Dry-type air-core reactor latent defect online diagnosis system and method based on multi-physics field cooperative monitoring

Abstract

The invention discloses a dry-type air-core reactor latent defect online diagnosis system and method based on multi-physics field cooperative monitoring. The system comprises a multi-dimensional sensing unit, an intelligent acquisition and triggering unit and a data analysis and intelligent diagnosis unit, the multi-dimensional sensing unit is used for collecting multiple physical quantity signals of the reactor in real time; the intelligent acquisition and trigger unit is used for synchronously acquiring all sensor signals after receiving the trigger signal; and the data analysis and intelligent diagnosis unit is used for early and accurate identification and early warning of latent defects. The system can accurately trigger and capture a switching transient process, through multi-physics-field cooperative monitoring penetration equipment encapsulation, perceive an internal state, and use a deep learning algorithm to construct an intelligent diagnosis model, early and accurate identification and early warning of latent defects such as insulation cracking, turn-to-turn short circuit and the like are realized. And finally, a complete solution integrating monitoring, diagnosis, early warning and life evaluation is formed.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power equipment state monitoring and fault diagnosis, and particularly relates to an online diagnosis system and method for latent defects of a dry-type air-core reactor based on multi-physical field cooperative monitoring. BACKGROUND

[0002] The dry-type air-core reactor is a key reactive power compensation device in a power system, and is widely used due to its advantages such as oil-free, fireproof and simple structure. However, its inherent laminated winding structure leads to relatively weak mechanical strength, and needs to bear huge electric force and thermal stress impact during switching operation. Statistics show that its faults mostly occur at the switching moment, and there are fewer problems in the stable operation stage.

[0003] At present, the detection of the dry-type air-core reactor mainly depends on periodic preventive tests (such as direct current resistance, inductance measurement, impedance measurement) and offline tests, which cannot capture the dynamic changes at the switching moment. The existing live detection technologies (such as infrared temperature measurement, ultraviolet imaging, vibration noise detection) also have obvious limitations:

[0004] 1. Timing matching problem: it is difficult to accurately start detection at the short moment (milliseconds to seconds) of switching operation, and key fault characteristic signals are easily missed;

[0005] 2. Space detection blind area: due to its dense and laminated structure, external detection means (such as infrared, ultraviolet) cannot effectively observe the internal interlayer and turn-to-turn insulation state changes, such as subtle insulation cracking, early turn-to-turn discharge and other latent defects;

[0006] 3. Safety risk: detection near the device at the switching moment poses a high safety risk to personnel and equipment.

[0007] Therefore, the existing technology has low detection rate for latent defects of the dry-type air-core reactor, cannot realize early warning of faults, and easily leads to device failure and even burning, causing unplanned power grid shutdown.

[0008] In order to overcome the above defects, a new online monitoring and intelligent diagnosis technology is needed, which can safely and accurately capture the slight changes in the internal state of the device at the switching moment, and effectively evaluate the interlayer insulation. SUMMARY

[0009] The application aims to provide a dry-type air-core reactor latent defect online diagnosis system and method based on multi-physical field cooperative monitoring.

[0010] To achieve the object of the application, the technical scheme provided by the application is as follows:

[0011] The first aspect

[0012] The application provides a dry-type air-core reactor latent defect online diagnosis system based on multi-physical field cooperative monitoring, comprising a multi-dimensional perception unit, an intelligent acquisition and triggering unit and a data analysis and intelligent diagnosis unit.

[0013] The multi-dimensional perception unit is used for acquiring multi-physical quantity signals of the reactor in real time.

[0014] The intelligent acquisition and triggering unit is used for synchronously acquiring all sensor signals after receiving a triggering signal.

[0015] The data analysis and intelligent diagnosis unit is used for early, accurate identification and early warning of latent defects.

[0016] The multi-dimensional perception unit comprises the following:

[0017] The detection coil array is arranged on the inner and outer surfaces of the reactor package and is used for monitoring the magnetic field distribution and distortion thereof in the radial and axial directions.

[0018] The high-frequency current sensor is sleeved on the incoming line end of the reactor and is used for acquiring high-frequency current signals flowing through the reactor to capture turn-to-turn and layer-to-layer discharge pulses.

[0019] The temperature sensor array adopts fiber Bragg gratings or distributed optical fiber temperature sensors and is attached to the surface of the reactor package and hot spot prone areas to monitor the temperature distribution and gradient change thereof.

[0020] The vibration sensor is installed on the reactor support and is used for monitoring mechanical vibration characteristics caused by electric power during switching.

[0021] The detection coil array adopts flexible PCB coils and is closely wound and attached to the surface of the outermost layer of the reactor package and the inner surface of the innermost layer of the reactor package, respectively, and one coil is arranged at the top, middle and bottom of each package to monitor the three-dimensional distribution of the magnetic field.

[0022] The high-frequency current sensor is a Rogowski coil with a bandwidth of 100 kHz-30 MHz, which is installed on the incoming cable below the reactor.

[0023] The vibration sensor is an industrial ICP acceleration sensor, which is installed on the steel structure support of the reactor base.

[0024] The intelligent acquisition and triggering unit comprises the following:

[0025] The precise triggering module communicates with the substation monitoring system or the auxiliary contact of the circuit breaker to obtain the circuit breaker opening / closing command or state signal, which is used as a precise triggering source to start high-speed sampling in advance.

[0026] The high-speed synchronous data acquisition card synchronously collects all sensor signals at a high sampling rate immediately after receiving the triggering signal, so as to ensure the complete capture of the transient characteristics of the switching process.

[0027] The data analysis and intelligent diagnosis unit comprises the following:

[0028] The signal preprocessing module pre-processes the collected original signals.

[0029] The deep learning feature extraction and diagnosis module is used to construct a deep neural network, with the pre-processed multi-source synchronous data as input, to automatically learn and filter sensitive features related to insulation defects, establish an intelligent diagnosis criterion model of insulation cracking and inter-turn discharge, and output the defect recognition result and confidence.

[0030] The high-frequency equivalent modeling and life prediction module is used to establish a high-frequency distributed parameter equivalent model of the dry-type air-core reactor, analyze the change law of the high-frequency characteristic parameters after insulation aging through simulation and multi-factor aging experiment, develop an insulation state evaluation method, and propose a residual life prediction method combined with the aging physical model.

[0031] The signal preprocessing module filters, denoises and normalizes the collected original signals.

[0032] The deep neural network is a one-dimensional CNN, LSTM or Transformer model.

[0033] The second aspect

[0034] The application provides a dry-type air-core reactor latent defect online diagnosis method based on multi-physical field collaborative monitoring.

[0035] Step S1: system deployment and initialization, installation of various sensors, configuration of acquisition parameters and trigger conditions;

[0036] Step S2: accurate triggering and data capture, monitoring of circuit breaker status, triggering of high-speed acquisition system at the moment before switching operation, complete recording of multi-physical field data during the whole switching process;

[0037] Step S3: multi-source data synchronization and fusion, aligning and fusing magnetic field, high-frequency current, temperature and vibration data under the same timestamp to form a multi-dimensional feature vector;

[0038] Step S4: intelligent diagnosis and defect identification, inputting the fused data into the trained deep learning model, the model automatically analyzes and judges whether there is a latent defect and its severity in the current state;

[0039] Step S5: state evaluation and early warning, generating a device state report according to the diagnosis result and the set threshold; if a defect is found, early warning is issued to prompt the operation and maintenance personnel to intervene;

[0040] Step S6: trend analysis and life prediction, long-term storage of data, tracking the change trend of key characteristic parameters, using high-frequency equivalent model and aging model to evaluate the degradation degree of insulation state and predict the remaining life.

[0041] Compared with the prior art, the beneficial effects of the present application are as follows:

[0042] (1) Breakthrough the detection timing bottleneck: through linkage with the controlled circuit breaker, accurate triggering and capture at the switching moment are realized, solving the problem that traditional methods cannot capture transient fault signals;

[0043] (2) Solve the detection space blind area: the magnetic field distortion is monitored by using a detection coil, which has strong penetration and can effectively reflect the state change of internal layers and turns, combined with a high-frequency current sensor, realizing the "perspective" diagnosis of internal defects;

[0044] (3) Improve the defect identification accuracy: multi-sensor information fusion technology is adopted, and magnetic, electric, thermal and vibration multi-physical field information is comprehensively utilized, which is mutually verified to avoid the uncertainty of single signal diagnosis. Further utilizing the powerful feature extraction and pattern recognition capability of deep learning, small and complex fault features can be automatically mined from massive data, significantly improving the identification accuracy and early warning capability of latent defects, especially early turn-to-turn short circuit;

[0045] (4) Realize the state depth evaluation: not only limited to fault alarm, but also through high frequency equivalent modeling and multi-factor aging experiment, the law of insulation aging is revealed from physical mechanism, the depth evaluation and life prediction from "whether fault" to "state how" and "how long can be used" are realized, and accurate decision basis is provided for condition-based maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 The system module schematic diagram provided for the embodiment of the application is shown. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the application will be clearly and completely described in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0048] The application belongs to the technical field of power equipment state monitoring and fault diagnosis, and particularly relates to an online monitoring system and method for a dry-type air-core reactor, and especially relates to early identification and diagnosis of latent defects in the switching process and operation of the air-core reactor through multi-sensor fusion and intelligent algorithms, which has significant practical value and broad application prospect.

[0049] As shown in Figure 1 The embodiment of the application provides a dry-type air-core reactor latent defect online diagnosis system, which comprises the following:

[0050] (1) Multi-dimensional perception unit: used for real-time acquisition of air-core reactor multi-physical quantity signals, and specifically comprising the following:

[0051] Probe coil array: arranged on the inner and outer surfaces of the air-core reactor, used for monitoring the magnetic field distribution and distortion thereof in the radial and axial directions. The magnetic field distortion is the most direct and sensitive feature of turn-to-turn short circuit.

[0052] High-frequency current sensor: sleeved on the incoming line end of the air-core reactor, used for acquiring high-frequency current signals (the frequency band is usually 10 kHz-10 MHz) flowing through the air-core reactor, so as to capture turn-to-turn and layer-to-layer discharge pulses.

[0053] Temperature sensor array: using fiber Bragg gratings or distributed optical fiber temperature sensors, attached to the surface of the air-core reactor package and hot spot prone areas, used for monitoring the temperature distribution and gradient change thereof.

[0054] Vibration sensor: installed on the air-core reactor support, used for monitoring the mechanical vibration characteristics caused by electric force during switching.

[0055] (2) Intelligent acquisition and triggering unit, specifically comprising the following:

[0056] Precise trigger module: communicate with the substation monitoring system or auxiliary contact of the circuit breaker to obtain the switching command or status signal, which is used as the precise trigger source to start the high-speed sampling in advance.

[0057] High-speed synchronous data acquisition card: after receiving the trigger signal, it synchronously acquires all sensor signals at a high sampling rate (usually ≥1 MS / s) to ensure the complete capture of the transient characteristics of the switching process (switching inrush current, transient process, steady-state process).

[0058] (3) Data analysis and intelligent diagnosis unit, specifically including the following:

[0059] Signal preprocessing module: filters, denoises, and normalizes the collected raw signals.

[0060] Deep learning feature extraction and diagnosis module: constructs a deep neural network (such as one-dimensional CNN, LSTM, or Transformer model) with preprocessed multi-source synchronous data as input, automatically learns and filters sensitive features related to insulation defects, establishes an intelligent diagnosis criterion model for insulation cracking and inter-turn discharge, and outputs defect recognition results and confidence.

[0061] High-frequency equivalent modeling and life prediction module: establishes a high-frequency distributed parameter equivalent model of the dry-type air-core reactor, analyzes the change law of high-frequency characteristic parameters (such as equivalent series resistance, distributed capacitance, and inductance) after insulation aging through simulation combined with multi-factor aging experiments (electrical-thermal-mechanical stress combined aging), develops an insulation state evaluation method, and proposes a residual life prediction method based on the aging physical model.

[0062] In addition, the application also provides an online diagnosis method based on the above-mentioned system, including the following steps:

[0063] Step S1: system deployment and initialization. Install various sensors, configure acquisition parameters and trigger conditions.

[0064] Step S2: precise triggering and data capture. Monitor the status of the circuit breaker and trigger the high-speed acquisition system at the moment before switching to record the multi-physical field data of the entire switching process.

[0065] Step S3: multi-source data synchronization and fusion. Align and fuse the magnetic field, high-frequency current, temperature, and vibration data under the same timestamp to form a multi-dimensional feature vector.

[0066] Step S4: intelligent diagnosis and defect recognition. Input the fused data into the trained deep learning model, and the model automatically analyzes and determines whether there is a latent defect (such as inter-turn short circuit or insulation cracking) and its severity.

[0067] Step S5: State evaluation and early warning. According to the diagnosis result and the set threshold, a device state report is generated. If a defect is found, an early warning is issued to prompt the operation and maintenance personnel to intervene.

[0068] Step S6: Trend analysis and life prediction. Long-term storage of data, tracking the change trend of key characteristic parameters, using high-frequency equivalent model and aging model to evaluate the degradation degree of insulation state and predict the remaining life.

[0069] Application example:

[0070] Take a 35kV voltage level, rated current 500A dry-type air core reactor as an example to realize the technical scheme of the application.

[0071] 1. System deployment:

[0072] Measurement coil array: flexible PCB coils are used, which are tightly wrapped around the surface of the outermost envelope and the inner surface of the innermost envelope of the reactor respectively (insulation and fixation need to be ensured during installation). One coil is arranged at the top, middle and bottom of each envelope to monitor the three-dimensional distribution of the magnetic field.

[0073] High-frequency current sensor: a Rogowski coil with a bandwidth of 100kHz-30MHz is selected and installed on the incoming cable below the reactor.

[0074] Temperature sensor: a distributed optical fiber temperature sensor (DTS) is used, and the temperature sensing optical fiber is tightly wound and attached to the surface of the outer envelope of the reactor.

[0075] Vibration sensor: an industrial ICP acceleration sensor is selected and installed on the steel structure support of the reactor base.

[0076] Intelligent acquisition and triggering unit: placed in the adjacent intelligent control cabinet. The triggering module directly accesses the opening / closing relay loop of the circuit breaker of the reactor through the cable to obtain the non-delayed hard contact signal.

[0077] Data analysis unit: can be integrated in the local server in the station or uploaded to the cloud platform.

[0078] 2. Diagnosis process:

[0079] When the substation background issues a "closing" instruction, the precise triggering module immediately captures the rising edge signal and sends a start command to the high-speed acquisition card.

[0080] The acquisition card synchronously records the data of all channels for 5 seconds at a sampling rate of 2.5MS / s (including 0.5s of background noise before closing and 4.5s of full process data after closing).

[0081] The data is transmitted to a data analysis unit through a network.

[0082] The signal preprocessing module integrates the magnetic field signal to restore the magnetic flux density and performs wavelet threshold denoising on the high-frequency current signal.

[0083] The preprocessed multi-channel time series data is spliced into a multi-dimensional array and input into a trained one-dimensional CNN model. The model is trained using historical data (including normal state and simulated inter-turn short circuit and insulation cracking fault data), and finally outputs a three-classification result: [normal, inter-turn short circuit suspicious, insulation cracking suspicious] and the corresponding probability.

[0084] If the diagnosis result is "normal", a report is generated and stored in the database. If it is "suspicious", an alarm message is immediately sent to the operation and maintenance personnel's mobile phone APP and monitoring background, and it is recommended to conduct an inspection or power-off inspection.

[0085] The system automatically performs trend analysis on the key features (such as magnetic field asymmetry, high-frequency pulse count, peak temperature) of all the switching data collected every month, and calls the life prediction module, combined with the running time of the reactor and the load history, to output the insulation health index and the estimated remaining life.

[0086] 3. Core algorithm and model:

[0087] Deep learning model: one-dimensional convolutional neural network (1D-CNN) combined with attention mechanism. The input layer receives the synchronized multi-channel signals. After several one-dimensional convolution and pooling operations, local features are extracted, and then different sensors and different time point features are given different weights through the attention layer. Finally, the classification result is output through the fully connected layer and the Softmax layer.

[0088] High-frequency equivalent model: each package of the reactor is equivalent to a π-type circuit, including series inductance L, series resistance R (representing loss, sensitive to aging) and capacitance C to ground. The initial parameters are obtained through sweep frequency measurement or simulation. In the multi-factor aging experiment, the changes of these parameters with aging time are monitored, and a mapping relationship library of parameter-aging degree-remaining life is established for on-site device evaluation.

[0089] Through the above embodiments, the present application successfully realizes online, accurate and intelligent diagnosis and prediction of latent defects of dry-type air-core reactors.

[0090] Finally, it should be noted that: the above embodiments are only used to illustrate and describe the present application, and are not intended to limit the present application to the scope of the described embodiments. In addition, those skilled in the art can understand that the present application is not limited to the above embodiments, and more variations and modifications can be made according to the teaching of the present application, which all fall within the scope of the present application.

Claims

1. A dry-type air-core reactor latent defect online diagnosis system based on multi-physical field cooperative monitoring, characterized in that, The application relates to a multi-dimensional sensing unit, an intelligent acquisition and triggering unit and a data analysis and intelligent diagnosis unit. The multi-dimensional sensing unit is used for collecting multi-physical quantity signals of a reactor in real time. The intelligent acquisition and triggering unit is used for synchronously collecting all sensor signals after receiving a triggering signal. The data analysis and intelligent diagnosis unit is used for early and accurate identification and early warning of latent defects. The multi-dimensional sensing unit comprises the following:

2. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 1, characterized in that, A detection coil array is arranged on the inner and outer surfaces of a reactor package and is used for monitoring the radial and axial magnetic field distribution and the distortion condition thereof; A high-frequency current sensor is sleeved on a wire inlet end of the reactor and is used for collecting high-frequency current signals flowing through the reactor so as to capture turn-to-turn and layer-to-layer discharge pulses; A temperature sensor array adopts fiber grating or a distributed fiber temperature sensor, is attached to the surface of the reactor package and a hot spot prone area, and is used for monitoring the temperature distribution and gradient change; A vibration sensor is installed on a reactor support and is used for monitoring mechanical vibration characteristics caused by electric power during switching. The detection coil array adopts flexible PCB coils, is closely wound on the surface of the outermost layer of the reactor package and the inner surface of the innermost layer of the reactor package respectively, one coil is arranged on the top, middle and bottom of each package respectively, and the three-dimensional distribution of the magnetic field is monitored.

3. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 2, characterized in that, The high-frequency current sensor selects a Rogowski coil with a bandwidth of 100 kHz to 30 MHz and is installed on the wire inlet cable below the reactor.

4. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 2, characterized in that, The vibration sensor selects an industrial ICP acceleration sensor and is installed on the steel structure support of the reactor base.

5. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 2, characterized in that, The intelligent acquisition and triggering unit comprises the following:

6. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 1, characterized in that, An accurate triggering module communicates with a substation monitoring system or an auxiliary contact of a circuit breaker, acquires circuit breaker opening / closing instructions or state signals, takes the accurate triggering source, and starts high-speed sampling in advance; A high-speed synchronous data acquisition card synchronously collects all sensor signals at a high sampling rate immediately after receiving a triggering signal, and ensures that the transient characteristics during the switching process are completely captured. The data analysis and intelligent diagnosis unit comprises the following:

7. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 1, characterized in that, A signal preprocessing module pre-processes the collected original signals; A deep learning feature extraction and diagnosis module is used for constructing a deep neural network, taking the pre-processed multi-source synchronous data as input, automatically learning and screening sensitive features related to insulation defects, establishing an intelligent diagnosis criterion model of insulation cracking and turn-to-turn discharge, and outputting a defect identification result and a confidence degree; A high-frequency equivalent modeling and life prediction module is used for establishing a high-frequency distributed parameter equivalent model of the dry-type air-core reactor, analyzing the change rule of the high-frequency characteristic parameters after insulation aging through simulation and multi-factor aging experiments, developing an insulation state evaluation method, and combining an aging physical model to propose a residual life prediction method. The signal preprocessing module filters, denoises and normalizes the collected original signals.

8. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 7, characterized in that, The deep neural network is a one-dimensional CNN, LSTM or Transformer model.

9. The online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to claim 7, characterized in that, ​ 10. A dry-type air-core reactor latent defect online diagnosis method based on multi-physical field cooperative monitoring, characterized in that, The method is realized based on the online diagnosis system for latent defects of dry-type air-core reactor based on multi-physical field cooperative monitoring according to any one of claims 1-9, and comprises the following steps: Step S1: system deployment and initialization, installing various sensors, configuring acquisition parameters and trigger conditions; Step S2: accurate triggering and data capturing, monitoring the state of the circuit breaker, triggering the high-speed acquisition system at the moment before the switching operation, and recording the multi-physical field data of the whole switching process; Step S3: multi-source data synchronization and fusion, aligning and fusing the magnetic field, high-frequency current, temperature and vibration data under the same timestamp to form a multi-dimensional feature vector; Step S4: intelligent diagnosis and defect identification, inputting the fused data into the trained deep learning model, and the model automatically analyzes and judges whether there is a latent defect and its severity in the current state; Step S5: state evaluation and early warning, generating a device state report according to the diagnosis result and the set threshold; if a defect is found, an early warning is issued to prompt the operation and maintenance personnel to intervene; Step S6: trend analysis and life prediction, long-term storage of data, tracking the change trend of key characteristic parameters, evaluating the degradation degree of insulation state and predicting the remaining life by using high-frequency equivalent model and aging model.

Images