CN118091270A - High-impedance fault detection method for distribution automation feeder terminal unit - Google Patents

Abstract

The method for detecting high impedance faults of feeder terminal units of distribution automation includes the following steps: Step 1: using wavelet transform to decompose the fault current signal of the distribution feeder into characteristic signals of different frequency bands and positions, and extracting the high impedance fault HIF characteristic signal; Step 2: according to the fault situation, using wavelet transform to extract the fluctuation characteristics of zero sequence fault current; Step 3: establishing a back propagation neural network BPNN as an identifier for high impedance fault HIF detection; Step 4: designing a feeder terminal unit FTU with built-in DSP; Step 5: performing high impedance fault HIF detection. The fault detection method of the present invention can reduce the power outage area and maintenance time, and can significantly improve the reliability of power supply.

Description

High impedance fault detection method for feeder terminal unit in distribution automation

Technical Field

The present invention relates to the technical field of power distribution fault detection, and in particular to a method for detecting high impedance faults of a power distribution automation feeder terminal unit.

Background technique

In the existing technology, the high impedance fault (HIF) problem has brought serious harm to the power distribution system, infrastructure, and especially residential electricity consumption. Although PC-based fault location and detection technology has been widely used in power distribution systems, high-precision detection of high impedance faults (HIF) is still challenging.

In order to more effectively detect highly nonlinear high impedance faults HIF, various wavelet transform technologies have attracted much attention. Feeder automation FA is the most basic and important part of the distribution network automation system DAS. It is an integrated system that realizes real-time monitoring, coordination and control of distribution line equipment. The feeder terminal unit FTU is an intelligent electronic device used for feeder automation FA, which is used to collect data from many feeders and convert them to the control center. With the rapid development of detection technology and embedded technology, the power system has put forward higher requirements on the reliability and accuracy of the feeder terminal unit FTU. However, the traditional feeder terminal unit FTU does not have a high impedance fault HIF detection DSP module and a high-bandwidth current sensor module. When a high impedance fault HIF occurs in the feeder, it is difficult to effectively detect and locate it. Many types of high impedance faults HIF cannot even be detected, making the reliability of the power system cannot be guaranteed.

In addition, the high impedance fault HIF signal detected by the existing feeder terminal unit FTU usually has the problem of incomplete detection, and when a high impedance fault HIF occurs, it is difficult for the feeder terminal unit FTU to quickly detect the fault and send the fault sign to the remote control center through the communication network. This is one of the biggest technical difficulties in the detection and positioning of high impedance fault HIF in actual distribution automation.

Summary of the invention

In order to solve the technical problems existing in high impedance fault HIF detection, the present invention provides a distribution automation feeder terminal unit high impedance fault detection method, which uses wavelet transform to extract the characteristics of high impedance fault HIF signal; at the same time, an enhanced feeder terminal unit FTU is developed to perform wavelet transform and high impedance fault HIF detection based on back propagation neural network BPNN. It can reduce the power outage area and maintenance time, and can significantly improve power reliability.

The technical solution adopted by the present invention is:

A method for detecting high impedance faults of a distribution automation feeder terminal unit comprises the following steps:

Step 1: Use wavelet transform to decompose the fault current signal of the distribution feeder into characteristic signals of different frequency bands and positions, and extract the high impedance fault HIF characteristic signal;

Step 2: According to the fault situation, the fluctuation characteristics of zero-sequence fault current are extracted by wavelet transform;

Step 3: Establish a back propagation neural network BPNN as an identifier for high impedance fault HIF detection;

Step 4: Design a feeder terminal unit FTU with built-in DSP;

Step 5: Perform high impedance fault HIF detection based on DSP.

In step 1, wavelet transform is used as a downsampling filter bank. Downsampling reduces the sampling frequency by half after each level of decomposition, so the number of samples will also be reduced by half, as follows:

The WT of four decomposition levels with sampling frequency _fs = 6kHz is detected, _c0 [n] is the original input signal, g[-n] is the high-pass filter, h[-n] is the low-pass filter, and ↓2 is downsampling by 2 times; through continuous low-pass and high-pass filtering, the input signal of each level of wavelet transform is decomposed into high-frequency and low-frequency signal components. At the first level, the original input signal _c0 [n] is decomposed into low-frequency signal component _c1 [n] and high-frequency signal component _d1 [n] by the filter; then, at the second level, the low-frequency signal component of the low-frequency signal component _c1 [n] is decomposed into low-frequency signal component _c2 [n] and high-frequency signal component _d2 [n], and so on until the end of the fourth level.

In the step 2, the fluctuation characteristics of the zero-sequence fault current 3I ₀ are extracted by wavelet transform for high impedance fault HIF detection.

In step 3, the back propagation neural network BPNN consists of 4 input neurons in the input layer, 8 neurons in a hidden layer and 1 neuron in the output layer;

The four input neurons d ₂ [n], d ₃ [n], d ₄ [n], and c ₄ [n] are the normalized wavelet coefficients of the zero-sequence fault current 3I ₀ respectively; the hidden layer contains 8 neurons, and the number of hidden neurons is determined by the network growth method, which depends on the tolerance value of the mean square error and is set to 0.001. Through a large number of simulation analyses, 8 hidden neurons are the best choice. The hidden layer uses the tan-sigmoid activation function to help the network better understand the data by learning and extracting the features of the input data;

The output layer has a neuron with a linear transfer function, which is responsible for converting the results of the hidden layer into specific outputs. Its output represents the predicted value of the input data.

The output of the back propagation neural network BPNN is rounded to 1 and 0, that is, when a fault occurs, the data above 0.5 is considered to be 1, and when it is operating normally, the data below 0.5 is considered to be 0.

In step 4, the feeder terminal unit FTU includes:

Central processing unit module, power module, digital communication module, input/output digital contact module, analog input module; each module communicates with each other through a serial communication interface, as shown in FIG4 .

The central processing unit module is respectively connected to the digital communication module, the input/output digital contact module, the analog input module, and the DSP module;

The power module is respectively connected to the power module, the digital communication module, the input/output digital contact module, and the DSP module to provide power for these modules.

Central processing unit module: used to improve the real-time performance and reliability of the device, the hardware model used is F8650X;

Power module: provides power for system operation, the hardware model used is GX-WK500-D48;

Digital communication module: realizes data transmission between devices, the hardware model used is XM1302E;

Input/output digital contact module: used to process and control digital signals, including switch signals, and the hardware model used is DQ 32x24VDC0.5A ST;

Analog input module: used to collect and process analog signals and convert analog signals into digital signals. The hardware model used is AI 16xI 2,4-wire ST.

The step 5 comprises the following steps:

Step (1), initialization: set parameters and variables, and reset initial values;

Step (2), input data: sample the instantaneous zero-sequence fault current 3I ₀ in the feeder every 10 seconds at a sampling frequency of 6 kHz;

Step (3), wavelet transform: performing wavelet transform on the sampled current signal;

Step (4), normalization: normalize the obtained wavelet coefficients to ensure that the normalized coefficients are consistent when input into the corresponding interval of NN;

Step (5), fault detection: send the fault judgment to the feeder terminal unit FTU through the trained neural network, otherwise, the process returns to step (2);

Step (6), activating the service recovery function FDIR: When a high impedance fault HIF occurs, the feeder terminal unit FTU will send a flag signal to the FRTU, and the feeder dispatch control center FDCC will immediately receive the fault information sent by the remote terminal FRTU of the distribution network, and the service recovery function FDIR mechanism will be activated, and the maintenance personnel will perform the following functions:

① Identify the fault area, ② Remotely disconnect the line switches before and after the fault point, ③ Restore power supply to the unaffected areas upstream and transmit power to downstream areas through other distribution lines.

An enhanced feeder terminal unit FTU, comprising: a central processing unit module, a power supply module, a digital communication module, an input/output digital contact module, and an analog input module;

The central processing unit module is respectively connected to the digital communication module, the input/output digital contact module, the analog input module, and the DSP module;

The power module is respectively connected to the power module, the digital communication module, the input/output digital contact module, and the DSP module to provide power to these modules;

The DSP module performs wavelet transform analysis and high impedance fault HIF judgment based on the back propagation neural network BPNN.

The present invention provides a method for detecting high impedance faults in a feeder terminal unit of a distribution automation system, and the technical effects are as follows:

1) Advantages of step 1 of the present invention: There are five main HIF features: ① low voltage and arc, ② irregular fault current, ③ nonlinearity of fault current, ④ accumulation and step of current, ⑤ asymmetry of current waveform; using wavelet transform as a downsampling filter can decompose the signal in time, extract the HIF features and detect its occurrence.

2) Advantages of step 2 of the present invention: Low-frequency analysis of the decomposed signal at the highest level can detect signals that change slowly and last for a long time, and wavelet transform can extract important features of fault voltage and fault current signals, even if they are very weak.

3) Advantages of step 3 of the present invention: Establishing a back propagation neural network BPNN as an identifier for high impedance fault HIF detection can reduce excessive input brought by wavelet signals, thereby achieving the effect of reducing convergence difficulty, fast training speed and high test efficiency.

4) Advantages of step 4 of the present invention: Designing a feeder terminal unit FTU with a built-in DSP improves its reliability and accuracy and can enhance the FDIR function.

5) Advantages of step 5 of the present invention: The HIF detection algorithm is written in C language. When HIF occurs, the FTU can quickly detect the fault and send a fault flag to the remote control center through the communication network to activate the FDIR mechanism. Subsequently, the maintenance personnel can not only understand the occurrence of the fault, but also determine the fault area according to the location of the FTU, thereby ensuring the safety and reliability of the power system operation.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a schematic diagram of the four decomposition levels of wavelet transform.

FIG2 is a schematic diagram corresponding to the zero-sequence HIF current.

FIG3 is a schematic diagram of the BPNN structure for HIF detection.

FIG4 is a structural diagram of a feeder terminal unit FTU.

FIG5 is a flow chart of HIF detection.

Detailed ways

A method for detecting high impedance faults in feeder terminal units of distribution automation is proposed, which includes two stages: first, wavelet transform is performed on the fault current signal of the distribution feeder to obtain the corresponding data signal; then, a properly trained neural network is used for fault identification processing. In the proposed method, wavelet transform is used to extract the features of high impedance fault HIF signals and back propagation neural network BPNN is used for high impedance fault HIF detection to develop a new type of HIF detector, called enhanced feeder terminal unit FTU. The designed enhanced enhanced feeder terminal unit FTU is a high-performance HIF detector, which is an improvement on the existing feeder terminal unit FTU with an embedded digital signal processor DSP and a high-frequency current sensor. The enhanced feeder terminal unit FTU monitors the operation of the distribution feeder system and sends a fault flag signal to the feeder dispatch control center FDCC when a high impedance fault HIF occurs. Based on the FA (Feeder Automation) mechanism, the high impedance fault HIF detection, isolation and service restoration (FDIR) function can be activated. Therefore, the power reliability can be significantly improved by reducing the power outage area and maintenance time. The effectiveness of the technology is verified by staged fault testing.

The following steps are involved:

Step 1. Use wavelet transform as a downsampling filter bank to timely decompose the signal into characteristic signals of different frequency bands and positions. Wavelet transform can decompose the sampled signal into sub-signals of different frequencies. Each sub-signal is in a different frequency band, and different features are highlighted at the corresponding positions of different frequency bands. In Figure 1, d ₁ [n], d ₂ [n], d ₃ [n], and d ₄ [n] are high-frequency signals after each decomposition, and c ₁ [n], c ₂ [n], c ₃ [n], and c ₄ [n] are low-frequency signals after each decomposition.

The high impedance fault HIF feature can be extracted and detected.

Downsampling halves the sampling frequency after each level of decomposition, so the number of samples will also be halved. The present invention uses wavelet DB10 as the mother wavelet function for current signal decomposition. In order to clearly analyze the high-frequency transient phenomenon of the arc, the WT of four decomposition levels with a sampling frequency of _fs = 6kHz is detected, as shown in Figure 1: _c0 [n] is the original input signal, g[-n] is a high-pass filter, h[-n] is a low-pass filter, ↓2 is downsampling by 2 times; through continuous low-pass and high-pass filtering, the input signal of each level of wavelet transform is decomposed into high-frequency and low-frequency signal components. In the first layer, the original input signal _c0 [n] is decomposed into a low-frequency signal component _c1 [n] and a high-frequency signal component _d1 [n] through a filter; then, in the second layer, the low-frequency signal component of the low-frequency signal component _c1 [n] is decomposed into a low-frequency signal component _c2 [n] and a high-frequency signal component _d2 [n], and so on.

Step 2: When a high impedance fault HIF occurs, the voltage and current of the fault phase change very little, so it is not easy to detect the fault. On the contrary, due to the increase in system imbalance, the fault will cause a large fluctuation in the zero-sequence current. According to the fault situation, the wavelet transform is used to extract the fluctuation characteristics of the zero-sequence fault current 3I ₀ for high impedance fault HIF detection. The waveform of the instantaneous zero-sequence fault current after 4 wavelet transform decompositions is shown in Figure 2: where c ₀ [n] is the original sampling signal, d _{1 [n]} ~ d ₄ [n] are details, and c ₄ [n] is an approximation. It can be observed that the amplitude of the waveform d ₁ [n] is small, but it has a spike and intermittent duration after the fault occurs. This phenomenon can be observed even at other levels. By analyzing the high frequency of the level 1 decomposition signal, transient disturbances can be easily observed. And by performing low-frequency analysis on the decomposed signal at the highest level, signals with slow changes and long duration can be detected. Wavelet transform can extract important features of fault voltage and fault current signals, even if they are very weak.

Step three, use the back propagation neural network BPNN as an identifier for HIF detection. The back propagation neural network BPNN structure for HIF detection proposed by the present invention is shown in Figure 3. The structure consists of 4 input neurons in the input layer, 8 neurons in 1 hidden layer and 1 neuron in the output layer. The 4 input neurons _d2 [n], _d3 [n], _d4 [n], _c4 [n] are the normalized wavelet coefficients of the zero-sequence fault current _3I0 respectively. The number of hidden neurons is determined by the network growth method, which depends on the tolerance value of the mean square error and is set to 0.001. The output layer has a neuron with a linear transfer function. The output of the NN is rounded to 1 and 0, that is, when a fault occurs, data above 0.5 is considered to be 1, and data below 0.5 is considered to be 0 during normal operation.

Step 4, a FTU with built-in DSP was designed to enhance the FDIR function. The enhanced FTU includes multiple modules, as shown in Figure 4: CPU module, power module, digital communication module, input/output digital contact module, analog input module, and DSP module for HIF detection;

The enhanced FTU utilizes the available FTU to perform HIF detection and location. The added functions and requirements are:

1): Use current sensor to simulate input interface. To process the measured current signal 3I ₀ , the input contact of the signal has the following characteristics: current input ±5A, noise resistance less than 15mA, current sensor bandwidth of 10kHz, and DSP sampling frequency must reach 10kHz.

2): The fault flag must be generated by the FTU to indicate that the HIF has been detected and identified. After the fault flag is generated, the FTU itself must automatically reset the flag and save the flag in the buffer to provide fault information to the FRTU. After the FTU reads the fault flag, it begins to clear the stored fault flag.

3): Use DSP TMS320C6713 chip for wavelet transform analysis and neural network fault judgment, and embed it into each FTU of the distribution system.

The TMS320C6713 DSP is used for HIF detection using wavelet transform and neural network. This DSP has an efficient cache configuration and high-speed computing capability, and is a low-cost development platform for real-time digital signal processing applications based on C language. It consists of a small circuit board containing the TMS320C6713 floating-point DSP and a TLV320AIC23 analog interface circuit, and is connected to the host through a USB interface. In addition, in order to detect and measure the current in the fault feeder, the ACS723LLCTR-05AB chip is used as a current sensor. It can detect ±5A AC or DC current and convert it into a 400mV/A voltage/current signal proportionally.

Step 5: The HIF detection algorithm is written in C language. The detection process is shown in Figure 5, and the specific steps are as follows:

Step (1), initialization:

Set parameters and variables, and reset initial values;

Step (2), input data:

The instantaneous zero-sequence current 3I ₀ in the feeder is sampled every 10 seconds at a sampling frequency of 6 kHz; Step (3), wavelet transform:

Perform wavelet transform on the sampled current signal;

Step (4), normalization:

Normalize the obtained wavelet coefficients to ensure that the normalized coefficients are consistent when input into the corresponding interval of NN;

Step (5), fault detection:

Through the trained neural network, the fault judgment is sent to the FTU, otherwise, the process returns to step (2);

Step (6)

Activate FDIR: When HIF occurs, FTU will send a flag signal to FRTU. The feeder dispatch center FDCC will receive the fault information from FRTU as soon as possible. The FDIR mechanism will be activated and the maintenance personnel will perform the following functions: ① Identify the fault area, ② Remotely disconnect the line switches before and after the fault point, ③ Restore power supply to the unaffected areas upstream and transmit power to the downstream areas through other distribution lines.

Claims (7)

1. The high-impedance fault detection method for the distribution automation feeder terminal unit is characterized by comprising the following steps of:

step 1: decomposing the fault current signal into characteristic signals of different frequency bands and positions by adopting wavelet transformation, and extracting high-impedance fault HIF characteristic signals;

Step 2: extracting fluctuation characteristics of zero sequence fault current by wavelet transformation aiming at fault conditions;

Step 3: establishing a Back Propagation Neural Network (BPNN) as an identifier for High Impedance Fault (HIF) detection;

Step 4: designing a Feeder Terminal Unit (FTU) with a built-in DSP;

step 5: high impedance fault HIF detection is performed based on the DSP.

2. The power distribution automation feeder termination unit high impedance fault detection method of claim 1, wherein: in the step 1, the wavelet transform is used as a downsampling filter bank, the downsampling frequency is halved after each stage of decomposition, and thus the number of samples is halved, specifically as follows:

Detecting the WT of four decomposition levels with sampling frequency f _s =6 kHz, c ₀ [ n ] being the original input signal, g < -n > being a high pass filter, h < -n > being a low pass filter, +.2 being downsampled by 2 times; the input signal of each stage of wavelet transformation is decomposed into high-frequency and low-frequency signal components through continuous low-pass and high-pass filtering, and at layer 1, the original input signal c ₀ [ n ] is decomposed into a low-frequency signal component c ₁ [ n ] and a high-frequency signal component d ₁ [ n ] through a filter; subsequently, at layer 2, the low-frequency signal component of the low-frequency signal component c ₁ [ n ] is decomposed into a low-frequency signal component c ₂ [ n ] and a high-frequency signal component d ₂ [ n ], and so on until the fourth layer ends.

3. The power distribution automation feeder termination unit high impedance fault detection method of claim 1, wherein: in the step 2, the fluctuation characteristic of the zero sequence fault current 3I ₀ is extracted by wavelet transformation and is used for high impedance fault HIF detection.

4. The power distribution automation feeder termination unit high impedance fault detection method of claim 1, wherein: in the step 3, the back propagation neural network BPNN is composed of 4 input neurons in the input layer, 8 neurons in the 1 hidden layer and 1 neuron in the output layer;

The 4 input neurons d ₂[n]、d₃[n]、d₄[n]、c₄ [ n ] are respectively normalized wavelet coefficients of the zero sequence fault current 3I ₀; the hidden layer contains 8 neurons, the output layer has a neuron with a linear transfer function, and the neuron is responsible for converting the result of the hidden layer into specific output, and the output represents a predicted value for input data;

The back propagation neural network BPNN has its output rounded to 1 and 0, i.e., data of 0.5 or more is considered to be 1 when a fault occurs, and data of 0.5 or less is considered to be 0 when normal operation occurs.

5. The power distribution automation feeder termination unit high impedance fault detection method of claim 1, wherein: in the step 4, the feeder terminal unit FTU includes:

The device comprises a central processing unit module, a power supply module, a digital communication module, an input/output digital contact module and an analog input module; the central processing unit module is respectively connected with the digital communication module, the input/output digital contact module, the analog input module and the DSP module;

The power module is respectively connected with the power module, the digital communication module, the input/output digital contact module and the DSP module to provide power for the modules.

6. The power distribution automation feeder termination unit high impedance fault detection method of claim 1, wherein: the step 5 comprises the following steps:

step (1), initializing: setting parameters and variables, and resetting initial values;

Step (2), inputting data: sampling the instantaneous zero sequence fault current 3I ₀ in the feeder line every 10 seconds at a sampling frequency of 6 kHz;

Step (3), wavelet transformation: performing wavelet transformation on the sampled current signal;

step (4), normalization: normalizing the obtained wavelet coefficient to ensure that the normalized coefficient is consistent when the corresponding interval of NN is input;

step (5), fault detection: transmitting the fault judgment to the Feeder Terminal Unit (FTU) through the trained neural network, otherwise, returning the process to the step (2);

Step (6), activating a service restoration function FDIR: when the high impedance fault HIF occurs, the feeder terminal unit FTU will send a flag signal to FRTU, and the feeder dispatch control center FDCC receives the fault information sent from the remote terminal FRTU of the power distribution network for the first time, and the service restoration function FDIR mechanism will be started.

7. An enhanced feeder terminal unit, FTU, characterized by comprising:

The device comprises a central processing unit module, a power supply module, a digital communication module, an input/output digital contact module and an analog input module; the central processing unit module is respectively connected with the digital communication module, the input/output digital contact module, the analog input module and the DSP module;

The power module is respectively connected with the power module, the digital communication module, the input/output digital contact module and the DSP module and provides power for the modules;

the DSP module performs any one of the high impedance fault detection methods as set forth in claims 1-6.