DTU terminal for high-reliability power distribution network

Abstract

The invention provides a DTU terminal for a high-reliability power distribution network, which belongs to the technical field of power distribution network equipment and comprises a main control chip, a multi-channel current and voltage acquisition module, a frequency phase detection module, an edge calculation processing module, a hierarchical storage module, a communication interface module, a real-time protection logic module and a power management module. A multi-band power quality disturbance identification and grading response module is arranged in the main control chip and is used for performing multi-band decomposition and feature extraction on the acquired current and voltage signals, identifying weak fault features, judging disturbance types and severity and triggering corresponding processing measures according to different disturbance grades; and finally, outputting electric energy quality monitoring data and a protection control instruction. The technical problem that the DTU terminal of the power distribution network is difficult to realize early accurate identification of weak fault characteristics and real-time grading response of multi-stage disturbance under limited computing resources is solved.

Description

A DTU terminal for high-reliability distribution networks Technical Field

[0001] This invention belongs to the field of power distribution network equipment technology, and more specifically, relates to a DTU terminal for high-reliability power distribution networks. Background Technology

[0002] As the core equipment of the distribution automation system, the distributed terminal unit (DTU) of the distribution network undertakes the key tasks of power quality monitoring, fault diagnosis and protection control. Traditional DTU terminals adopt fault detection methods based on single-band analysis, which identify abnormal states of the distribution network by monitoring the amplitude changes and frequency shifts of current and voltage. It performs well in scenarios where the grid load is stable and the disturbance characteristics are obvious.

[0003] However, in actual operation, the current changes caused by early faults such as insulation degradation and increased contact resistance are only at the milliampere level, and the voltage anomalies are only at the millivolt level. These weak characteristic signals are often submerged in the normal operation noise, and traditional single-band analysis methods are difficult to effectively extract and identify them. In addition, the types of disturbances in the distribution network are diverse, ranging from slight voltage flicker to severe phase-to-phase short circuits. Different disturbances require different response strategies, while traditional methods use fixed threshold judgments and lack the ability to finely classify the severity of disturbances, resulting in a high false alarm rate or untimely response.

[0004] In other words, existing technologies present technical challenges in enabling distribution network DTU terminals to accurately identify subtle fault characteristics early and provide real-time hierarchical responses to multi-level disturbances with limited computing resources. Summary of the Invention

[0005] In view of this, the present invention provides a DTU terminal for high-reliability distribution networks, which can solve the technical problem in the prior art that distribution network DTU terminals are unable to achieve early and accurate identification of weak fault characteristics and real-time hierarchical response to multi-level disturbances under limited computing resources.

[0006] This invention is implemented as follows: This invention provides a DTU terminal for high-reliability distribution networks, including a main control chip, a multi-channel current and voltage acquisition module, a frequency and phase detection module, a hierarchical storage module, a communication interface module, and a real-time protection logic module. The multi-channel current and voltage acquisition module is used to acquire three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals from the distribution network. The frequency and phase detection module is used to monitor the grid frequency, three-phase voltage phase angle, and three-phase current phase angle. The multi-channel current and voltage acquisition module, the frequency and phase detection module, the hierarchical storage module, the communication interface module, and the real-time protection logic module are all electrically connected to the main control chip. The main control chip is equipped with a multi-band power quality disturbance identification and hierarchical response module, which performs multi-band decomposition and feature extraction on the acquired current and voltage signals, identifies weak fault characteristics, determines the disturbance type and severity, and triggers corresponding processing measures according to different disturbance levels, ultimately outputting power quality monitoring data and protection control commands.

[0007] The multi-channel current and voltage acquisition module uses a 24-bit analog-to-digital converter with a sampling rate of 10240Hz, and the frequency phase detection module uses phase-locked loop technology to achieve frequency tracking.

[0008] The hierarchical storage module includes a high-speed circular buffer memory, an event-triggered memory, and a historical data compressed memory. The high-speed circular buffer memory is used to store high-frequency sampled data from the most recent 10 seconds, and the event-triggered memory is used to store complete waveform data before and after a fault event.

[0009] The multi-band power quality disturbance identification and graded response module is used to perform the following steps: acquiring the instantaneous values ​​of three-phase current, three-phase voltage, zero-sequence current, and zero-sequence voltage from the multi-channel current and voltage acquisition module; acquiring the current grid frequency and three-phase voltage and current phase angle sequences from the frequency and phase detection module; performing 8-level wavelet packet decomposition on the instantaneous current and voltage values ​​to obtain frequency band sub-band signals and calculating frequency band energy values ​​to construct a disturbance degree vector; performing Hilbert transform on the frequency band sub-band signals to extract the instantaneous frequency change rate and construct a continuity evaluation matrix; weightedly fusing the disturbance degree vector and the continuity evaluation matrix to obtain a comprehensive disturbance feature value; determining the disturbance state type based on a threshold; constructing a discrete feature vector for minor disturbance states and using a support vector machine classifier to identify the disturbance type; constructing a mutation detection matrix for moderate disturbance states and calculating a moderate disturbance evaluation index; constructing a balance matrix for severe disturbance states; constructing an emergency degree vector and calculating the Euclidean distance between it and a preset standard emergency degree vector; and triggering corresponding processing measures based on the evaluation results.

[0010] Specifically, the construction of the disorder vector involves performing 8-level wavelet packet decomposition on the instantaneous value sequences of three-phase current and three-phase voltage to obtain 256 frequency band sub-band signals. The frequency energy value of each frequency band sub-band signal within the time window is calculated, and the 1536 frequency band energy values ​​are arranged in order to form the disorder vector.

[0011] Specifically, the calculation of the frequency band energy value involves summing the squares of the amplitudes of all sampling points of the frequency band sub-band signal and then taking the square root.

[0012] Specifically, the construction of the continuity evaluation matrix involves performing a Hilbert transform on the frequency band sub-band signal to obtain an instantaneous frequency sequence, calculating the instantaneous frequency change rate between adjacent sampling points, taking the absolute value of the instantaneous frequency change rate and then averaging it, and stacking the average instantaneous frequency change rates in chronological order to form a continuity evaluation matrix.

[0013] Specifically, the weighted fusion calculation involves multiplying the frequency band energy value by a weighting coefficient of 0.6 and then adding it to the instantaneous frequency change rate multiplied by a weighting coefficient of 0.4 to obtain the weighted fusion value. The average value of the weighted fusion value is then used to obtain the comprehensive disturbance characteristic value.

[0014] Specifically, the determination of the disturbance state type is based on the relationship between the comprehensive disturbance characteristic value and three thresholds, classifying the disturbance into normal operation state, slight disturbance state, moderate disturbance state, and severe disturbance state.

[0015] Specifically, the construction of the discrete feature vector involves extracting the amplitude standard deviation, amplitude skewness, and amplitude kurtosis of the instantaneous values ​​of the three-phase current and the three-phase voltage within a time window, and combining the average values ​​of the instantaneous values ​​of the zero-sequence current and the zero-sequence voltage within the time window to arrange the 20 discrete feature parameters into a discrete feature vector.

[0016] The beneficial effects of the present invention are: (1) The present invention decomposes the current and voltage signals into 256 frequency band sub-band signals by using 8-layer wavelet packet decomposition, and extracts the instantaneous frequency change rate of each frequency band by combining Hilbert transform, and constructs a disorder degree vector and a continuity evaluation matrix containing 1536 frequency band energy values, thereby realizing the multi-band fine capture of weak fault characteristics and solving the technical problem that traditional single-band analysis methods are difficult to identify milliampere-level current changes and millivolt-level voltage anomalies; (2) The present invention designs a three-level classification judgment mechanism for slight disturbance state, moderate disturbance state and severe disturbance state, and constructs a discrete feature vector, a mutation detection matrix, a balance matrix and an emergency vector for different disturbance levels, respectively. Combined with a support vector machine classifier and multi-threshold judgment, it realizes a differentiated response strategy from data recording, alarm notification to trip protection, and avoids the problems of high false alarm rate and untimely response caused by traditional fixed threshold judgment.

[0017] In summary, this invention, through a collaborative mechanism of multi-band feature extraction and multi-level classification response, achieves early and accurate identification of weak fault characteristics and real-time hierarchical response to multi-level disturbances under the constraint of limited computing resources in the DTU terminal. This solves the technical problem mentioned in the background art that distribution network DTU terminals are unable to achieve early and accurate identification of weak fault characteristics and real-time hierarchical response to multi-level disturbances under limited computing resources. Attached Figure Description

[0018] Figure 1 is a schematic diagram of the composition of a DTU terminal.

[0019] Figure 2 shows the curves of the effective value of the three-phase current at different monitoring times.

[0020] Figure 3 shows the frequency band energy distribution under slight disturbance conditions.

[0021] Figure 4 shows the phase angle variation of the three-phase voltage under moderate disturbance conditions. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0023] Figure 1 shows a schematic diagram of the composition of a DTU terminal for a high-reliability distribution network provided by the present invention, including: a main control chip, a multi-channel current and voltage acquisition module, a frequency and phase detection module, an edge computing processing module, a hierarchical storage module, a communication interface module, a real-time protection logic module, and a power management module.

[0024] The multi-channel current and voltage acquisition module is located at the input end of the DTU terminal. This module acquires three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals from the distribution network. It employs a 24-bit analog-to-digital converter with a sampling rate of 10240Hz. The output of this module is electrically connected to the first data interface of the main control chip. The frequency and phase detection module is electrically connected to the second data interface of the main control chip. This module monitors the grid frequency, three-phase voltage phase angle, and three-phase current phase angle in real time. It uses phase-locked loop (PLL) technology for frequency tracking, with a frequency detection range of 45Hz to 65Hz and a phase detection accuracy of 0.1 degrees. The edge computing processing module is located inside the main control chip. This module includes a field-programmable gate array (FPGA) acceleration unit and a lightweight computing unit. The FPGA acceleration unit performs hardware logic operations for emergency protection actions, while the lightweight computing unit performs complex power quality analysis and fault location calculations. The hierarchical storage module is electrically connected to the third data interface of the main control chip. The hierarchical storage module includes a high-speed circular buffer memory, an event-triggered memory, and a historical data compression memory. The high-speed circular buffer memory stores high-frequency sampling data from the most recent 10 seconds. The event-triggered memory stores complete waveform data for 5 seconds before and after a fault event. The historical data compression memory stores annual historical data compressed using wavelet compression. The real-time protection logic module is electrically connected to the fourth data interface of the main control chip. This real-time protection logic module is implemented in hardware and is used to trigger protection actions and output trip signals within 5 milliseconds when overcurrent, undervoltage, overvoltage, or frequency exceeding limits are detected. The communication interface module is electrically connected to the fifth data interface of the main control chip. This communication interface module includes an Ethernet interface, an RS485 interface, and a wireless communication interface for data interaction with the distribution automation master station system. The power management module is electrically connected to the power input terminal of the main control chip. This power management module is used to convert the AC power supply of the distribution network into the DC operating power required by the DTU terminal. This power management module includes an isolation transformer, a rectifier filter circuit, and a multi-channel voltage regulator output circuit, with output voltages of 12V, 5V, and 3.3V.

[0025] The main control chip is equipped with a multi-band power quality disturbance identification and graded response module, which is used to perform multi-band decomposition and feature extraction on the collected current and voltage signals, identify weak fault characteristics, determine the disturbance type and severity, trigger corresponding processing measures according to different disturbance levels, and finally output power quality monitoring data and protection control commands.

[0026] The multi-band power quality disturbance identification and graded response module is used to perform the following steps: S01, acquiring the instantaneous values ​​of phase A current, phase B current, phase C current, phase A voltage, phase B voltage, phase C voltage, zero-sequence current, and zero-sequence voltage from the multi-channel current and voltage acquisition module, and simultaneously acquiring the current grid frequency, phase A voltage phase angle sequence, phase B voltage phase angle sequence, phase C voltage phase angle sequence, phase A current phase angle sequence, phase B current phase angle sequence, and phase C current phase angle sequence from the frequency and phase detection module, with a data acquisition period of 97.66 microseconds; S02, processing the acquired phase A current instantaneous values... The time value sequences, instantaneous B-phase current sequence, instantaneous C-phase current sequence, instantaneous A-phase voltage sequence, instantaneous B-phase voltage sequence, and instantaneous C-phase voltage sequence are each subjected to 8-level wavelet packet decomposition. The current and voltage signals of each phase are decomposed into 256 frequency band sub-band signals. The frequency band energy value of each frequency band sub-band signal is calculated. The 256 frequency band energy values ​​of A-phase current, B-phase current, C-phase current, A-phase voltage, B-phase voltage, and C-phase voltage are arranged sequentially to construct a disorder vector containing 1536 frequency band energy values; S03, Perform Hilbert transform on each frequency band sub-band signal obtained in step S02 to extract the instantaneous amplitude sequence and instantaneous frequency sequence of each frequency band sub-band signal. Calculate the instantaneous frequency change rate between adjacent sampling points for the instantaneous frequency sequence of each frequency band sub-band signal. Arrange the instantaneous frequency change rates of all frequency band sub-band signals within the current 200-millisecond time window according to time order and frequency band order to construct a continuity evaluation matrix. In step S04, perform weighted fusion of the 1536 frequency band energy values ​​in the disorder vector with the corresponding instantaneous frequency change rates of the frequency band sub-band signals in the continuity evaluation matrix. The weighted fusion method is to multiply the frequency band energy value by a weighting coefficient of 0.6 and then multiply the corresponding instantaneous frequency change rate by a weighting coefficient of 0. The values ​​are summed after step 4 to obtain 1536 weighted fusion values. The average of these 1536 weighted fusion values ​​is used to obtain the comprehensive disturbance characteristic value. A preliminary judgment is made based on the preset first threshold of 0.03, second threshold of 0.12, and third threshold of 0.28. If the comprehensive disturbance characteristic value is less than the first threshold of 0.03, it is determined to be in normal operation, and the process returns to step S01 to continue monitoring. If the comprehensive disturbance characteristic value is between the first threshold of 0.03 and the second threshold of 0.12, it is determined to be in a slight disturbance state, and step S05 is executed. If the comprehensive disturbance characteristic value is between the second threshold of 0.12 and the third threshold of 0.28, it is determined to be in a moderate disturbance state, and step S06 is executed. If the comprehensive disturbance characteristic value is greater than the third threshold of 0.12, the process continues.28. If the condition is determined to be a severe disturbance, proceed to step S07; S05. For a slight disturbance, extract the amplitude standard deviation, amplitude skewness, and amplitude kurtosis of the instantaneous values ​​of phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage within a 200-millisecond time window. Combine these with the instantaneous values ​​of zero-sequence current and zero-sequence voltage to construct a discrete feature vector containing 20 discrete feature parameters. Use a support vector machine classifier to classify and identify the disturbance type as voltage flicker, slight harmonics, or load fluctuation. The identification results and the discrete feature vector are stored in the event trigger memory, and the process returns to step S01 to continue monitoring; S06, for moderate disturbance states, the phase jump amplitudes of the A-phase voltage phase angle sequence, B-phase voltage phase angle sequence, and C-phase voltage phase angle sequence within 5 consecutive power frequency cycles are extracted, the deviation value of the phase jump amplitude from the standard power frequency cycle phase change of 360 degrees is calculated, and a mutation detection matrix containing 15 phase jump deviation values ​​is constructed. Simultaneously, the total harmonic distortion rate of the A-phase current instantaneous value sequence, B-phase current instantaneous value sequence, and C-phase current instantaneous value sequence is calculated. The maximum eigenvalue of the mutation detection matrix is ​​then compared with the... After normalizing the total harmonic distortion rate, multiply the results to obtain the moderate disturbance assessment index. If the moderate disturbance assessment index is less than 0.55, the disturbance is classified as voltage sag or harmonic pollution. Only the data is recorded, and an alarm message is sent to the distribution automation master station system through the communication interface module. Return to step S01 to continue monitoring. If the moderate disturbance assessment index is greater than or equal to 0.55, the disturbance is classified as a precursor to a phase-to-phase short circuit or a precursor to a ground fault. Execute step S08. S07: For severe disturbance conditions, immediately activate the real-time protection logic module, and simultaneously extract the instantaneous value sequence of phase A current from 3 seconds before the fault occurs to 2 seconds after the fault occurs, and the phase B current... The instantaneous values ​​of phase C current, phase A voltage, phase B voltage, and phase C voltage are stored in the event trigger memory. The effective values ​​of phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage are calculated, and a balance matrix containing six unbalance parameters is constructed. The degree of three-phase unbalance is determined by this balance matrix. If the determinant of the balance matrix is ​​less than 0.15, a severe three-phase unbalance fault is identified, triggering the real-time protection logic module to output a trip signal. If the determinant of the balance matrix is ​​greater than or equal to 0...15. If the fault is determined to be a single-phase or two-phase fault, proceed to step S08; S08: Extract the frequency deviation between the current grid frequency and the rated frequency of 50Hz, the ratio of the minimum voltage value among the effective values ​​of phase A, phase B, and phase C voltages to the rated voltage of 220V, and the ratio of the maximum current value among the effective values ​​of phase A, phase B, and phase C currents to the rated current of 100A. Divide the frequency deviation by 50Hz to obtain the frequency deviation rate. Take the absolute value of the frequency deviation rate. Arrange the voltage ratio and the current ratio in order to construct an emergency vector containing three emergency parameters. Calculate the Euclidean distance between the emergency vector and the preset standard emergency vector. If the Euclidean distance is... If the Euclidean distance is less than 0.32, the fault is considered minor. An alarm message is sent to the distribution automation master station system via the communication interface module, and the data acquisition frequency of the multi-channel current and voltage acquisition module is increased to 20480Hz for continuous monitoring for 30 seconds. The process returns to step S01. If the Euclidean distance is between 0.32 and 0.76, the fault is considered moderate. The real-time protection logic module is triggered to enter an early warning state and activate the backup power supply path. Steps S01 to S08 are continued until the fault is cleared. If the Euclidean distance is greater than 0.76, the fault is considered severe. The real-time protection logic module is immediately triggered to output a trip signal to disconnect the faulty line, and the fault data is uploaded to the distribution automation master station system via the communication interface module.

[0027] The process of establishing the disorder vector is as follows: After performing 8-level wavelet packet decomposition on the instantaneous current sequences of phase A, phase B, phase C, phase A, phase B, and phase C, respectively, the current and voltage signals of each phase are each given 256 frequency band sub-band signals at the 8th level. For the j-th frequency band sub-band signal, its frequency band energy value within the current 200-millisecond time window is calculated. The frequency band energy value calculation method is to sum the squares of the amplitudes of all sampling points of the frequency band sub-band signal and then take the square root. The value of j ranges from 1 to 256. The 256 frequency band energy values ​​of phase A current are arranged in ascending order of frequency, from the 1st to the 256th position. The 256 frequency band energy values ​​of phase B current are arranged in ascending order of frequency, from the 257th to the 512th position. The 256 frequency band energy values ​​of phase C current are arranged in ascending order of frequency, from the 1st to the 256th position. The frequency band energy values ​​are arranged in ascending order of frequency from position 513 to position 768. The 256 frequency band energy values ​​of phase A voltage are arranged in ascending order of frequency from position 769 to position 1024. The 256 frequency band energy values ​​of phase B voltage are arranged in ascending order of frequency from position 1025 to position 1280. The 256 frequency band energy values ​​of phase C voltage are arranged in ascending order of frequency from position 1281 to position 1536. This forms a disorder vector containing 1536 frequency band energy values. The j-th element of the disorder vector represents the frequency band energy value of the j-th frequency band sub-band signal. When the distribution network is operating normally, the values ​​of each element of the disorder vector are relatively uniform and small. When the distribution network experiences disturbances, the frequency band energy values ​​of some frequency band sub-band signals will increase significantly, resulting in an uneven distribution of the values ​​of each element of the disorder vector.

[0028] The establishment process of the continuity evaluation matrix is ​​as follows: Hilbert transform is performed on the 1536 frequency band sub-band signals obtained from 8-level wavelet packet decomposition to obtain the instantaneous amplitude sequence and instantaneous frequency sequence of each frequency band sub-band signal. For the instantaneous frequency sequence of the j-th frequency band sub-band signal, the instantaneous frequency change rate between adjacent sampling points is calculated. The instantaneous frequency change rate is calculated by subtracting the instantaneous frequency value of the previous sampling point from the instantaneous frequency value of the current sampling point, then dividing by the sampling period of 97.66 microseconds. The absolute value of all instantaneous frequency change rates of the frequency band sub-band signal within the current 200-millisecond time window is taken, and the average value is calculated to obtain the average value of the j-th frequency band sub-band signal. The instantaneous frequency change rate is calculated by arranging the average instantaneous frequency change rates of the 1536 frequency band sub-band signals in ascending order of frequency to form a 1536-dimensional row vector. This calculation process is repeated for each sampling moment within a 200-millisecond time window to obtain several 1536-dimensional row vectors. These 1536-dimensional row vectors are then stacked in chronological order to form a matrix, which is the continuity evaluation matrix. The number of rows in the continuity evaluation matrix is ​​equal to the number of sampling points within the 200-millisecond time window, which is 2048, and the number of columns is 1536. The element in the i-th row and j-th column of the continuity evaluation matrix represents the average instantaneous frequency change rate of the j-th frequency band sub-band signal at the i-th sampling moment.

[0029] The process of establishing the discreteness feature vector is as follows: For slight disturbances, the amplitude standard deviation, amplitude skewness, and amplitude kurtosis of the instantaneous value sequence of phase A current within a 200-millisecond time window are extracted. The amplitude standard deviation is calculated by taking the square root of the variance of all current sample values ​​within the 200-millisecond time window. The amplitude skewness is calculated by subtracting the average value from all current sample values ​​within the 200-millisecond time window, calculating the cube of the result, calculating the average value again, and finally dividing by the cube of the standard deviation. The amplitude kurtosis is calculated by subtracting the average value from all current sample values ​​within the 200-millisecond time window. The values ​​are then raised to the fourth power, averaged, and finally divided by the fourth power of the standard deviation. This yields the standard deviation, skewness, and kurtosis of the phase A current amplitude for the instantaneous value series. Similarly, the same methods are used to calculate the standard deviation, skewness, and kurtosis of the phase B current amplitude for the instantaneous value series. For the instantaneous value sequence of phase B voltage, the standard deviation, skewness, and kurtosis of phase B voltage amplitude are calculated. Similarly, for the instantaneous value sequence of phase C voltage, the standard deviation, skewness, and kurtosis of phase C voltage amplitude are calculated. The average value of the instantaneous zero-sequence current within a 200-millisecond time window is taken as the average value of the zero-sequence current, and the average value of the instantaneous zero-sequence voltage within a 200-millisecond time window is taken as the average value of the zero-sequence voltage. These 20 dispersion characteristic parameters are then calculated according to the standard deviation, skewness, and kurtosis of phase A current amplitude. The following parameters are arranged in sequence: current amplitude kurtosis, B-phase current amplitude standard deviation, B-phase current amplitude skewness, B-phase current amplitude kurtosis, C-phase current amplitude standard deviation, C-phase current amplitude skewness, C-phase current amplitude kurtosis, A-phase voltage amplitude standard deviation, A-phase voltage amplitude skewness, A-phase voltage amplitude kurtosis, B-phase voltage amplitude standard deviation, B-phase voltage amplitude skewness, B-phase voltage amplitude kurtosis, C-phase voltage amplitude standard deviation, C-phase voltage amplitude skewness, C-phase voltage amplitude kurtosis, zero-sequence current average value, and zero-sequence voltage average value, forming a discrete feature vector containing 20 discrete feature parameters.

[0030] The process of establishing the mutation detection matrix involves extracting the phase angle sequences of phase A, phase B, and phase C voltages over five consecutive power frequency cycles. For the phase A voltage phase angle sequence, the initial value of the phase A voltage phase angle is recorded at the beginning of the k-th power frequency cycle, and the ending value is recorded at the end of the k-th power frequency cycle. The difference between the ending value and the initial value is calculated, which represents the phase change of phase A voltage in the k-th power frequency cycle. The standard power frequency cycle phase change is 360 degrees. The deviation between the phase change and the standard power frequency cycle phase change (360 degrees) is calculated as the phase jump deviation value of phase A voltage in the k-th power frequency cycle. The above calculation is repeated for 5 power frequency cycles to obtain 5 phase jump deviation values ​​for phase A voltage. Similarly, 5 phase jump deviation values ​​for phase B voltage phase angle sequence are calculated, and 5 phase jump deviation values ​​for phase C voltage phase angle sequence are calculated. The 5 phase jump deviation values ​​for phase A voltage are used as the first row of the mutation detection matrix, the 5 phase jump deviation values ​​for phase B voltage are used as the second row of the mutation detection matrix, and the 5 phase jump deviation values ​​for phase C voltage are used as the third row of the mutation detection matrix, forming a 3-row, 5-column mutation detection matrix. The element in the i-th row and j-th column of the mutation detection matrix represents the phase jump deviation value of phase i voltage in the j-th power frequency cycle. The mutation detection matrix contains a total of 15 phase jump deviation values.

[0031] The process of establishing the balance matrix is ​​as follows: Calculate the effective values ​​of phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage; calculate the average value of the three-phase currents (the method for calculating the average value of the three-phase currents is to add the effective values ​​of phase A current, phase B current, and phase C current, and then divide by 3); calculate the average value of the three-phase voltages (the method for calculating the average value of the three-phase voltages is to add the effective values ​​of phase A voltage, phase B voltage, and phase C voltage, and then divide by 3); calculate the phase A current imbalance (the method for calculating the phase A current imbalance is to subtract the average value of the three-phase current from the effective value of phase A current, and then divide by the average value of the three-phase current); similarly, calculate the phase B current imbalance (the method for calculating the phase B current imbalance is to subtract the average value of the three-phase current from the effective value of phase B current, and then divide by the average value of the three-phase current); calculate the phase C current imbalance (the method for calculating the phase C current imbalance is to subtract the average value of the three-phase current from the effective value of phase C current, and then divide by the average value of the three-phase current); similarly, calculate the phase A voltage imbalance (the method for calculating the phase A voltage imbalance is to subtract the average value of the three-phase current from the effective value of phase B current, and then divide by the average value of the three-phase current); similarly, calculate the phase A voltage imbalance (the method for calculating the phase A voltage imbalance is to subtract the average value of the three-phase current from the effective value of phase B current, and then divide by the average value of the three-phase current). The voltage imbalance of phase B is calculated by subtracting the average value of the three-phase voltage from the effective value of the voltage and then dividing by the average value of the three-phase voltage. Similarly, the voltage imbalance of phase C is calculated by subtracting the average value of the three-phase voltage from the effective value of the voltage and then dividing by the average value of the three-phase voltage. The current imbalance of phase A, phase B, and phase C are used as the first row of a balance matrix, and the voltage imbalance of phase A, phase B, and phase C are used as the second row, forming a 2x3 balance matrix. The element in the first row and first column of the balance matrix is ​​the current imbalance of phase A, the element in the second row and second column is the current imbalance of phase B, and the element in the third row and first column is the current imbalance of phase C. The element in the second row and first column is the voltage imbalance of phase A, the element in the second row and second column is the voltage imbalance of phase B, and the element in the second row and third column is the voltage imbalance of phase C. The balance matrix contains a total of 6 imbalance parameters.

[0032] The process of establishing the urgency vector is as follows: Extract the current grid frequency value; calculate the frequency deviation as the current grid frequency value minus the rated frequency of 50Hz; calculate the frequency deviation rate by dividing the frequency deviation by 50Hz; extract the minimum value among the effective values ​​of phase A, phase B, and phase C voltages as the minimum voltage value; calculate the voltage ratio as the minimum voltage value divided by the rated voltage of 220V; extract the maximum value among the effective values ​​of phase A, phase B, and phase C currents as the maximum current value; calculate the current ratio as the maximum current value divided by the rated current of 100A; arrange the absolute value of the frequency deviation rate, the voltage ratio, and the current ratio in sequence to form an urgency vector containing three urgency parameters. The first element of the urgency vector is the absolute value of the frequency deviation rate, the second element is the voltage ratio, and the third element is the current ratio. The preset standard urgency vector is a 3-dimensional vector, with the three elements being 0, 1, and 1, representing that under ideal conditions, the frequency deviation rate is 0, the voltage ratio is 1, and the current ratio is 1.

[0033] The method for calculating the total harmonic distortion (THD) is as follows: A fast Fourier transform is performed on the instantaneous value sequence of phase A current to obtain the spectrum of the phase A current signal. The amplitude of the fundamental component and the amplitude of each harmonic component are extracted. Each harmonic component refers to a component whose frequency is an integer multiple of the fundamental frequency. The 2nd to 50th harmonics are extracted. The square root of the sum of the squares of the amplitudes of all harmonic components is taken to obtain the total harmonic amplitude. The total harmonic amplitude is divided by the amplitude of the fundamental component to obtain the THD of the phase A current. The same method is used to calculate the THD of the phase B current instantaneous value sequence and the THD of the phase C current instantaneous value sequence. The largest THD among the phase A, phase B, and phase C current THD is taken as the THD used in step S06.

[0034] The training process of the Support Vector Machine (SVM) classifier involves collecting a large amount of historical power quality disturbance data, extracting discrete feature vectors for each historical data sample, and labeling the disturbance type with three types: voltage flicker, minor harmonics, and load fluctuation. The historical data samples are randomly divided into a training set and a test set, with the training set accounting for 80% of the total samples and the test set accounting for 20%. A radial basis function (RBF) kernel is used as the kernel function for the SVM, with the initial value of the kernel parameter in the RBF kernel function expression set to 0.5. The SVM classifier is trained using a sequential minimum optimization algorithm. During training, the kernel parameter and penalty parameter are optimized using cross-validation. After training, the accuracy of the classifier is evaluated using the test set. If the accuracy is below 90%, the kernel parameter and penalty parameter are adjusted, and the classifier is retrained until the accuracy reaches above 90%.

[0035] The wavelet compression algorithm involves performing a 5-level wavelet decomposition on historical data, using the Daubechies4 wavelet as the wavelet basis function. After decomposition, low-frequency approximation coefficients and high-frequency detail coefficients are obtained. A threshold denoising method is applied to the high-frequency detail coefficients. The threshold is calculated by multiplying the standard deviation of the high-frequency detail coefficients by a coefficient of 0.6745, and then by the square root of the natural logarithm of the number of sampling points (2048) within a 200-millisecond time window. High-frequency detail coefficients with absolute values ​​less than the threshold are set to zero, while those with absolute values ​​greater than or equal to the threshold are retained. The processed wavelet coefficients are then encoded and stored. A run-length encoding method is used to compress continuous zero-value coefficients, reducing storage space usage.

[0036] The Euclidean distance calculation method is as follows: Let the urgency vector be vector X, which contains three urgency parameters. Let the preset standard urgency vector be vector Y, which contains three elements. Calculate the difference between the first urgency parameter of vector X and the first element of vector Y, and then square it. Calculate the difference between the second urgency parameter of vector X and the second element of vector Y, and then square it. Calculate the difference between the third urgency parameter of vector X and the third element of vector Y, and then square it. Add the three squared values ​​together and take the square root to obtain the Euclidean distance between vector X and vector Y.

[0037] The 200-millisecond time window is a time interval of 200 milliseconds calculated backward from the current sampling time. Under the condition of a sampling rate of 10240Hz, the 200-millisecond time window contains 2048 sampling points.

[0038] The normalization method involves dividing the maximum eigenvalue of the mutation detection matrix by the maximum value of all elements in the mutation detection matrix to obtain the normalized maximum eigenvalue, dividing the total harmonic distortion rate by 1 to obtain the normalized total harmonic distortion rate, and multiplying the normalized maximum eigenvalue by the normalized total harmonic distortion rate to obtain the moderate disturbance assessment index.

[0039] The method for obtaining the first threshold of 0.03 is as follows: collect 1000 sets of current and voltage data under normal operating conditions of the distribution network, calculate the comprehensive disturbance characteristic value for each set of data according to the method of steps S02 to S04, perform statistical analysis on the comprehensive disturbance characteristic values ​​of the 1000 sets of data, calculate the average value and standard deviation of the comprehensive disturbance characteristic values, and add 1 standard deviation to the average value as the first threshold of 0.03. The first threshold of 0.03 is used to distinguish between normal operating conditions and slight disturbance conditions.

[0040] The method for obtaining the second threshold 0.12 is as follows: collect 500 sets of current and voltage data under slight disturbance conditions in the distribution network. The slight disturbance conditions include voltage flicker, slight harmonics, and load fluctuations. Calculate the comprehensive disturbance characteristic value for each set of data according to the methods in steps S02 to S04. Perform statistical analysis on the comprehensive disturbance characteristic values ​​of the 500 sets of data, calculate the average value and standard deviation of the comprehensive disturbance characteristic values, and add twice the standard deviation to the average value as the second threshold 0.12. The second threshold 0.12 is used to distinguish between slight disturbance conditions and moderate disturbance conditions.

[0041] The method for obtaining the third threshold of 0.28 is as follows: collect 500 sets of current and voltage data under moderate disturbance conditions in the distribution network. The moderate disturbance conditions include voltage sag, harmonic pollution, phase-to-phase short circuit precursors, and ground fault precursors. For each set of data, calculate the comprehensive disturbance characteristic value according to the methods in steps S02 to S04. Perform statistical analysis on the comprehensive disturbance characteristic values ​​of the 500 sets of data, calculate the average value and standard deviation of the comprehensive disturbance characteristic values, and add twice the standard deviation to the average value as the third threshold of 0.28. The third threshold of 0.28 is used to distinguish between moderate disturbance conditions and severe disturbance conditions.

[0042] The method for obtaining the moderate disturbance assessment index threshold of 0.55 is as follows: 300 sets of current and voltage data on voltage sags and harmonic pollution in the distribution network are collected; 200 sets of current and voltage data on phase-to-phase short-circuit precursors and ground fault precursors in the distribution network are collected; for each set of data, the moderate disturbance assessment index is calculated according to the method in step S06; statistical analysis is performed on the moderate disturbance assessment indices of the 300 sets of voltage sags and harmonic pollution data; the average value and standard deviation of the moderate disturbance assessment index are calculated to obtain the moderate disturbance assessment index of voltage sags and harmonic pollution. The upper limit of the disturbance assessment index is determined by statistically analyzing the moderate disturbance assessment index of the 200 sets of phase-to-phase short circuit precursors and ground fault precursors. The average value and standard deviation of the moderate disturbance assessment index are calculated to obtain the lower limit of the moderate disturbance assessment index for phase-to-phase short circuit precursors and ground fault precursors. The average value of the upper limit value and the lower limit value is taken as the moderate disturbance assessment index threshold of 0.55. The moderate disturbance assessment index threshold of 0.55 is used to distinguish between voltage sags or harmonic pollution and phase-to-phase short circuit precursors or ground fault precursors.

[0043] The method for obtaining the balance matrix determinant threshold of 0.15 is as follows: 400 sets of current and voltage data for severe three-phase imbalance faults in the distribution network are collected, and 400 sets of current and voltage data for single-phase or two-phase faults in the distribution network are collected. For each set of data, the determinant value of the balance matrix is ​​calculated according to the method in step S07. Statistical analysis is performed on the determinant values ​​of the 400 sets of severe three-phase imbalance fault data to calculate the average and standard deviation of the determinant values, obtaining the upper limit value of the determinant value for severe three-phase imbalance faults. Statistical analysis is also performed on the determinant values ​​of the 400 sets of single-phase or two-phase fault data to calculate the average and standard deviation of the determinant values, obtaining the lower limit value of the determinant value for single-phase or two-phase faults. The average of the upper limit value and the lower limit value is used as the balance matrix determinant threshold of 0.15. The balance matrix determinant threshold of 0.15 is used to distinguish between severe three-phase imbalance faults and single-phase or two-phase faults.

[0044] The method for obtaining the first Euclidean distance threshold of 0.32 is as follows: 600 sets of current and voltage data for minor faults in the distribution network are collected. For each set of data, the Euclidean distance between the urgency vector and the preset standard urgency vector is calculated according to the method in step S08. The Euclidean distances of the 600 sets of data are statistically analyzed to calculate the average value and standard deviation of the Euclidean distances. The average value plus 1.5 times the standard deviation is used as the first Euclidean distance threshold of 0.32. The first Euclidean distance threshold of 0.32 is used to distinguish between minor and moderate faults.

[0045] The method for obtaining the second Euclidean distance threshold of 0.76 is as follows: 500 sets of current and voltage data with moderate fault severity in the distribution network are collected, and 400 sets of current and voltage data with severe fault severity in the distribution network are collected. For each set of data, the Euclidean distance between the urgency vector and the preset standard urgency vector is calculated according to the method in step S08. The Euclidean distances of the 500 sets of data with moderate fault severity are statistically analyzed, and the average and standard deviation of the Euclidean distances are calculated to obtain the upper limit value of the Euclidean distance for moderate fault severity. The Euclidean distances of the 400 sets of data with severe fault severity are statistically analyzed, and the average and standard deviation of the Euclidean distances are calculated to obtain the lower limit value of the Euclidean distance for severe fault severity. The average of the upper limit value and the lower limit value is taken as the second Euclidean distance threshold of 0.76. The second Euclidean distance threshold of 0.76 is used to distinguish between moderate and severe fault severity.

[0046] The weighting coefficients 0.6 and 0.4 are obtained as follows: 2000 sets of current and voltage data under different disturbance states in the distribution network are collected. For each set of data, the frequency band energy value in the disturbance degree vector and the instantaneous frequency change rate in the continuity evaluation matrix are calculated. Principal component analysis is used to analyze the contribution rate of the frequency band energy value and the instantaneous frequency change rate to disturbance identification. The principal component contribution rate of the frequency band energy value to disturbance identification is calculated to be 60%, and the principal component contribution rate of the instantaneous frequency change rate to disturbance identification is calculated to be 40%. The principal component contribution rate of the frequency band energy value (60%) is used as the weighting coefficient 0.6, and the principal component contribution rate of the instantaneous frequency change rate (40%) is used as the weighting coefficient 0.4.

[0047] The rated voltage of 220V is the effective value of the standard phase voltage of the distribution network, and the rated current of 100A is the effective value of the rated current of the standard line of the distribution network. The rated voltage of 220V and the rated current of 100A are determined according to the design parameters of the distribution network.

[0048] The method for obtaining the parameter of increasing the data acquisition frequency to 20480Hz is to multiply the original data acquisition frequency of 10240Hz by 2 to obtain the increased data acquisition frequency of 20480Hz. The increased data acquisition frequency of 20480Hz is used to improve the sampling accuracy when the fault is relatively minor so as to more accurately monitor the fault development trend.

[0049] The method for obtaining the 30-second continuous monitoring time parameter is as follows: based on the statistical analysis of the development cycle of minor faults in the distribution network, 800 sets of minor fault data are collected, the time interval from the occurrence of the fault to its elimination or aggravation is analyzed, the average value of the time interval is calculated to be 25 seconds, and the average value is added to the standard deviation of 5 seconds to obtain the continuous monitoring time of 30 seconds. The continuous monitoring time of 30 seconds is used to ensure that there is enough time to observe the development trend of the fault when the fault is relatively minor.

[0050] Specifically, the principle of this invention is as follows: This invention can solve the technical problem that distribution network DTU terminals are unable to achieve early and accurate identification of weak fault characteristics and real-time hierarchical response to multi-level disturbances under limited computing resources. Its principle lies in combining multi-scale decomposition of signal processing with a disturbance hierarchical response mechanism. Eight-layer wavelet packet decomposition expands the current and voltage signals simultaneously in the time and frequency domains. Compared with traditional single-band analysis, it can simultaneously capture the energy distribution of different frequency components in 256 frequency bands. Local frequency band energy anomalies caused by weak faults can be amplified and displayed in the disorder vector, while normal operation noise is relatively suppressed because it is distributed in multiple frequency bands. The instantaneous frequency change rate extracted by Hilbert transform reflects the non-stationary characteristics of the signal. After weighted fusion with the frequency band energy value, the resulting comprehensive disturbance feature value contains fault information in both the energy domain and the frequency domain, improving the sensitivity of weak feature identification. The three-level classification and judgment mechanism quantifies and classifies the severity of disturbances by setting a first threshold, a second threshold, and a third threshold. Under the condition of slight disturbance, only data is recorded and alarms are sent, avoiding unnecessary protection actions. Under the condition of moderate disturbance, the judgment is further refined by combining phase jump and harmonic analysis. Under the condition of severe disturbance, the protection logic is immediately activated and a trip signal is triggered. This achieves a precise match between response measures and the degree of disturbance, reduces the false alarm rate, and ensures timely handling of critical faults. The entire scheme adopts a lightweight matrix operation and lookup table method, and the computational complexity is controlled within the capability of the DTU terminal processor, meeting the real-time requirements.

[0051] The following provides a specific embodiment 1 of the present invention, and the specific implementation of multiple steps in this embodiment 1 is described in detail below.

[0052] The specific implementation of step S02 is as follows: The acquired instantaneous current and voltage value sequences of each phase are subjected to 8-level wavelet packet decomposition, decomposing the current and voltage signals of each phase into 256 frequency band sub-band signals. After 8-level wavelet packet decomposition, the following is obtained: Each frequency band sub-band signal corresponds to a certain frequency range, which is determined by the sampling rate of 10240 Hz. The center frequency of the sub-band signal in each frequency band is ,in For the first The center frequency of each sub-band signal, measured in Hertz (Hz). The value range is from 1 to 256. The frequency energy value of each sub-band signal is calculated, and a disorder vector containing 1536 frequency band energy values ​​is constructed. The formula for calculating the frequency band energy value of a sub-band signal in a frequency band is expressed as follows: In the formula, For the first The frequency band energy value of each sub-band signal, in amperes or volts; For the first The sub-band signal of each frequency band in the first The amplitude of each sampling point, in amperes or volts; This represents the number of sampling points within a 200-millisecond time window, with a value of 2048. This represents the sampling point number within the time window, ranging from 1 to 2048. Disorder vector. The construction formula is expressed as follows: In the formula, This is the disorder vector, with a dimension of 1536; The current of phase A is the first The frequency band energy value of each sub-band signal, in amperes; For the B-phase current The frequency band energy value of each sub-band signal, in amperes; For the C-phase current The frequency band energy value of each sub-band signal, in amperes; The voltage of phase A is the first The frequency band energy value of each sub-band signal, in volts; For phase B voltage The frequency band energy value of each sub-band signal, in volts; For the C-phase voltage The formula calculates the frequency energy value of each frequency band sub-band signal, in volts. It calculates the frequency energy value of each of the 256 frequency band sub-band signals after 8-layer wavelet packet decomposition, reflecting the energy distribution characteristics of the distribution network in different frequency bands. When disturbances occur in the distribution network, the energy in certain frequency bands will increase significantly, resulting in uneven distribution of the elements of the disturbance degree vector, thus enabling preliminary detection of the disturbance.

[0053] The specific implementation of step S03 is as follows: Perform a Hilbert transform on each frequency band sub-band signal to extract the instantaneous amplitude sequence and instantaneous frequency sequence. The Hilbert transform is used to extract the analytic signal of the signal; for real signals... Its Hilbert transformation is The analytic signal is represented as ,in The result of the Hilbert transform, units and same; The original signal, The variable is for integration, and the unit is seconds. For time variables, the unit is seconds; To analyze the signal; This is an instantaneous amplitude sequence, with units of 1 and 2. same; The instantaneous phase is expressed in radians. The imaginary unit. Instantaneous frequency sequence. ,in Let be the instantaneous frequency, in Hertz. Calculate the rate of change of the instantaneous frequency and construct a continuity evaluation matrix. The formula for calculating the instantaneous frequency change rate of a sub-band signal in each frequency band is as follows: In the formula, For the first The sub-band signal of each frequency band in the first The instantaneous rate of change of frequency at each sampling moment, in Hertz per second; For the first The sub-band signal of each frequency band in the first The instantaneous frequency value at each sampling moment, in Hertz; For the first The sub-band signal of each frequency band in the first The instantaneous frequency value at each sampling moment, in Hertz; The sampling period is 97.66 microseconds. This represents the sampling time sequence number within the time window, with a value ranging from 2 to 2048. The formula for calculating the average instantaneous frequency change rate of each frequency band sub-band signal is as follows: In the formula, For the first The average instantaneous frequency change rate of each frequency band sub-band signal, in Hertz per second; This represents the number of sampling points within a 200-millisecond time window, with a value of 2048. Continuity evaluation matrix. The construction formula is expressed as follows: In the formula, It is a continuous evaluation matrix with 2048 rows and 1536 columns; For the first The sampling time of the first sampling moment The average instantaneous frequency change rate of each frequency band sub-band signal, in Hertz per second; For the first Each sampling time, The value range is from 1 to 2048. This matrix extracts instantaneous frequency information through Hilbert transform and calculates the rate of frequency change between adjacent sampling points, reflecting the continuity characteristics of the signal in the time and frequency domain, and is used to identify signal abrupt changes and oscillation characteristics.

[0054] The specific implementation of step S04 is as follows: The frequency band energy value in the disorder vector is weighted and fused with the instantaneous frequency change rate of the corresponding frequency band sub-band signal in the continuity evaluation matrix. The frequency band energy value and the average instantaneous frequency change rate are normalized by dividing by their respective maximum values, ensuring the normalized value is between 0 and 1. The comprehensive disturbance characteristic value is calculated, and a preliminary judgment is made based on a preset threshold. The formula for calculating the weighted fusion value is as follows: In the formula, For the first The weighted fusion value of each frequency band, dimensionless; For the first The frequency band energy value of each sub-band signal, in amperes or volts; This represents the maximum energy value across all frequency bands, expressed in amperes or volts. For the first The average instantaneous frequency change rate of each frequency band sub-band signal, in Hertz per second; The maximum value of the average instantaneous rate of change of frequency across all frequency bands, expressed in Hertz per second; The value range is from 1 to 1536. The formula for calculating the comprehensive disturbance characteristic value is as follows: In the formula, The comprehensive disturbance characteristic value is dimensionless. This formula is a weighted fusion of normalized frequency band energy values ​​and instantaneous frequency change rates. Weighting coefficients of 0.6 and 0.4 reflect the contribution rates of frequency band energy and frequency change to disturbance identification, respectively. The weighting coefficients are obtained using principal component analysis (PCA). PCA extracts the main features of the data by calculating the eigenvalues ​​and eigenvectors of the covariance matrix. The contribution rate calculation formula is as follows: In the formula, For the first The contribution rate of each principal component is dimensionless. For the first One eigenvalue; The total number of features, with a value of 2; For feature serial number; The value can be either 1 or 2. The principal component contribution rate of the frequency band energy value is 60%, i.e. The principal component contribution rate of the instantaneous frequency change rate is 40%, i.e. The magnitude of the comprehensive disturbance characteristic value reflects the overall disturbance level of the distribution network. By comparing it with preset thresholds of 0.03, 0.12, and 0.28, the system can classify and determine the normal operation state, slight disturbance state, moderate disturbance state, and severe disturbance state.

[0055] The specific implementation of step S05 is as follows: For slight disturbances, extract the statistical feature parameters of the instantaneous current and voltage value sequences, combine the zero-sequence current and zero-sequence voltage to construct a discrete feature vector, and use a support vector machine classifier to identify the disturbance type. The formula for calculating the amplitude standard deviation is as follows: In the formula, The standard deviation of the amplitude is expressed in amperes or volts. For the first The current or voltage value at each sampling point is expressed in amperes or volts. This is the average of all sampled values ​​within a 200-millisecond time window, expressed in amperes or volts. This represents the number of sampling points, with a value of 2048. Here is the sampling point number, ranging from 1 to 2048. The formula for calculating amplitude skewness is as follows: In the formula, This refers to amplitude skewness, which is dimensionless. The formula for calculating amplitude kurtosis is as follows: In the formula, This represents the amplitude kurtosis, which is dimensionless. It is the discreteness eigenvector. The construction formula is expressed as follows: In the formula, This is a discrete feature vector with a dimension of 20; These are the standard deviation of the amplitude, amplitude skewness, and amplitude kurtosis of the phase A current, respectively. These are the standard deviation of the amplitude, amplitude skewness, and amplitude kurtosis of the B-phase current, respectively. These are the standard deviation of the amplitude, amplitude skewness, and amplitude kurtosis of the C-phase current, respectively. These are the standard deviation of the amplitude, amplitude skewness, and amplitude kurtosis of the phase A voltage, respectively. These are the amplitude standard deviation, amplitude skewness, and amplitude kurtosis of the B-phase voltage, respectively. These are the standard deviation of the amplitude, amplitude skewness, and amplitude kurtosis of the C-phase voltage, respectively. This represents the average zero-sequence current, in amperes. The zero-sequence voltage average value is expressed in volts. This feature vector, by extracting statistical characteristic parameters of the signal, reflects the distribution characteristics of the current and voltage signals under slight disturbances. A support vector machine (SVM) classifier is used to classify and identify this feature vector. The radial basis function kernel function of the SVM is expressed as follows: ,in The kernel function value is dimensionless. and There are two feature vectors; This is a kernel parameter, with a default value of 0.5. The Euclidean distance between the two vectors is given. Support Vector Machine (SVM) classifiers can effectively distinguish between three types of disturbances: voltage flicker, minor harmonics, and load fluctuations.

[0056] The specific implementation of step S06 is as follows: For a moderate disturbance state, phase jump information and total harmonic distortion rate are extracted, a sudden change detection matrix is ​​constructed, and a moderate disturbance assessment index is calculated. The formula for calculating the phase jump deviation value is expressed as follows: In the formula, For the first Phase voltage at the Phase jump deviation value per power frequency cycle, in degrees; For the first Phase voltage at the The phase angle at the end of each power frequency cycle, in degrees; For the first Phase voltage at the The phase angle at the start of each power frequency cycle, in degrees; The value can be A, B, or C, representing one of the three phases; The value ranges from 1 to 5, representing the nth power frequency cycle. (Mutation detection matrix) The construction formula is expressed as follows: In the formula, This is a mutation detection matrix with 3 rows and 5 columns. (Mutation Detection Matrix) Maximum eigenvalue By solving the characteristic equation Obtain, among which For matrix Transpose of; It is the identity matrix; For eigenvalues; This represents a determinant operation, selecting the maximum value among all eigenvalues ​​as the determinant. The calculation of total harmonic distortion (THD) requires first performing a Fast Fourier Transform (FFT) on the instantaneous current value sequence. The FFT converts the time-domain signal into a frequency-domain signal. For a sequence of length... discrete signals Its fast Fourier transform is ,in It is a frequency domain signal; This is a time-domain signal, measured in amperes. This is a frequency index, with values ​​ranging from 0 to... ; This is a time index, with values ​​ranging from 0 to... ; This represents the number of sampling points, with a value of 2048. The unit is imaginary. The formula for calculating the total harmonic distortion (THD) is as follows: In the formula, The total harmonic distortion (THD) is dimensionless. For the first The amplitude of the second harmonic component, in amperes; This represents the amplitude of the fundamental component, measured in amperes. The value ranges from 2 to 50, representing the harmonic order. The formula for calculating the moderate disturbance assessment index is as follows: In the formula, This is a dimensionless index for assessing moderate disturbances. The largest eigenvalue of the mutation detection matrix; This represents the maximum value among all elements of the mutation detection matrix, expressed in degrees. In this formula, The largest eigenvalue after normalization. It is a dimensionless quantity, therefore It is dimensionless. This formula reflects the phase stability and waveform quality of the power grid under moderate disturbance conditions by normalizing and fusing phase jump information and harmonic distortion rate. When the evaluation index is less than 0.55, it is judged as voltage sag or harmonic pollution. When the evaluation index is greater than or equal to 0.55, it is judged as a precursor to phase-to-phase short circuit or ground fault.

[0057] The specific implementation of step S07 is as follows: For severe disturbance conditions, the real-time protection logic module is immediately activated to extract current and voltage waveform data before and after the fault, calculate the three-phase unbalance, and construct a balance matrix to determine the fault type. Balance matrix The construction formula is expressed as follows: In the formula, This is a balance matrix with 2 rows and 3 columns, and all elements are dimensionless. These are the effective values ​​of the currents in phases A, B, and C, respectively, in amperes. These are the effective values ​​of the voltages for phases A, B, and C, respectively, in volts. This is the average value of the three-phase current, in amperes. This represents the average three-phase voltage, in volts. Balance matrix. The determinant value is calculated using The generalized determinant of a matrix is ​​defined as follows: ,in For matrix Transpose of; for A matrix, whose standard determinant is calculated as follows: ,in Representation matrix The Line number Column elements, and The values ​​are all 1 or 2. This matrix reflects the degree of three-phase imbalance by calculating the relative deviation of the current and voltage of each phase from the average value. When the determinant value of the balance matrix is ​​less than 0.15, it is judged as a serious three-phase imbalance fault and a trip signal is triggered. When the determinant value is greater than or equal to 0.15, it is judged as a single-phase or two-phase fault and step S08 is executed.

[0058] The specific implementation of step S08 involves extracting the frequency deviation, voltage ratio, and current ratio to construct an emergency vector, calculating the Euclidean distance between this vector and a preset standard emergency vector, and determining the severity of the fault based on the distance and taking corresponding measures. Emergency Vector The construction formula is expressed as follows: In the formula, This is an urgency vector with a dimension of 3, and all elements are dimensionless. This is the current power grid frequency value, in Hertz. This is the minimum effective value of the three-phase voltage, expressed in volts. This represents the maximum effective value of the three-phase current, measured in amperes. Preset standard urgency vector. ,in The preset standard urgency vector has a dimension of 3, and all elements are dimensionless. The three elements represent, under ideal conditions, a frequency deviation rate of 0, a voltage ratio of 1, and a current ratio of 1, respectively. The Euclidean distance calculation formula is as follows: In the formula, The distance is Euclidean and dimensionless. These are the urgency vectors. All three elements are dimensionless. This formula quantifies the severity of a fault by calculating the Euclidean distance between the actual urgency vector and the ideal state vector. When the distance is less than 0.32, the fault is considered minor; when the distance is between 0.32 and 0.76, the fault is considered moderate; and when the distance is greater than 0.76, the fault is considered severe, and a trip signal is immediately triggered to disconnect the faulty line.

[0059] In the explanation section, the method for obtaining the first threshold of 0.03 is as follows: 1000 sets of current and voltage data under normal operating conditions of the distribution network are collected, and the comprehensive disturbance characteristic value is calculated for each set of data according to steps S02 to S04. The formula for calculating the average value of the comprehensive disturbance characteristic value is expressed as follows: In the formula, The average value of the comprehensive disturbance characteristics is dimensionless. For the first The comprehensive perturbation characteristic value of the dataset is dimensionless. Here is the data set number, ranging from 1 to 1000. The formula for calculating the standard deviation is as follows: In the formula, The standard deviation is dimensionless. The formula for calculating the first threshold is as follows: In the formula, The first threshold is 0.03, which is dimensionless. This first threshold is used to distinguish between normal operation and slight disturbance.

[0060] The second threshold of 0.12 is obtained by collecting 500 sets of current and voltage data under slight disturbance conditions in the distribution network. The formula for calculating the average value of the comprehensive disturbance characteristic value is as follows: In the formula, The average value of the comprehensive disturbance characteristics is dimensionless. For the first The comprehensive perturbation characteristic value of the dataset is dimensionless. Here is the data set number, ranging from 1 to 500. The formula for calculating the standard deviation is as follows: In the formula, The standard deviation of the comprehensive disturbance eigenvalues ​​is dimensionless. The formula for calculating the second threshold is as follows: In the formula, The second threshold, with a value of 0.12, is dimensionless. This second threshold is used to distinguish between slightly disturbed and moderately disturbed states.

[0061] The third threshold of 0.28 was obtained by collecting 500 sets of current and voltage data under moderate disturbance conditions in the distribution network. The formula for calculating the average value of the comprehensive disturbance characteristic value is as follows: In the formula, The average value of the comprehensive disturbance characteristics is dimensionless. For the first The comprehensive perturbation characteristic value of the dataset is dimensionless. Here is the data set number, ranging from 1 to 500. The formula for calculating the standard deviation is as follows: In the formula, The standard deviation is dimensionless. The formula for calculating the third threshold is as follows: In the formula, The third threshold, with a value of 0.28, is dimensionless. This third threshold is used to distinguish between moderate and severe disturbance states.

[0062] To better understand and implement this invention, a specific application scenario is provided below as Example 2: A technical team deployed a DTU terminal system based on this invention in an urban power distribution network renovation project for real-time monitoring and protection of the operational safety of a 10kV distribution line. The distribution line has a rated voltage of 10kV and a rated current of 100A, supplying power to a mixed load area encompassing commercial areas and industrial parks. The technical team installed a DTU terminal on the outgoing line side of the substation, configuring a multi-channel current and voltage acquisition module with a sampling rate of 10240Hz and a data acquisition cycle of 97.66 microseconds.

[0063] During a monitoring operation, as shown in Figure 2, the multi-channel current and voltage acquisition module of the DTU terminal continuously acquired three-phase current and voltage signals. At monitoring time T=0, the system was in normal operation. The effective value of phase A current was 85A, phase B current was 82A, and phase C current was 88A. The effective values ​​of phase A voltage were 5773V, phase B voltage was 5810V, and phase C voltage was 5795V. The frequency and phase detection module measured the grid frequency to be 50.02Hz, and the phase angles of the three-phase voltages were 0 degrees, 120 degrees, and 240 degrees, respectively, which met the normal operation requirements.

[0064] After acquiring all collected data in step S01, the multi-band power quality disturbance identification and hierarchical response module within the main control chip performs 8-level wavelet packet decomposition in step S02. After decomposing the instantaneous value sequence of phase A current, 256 frequency band sub-band signals are obtained, and the frequency energy value of each band is calculated. The first frequency band corresponds to the low-frequency component from 0 to 20 Hz, with a frequency energy value of 1.24; the 128th frequency band corresponds to the mid-frequency component from 2560 to 2580 Hz, with a frequency energy value of 0.08; and the 256th frequency band corresponds to the high-frequency component from 5100 to 5120 Hz, with a frequency energy value of 0.03. The phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage are processed in the same way, ultimately constructing a disturbance degree vector containing 1536 frequency band energy values.

[0065] In step S03, the system performs a Hilbert transform on each frequency band sub-band signal to extract the instantaneous amplitude sequence and instantaneous frequency sequence. For the instantaneous frequency sequence of the first frequency band sub-band signal, the instantaneous frequency change rate between adjacent sampling points is calculated. The instantaneous frequency of the current sampling point is 49.98 Hz, and the instantaneous frequency of the previous sampling point is 50.01 Hz, resulting in an instantaneous frequency change rate of 307.58 Hz / s. Within a 200-millisecond time window containing 2048 sampling points, the absolute average of all instantaneous frequency change rates is calculated to be 285.42 Hz / s, which is taken as the average instantaneous frequency change rate of the first frequency band sub-band signal. The average instantaneous frequency change rates of the 1536 frequency band sub-band signals are stacked in chronological order to form a continuity evaluation matrix of 2048 rows and 1536 columns.

[0066] In step S04, the energy value of the first frequency band in the disorder vector (1.24) is multiplied by a weighting coefficient of 0.6 to obtain 0.744. The corresponding average instantaneous frequency change rate in the continuity evaluation matrix (285.42 Hz / s) is multiplied by a weighting coefficient of 0.4 to obtain 114.168. The two values ​​are added together to obtain a weighted fusion value of 114.912. The average of the 1536 weighted fusion values ​​is calculated to obtain a comprehensive disturbance characteristic value of 0.028. Since the comprehensive disturbance characteristic value of 0.028 is less than the first threshold of 0.03, the system is determined to be in normal operating condition, and the process returns to step S01 to continue monitoring.

[0067] At monitoring time T=15s, as shown in Figure 3, the load on the distribution line suddenly increases. The effective value of the current in phase A rises to 92A, the effective value of the current in phase B rises to 89A, and the effective value of the current in phase C rises to 95A. The voltage drops slightly, with the effective values ​​of the voltage in phase A being 5690V, the voltage in phase B being 5715V, and the voltage in phase C being 5702V. The system recalculates the disturbance vector and the continuity evaluation matrix. At this time, the energy value of the first frequency band increases to 1.68, the average instantaneous frequency change rate increases to 412.35Hz / s, and the calculated result of the comprehensive disturbance characteristic value is 0.085. Since the comprehensive disturbance characteristic value of 0.085 is between the first threshold of 0.03 and the second threshold of 0.12, the system determines it to be a slight disturbance state and executes step S05.

[0068] In step S05, the system extracts the statistical characteristic parameters of the current and voltage of each phase within a 200-millisecond time window. The standard deviation of the phase A current amplitude is 8.52A, the amplitude skewness is 0.24, and the amplitude kurtosis is 3.18. The standard deviation of the phase B current amplitude is 7.89A, the amplitude skewness is 0.19, and the amplitude kurtosis is 3.05. The standard deviation of the phase C current amplitude is 9.14A, the amplitude skewness is 0.28, and the amplitude kurtosis is 3.32. The standard deviation of the phase A voltage amplitude is 185V, the amplitude skewness is 0.12, and the amplitude kurtosis is 2.95. The standard deviation of the phase B voltage amplitude is 172V, the amplitude skewness is 0.08, and the amplitude kurtosis is 2.88. The standard deviation of the phase C voltage amplitude is 198V, the amplitude skewness is 0.15, and the amplitude kurtosis is 3.02. The average zero-sequence current is 2.5A, and the average zero-sequence voltage is 85V. The 20 discrete feature parameters are used to construct a discrete feature vector, which is then input into a support vector machine classifier for classification and identification. The identification result is load fluctuation. The system stores the result in the event trigger memory and returns to step S01 to continue monitoring.

[0069] At monitoring time T=42s, as shown in Figure 4, an abnormal disturbance occurred in the power distribution line. The effective value of the current in phase A jumped to 135A, the effective value of the current in phase B was 128A, and the effective value of the current in phase C was 142A. The voltage dropped significantly, with the effective value of the voltage in phase A dropping to 4820V, the effective value of the voltage in phase B dropping to 4905V, and the effective value of the voltage in phase C dropping to 4865V. The system calculated a comprehensive disturbance characteristic value of 0.186, which is between the second threshold of 0.12 and the third threshold of 0.28, and determined to be a moderate disturbance state. Step S06 was then executed.

[0070] In step S06, the system extracts the phase jump amplitude of the three-phase voltage phase angle sequence over five consecutive power frequency cycles. In the first power frequency cycle, the phase change of phase A is 362.5 degrees, deviating from the standard power frequency cycle phase change of 360 degrees by 2.5 degrees. In the second power frequency cycle, the phase change of phase A is 358.8 degrees, with a deviation of -1.2 degrees. The phase jump deviations of phase A in the third to fifth power frequency cycles are 3.1 degrees, -0.8 degrees, and 1.9 degrees, respectively. The phase jump deviations of phases B and C are calculated using the same method, constructing a 3x5 abrupt change detection matrix, as shown in Table 1.

[0071] Table 1. Phase jump deviation values ​​of the mutation detection matrix

[0072] The system calculates the maximum eigenvalue of the mutation detection matrix to be 8.64. Fast Fourier Transforms are performed on the A-phase, B-phase, and C-phase currents to extract the 2nd to 50th harmonic components. The fundamental amplitude of the A-phase current is 190.53A, the total harmonic amplitude is 28.58A, and the total harmonic distortion (THD) of the A-phase current is 0.150. The THD of the B-phase current is 0.142, and the THD of the C-phase current is 0.165. The maximum THD of 0.165 is used for subsequent calculations. Dividing the maximum eigenvalue of the mutation detection matrix (8.64) by the maximum value of all elements (3.8) yields the normalized maximum eigenvalue of 2.274. Dividing the THD of 0.165 by 1 yields the normalized THD of 0.165. Multiplying these two values ​​gives the moderate disturbance assessment index of 0.375. Since the moderate disturbance assessment index of 0.375 is less than 0.55, the system classifies the disturbance as a voltage sag, records only the data, and sends the alarm information to the distribution automation master station system through the communication interface module, returning to step S01 to continue monitoring.

[0073] At monitoring time T=68s, a serious fault occurred in the power distribution line. The effective value of the current in phase A suddenly increased to 285A, the effective value of the current in phase B was 125A, and the effective value of the current in phase C was 118A. The effective value of the voltage in phase A plummeted to 2850V, the effective value of the voltage in phase B was 5520V, and the effective value of the voltage in phase C was 5605V. The system calculated a comprehensive disturbance characteristic value of 0.412, which is greater than the third threshold of 0.28, and determined it to be a serious disturbance state. The real-time protection logic module was immediately activated, and step S07 was executed.

[0074] In step S07, the system extracts complete waveform data from 3 seconds before the fault occurs to 2 seconds after the fault occurs and stores it in the event trigger memory. The average three-phase current is calculated to be 176A, and the average three-phase voltage is 4658.33V. The phase A current imbalance is 0.619, the phase B current imbalance is -0.290, and the phase C current imbalance is -0.330. The phase A voltage imbalance is -0.388, the phase B voltage imbalance is 0.185, and the phase C voltage imbalance is 0.203. A 2x3 balance matrix is ​​constructed, and its determinant is calculated to be 0.268, which is greater than or equal to 0.15, thus determining it to be a single-phase fault, and step S08 is executed.

[0075] In step S08, the system extracts the current grid frequency value as 49.65Hz, calculates the frequency deviation as -0.35Hz, the frequency deviation rate as -0.007, and the absolute value of the frequency deviation rate as 0.007. The minimum value of 2850V among the effective values ​​of phase A voltage (2850V), phase B voltage (5520V), and phase C voltage (5605V) is extracted as the minimum voltage value. The rated line voltage is... The kV phase voltage rating is 5770V, and the calculated voltage ratio is 0.494. The maximum value of 285A among the effective values ​​of phase A current (285A), phase B current (125A), and phase C current (118A) is extracted as the maximum current value, and the calculated current ratio is 2.85. Emergency vectors are constructed with values ​​of 0.007, 0.494, and 2.85, while the preset standard emergency vectors are 0, 1, and 1. The calculated Euclidean distance is 1.912, which is greater than 0.76, indicating a severe fault. The real-time protection logic module is immediately triggered to output a trip signal to disconnect the faulty line, and the fault data is simultaneously uploaded to the distribution automation master station system via Ethernet interface.

[0076] The entire fault handling process, from detecting a severe disturbance to outputting a trip signal, took 4.2 milliseconds, meeting the requirement of triggering protection action within 5 milliseconds. The high-speed circular buffer of the hierarchical storage module retained high-frequency sampling data from the 10 seconds prior to the fault, while the event-triggered memory fully recorded waveform data for 5 seconds before and after the fault, providing detailed data support for subsequent fault analysis. The historical data compression memory used a 5-layer Daubechies 4 wavelet decomposition and run-length encoding method to compress the fault data to 18% of the original data volume, significantly reducing storage space usage.

[0077] This invention achieves refined monitoring and intelligent protection of the distribution network's operating status through multi-band power quality disturbance identification and graded response technology. Compared to traditional DTU terminals that rely solely on a single threshold for judgment, this invention employs 8-layer wavelet packet decomposition to decompose current and voltage signals into 256 frequency band sub-band signals. This enables the capture of energy distribution variations of weak fault characteristics across different frequency bands. Combined with the instantaneous frequency change rate extracted by Hilbert transform, a comprehensive disturbance characteristic value evaluation system is constructed, significantly improving the sensitivity and accuracy of disturbance identification. Traditional protection devices typically use single electrical quantity criteria such as overcurrent and undervoltage, which are easily affected by load fluctuations and voltage flicker, leading to malfunctions. In contrast, this invention establishes a multi-dimensional feature space, including a disorder vector, a continuity evaluation matrix, a discreteness feature vector, a mutation detection matrix, a balance matrix, and an emergency vector, to achieve graded identification and response for normal operation, minor disturbances, moderate disturbances, and severe disturbances, effectively avoiding malfunctions and failures to operate. Traditional protection devices have relatively simple response strategies and cannot flexibly adjust their handling measures according to the severity of disturbances. In contrast, this invention adopts differentiated response strategies for different disturbance levels. Under mild disturbance conditions, it only records data and continues to monitor. Under moderate disturbance conditions, it sends alarm information and determines whether to enter the early warning state. Under severe disturbance conditions, it immediately activates the real-time protection logic module to output a trip signal. This not only ensures the safe and reliable operation of the power grid but also avoids unnecessary power outage losses and improves power supply continuity and quality.

[0078] It should be noted that the variables involved in this invention are explained in detail in Tables 2, 3, and 4.

[0079] Table 2. Variable Explanation Table (Part 1)

[0080] Table 3. Variable Explanation Table (Part Two)

[0081] Table 4. Variable Explanation Table (Part 3)

[0082] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A DTU terminal for high-reliability distribution networks, comprising a main control chip, a multi-channel current and voltage acquisition module, a frequency and phase detection module, a hierarchical storage module, a communication interface module, and a real-time protection logic module, characterized in that, The multi-channel current and voltage acquisition module is used to acquire three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals of the distribution network. The frequency and phase detection module is used to monitor the grid frequency, three-phase voltage phase angle, and three-phase current phase angle. The multi-channel current and voltage acquisition module, the frequency and phase detection module, the hierarchical storage module, the communication interface module, and the real-time protection logic module are electrically connected to the main control chip. The main control chip is equipped with a multi-band power quality disturbance identification and hierarchical response module, which is used to perform multi-band decomposition and feature extraction on the acquired current and voltage signals, identify weak fault characteristics, determine the disturbance type and severity, and trigger corresponding processing measures according to different disturbance levels, and finally output power quality monitoring data and protection control commands.

2. The DTU terminal according to claim 1, characterized in that, The multi-channel current and voltage acquisition module uses a 24-bit analog-to-digital converter with a sampling rate of 10240Hz, and the frequency phase detection module uses phase-locked loop technology to achieve frequency tracking.

3. The DTU terminal according to claim 2, characterized in that, The hierarchical storage module includes a high-speed circular buffer memory, an event-triggered memory, and a historical data compressed memory. The high-speed circular buffer memory is used to store high-frequency sampled data from the most recent 10 seconds, and the event-triggered memory is used to store complete waveform data before and after a fault event.

4. The DTU terminal according to claim 3, characterized in that, The multi-band power quality disturbance identification and graded response module is used to perform the following steps: acquiring the instantaneous values ​​of three-phase current, three-phase voltage, zero-sequence current, and zero-sequence voltage from the multi-channel current and voltage acquisition module; acquiring the current grid frequency and three-phase voltage and current phase angle sequences from the frequency and phase detection module; performing 8-level wavelet packet decomposition on the instantaneous current and voltage values ​​to obtain the frequency band sub-band signals and calculating the frequency band energy values ​​to construct a disturbance degree vector; performing Hilbert transform on the frequency band sub-band signals to extract the instantaneous frequency change rate and construct a continuity evaluation matrix; weightedly fusing the disturbance degree vector and the continuity evaluation matrix to obtain a comprehensive disturbance feature value; determining the disturbance state type based on a threshold; constructing a discrete feature vector for minor disturbance states and using a support vector machine classifier to identify the disturbance type; constructing a mutation detection matrix for moderate disturbance states and calculating a moderate disturbance evaluation index; constructing a balance matrix for severe disturbance states; constructing an emergency degree vector and calculating the Euclidean distance between it and a preset standard emergency degree vector; and triggering corresponding processing measures based on the evaluation results.

5. The DTU terminal according to claim 4, characterized in that, The construction of the disorder vector involves performing 8-level wavelet packet decomposition on the instantaneous value sequences of three-phase current and three-phase voltage to obtain 256 frequency band sub-band signals. The frequency energy value of each frequency band sub-band signal within the time window is calculated, and the 1536 frequency band energy values ​​are arranged in order to form the disorder vector.

6. The DTU terminal according to claim 5, characterized in that, The calculation of the energy value of the frequency band is specifically to sum the squares of the amplitudes of all sampling points of the frequency band sub-band signal and then take the square root.

7. The DTU terminal according to claim 6, characterized in that, The construction of the continuity evaluation matrix specifically involves performing a Hilbert transform on the frequency band sub-band signal to obtain an instantaneous frequency sequence, calculating the instantaneous frequency change rate between adjacent sampling points, taking the absolute value of the instantaneous frequency change rate and then averaging it, and stacking the average instantaneous frequency change rates in chronological order to form a continuity evaluation matrix.

8. The DTU terminal according to claim 7, characterized in that, The weighted fusion calculation specifically involves multiplying the frequency band energy value by a weighting coefficient of 0.6 and then adding it to the instantaneous frequency change rate multiplied by a weighting coefficient of 0.4 to obtain the weighted fusion value. The average value of the weighted fusion value is then used to obtain the comprehensive disturbance characteristic value.

9. The DTU terminal according to claim 8, characterized in that, The determination of the disturbance state type is specifically based on the relationship between the comprehensive disturbance characteristic value and three thresholds, classifying the disturbance into normal operation state, slight disturbance state, moderate disturbance state, and severe disturbance state.

10. The DTU terminal according to claim 9, characterized in that, The construction of the discrete feature vector specifically involves extracting the amplitude standard deviation, amplitude skewness, and amplitude kurtosis of the three-phase current instantaneous value sequence and the three-phase voltage instantaneous value sequence within a time window, and combining the average values ​​of the zero-sequence current instantaneous value and the zero-sequence voltage instantaneous value within the time window to arrange the 20 discrete feature parameters into a discrete feature vector.

Images