evaluate the time-frequency information, and CNN accurately defines the amplitude matrix of KBN. The downsampling factor and the convolution kernel size are both set to 6.

[0119] Table 3 gives the accuracy results of the recognition and classification of the AC voltage signal of the secondary cabinet terminal.

[0120] Table 3 Comparison of recognition and classification results accuracy

[0121]

[0122] As shown in Table 3, GADF-CNN involves fewer model parameters and layers, takes the shortest training time, and has a smaller standard deviation than other methods. Analysis of Table 3 shows that GADF-CNN has improved in accuracy and noise resistance, and GADF-CNN performs better than CNN-LSTM and KBN-CNN at different noise levels.

[0123] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants 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.

[0124] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for identifying and classifying AC and DC voltage signals of secondary cabinet terminals, characterized in that: The steps include: S1: The AC voltage signal time series is converted into image data through the Gram angle difference field. Based on the principle of the Gram angle difference field matrix, the Gram angle difference field is an image representing the time series in the polar coordinate system. Therefore, each element of the Gram angle difference field matrix represents the relationship between each time point. S2: For a typical AC voltage signal with n observations, the time series is represented as Z = {z1, z2, ..., z n }, standardize the time series Z to between -1 and 1, and the calculation formula is as follows: Where zi represents the typical AC voltage signal of the i-th observation value, Represents the value of the typical AC voltage signal of the i-th observation value after being standardized to between -1 and 1; S3: Convert the one-dimensional time series Z = {z1,z2,......,z n }Convert to polar coordinate system, the conversion formula is as follows: The angle φ is obtained by taking the standardized observed value The arc cosine is determined by the time label OK, i is the timestamp, Q is a constant factor used to regularize the span of polar coordinates; S4: The change of the original amplitude sequence value will cause the angle change in the polar coordinate system over time. The time correlation between different time intervals is calculated by the trigonometric difference between each sampling point. The calculated polar coordinates are used to evaluate the GADF matrix as follows: S5: The AC voltage signal time series is converted into a GADF image through a GADF matrix, and its features are extracted through a convolutional neural network. The polar coordinates are extracted from the time signal using the formula in S3. The GADF matrix is ​​calculated using the formula in S4 using the obtained polar coordinates. The obtained matrix is ​​visualized as an image, which is consistent near the main diagonal. Different AC voltage signals are represented by different images, thereby generating different features, thereby improving classification accuracy. S6: Use convolutional neural networks to assign weights and biases to each feature in the image to distinguish different features; convolutional neural networks include multiple layers of filtering stages to extract features from the input image and a classification stage; The convolution layer uses a two-dimensional convolution method to transform lower-level features into relatively higher-level features; The convolutional layer consists of multiple filters, each of which outputs a feature map; The kernel function generates a feature map based on the provided image; The convolution operation uses the following formula: Among them, W k It is c k The weight matrix of the feature, the input data is represented as y, the symbol * represents a two-dimensional convolution operation, f(.) is a nonlinear activation function, B is a bias term, and k is the number of kernels; this convolution layer extracts all hidden features; S7: In order to improve the convergence speed and reduce overfitting, batch normalization is used after each convolution layer. Then, the convolution layer and batch normalization layer are followed by activation operations. The purpose is to add nonlinear characteristics to the CNN framework to improve the learning effect. Due to the general short training time and the existence of gradient reduction problems, the present invention selects the rectified linear unit (ReLU) activation function to implement the nonlinear activation operation. The ReLU function is as follows: Where N is the total set of AC voltage categories, H is the output of the fully connected layer (FC layer), I is the total number of AC voltage categories, and the exp(x) function calculates the exponential value e of the floating point parameter x. x , where e is equal to 2.17128128…); The correct category is determined by the maximum probability value, and the category with the largest probability value is the final judgment result of the category.

2. A method for identifying and classifying AC and DC voltage signals of secondary cabinet terminals according to claim 1, characterized in that: The classification steps of S5 and S6 are as follows: S51: convert the AC voltage signal time series data into a polar coordinate system, and then apply the S4 formula to calculate the corresponding GADF matrix; S52: Set random values ​​for CNN parameters such as learning rate, batch size, and number of cores per layer; S53: All original images are divided into two categories: training set and test set. Through the training images, the model performs forward propagation to evaluate the performance probability of each type; In order to determine the output layer, the cross entropy function is used as the loss function. The specific formula is as follows: in, represents the true distribution, b i Represents the network output distribution, n represents the total number of categories; S54: Estimates the error gradient of all parameters and weights in the model. In order to minimize the output error, the gradient descent method is used to adjust all convolution kernel parameters; S55: for each image in the training set, repeat steps S52 to S54 until the error is within a predetermined range; S56: After the model training is completed, the test set is input into the model to determine its accuracy.

3. The method for identifying and classifying AC and DC voltage signals of secondary cabinet terminals according to claim 1 is characterized in that: The framework using the convolutional neural network includes three convolutional layers, three maximum pooling layers, a fully connected layer, and an output layer. In practical applications, the learning rate, the number of layers in the network, and the size of the convolution kernel have a significant impact on the accuracy of the framework using the convolutional neural network. Appropriate parameters are selected by a heuristic method. The convolutional layers include 0, 1, 2, 3, 4, 5, 6, 7, and 8 layers. The fully connected layers include 9, 10, and 11 layers. The maximum pooling layers include 3, 4, 5, 6, 7, 8, 9, 10, and 11 layers. The types of the 0th layer, the 1st layer, and the 2nd layer are batch normalization of the input image convolution, and the activation function is ReLU. The convolution kernel size and stride of the 1st layer are 6x6x6 and 2 respectively. The output size of the 0th layer is 400x400, the output size of the 1st layer is 400x400x32, and the output size of the 2nd layer is 400x400x32. The types of the 3rd, 4th and 5th layers are maximum pooling convolution batch normalization, the activation function is ReLU, the convolution kernel size and stride of the 3rd layer are 4x4 and 4 respectively, the convolution kernel size and stride of the 4th layer are 6x6x32 and 2 respectively, the output size of the 3rd layer is 200x200x32, the output size of the 4th layer is 200x200x64, and the output size of the 5th layer is 200x200x64; The types of the 6th, 7th and 8th layers are maximum pooling convolution batch normalization, the activation function is ReLU, the convolution kernel size and stride of the 6th layer are 4x4 and 4 respectively, the convolution kernel size and stride of the 7th layer are 6x6x64 and 2 respectively, the output size of the 6th layer is 100x100x64, the output size of the 7th layer is 100x100x128, and the output size of the 8th layer is 100x100x128; The types of the 9th, 10th and 11th layers are maximum pooling fully connected normalization, the convolution kernel size and stride of the 9th layer are 4x4 and 4 respectively, and the output size of the 9th layer is 50x50x128; the output size of the 10th layer is 2x2x24, and the output size of the 11th layer is 2x2x24.

4. A secondary panel cabinet terminal AC / DC voltage signal separation circuit, characterized in that: It includes a preamplifier and AC / DC separation module, which includes a preamplifier module, a high-pass filter module and a low-pass filter module. The preamplifier module and the AC / DC separation module are used to preliminarily amplify and filter the input signal obtained from the secondary panel cabinet terminal, and separate the AC / DC signal; The preamplifier module includes an operational amplifier U1, a resistor R1, a resistor R2, and a resistor R3. The positive input terminal of the operational amplifier U1 is connected to the input signal INPUT. The resistor R1 is arranged in series between the positive input terminal of the operational amplifier U1 and the input signal INPUT. The negative input terminal of the operational amplifier U1 is grounded. The resistor R2 is arranged in series between the negative input terminal of the operational amplifier U1 and the ground. The resistor R3 is connected in parallel between the output terminal of the operational amplifier U1 and the negative input terminal of the operational amplifier U1. The output terminal of the operational amplifier U1 is connected to the high-pass filter module and the low-pass filter module. The high-pass filter module includes an operational amplifier U2, a capacitor C1, a capacitor C2, a capacitor C3, a resistor R4, and a resistor R5. The positive input terminal of the operational amplifier U2 is connected to the output terminal of the operational amplifier U1. The capacitors C1, C2, and C3 are sequentially connected in series between the positive input terminal of the operational amplifier U2 and the output terminal of the operational amplifier U1. The resistor R4 is a resistor grounded between the capacitor C1 and the capacitor C2. The resistor R5 is a resistor grounded between the capacitor C2 and the capacitor C3. The negative input terminal of the operational amplifier U2 is connected to the output terminal of the operational amplifier U2 and then outputs an AC voltage after the AC and DC voltages of the secondary cabinet wiring terminal are separated. The low-pass filter module is mainly composed of an operational amplifier U3, a resistor R6, a resistor R7, a resistor R8, a capacitor C4, and a capacitor C5. The positive input terminal of the operational amplifier U3 is connected to the output terminal of the operational amplifier U1. The resistors R6, R7, and R8 are connected in series between the positive input terminal of the operational amplifier U3 and the output terminal of the operational amplifier U1. The capacitor C4 is a resistor grounded between the resistor R6 and the resistor R7. The capacitor C5 is a resistor grounded between the resistor R7 and the resistor R8. The negative input terminal of the operational amplifier U3 is connected to the output terminal of the operational amplifier U3. The negative input terminal of the operational amplifier U3 is connected to the output terminal of the operational amplifier U3, and the DC voltage after the AC and DC voltages of the secondary panel cabinet terminal are separated is output after the connection point between the negative input terminal of the operational amplifier U3 and the output terminal of the operational amplifier U3.

5. The AC / DC voltage signal separation circuit for secondary panel cabinet terminals according to claim 4 is characterized in that: The amplification factor of the input voltage signal by the preamplifier module is the ratio of the resistor R3 to the resistor R2 plus one.

6. The AC / DC voltage signal separation circuit for secondary cabinet wiring terminals according to claim 4 is characterized in that: The ratio of the capacitance of the capacitor C1, the capacitor C2 and the capacitor C3 of the high-pass filter module is 1:2:1, and the resistance of the resistor R4 is equal to that of the resistor R5.

7. The AC / DC voltage signal separation circuit for secondary panel cabinet terminals according to claim 4 is characterized in that: The resistance ratio of the resistor R6, the resistor R7 and the resistor R8 in the low-pass filter module is 1:2:1, and the capacitance of the capacitor C4 is equal to that of the capacitor C5.

8. A separation method using the secondary panel cabinet terminal AC and DC voltage signal separation circuit as claimed in any one of claims 4 to 7, characterized in that: The separation method comprises the following steps: S1A: The preamplifier module is used to amplify the input signal to increase the signal strength and ensure the signal stability and accuracy in the subsequent processing links; S2A: The operational amplifier U1 is a core component, which is used to amplify the input signal. The gain of the amplifier is set by the resistor network R1, R2 and R3 to ensure that the signal strength meets the requirements of the subsequent processing module; the resistor R1 connects the input signal INPUT and the positive input terminal of U1, which is used to guide the signal to the amplifier and appropriately attenuate or adjust the signal voltage; one end of the resistor R2 is grounded, and the other end is connected to the negative input terminal of U1, which is used to form the feedback circuit of the amplifier; the resistor R3 connects the output terminal and the negative input terminal of U1 as part of the feedback path to ensure the linearity and stability of the signal after amplification; through this pre-amplification circuit, the input signal is sufficiently amplified without distortion or noise introduction, so that the subsequent filtering process can more effectively separate the AC and DC signals; S3A: The high-pass filter module is used to extract the AC component from the amplified signal and remove the DC part. This module is based on the operational amplifier U2 and combines the filter network composed of capacitors and resistors to separate the AC signal; S4A: Capacitors C1, C2 and C3 form a three-stage series capacitor network, which, as the core component of the high-pass filter, can block the passage of DC signals and allow AC components with higher frequencies to pass smoothly; resistor R4 connects capacitors C1 and C2 and is grounded to provide a signal discharge path to prevent DC signal accumulation; resistor R5 is similar to resistor R4, connecting capacitors C2 and C3 and is grounded to further enhance the filtering effect; the positive input terminal of the operational amplifier U2 is connected to the capacitor network to extract the filtered AC signal and perform appropriate amplification; The negative input terminal of U2 is directly connected to its output terminal to form a unit gain configuration to ensure that the output signal is not distorted; finally, after passing through the high-pass filter module, the AC signal in the secondary cabinet wiring terminal is separated and output for subsequent identification and classification; S5A: The low-pass filter module is specifically used to extract the DC component in the amplified signal. The low-pass filter module is based on the operational amplifier U3 and a filter network composed of resistors and capacitors, which can effectively remove AC noise and extract a stable DC voltage. S6A: Resistors R6, R7 and R8 form a resistor network for DC signal filtering. R6 is connected to the output end of the operational amplifier U1. Resistors R7 and R8 are connected in series to form a filtering path, which adjusts the time constant of the circuit to determine the passing range of the low-frequency signal. Capacitors C4 and C5 are connected in parallel with resistors R6 and R7 to remove high-frequency AC components so that low-frequency DC signals can pass through the resistor network smoothly. The positive input end of the operational amplifier U3 is connected to the filtered signal to amplify and buffer the DC signal. The negative input end of U3 is directly connected to the output end to maintain unit gain. Finally, the output of the low-pass filter module is a DC voltage signal extracted from the input signal, which can be used for subsequent detection and analysis.

Images