CN-121347993-B - Method, device, system, medium and equipment for detecting power distribution network ground fault

CN121347993BCN 121347993 BCN121347993 BCN 121347993BCN-121347993-B

Abstract

The invention relates to the technical field of power distribution network fault detection, and discloses a power distribution network ground fault detection method, a device, a system, a medium and equipment, wherein the method comprises the steps of collecting three-phase current signals on a power distribution network line and judging whether a fault detection starting condition is met; when the fault prediction probability is larger than a probability threshold, a tripping instruction for controlling the circuit breaker connected with the fault line in the power distribution network line to open is generated, and if the ground fault is not detected to be isolated within the preset time, the upstream protection node is triggered to execute backup tripping operation based on time-step delay. The method combines pure electric current driving, intelligent identification, on-site quick action and differential backup, and realizes high-precision identification of the ground fault reduction dependent condition of the power distribution network and high-reliability small-range isolation of the ground fault.

Inventors

ZOU YU
CHEN LIJIANG
SU YIFENG
CHEN HENGLONG
CHEN JIANTAO
SONG YINGUAN
Bin Mingzhi
Liao Maotian
XU KUNNAN
Wang Daigao
FU SHUANG
FENG ZHIDONG
YAO MINFANG

Assignees

广西电网能源科技有限责任公司
广西电网有限责任公司钦州供电局

Dates

Publication Date: 20260508
Application Date: 20251219

Claims (9)

1. The method for detecting the ground fault of the power distribution network is characterized by comprising the following steps of: Collecting three-phase current signals on a power distribution network line, and judging whether the three-phase current signals meet fault detection starting conditions or not; When the three-phase current signal meets the fault detection starting condition, three-phase current waveform data in a preset time period before and after a starting time is input into a preset convolutional neural network model, and the convolutional neural network model directly outputs the fault prediction probability of the power distribution network line in a ground fault state according to an input tensor formed based on the three-phase current waveform data; comparing the fault prediction probability with a preset probability threshold, and generating a tripping instruction when the fault prediction probability is larger than the probability threshold, wherein the tripping instruction is used for controlling a breaker connected with a fault line in the power distribution network line to break; If the ground fault is not detected to be isolated within the preset time, triggering the current protection node based on the preset time level difference delay, and executing backup tripping operation; If the ground fault is not detected to be isolated within the preset time, triggering the current protection node based on the preset time level difference delay, and executing the backup trip operation, wherein the backup trip operation comprises the following steps: Sequentially arranging a plurality of protection nodes along the direction from the load side to the power supply side of the power distribution network, wherein the protection nodes are configured with action delays which are gradually increased, and the action delays of adjacent protection nodes positioned at the downstream of the current protection node are smaller than the action delays of the current protection node; continuously detecting whether the power distribution network ground fault of the area downstream of the current protection node is isolated or not during the time delay and timing of the current protection node; if the current protection node detects that the ground fault of the power distribution network is isolated before the delay timing of the current protection node is finished, automatically resetting the current protection node; and if the current protection node does not detect that the ground fault of the power distribution network is isolated after the delay time of the current protection node is finished, the current protection node executes tripping operation.
2. The method of claim 1, wherein the collecting three-phase current signals on a distribution network line and determining whether the three-phase current signals satisfy a fault detection initiation condition comprises: sampling three-phase current signals on a power distribution network line in real time, and calculating current variation of adjacent sampling points; Counting a steady-state average value and a steady-state standard deviation of the current variation in a preset sliding time window, and determining a dynamic starting threshold value based on the steady-state average value, the steady-state standard deviation and a preset sensitivity coefficient; When the current variation of the current sampling point is larger than the dynamic starting threshold value, judging that the three-phase current signal meets the fault detection starting condition.
3. The method of claim 2, wherein the determining a dynamic start threshold based on the steady state average value, the steady state standard deviation, and a preset sensitivity coefficient comprises: The following formula is constructed based on the steady-state average value, the steady-state standard deviation and a preset sensitivity coefficient to determine the dynamic start threshold: wherein T is the dynamic start threshold, Steady state As a result of the steady-state average value, Steady state As a function of the steady-state standard deviation, I k is the current variation, K is the sensitivity coefficient, N is the sampling point number of a sliding time window, and K is the sampling point.
4. The method according to claim 1, wherein the inputting the three-phase current waveform data within a preset time period before and after the start time into a preset convolutional neural network model, the convolutional neural network model directly outputting the fault prediction probability that the power distribution network line is in the ground fault state according to an input tensor formed based on the three-phase current waveform data, includes: Acquiring a time sequence representation corresponding to the three-phase current signal, obtaining three-phase current waveform data, carrying out normalization processing on the three-phase current waveform data, and intercepting data in preset time periods before and after a starting time to form an input tensor; inputting the input tensor into a preset convolutional neural network model, wherein the convolutional neural network model comprises a one-dimensional convolutional layer, a full-connection layer and an output layer; extracting coupling transient characteristics of the three-phase current signals in a time dimension through the one-dimensional convolution layer, and outputting a linear score through the full-connection layer; And converting the linear score into a normal prediction probability of the power distribution network in a normal running state and a fault prediction probability of the power distribution network in a ground fault state by adopting a Softmax function through the output layer, wherein the sum of the normal prediction probability and the fault prediction probability is 1.
5. The method of claim 1, wherein the determination of whether the distribution network ground fault is isolated is made by: And whether the fault current detected by the current protection node disappears or not.
6. An apparatus for detecting a ground fault of an electrical distribution network, wherein the apparatus is configured to implement the method for detecting a ground fault of an electrical distribution network according to any one of claims 1 to 5, and comprises: the current sampling module is used for synchronously collecting three-phase current signals; A mass memory for storing current waveform data before and after a fault occurrence time when a ground fault is detected; the main control processing module is used for running a preset convolutional neural network model to identify the grounding fault and outputting a fault judgment result; the relay outlet module is used for outputting a tripping instruction according to the fault judgment result; The power supply module is used for converting an externally provided AC power supply or DC power supply into a digital power supply for working a digital circuit, an analog power supply for working an analog circuit and an operation power supply of a relay port; The wireless communication module is used for uploading running state data to a remote power distribution network master station and receiving a fixed value adjustment instruction, a remote control opening instruction and a closing instruction issued by the remote power distribution network master station; And the ad hoc network communication module is used for networking a plurality of fault detection devices and performing data interaction with a remote computer.
7. A power distribution network ground fault detection system, the system comprising: the fault detection starting unit is used for collecting three-phase current signals on a power distribution network line and judging whether the three-phase current signals meet fault detection starting conditions or not; the prediction probability output unit is used for inputting three-phase current waveform data in a preset duration before and after the starting time into a preset convolutional neural network model when the three-phase current signal meets the fault detection starting condition, and the convolutional neural network model directly outputs the fault prediction probability of the power distribution network circuit in a ground fault state according to an input tensor formed based on the three-phase current waveform data; The fault line cutting unit is used for comparing the fault prediction probability with a preset probability threshold value, and generating a tripping instruction when the fault prediction probability is larger than the probability threshold value, wherein the tripping instruction is used for controlling a breaker connected with a fault line in the power distribution network line to break; The step difference backup tripping unit is used for triggering the current protection node based on the preset time step difference delay and executing backup tripping operation if the ground fault is not detected to be isolated within the preset time; The step-difference backup tripping unit is further used for sequentially arranging a plurality of protection nodes along the direction from the load side to the power side of the power distribution network, wherein the protection nodes are configured with action delays which are gradually increased, and the action delays of adjacent protection nodes positioned at the downstream of the current protection node are smaller than the action delays of the current protection node; continuously detecting whether the power distribution network ground fault of the area downstream of the current protection node is isolated or not during the time delay and timing of the current protection node; if the current protection node detects that the ground fault of the power distribution network is isolated before the delay timing of the current protection node is finished, automatically resetting the current protection node; and if the current protection node does not detect that the ground fault of the power distribution network is isolated after the delay time of the current protection node is finished, the current protection node executes tripping operation.
8. A storage medium having stored thereon a computer program, which when executed by a processor, implements the steps of the method of any of claims 1 to 5.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the computer program when executed by the processor implements the steps of the method according to any one of claims 1 to 5.

Description

Method, device, system, medium and equipment for detecting power distribution network ground fault Technical Field The invention relates to the technical field of power distribution network fault detection, in particular to a power distribution network ground fault detection method, a device, a system, a medium and equipment. Background In a 10kV power distribution network system, a neutral point grounding mode is usually in the forms of ungrounded, arc suppression coil grounding or small-resistance grounding and the like, wherein the ungrounded and arc suppression coil grounding system is widely applied to rural and partial urban distribution networks due to simple structure and good power supply continuity, the urban power grid is capable of improving the rapid fault removal capacity, the neutral point grounding mode is increasingly adopted, however, no matter the grounding mode is adopted, single-phase grounding faults are the most common fault types in the power distribution network, in the ungrounded or arc suppression coil grounding system, a single-phase grounding circuit is not formed, fault current is mainly system capacitance current with small amplitude and is easily influenced by the length of a line, the operation working condition and the compensation effect of the arc suppression coil, and in the small-resistance grounding system, high-resistance grounding faults can also lead to weak fault current characteristics, so that the problem of accurately and rapidly identifying grounding faults and realizing accurate isolation under complex operation conditions becomes a key problem of safe and reliable operation of the power distribution network. The method has the advantages that the method mainly depends on a small current line selection device at a transformer station end, zero sequence voltage/current power direction criteria, fault indicators or manual trial pulling and the like, however, the method has different defects that firstly, the zero sequence protection method is required to depend on a zero sequence voltage transformer, most lines are not configured, false zero sequence voltage is easy to generate when three phases are unbalanced, false starting is caused, the compensation effect of an arc suppression coil can weaken zero sequence current characteristics, the criterion reliability is further reduced, secondly, the accuracy of the traditional line selection device is generally low in an arc suppression coil system or high-resistance grounding scene, the fault indicators are often required to be checked through round-pull non-fault lines, large-range power failure is caused, normal loads are possibly cut by mistake, finally, the fault indicators are limited by installation environments, are difficult to be deployed in a cable channel or a closed circuit, are easy to be influenced by transient disturbances such as lightning stroke and switching operation, the fault section is easy to be indicated, and the fault section can not be accurately positioned. In summary, in the prior art, the problem that the ground fault detection of the power distribution network generally depends on an additional sensor, has weak anti-interference capability, has poor adaptability to a high-resistance/compensation system, has low fault recognition accuracy, has a large power failure range and the like exists. Therefore, a power distribution network ground fault detection method which does not depend on zero sequence voltage, is only based on conventional three-phase current and has high identification accuracy is urgently needed to be researched. Disclosure of Invention In view of the above, the application provides a method, a device, a system, a medium and equipment for detecting a power distribution network ground fault, which mainly aims to solve the technical problems that in the prior art, the power distribution network ground fault detection depends on an additional sensor, has weak anti-interference capability, has poor adaptability to a high-resistance/compensation system, has low fault recognition accuracy and the like. According to a first aspect of the present invention, there is provided a method for detecting a ground fault of a power distribution network, including: Collecting three-phase current signals on a power distribution network line, and judging whether the three-phase current signals meet fault detection starting conditions or not; When the three-phase current signal meets the fault detection starting condition, three-phase current waveform data in a preset time length before and after a starting time are input into a preset convolutional neural network model, and the fault prediction probability of the power distribution network line in a ground fault state is obtained; comparing the fault prediction probability with a preset probability threshold, and generating a tripping instruction when the fault prediction probability is larger than the probability threshold, wherein the trip