Troubleshooting method for high-voltage switch cabinet

Abstract

The invention discloses a high-voltage switch cabinet fault elimination method, which comprises a sensor setting step, a feature learning step, a first model construction step, a fault classification step, a second model construction step and a fault confirmation step, and is characterized in that a plurality of groups of sensors are set to carry out real-time monitoring on operation data, sound data and image data of a high-voltage switch cabinet; the operation data, the sound data and the image data are preprocessed and fused to form multi-modal data, the multi-modal data are input into a multi-modal data set, and the multi-modal data set is analyzed through three models, so that multi-aspect fault detection and diagnosis on the high-voltage switch cabinet are realized on line. The method for performing periodic inspection or diagnosis by the inspector is relatively high in diagnosis capability, the diagnosis effect on some complex switch cabinet faults is improved, and the diagnosis accuracy of the high-voltage switch cabinet is improved.

Description

Technical Field

[0001] The invention relates to the technical field of online fault diagnosis of high-voltage switch cabinets, and in particular to a fault elimination method for high-voltage switch cabinets. Background Art

[0002] High-voltage switchgear is a very critical equipment in the distribution network system. It is responsible for putting some power lines into operation or cutting them out. It is related to the operation level, safety level and power supply service quality of the distribution network. Once a high-voltage switchgear fails, it will first cause damage to related power equipment and lines, and then cause large-scale power outages for users and other power supply accidents that affect the normal production and life of the people.

[0003] For high-voltage switchgear and other important electrical equipment, the distribution network equipment is generally inspected and maintained regularly. This can easily lead to the failure to timely inspect and maintain the equipment that should be repaired, thus affecting the technical performance and service life of the equipment that should be repaired. In severe cases, it may cause failures or accidents, affecting the reliability of power supply. At the same time, there are also many power supply equipment that are regularly inspected without any failure, which can easily lead to a waste of manpower, material and financial resources.

[0004] At present, the main types of high-voltage switchgear failures are the following common faults: insulation discharge failure, refusal to operate, malfunction failure, overheating failure, opening and closing failure, current-carrying failure, external force and other failures. In the condition inspection and maintenance work, the inspection personnel usually carry out periodic inspection and maintenance to monitor whether the equipment is aging, and monitor the equipment status by using ultrasonic detectors, infrared thermal imagers, and partial discharge testers on the switchgear; secondly, the maintenance personnel obtain relevant operation information of the equipment through online monitoring systems and daily live detection and other technologies, and quantitatively evaluate the equipment status in combination with mathematical analysis algorithms, and use this as auxiliary decision-making information for formulating maintenance plans, so as to minimize the power outage time caused by equipment maintenance, extend the service life of the equipment, reduce the maintenance cost of the equipment, realize the intelligentization, automation and dynamicization of the distribution network equipment status maintenance work, and improve the timeliness and accuracy of the status maintenance. Summary of the invention

[0005] The present invention provides a high-voltage switch cabinet troubleshooting method to solve the problem that the existing online diagnostic systems are mostly single systems, the diagnostic ability of the method of periodic inspection or diagnosis by patrol personnel is relatively weak, and the diagnostic effect for some complex switch cabinet faults is not good.

[0006] In order to solve the above technical problems, the present invention discloses the following technical solutions: The present invention provides a fault troubleshooting method for a high-voltage switch cabinet, comprising a sensor setting step, a feature learning step, a first model building step, a fault classification step, a second model building step and a fault confirmation step.

[0007] The sensor setting step is used to set multiple groups of sensors to monitor the operation data, sound data and image data of the high-voltage switchgear in real time, pre-process the operation data, the sound data and the image data, fuse them to form multimodal data and input them into the multimodal data set; the feature learning step is to use the ARIMA model to perform sequence modeling on the multimodal data to form a first model; the first model construction step is to construct a neural network model, train and optimize the model; the fault classification step is to classify the fault according to the trained neural network model; the second model construction step is to construct a second model by combining the parameter information of the high-voltage switchgear to obtain diagnostic information; the fault confirmation step is to obtain a diagnostic plan through the diagnostic information to determine the specific online fault information of the high-voltage switchgear.

[0008] Furthermore, in the sensor setting step, pre-processing of the operating data, the sound data and the image data includes a data cleaning step and a denoising step. The data cleaning step is to normalize the obtained operating data, the sound data and the image data; and the denoising step is to denoise the cleaned data through filtering technology.

[0009] Furthermore, the feature learning step includes a feature extraction step, a feature extraction step and an input step. The feature extraction step is to extract important features of multimodal data through an attention mechanism, and the extraction target is information related to fault mode classification; the feature conversion step is to convert the important features of multimodal data into feature vector representation; the input step is to input the feature vector into the first model to capture the timing information and long-term dependencies in the data to better process multimodal data.

[0010] Furthermore, in the fault confirmation step, the diagnostic information is transmitted to a human-machine intelligent interface, and the power grid operation and maintenance personnel determine the diagnostic plan, input the diagnostic plan data into the data for diagnostic work, and complete the maintenance of the high-voltage switch cabinet.

[0011] Furthermore, in the fault confirmation step, the second model is used to quantitatively describe the high-voltage switchgear fault to obtain the defect degree of the fault.

[0012] Compared with the prior art, the present invention has at least the following technical effects: The present invention provides a troubleshooting method for a high-voltage switch cabinet. By establishing a first model, a second model and a neural network model, the high-voltage switch cabinet can be detected and diagnosed in various aspects online. The method of periodic inspection or diagnosis by patrol personnel has a relatively strong diagnostic capability, improves the diagnostic effect of some complex switch cabinet faults, and increases the accuracy of high-voltage switch cabinet diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0014] Figure 1 A flowchart of a method for troubleshooting a high-voltage switchgear according to an embodiment of the present invention; Figure 2 This is a flowchart of preprocessing data according to an embodiment of the present invention; Figure 3 The figure is a flow chart of the feature learning steps described in the embodiment of the present invention. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0016] like Figure 1 As shown, the present invention provides a troubleshooting method for a high-voltage switch cabinet, comprising step S1) a sensor setting step, step S2) a feature learning step, step S3) a first model building step, step S4) a fault classification step, step S5) a second model building step and step S6) a fault confirmation step.

[0017] Step S1) Sensor setting step, setting multiple groups of sensors to monitor the operation data, sound data and image data of the high-voltage switchgear in real time, pre-processing and fusing the operation data, the sound data and the image data to form multimodal data and inputting them into the multimodal data set.

[0018] like Figure 2 As shown, in the sensor setting step, preprocessing the operation data, the sound data and the image data includes step S11) a data cleaning step and step S12) a denoising step.

[0019] Step S11) Data cleaning step, normalizing the obtained operation data, the sound data and the image data to ensure the accuracy and consistency of the data.

[0020] Step S12) De-noising step, in which the cleaned data is de-noised by filtering technology to improve the reliability and accuracy of the data.

[0021] Step S2) feature learning step, using ARIMA model to perform sequence modeling on multimodal data to form a first model.

[0022] like Figure 3 As shown, step S2) the feature learning step includes step S21) the feature extraction step, step S22) the feature extraction step and step S23) the input step.

[0023] Step S21) Feature extraction step, through the attention mechanism, extracts important features of multimodal data, and the extraction target is information related to fault mode classification.

[0024] Step S22) Feature extraction step, converting important features of multimodal data into feature vector representation.

[0025] Step S23) Input step, inputting the feature vector into the first model to capture the temporal information and long-term dependencies in the data to better process multimodal data.

[0026] Step S3) The first model building step is to build a neural network model, train and optimize the model.

[0027] Step S4) Fault classification step, classifying the faults according to the trained neural network model.

[0028] Step S5) A second model building step is to build a second model by combining parameter information of the high-voltage switch cabinet to obtain diagnostic information.

[0029] In this embodiment, during the training process, the first model, the second model and the neural network model update the parameters of the model through an optimization algorithm to minimize the loss function; then the parameters of the model are updated according to the gradient of the loss function, and multiple training iterations are repeated. In each training iteration, the output of the model is calculated, the loss function is calculated, and then the gradient of the loss function is calculated, and finally the optimization algorithm is used to update the parameters of the model.

[0030] Step S6) Fault confirmation step, by obtaining a diagnostic solution through diagnostic information, to determine the specific online fault information of the high-voltage switch cabinet.

[0031] Furthermore, in step S6) the fault confirmation step, the diagnostic information is transmitted to a human-machine intelligent interface, the power grid operation and maintenance personnel determine the diagnostic plan, input the diagnostic plan data into the data for diagnostic work, and complete the maintenance of the high-voltage switch cabinet.

[0032] Furthermore, in step S6) the fault confirmation step, the second model is used to quantitatively describe the high-voltage switch cabinet fault to obtain the defect degree of the fault.

[0033] The advantage of the present application is that a method for troubleshooting a high-voltage switchgear is provided. By establishing a first model, a second model and a neural network model, various fault detection and diagnosis of the high-voltage switchgear can be performed online. The method of periodic inspection or diagnosis by patrol personnel has a relatively strong diagnostic capability, which improves the diagnostic effect of some complex switchgear faults and increases the accuracy of diagnosis of high-voltage switchgear.

[0034] The above is a detailed introduction to a troubleshooting method for a high-voltage switchgear provided in an embodiment of the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the technical scheme and core idea of ​​the present invention. Ordinary technicians in this field should understand that they can still modify the technical schemes recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents; and these modifications or replacements do not make the essence of the corresponding technical scheme deviate from the scope of the technical scheme of the embodiments of the present invention.

Claims

1. A method for troubleshooting a high-voltage switch cabinet, characterized in that: include: a sensor setting step, setting a plurality of groups of sensors to monitor the operation data, sound data and image data of the high-voltage switch cabinet in real time, pre-processing and fusing the operation data, the sound data and the image data to form multimodal data and inputting the data into a multimodal data set; In the feature learning step, the ARIMA model is used to perform sequence modeling on the multimodal data to form the first model; The first model building step is to build a neural network model, train and optimize the model; Fault classification step, classifying faults according to the trained neural network model; A second model building step, building a second model by combining parameter information of the high-voltage switchgear to obtain diagnostic information; The fault confirmation step obtains the diagnostic plan through the diagnostic information and determines the specific online fault information of the high-voltage switchgear.

2. The troubleshooting method for high-voltage switchgear according to claim 1, characterized in that: In the sensor setting step, preprocessing the operation data, the sound data and the image data includes: a data cleaning step of normalizing the obtained operation data, the sound data, and the image data; and In the denoising step, the cleaned data is denoised by filtering technology.

3. The troubleshooting method for high-voltage switchgear according to claim 1, characterized in that: The feature learning step includes: Feature extraction step, through the attention mechanism, extracts important features of multimodal data, and the target is to extract information related to fault mode classification; A feature conversion step, which converts the important features of the multimodal data into feature vector representations; and The input step inputs the feature vector into the first model to capture the temporal information and long-term dependencies in the data to better process multimodal data.

4. The troubleshooting method for high-voltage switchgear according to claim 1, characterized in that: In the fault confirmation step, The diagnostic information is transmitted to a human-machine intelligent interface, and the power grid operation and maintenance personnel determine the diagnostic plan, input the diagnostic plan data into the data for diagnostic work, and complete the maintenance of the high-voltage switch cabinet.

5. The troubleshooting method for high-voltage switchgear according to claim 1, characterized in that: In the fault confirmation step, The second model is used to quantitatively describe the high-voltage switch cabinet fault and obtain the defect degree of the fault.

Images