Distribution automation equipment supervision and acquisition system

Abstract

The invention provides a power distribution automation equipment supervision and acquisition system. According to the system, a space-time alignment algorithm is adopted for infrared signals, vibration signals and electrical signals to achieve time synchronization and space registration; thermal imaging analysis is carried out on the infrared signals after synchronous registration, an overheating area is automatically selected, the temperature gradient is calculated, vibration signal analysis is carried out on the vibration signals after synchronous registration, the vibration entropy is calculated, partial discharge analysis is carried out on the electrical signals, and the discharge pulse number is obtained; and carrying out weighted fusion on the temperature gradient, the vibration entropy and the discharge pulse number, inputting the fused features into an LSTM model to carry out time sequence correlation analysis, obtaining a fault type and a fault type probability, and carrying out early warning on the 750KV transformer substation according to the fault type probability. The invention provides an automatic equipment supervision and acquisition system for a 750KV transformer substation, and the system has the advantages of low omission ratio and high accuracy.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a distribution automation equipment monitoring and acquisition system. Background Art

[0002] The SCADA (Supervisory Control And Data Acquisition) system is a computer-based DCS (Distributed Control System) and power automation monitoring system that can be applied to data acquisition and monitoring control in the power field.

[0003] In power systems, SCADA systems are the most widely used and technologically mature. They play a crucial role in telecontrol systems, monitoring and controlling on-site equipment to implement functions such as data acquisition, equipment control, measurement, parameter adjustment, and various signal alarms. RTUs (Remote Terminal Units) and FTUs (Feeder Terminal Units) are key components of SCADA systems and play a crucial role in today's integrated substation automation.

[0004] The SCADA system is a computer-based production process control and scheduling automation system that can monitor and control on-site operating equipment, and realize functions such as data collection, equipment control, measurement, parameter adjustment, and various signal alarms. Summary of the Invention

[0005] The present invention provides a distribution automation equipment supervision and collection system, the main purpose of which is to provide an automation equipment supervision and collection system for 750KV substations, improve supervision accuracy and reduce false alarm rate.

[0006] An embodiment of the present invention provides a distribution automation equipment monitoring and acquisition system. An infrared thermal imager, an acoustic vibration sensor, a current sensor, and a voltage sensor are installed at a 750 kV substation. The monitoring and acquisition system implements an early warning function for the 750 kV substation through the following steps:

[0007] Acquire an infrared signal through the infrared thermal imager, acquire a vibration signal through the acoustic vibration sensor, and acquire an electrical signal through the current sensor and the voltage sensor;

[0008] Using a time-space alignment algorithm to achieve time synchronization and spatial registration of the infrared signal, the vibration signal, and the electrical signal;

[0009] Perform thermal imaging analysis on the synchronously registered infrared signal, automatically select the overheated area and calculate the temperature gradient, perform vibration signal analysis on the synchronously registered vibration signal, calculate the vibration entropy, and perform partial discharge analysis on the electrical signal to obtain the number of discharge pulses;

[0010] The temperature gradient, the vibration entropy, and the number of discharge pulses are input into the LSTM model for time series correlation analysis to obtain the fault type and the fault type probability, and an early warning is issued for the 750KV substation based on the fault type probability.

[0011] Furthermore, the infrared signal, the vibration signal and the electrical signal are subjected to time-space alignment algorithm to achieve time synchronization and space registration, the steps comprising:

[0012] Taking the sampling rate of the electrical signal as a reference frequency, cubic spline interpolation is performed on the infrared signal to increase the sampling rate of the infrared signal to the reference frequency, and a dynamic time warping algorithm is applied to the vibration signal and the electrical signal to compensate for mechanical vibration transmission delay and achieve time synchronization function;

[0013] The temperature matrix output by the infrared thermal imager is converted into a three-dimensional point cloud, and based on the three-dimensional point cloud and the transformation matrix, the point cloud correspondence between the infrared signal and the real equipment in the 750KV substation is realized, wherein the transformation matrix is ​​obtained according to the installation position of the infrared thermal imager. The transformation matrix represents the correspondence between the three-dimensional point cloud and the real physical space coordinates, realizing the spatial synchronization function.

[0014] Furthermore, the thermal imaging analysis is performed on the synchronously registered infrared signal, the overheated area is automatically selected and the temperature gradient is calculated, and the steps include:

[0015] For transformers, GIS equipment, and overhead lines, time difference is performed on 10 consecutive frames of thermal images of each device to obtain the temperature change matrix, binary mask, and time domain energy map of each device;

[0016] The temperature change matrix of each device is superimposed with the infrared signal of each device. The superimposed data is input into the YOLOv5 model to automatically select the overheating area of ​​each device and output the temperature gradient of each device.

[0017] According to the overheating area of ​​each device and the transformation matrix and the rotation and translation matrix, the backup infrared fault point of each device is matched.

[0018] Furthermore, the vibration signal after synchronous registration is subjected to vibration signal analysis to calculate vibration entropy, the steps comprising:

[0019] Perform wavelet packet decomposition on the vibration signal after synchronous registration to extract the power frequency component and high-frequency residual signal;

[0020] Performing EMD decomposition on the high-frequency residual signal to obtain intrinsic mode function components representing mechanical looseness;

[0021] Calculating the nonlinear dynamic entropy value of the screened intrinsic mode function components to obtain the vibration entropy and sideband energy ratio;

[0022] The instantaneous frequency of the filtered intrinsic mode function component is calculated by Hilbert transform, and the presence of modulation phenomenon and the sideband energy ratio are detected according to the instantaneous frequency, so as to obtain the backup vibration fault point.

[0023] Furthermore, the filtered intrinsic mode function components are obtained through the following steps:

[0024] The false components of the eigenmode function are eliminated by a correlation coefficient method, and the first N eigenmode function components with cumulative energy greater than 85% are retained, where N is a positive integer.

[0025] Furthermore, the step of performing partial discharge analysis on the electrical signal to obtain the number of discharge pulses includes:

[0026] Performing pulse detection on the electrical signal to generate the discharge pulse number and PRPD spectrum;

[0027] The PRPD map is input into the improved VGG16 model to obtain the discharge type probability and the pulse density prediction value of each phase interval. The fully connected layer in the improved VGG16 model is replaced by the attention mechanism module;

[0028] A backup electrical fault point is generated according to the discharge type probability and the pulse density prediction value of each phase interval.

[0029] Furthermore, the temperature gradient, the vibration entropy, and the number of discharge pulses are input into the LSTM model for time series correlation analysis to obtain the fault type and the probability of the fault type, the steps include:

[0030] The temperature gradient, the vibration entropy, and the number of discharge pulses are input into the LSTM model. If the temperature gradient exceeds a preset gradient threshold and the vibration entropy is within a normal range, the fault type and fault type probability corresponding to the temperature gradient are output, and a secondary warning is issued.

[0031] If the vibration entropy exceeds a preset baseline and the temperature gradient is within a normal range, the fault type and fault type probability corresponding to the vibration entropy are output, and a secondary warning is issued.

[0032] Furthermore, the step of inputting the temperature gradient, the vibration entropy, and the number of discharge pulses into the LSTM model for time series correlation analysis to obtain the fault type and the fault type probability also includes:

[0033] If the temperature gradient exceeds the preset gradient threshold and the vibration entropy exceeds the preset baseline, the fault type is mechanical looseness resulting in increased contact resistance, and a first-level warning is issued;

[0034] If the temperature gradient exceeds the preset gradient threshold and the number of discharge pulses exceeds the preset pulse threshold, the fault type is insulation degradation causing partial discharge, and a first-level warning is issued;

[0035] If the temperature gradient exceeds the preset gradient threshold, the vibration entropy exceeds the preset baseline, and the number of discharge pulses exceeds the preset pulse threshold, the fault type is an arc fault precursor, and a first-level warning is issued.

[0036] Furthermore, the infrared thermal imager is installed at the transformers, circuit breakers, disconnectors, and busbar joints of the 750KV substation and transmission lines, the acoustic vibration sensor is installed on the large equipment casing of the 750KV substation, the current sensor is installed at the current transformer, and the voltage sensor is installed at the busbar, incoming line, and outgoing line intervals of the 750KV substation.

[0037] This invention proposes a distribution automation equipment monitoring and acquisition system. By installing infrared thermal imagers, acoustic vibration sensors, current sensors, and voltage sensors at various locations throughout a 750kV substation and processing the collected data, it can provide early warning for the 750kV substation. First, a spatiotemporal alignment algorithm is used to synchronize and spatially align the acquired infrared, vibration, and electrical signals. Thermal imaging analysis is then performed on the synchronized infrared signals, automatically identifying overheated areas and calculating temperature gradients. The synchronized vibration signals are analyzed to obtain vibration entropy, and the electrical signals are analyzed for partial discharge to determine the number of discharge pulses. Finally, a weighted fusion of the temperature gradient, vibration entropy, and discharge pulse number is performed, and the fused features are input into an LSTM model for temporal correlation analysis to determine the fault type and probability, enabling early warning.

[0038] The advantages of the embodiments of the present invention are as follows:

[0039] (1) In the embodiment of the present invention, the detection sensitivity of weak temperature rise signals can be increased by 4 times through time difference, thereby improving the accuracy of early warning.

[0040] (2) In the embodiment of the present invention, the temperature change matrix can suppress reflection interference, solving the problem that the traditional method easily misjudges surface discharge heating as sunlight reflection, thereby reducing the missed detection rate.

[0041] (3) The embodiment of the present invention enhances the significance of key alarms through weight fusion and uses LSTM to mine the temporal causal relationship between multiple features, thereby achieving an upgrade from "single point alarm" to "state evolution prediction" and improving accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of the execution process of a distribution automation equipment supervision and collection system provided by an embodiment of the present invention.

[0043] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0044] The embodiments of the present application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and are not to be construed as limiting the present application.

[0045] In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0046] In the embodiments of the present application, at least one refers to one or more; a plurality refers to two or more. In the description of the present application, words such as "first", "second", and "third" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0047] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, the terms "including," "comprising," "having," and their variations in this specification all mean "including but not limited to," unless otherwise specifically stated.

[0048] It should be noted that in the embodiments of the present application, "connection" can be understood as electrical connection, and the connection between two electrical components can be a direct or indirect connection between the two electrical components. For example, the connection between A and B can be either a direct connection between A and B or an indirect connection between A and B through one or more other electrical components.

[0049] Figure 1 A flow chart of the execution process of a distribution automation equipment supervision and acquisition system provided by an embodiment of the present invention is as follows: Figure 1 As shown, an infrared thermal imager, an acoustic vibration sensor, a current sensor, and a voltage sensor are installed at a 750KV substation. The monitoring and acquisition system implements an early warning function for the 750KV substation through the following steps:

[0050] Specifically, the infrared thermal imager is installed at the transformers, circuit breakers, disconnectors, and busbar joints of the 750KV substation and transmission lines; the acoustic vibration sensor is installed on the large equipment casing of the 750KV substation; the current sensor is installed at the current transformer; and the voltage sensor is installed at the busbar, incoming line, and outgoing line intervals of the 750KV substation.

[0051] S10, acquiring an infrared signal through the infrared thermal imager, acquiring a vibration signal through the acoustic vibration sensor, and acquiring an electrical signal through the current sensor and the voltage sensor;

[0052] The infrared signal of the infrared thermal imager, the vibration signal of the acoustic vibration sensor, the current signal of the current sensor, and the voltage signal of the voltage sensor are collected according to the collection frequency, where the current signal and the voltage signal are collectively referred to as electrical signals.

[0053] In order to solve the problem of time synchronization and spatial registration of data from sensors with different sampling rates (thermal imaging 1Hz, vibration 10kHz, electrical quantity 4kHz), such as infrared, vibration, and electrical quantity sensors, it is necessary to first perform time-space synchronization and spatial registration on these data.

[0054] S20, using a spatiotemporal alignment algorithm to achieve time synchronization and spatial registration of the infrared signal, the vibration signal, and the electrical signal;

[0055] S21, using the sampling rate of the electrical signal as a reference frequency, performing cubic spline interpolation on the infrared signal to increase the sampling rate of the infrared signal to the reference frequency, and applying a dynamic time warping algorithm to the vibration signal and the electrical signal to compensate for mechanical vibration transmission delay and achieve time synchronization;

[0056] In the embodiment of the present invention, the sampling rate of the electrical signal is used as the reference frequency, and cubic spline interpolation is used for low-frequency signals such as infrared signals to increase them from 1 Hz to 4 kHz, maintaining the same frequency as the electrical quantity; then, the vibration signal and the electrical signal are dynamically bent and matched, and the Dynamic Time Warping (DTW) algorithm is applied to compensate for the mechanical vibration transmission delay.

[0057] S22, converting the temperature matrix output by the infrared thermal imager into a three-dimensional point cloud, and realizing the point cloud correspondence between the infrared signal and the real equipment in the 750KV substation based on the three-dimensional point cloud and the transformation matrix, wherein the transformation matrix is ​​obtained according to the installation position of the infrared thermal imager, and the transformation matrix represents the correspondence between the three-dimensional point cloud and the real physical space coordinates, thereby realizing the spatial synchronization function.

[0058] The temperature matrix (640*512 pixels) output by an infrared thermal imager is essentially two-dimensional image data, consisting solely of pixel coordinates (x, y) and temperature values ​​(T). Only after conversion to a three-dimensional point cloud (x, y, T) can a mapping relationship be established with the device's real-world physical coordinates. For example, if a high-temperature point is located at (120, 300) in the point cloud, the corresponding flange bolt position on the actual device can be determined using the transformation matrix.

[0059] Point cloud data preserves the geometric topology of the equipment, such as the temperature distribution of raised areas like bushings and flanges. This serves as the basis for spatial registration with vibration sensors. For example, iterative optimization requires calculating the Euclidean distance between high-temperature areas in the point cloud and the vibration sensor's monitoring points. Using only a two-dimensional image, this distance cannot be accurately determined. Vibration sensor monitoring points are typically annotated with three-dimensional coordinates (e.g., the side of a transformer tank 1.5 meters above the ground). Therefore, the infrared data must be converted to three-dimensional space to establish the association.

[0060] Through the joint calibration of the thermal imager and the total station, the transformation relationship between the thermal imager coordinate system and the substation global coordinate system is determined; during the operation of the equipment, the rotation matrix R and the translation matrix t are iteratively optimized through the ICP (Iterated Closest Points) algorithm to minimize the distance between the feature points of the point cloud and the equipment model.

[0061] The pixel coordinates in a two-dimensional thermal image cannot directly reflect the actual spatial distance. After point cloud conversion, the temperature gradient in real space can be calculated, which is a key indicator for assessing equipment overheating. The ICP algorithm requires that the input must be a three-dimensional spatial point set, and two-dimensional data cannot be used for rotation and translation calculations. The acceleration direction of the vibration sensor needs to be aligned with the spatial direction of the temperature distribution and must rely on three-dimensional coordinates.

[0062] S30, performing thermal imaging analysis on the synchronously registered infrared signal, automatically selecting the overheated area and calculating the temperature gradient, performing vibration signal analysis on the synchronously registered vibration signal, calculating the vibration entropy, and performing partial discharge analysis on the electrical signal to obtain the number of discharge pulses;

[0063] Specifically, the thermal imaging analysis is performed on the synchronously registered infrared signal, the overheated area is automatically selected, and the temperature gradient is calculated. The steps include:

[0064] S31A, for transformers, GIS equipment, and overhead lines, time-difference is performed on 10 consecutive frames of thermal images of each device to obtain the temperature change matrix, binary mask, and time-domain energy map of each device;

[0065] In the context of power equipment monitoring, a device is an electrical unit with independent functions that is continuously monitored, and the same spatial reference point must be maintained when imaging the device. A time-differentiation algorithm is performed on 10 consecutive frames of each device (typically spanning 30 seconds and sampling at 1 Hz) to generate a temperature change matrix, a binary mask, and a time-domain energy map. The temperature change matrix represents the rate of temperature rise or fall at each pixel and is often used to capture rapidly developing hotspots. The binary mask indicates areas of significant change and can be represented by a temperature change greater than a threshold, often used to reduce background interference. The time-domain energy map represents the cumulative temperature change and is often used to identify persistent degradation, such as contact oxidation.

[0066] For example, in a typical differential result example, under normal conditions, the temperature change matrix is ​​randomly distributed within ±0.3°C; the fault state is: a continuous positive growth area of ​​ΔT>1°C / s appears locally (bolts are loose), or an annular diffusion ΔT distribution (casing oil shortage).

[0067] In the embodiment of the present invention, the detection sensitivity of a weak temperature rise signal can be increased by 4 times through time difference.

[0068] S32A superimposes the temperature change matrix of each device with the infrared signal of each device, inputs the superimposed data into the YOLOv5 model, automatically selects the overheating area of ​​each device and outputs the temperature gradient of each device;

[0069] The infrared signal of each device is the original thermal image collected by the infrared thermal imager. The temperature change matrix of each device is superimposed on the infrared signal to form a 4-channel input, namely [R, G, B, ΔT], where RGB is the pseudo-color thermal image and ΔT is the normalized value of the difference matrix.

[0070] The superimposed data is fed into the YOLOv5 model, which includes two detection heads: one outputs a predicted bounding box and category, and the other directly outputs the temperature rise rate. Specifically, the YOLOv5 model outputs the overheated area, defect type, and temperature rise rate. The overheated area data format is [x_center, y_center, w, h], and the defect type is a classification probability vector, where x_center represents the horizontal coordinate of the bounding box center, y_center represents the vertical coordinate of the bounding box center, w represents the width of the bounding box, and h represents the height of the bounding box.

[0071] The YOLOv5 model is a deep learning-based object detection model that uses modules such as a backbone network, a detection head, and a loss function to quickly detect and locate multiple objects in an image. The YOLOv5 model consists of a backbone network, a feature fusion network, and a detection head.

[0072] In the embodiment of the present invention, the temperature change matrix can suppress reflection interference, thereby solving the problem that surface discharge heating is easily misjudged as sunlight reflection in traditional methods, thereby reducing the missed detection rate.

[0073] S33A: Matching the backup infrared fault point of each device according to the overheating area of ​​each device, the transformation matrix, and the rotation and translation matrix.

[0074] For example, in the time difference stage, the contact position ΔT was found to be +1.2°C (the average of adjacent pixels +0.3°C) in the 5th frame, and ΔT increased to +3.8°C in the 8th frame, forming a significant hot spot area.

[0075] YOLOv5 model analysis: output bounding box (0.52, 0.71, 0.05, 0.04);

[0076] Classified as “poor contact” (92% confidence level);

[0077] The temperature rise rate is 5.4°C / min (exceeding the threshold by 2°C / min).

[0078] It can be determined that the isolating switch is overheated and the backup infrared fault point is the isolating switch.

[0079] Therefore, temporal difference processing upgrades thermal image analysis from "static temperature distribution observation" to "dynamic thermal process capture." Combined with an improved YOLOv5 model, this enables closed-loop analysis from pixel-level changes to equipment defect diagnosis. While the 3D point cloud is not directly involved in this process, it provides a spatial reference for subsequent multi-sensor data fusion.

[0080] Specifically, the steps of performing vibration signal analysis on the synchronously registered vibration signal and calculating the vibration entropy include:

[0081] S31B, performing wavelet packet decomposition on the vibration signal after synchronous registration to extract the power frequency component and high-frequency residual signal;

[0082] The registered vibration signal is decomposed by wavelet packet to extract the power frequency component and high-frequency residual signal. The power frequency component can be 50HZ and 100HZ, and the signal other than the power frequency component is called the high-frequency residual signal.

[0083] S32B, performing EMD decomposition on the high-frequency residual signal to obtain an intrinsic mode function component representing mechanical looseness;

[0084] Empirical mode decomposition (EMD) is performed on the high-frequency residual signal to obtain intrinsic mode function (IMF) components that characterize mechanical looseness. Each IMF component represents a vibration mode at a specific time scale in the original signal, and their frequencies are arranged from high to low, with IMF1 component having the highest frequency. For example, when a transformer winding is loose, IMF3 component typically corresponds to the local resonant frequency of the mechanical structure.

[0085] S33B, calculating the nonlinear dynamic entropy value of the screened intrinsic mode function components to obtain the vibration entropy and sideband energy ratio;

[0086] The IMF components that are screened and stored are usually IMF2, IMF3, and IMF4. The nonlinear dynamic entropy value of the screened IMF components is calculated to quantify the signal complexity and obtain the vibration entropy. The normal value range of the vibration entropy is 0.6 to 0.8. If the vibration entropy is greater than 1.2, it indicates that a fault state has been entered, and the complexity increases due to the increase in impact components.

[0087] The IMF components and corresponding fault types are as follows: IMF1 component, the typical frequency range is 1kHz-5kHz, and the associated fault type is transient vibration caused by partial discharge; IMF3 component, the typical frequency range is 200Hz-800Hz, and the associated fault type is mechanical looseness of structural parts; IMF5 component, the typical frequency range is 50Hz-150Hz, and the associated fault type is magnetostrictive vibration of the iron core.

[0088] By screening out the IMF components, the sampling correlation coefficient method can be used to eliminate false components. For example, the IMF with a correlation coefficient of less than 0.3 of the original signal is discarded, and the top N IMF components with a cumulative energy greater than 85% are retained. N is usually 4-6, and the screened IMF components can be obtained.

[0089] S34B, calculating the instantaneous frequency of the screened intrinsic mode function component through Hilbert transform, and detecting whether there is a modulation phenomenon and the sideband energy ratio based on the instantaneous frequency, thereby obtaining the backup vibration fault point.

[0090] The instantaneous frequency of the IMF component is screened out through Hilbert transform to detect whether there is a modulation phenomenon, such as the 50Hz±fn sideband that appears when the bolt is loose.

[0091] For example, the original signal: vibration acceleration 0.8m / s 2 (Normal baseline 0.3m / s 2 ).

[0092] IMF3 analysis: Center frequency: 573Hz; Sample entropy: 1.15 (normal 0.72); Sideband energy ratio: 0.28 (normal <0.1)

[0093] Diagnostic conclusion: Insufficient bolt preload leads to mechanical resonance. This bolt is the backup vibration fault point.

[0094] Specifically, the step of performing partial discharge analysis on the electrical signal to obtain the number of discharge pulses includes:

[0095] S31C, performing pulse detection on the electrical signal to generate the discharge pulse number and PRPD spectrum;

[0096] The short-time energy over-threshold method is used to extract pulses from the electrical signal to generate the discharge pulse count and PRPD (Phase Resolved Partial Discharge) spectrum. The horizontal axis of the PRPD spectrum is the power frequency phase, the vertical axis is the pulse amplitude, and the chromaticity is the normalized count pulse density.

[0097] S32C, inputting the PRPD map into the improved VGG16 model to obtain the discharge type probability and the pulse density prediction value of each phase interval, and the fully connected layer in the improved VGG16 model is replaced with an attention mechanism module;

[0098] The PRPD map is input into the improved VGG16 model. The VGG16 model replaces the fully connected layer of the traditional VGG16 model with an attention module + regression head, which outputs the discharge type probability and pulse density distribution. The discharge type probability includes surface discharge, internal discharge and corona.

[0099] S33C: Generate a backup electrical fault point based on the discharge type probability and the pulse density prediction value of each phase interval.

[0100] Then, the backup electrical fault points that may have faults are screened out based on the discharge type probability and the pulse density prediction value of each phase interval.

[0101] S40: Input the temperature gradient, the vibration entropy, and the number of discharge pulses into an LSTM model for time series correlation analysis to obtain a fault type and a fault type probability, and issue an early warning to the 750KV substation based on the fault type probability.

[0102] Specifically, the weighted fusion of the temperature gradient, the vibration entropy and the number of discharge pulses comprises the following steps:

[0103] S41, inputting the temperature gradient, the vibration entropy, and the number of discharge pulses into an LSTM model. If the temperature gradient exceeds a preset gradient threshold and the vibration entropy is within a normal range, outputting the fault type and fault type probability corresponding to the temperature gradient, and performing a secondary warning.

[0104] S42, if the vibration entropy exceeds a preset baseline and the temperature gradient is within a normal range, outputting the fault type and fault type probability corresponding to the vibration entropy and issuing a secondary warning;

[0105] S43, if the temperature gradient exceeds the preset gradient threshold and the vibration entropy exceeds the preset baseline, the fault type is mechanical looseness leading to increased contact resistance, and a level one warning is issued;

[0106] S44, if the temperature gradient exceeds the preset gradient threshold and the number of discharge pulses exceeds the preset pulse threshold, the fault type is insulation degradation causing partial discharge, and a first-level warning is issued;

[0107] S45, if the temperature gradient exceeds the preset gradient threshold, the vibration entropy exceeds the preset baseline, and the number of discharge pulses exceeds the preset pulse threshold, the fault type is an arc fault precursor, and a first-level warning is issued.

[0108] If only the temperature gradient exceeds the threshold while the vibration entropy and the number of discharge pulses are within the normal range, for example, the temperature gradient is >2°C / min, specifically, a continuous positive growth area of ​​ΔT>1°C / s appears locally, the fault type is loose bolts, or if there is an annular diffusion ΔT distribution, the fault type is casing oil shortage, which can be an independent second-level alarm.

[0109] If only the vibration entropy exceeds 30% of the baseline and there is a characteristic frequency of 573HZ sideband, while other temperature gradients and discharge pulse numbers are within the normal range, it can be determined that the fault type is overheating of the disconnector head, and an independent second-level alarm can be issued.

[0110] When both alarms occur on the same device within a time window (±10 seconds), a weighting mechanism is activated. If the temperature gradient exceeds the preset gradient threshold and the vibration entropy exceeds the preset baseline, while the number of discharge pulses is within the normal range, the fault type is mechanical looseness leading to increased contact resistance, and a Level 1 warning is issued. If the temperature gradient exceeds the preset gradient threshold and the number of discharge pulses exceeds the preset pulse threshold, while the vibration entropy is within the normal range, the fault type is insulation degradation leading to partial discharge, and a Level 1 warning is issued. If the temperature gradient exceeds the preset gradient threshold, the vibration entropy exceeds the preset baseline, and the number of discharge pulses exceeds the preset pulse threshold, the fault type is an arc fault precursor.

[0111] The input layer of the LSTM model inputs the time step and feature dimensions, which include temperature gradient, vibration entropy, and number of discharge pulses. The hidden layer of the LSTM model includes two LSTM layers, each with 64 neurons, and Dropout = 0.2 to prevent overfitting. The output layer of the LSTM model outputs the failure probability and predicts the feature change trend in the next three steps (15 seconds).

[0112] When the fault probability output by LSTM is greater than 0.7, the following actions are triggered: automatically generating a work order containing the fault location (such as "transformer phase A bushing") and fault type, and adjusting the protection setting (such as shortening the overcurrent protection action time by 20%).

[0113] The embodiment of the present invention enhances the significance of key alarms through weight fusion and uses LSTM to mine the temporal causal relationship between multiple features, thereby realizing the upgrade from "single point alarm" to "state evolution prediction" and improving accuracy.

[0114] Each module in the aforementioned distribution automation equipment monitoring and data collection system can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0115] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0116] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0117] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A distribution automation equipment supervision and collection system, characterized in that: An infrared thermal imager, an acoustic vibration sensor, a current sensor, and a voltage sensor are installed at the 750KV substation. The monitoring and acquisition system implements an early warning function for the 750KV substation through the following steps: Acquire an infrared signal through the infrared thermal imager, acquire a vibration signal through the acoustic vibration sensor, and acquire an electrical signal through the current sensor and the voltage sensor; Using a time-space alignment algorithm to achieve time synchronization and spatial registration of the infrared signal, the vibration signal, and the electrical signal; Perform thermal imaging analysis on the synchronously registered infrared signal, automatically select the overheated area and calculate the temperature gradient, perform vibration signal analysis on the synchronously registered vibration signal, calculate the vibration entropy, and perform partial discharge analysis on the electrical signal to obtain the number of discharge pulses; The temperature gradient, the vibration entropy, and the number of discharge pulses are input into the LSTM model for time series correlation analysis to obtain the fault type and the fault type probability, and an early warning is issued for the 750KV substation based on the fault type probability.

2. The distribution automation equipment supervision and collection system according to claim 1, characterized in that: The steps of using a time-space alignment algorithm to achieve time synchronization and spatial registration of the infrared signal, the vibration signal, and the electrical signal include: Taking the sampling rate of the electrical signal as a reference frequency, cubic spline interpolation is performed on the infrared signal to increase the sampling rate of the infrared signal to the reference frequency, and a dynamic time warping algorithm is applied to the vibration signal and the electrical signal to compensate for mechanical vibration transmission delay and achieve time synchronization function; The temperature matrix output by the infrared thermal imager is converted into a three-dimensional point cloud, and based on the three-dimensional point cloud and the transformation matrix, the point cloud correspondence between the infrared signal and the real equipment in the 750KV substation is realized, wherein the transformation matrix is ​​obtained according to the installation position of the infrared thermal imager. The transformation matrix represents the correspondence between the three-dimensional point cloud and the real physical space coordinates, realizing the spatial synchronization function.

3. The distribution automation equipment supervision and collection system according to claim 2, characterized in that: The thermal imaging analysis of the synchronously registered infrared signal is performed to automatically select the overheated area and calculate the temperature gradient, including the following steps: For transformers, GIS equipment, and overhead lines, time difference is performed on 10 consecutive frames of thermal images of each device to obtain the temperature change matrix, binary mask, and time domain energy map of each device; The temperature change matrix of each device is superimposed with the infrared signal of each device. The superimposed data is input into the YOLOv5 model to automatically select the overheating area of ​​each device and output the temperature gradient of each device. According to the overheating area of ​​each device and the transformation matrix and the rotation and translation matrix, the backup infrared fault point of each device is matched.

4. The distribution automation equipment supervision and collection system according to claim 1, characterized in that: The steps of performing vibration signal analysis on the vibration signal after synchronous registration and calculating vibration entropy include: Perform wavelet packet decomposition on the vibration signal after synchronous registration to extract the power frequency component and high-frequency residual signal; Performing EMD decomposition on the high-frequency residual signal to obtain intrinsic mode function components representing mechanical looseness; Calculating the nonlinear dynamic entropy value of the screened intrinsic mode function components to obtain the vibration entropy and sideband energy ratio; The instantaneous frequency of the filtered intrinsic mode function component is calculated by Hilbert transform, and the presence of modulation phenomenon and the sideband energy ratio are detected according to the instantaneous frequency, so as to obtain the backup vibration fault point.

5. The distribution automation equipment supervision and collection system according to claim 4 is characterized in that: The selected intrinsic mode function components are obtained through the following steps: The false components of the eigenmode function are eliminated by a correlation coefficient method, and the first N eigenmode function components with cumulative energy greater than 85% are retained, where N is a positive integer.

6. The distribution automation equipment supervision and collection system according to claim 1, characterized in that: The steps of performing partial discharge analysis on the electrical signal to obtain the number of discharge pulses include: Performing pulse detection on the electrical signal to generate the discharge pulse number and PRPD spectrum; The PRPD map is input into the improved VGG16 model to obtain the discharge type probability and the pulse density prediction value of each phase interval. The fully connected layer in the improved VGG16 model is replaced by the attention mechanism module; A backup electrical fault point is generated according to the discharge type probability and the pulse density prediction value of each phase interval.

7. The distribution automation equipment supervision and collection system according to claim 1, characterized in that: The step of inputting the temperature gradient, the vibration entropy, and the number of discharge pulses into the LSTM model for time series correlation analysis to obtain the fault type and the fault type probability includes: The temperature gradient, the vibration entropy, and the number of discharge pulses are input into the LSTM model. If the temperature gradient exceeds a preset gradient threshold and the vibration entropy is within a normal range, the fault type and fault type probability corresponding to the temperature gradient are output, and a secondary warning is issued. If the vibration entropy exceeds a preset baseline and the temperature gradient is within a normal range, the fault type and fault type probability corresponding to the vibration entropy are output, and a secondary warning is issued.

8. The distribution automation equipment supervision and collection system according to claim 7, characterized in that: The step of inputting the temperature gradient, the vibration entropy, and the number of discharge pulses into the LSTM model for time series correlation analysis to obtain the fault type and the fault type probability further includes: If the temperature gradient exceeds the preset gradient threshold and the vibration entropy exceeds the preset baseline, the fault type is mechanical looseness resulting in increased contact resistance, and a first-level warning is issued; If the temperature gradient exceeds the preset gradient threshold and the number of discharge pulses exceeds the preset pulse threshold, the fault type is insulation degradation causing partial discharge, and a first-level warning is issued; If the temperature gradient exceeds the preset gradient threshold, the vibration entropy exceeds the preset baseline, and the number of discharge pulses exceeds the preset pulse threshold, the fault type is an arc fault precursor, and a first-level warning is issued.

9. The distribution automation equipment monitoring and collection system according to any one of claims 1 to 8, characterized in that: The infrared thermal imager is installed on the transformers, circuit breakers, disconnectors and busbar joints of the 750KV substation and transmission lines. The acoustic vibration sensor is installed on the large equipment casing of the 750KV substation. The current sensor is installed on the current transformer. The voltage sensor is installed on the busbar, incoming line and outgoing line intervals of the 750KV substation.

Images