Discharge monitoring system and method based on dynamic frequency band selection and network identification

Abstract

The present invention discloses a discharge monitoring system and method based on dynamic frequency band selection and network identification. The system includes a signal acquisition and processing module, a dynamic frequency band optimization module, and an intelligent identification and analysis module. Based on the dynamic spectrum perception capability of a programmable multi-bandpass filter group, combined with real-time environmental noise spectrum analysis, the system actively avoids interference from communication frequency bands and harmonics of industrial equipment, and improves the background purity of partial discharge signals. In the strong electromagnetic coupling scenario of a substation, the system solves the inherent sensitivity defects of fixed frequency band monitoring through a millisecond-level frequency band seamless switching mechanism, achieves a leap in the ability to capture weak discharge signals, suppresses missed detection, provides a reliable data basis for the insulation status assessment of power equipment, and forms a rapid immune response and dynamic optimization capability for new electromagnetic interference. The system autonomously improves the noise knowledge base and identification boundaries during long-term operation to ensure that the monitoring performance remains stable and reliable throughout its life cycle.

Description

Technical Field

[0001] The present invention relates to the technical field of online monitoring of electric power equipment, and in particular to a discharge monitoring system and method based on dynamic frequency band selection and network identification. Background Art

[0002] The power equipment monitoring device is a device used to monitor, detect and manage the status of power system equipment. It plays an important role in the operation and maintenance of the power system and helps to improve the reliability, safety and economy of the power system. The power equipment online monitoring device can monitor the operating status of the power equipment in real time, promptly detect possible signs of equipment failure, and issue an alarm in advance to avoid downtime and losses caused by equipment failure.

[0003] However, existing technologies have some problems: 1. Limitations of fixed frequency band monitoring; 2. Poor adaptability to noise interference; 3. Existing technologies have not yet solved the problem of coordinated optimization of dynamic frequency band selection and real-time signal validity verification.

[0004] Therefore, a discharge monitoring system and method based on dynamic frequency band selection and network identification are proposed to solve the above problems. Summary of the Invention

[0005] The main purpose of the present invention is to provide a discharge monitoring system and method based on dynamic frequency band selection and network identification to solve the problems raised in the above background.

[0006] To achieve the above objectives, the present invention adopts a technical solution: a discharge monitoring method based on dynamic frequency band selection and network identification, the method comprising the following steps:

[0007] Step S1: collecting original electromagnetic signals of power equipment through a UHF sensor array to generate an initial UHF signal data set;

[0008] Step S2: inputting the initial UHF signal data set into a multi-band filtering module for parallel frequency band processing to generate a low-frequency bandpass signal, a high-frequency bandpass signal, a full-frequency direct communication signal, and a selective dual-band combination signal;

[0009] Step S3, performing time domain feature extraction, frequency domain feature extraction, and statistical feature extraction on the signals of each frequency band outputted in step S2 to generate a multi-dimensional feature vector set;

[0010] Step S4: inputting the multidimensional feature vector set into a pre-trained neural network classification model to verify signal validity, outputting a partial discharge signal probability value and a noise probability value, and generating signal classification identification data;

[0011] Step S5: Calculate the real-time signal-to-noise ratio (SNR) of each frequency band based on the signal classification identification data, and generate a frequency band performance evaluation matrix in combination with a preset pulse density threshold;

[0012] Step S6: Execute dynamic frequency band switching decision according to the frequency band performance evaluation matrix;

[0013] Step S7: adjusting the frequency band polling period according to the switching mode parameters configured by the user, generating an optimized partial discharge monitoring result and outputting an alarm instruction;

[0014] Preferably, the step S2 includes:

[0015] Step S21: Configure a four-way parallel filtering chain through a programmable filter module:

[0016] Link 1: Use a 550MHz bandpass filter to generate a low-frequency bandpass signal;

[0017] Link 2: Use a 1300MHz bandpass filter to generate a high-frequency bandpass signal;

[0018] Link 3: Uses a 900MHz all-pass filter to generate a full-band direct-through signal;

[0019] Link 4: Dynamically selects two non-overlapping frequency bands for combined filtering to generate a dual-band combined signal;

[0020] Step S22: Control the signal output timing of each filter chain through a multi-way selection switch.

[0021] Preferably, the operation of frequency domain feature extraction in step S3 is specifically as follows:

[0022] Step S31: performing discrete wavelet transform on each frequency band signal to decompose it into 8 sub-bands;

[0023] Step S32: Calculate the energy proportion of each sub-band:

[0024] ;

[0025] in, is the energy proportion of the kth sub-band, is the wavelet coefficient of the kth sub-band, is the sampling point index, N is the number of sampling points, is the sub-band index;

[0026] Step S33: Use the energy proportion distribution vector as a frequency domain feature.

[0027] Preferably, the neural network classification model construction in step S4 includes:

[0028] Step S41: The model structure adopts a CNN-LSTM hybrid architecture:

[0029] Input layer: receives a feature vector of dimension 12;

[0030] Convolutional layer: 3 layers of 1D-CNN with a convolution kernel size of 3×1;

[0031] LSTM layer: bidirectional LSTM unit number 64;

[0032] Output layer: Softmax activation function outputs partial discharge / noise probability;

[0033] Step S42: The training data includes:

[0034] Positive samples: pulse waveforms of corona discharge and suspension discharge;

[0035] Negative samples: mobile phone signals, arc noise, and switch operation interference.

[0036] Preferably, the logic of the dynamic frequency band switching decision in step S6 is:

[0037] Step S61: Initialize the polling cycle T and activate each frequency band in sequence to collect signals;

[0038] Step S62: Calculate the comprehensive performance score of each frequency band:

[0039] ;

[0040] in, is the comprehensive performance score of the i-th frequency band, is the current signal-to-noise ratio, The switching time is is the weight coefficient;

[0041] Step S63: Select The highest frequency band serves as the main monitoring channel.

[0042] Preferably, the switching mode parameter configuration in step S7 includes:

[0043] Emergency mode: polling period T = 1s, SNR drop threshold = 15%;

[0044] Steady-state mode: polling period T = 15 minutes, SNR drop threshold = 25%.

[0045] Preferably, the method further comprises an online learning mechanism:

[0046] Step S71: When unidentified signal types appear continuously, trigger the data enhancement module to generate synthetic samples;

[0047] Step S72: Update the neural network classification model using an incremental learning algorithm:

[0048] ;

[0049] in, are the updated model parameters, are the model parameters before updating, is the learning rate, is the gradient of the loss function L with respect to the parameter θ, Adding samples to the model The predicted output of is the true label of the newly added sample.

[0050] Preferably, the calculation formula of the signal-to-noise ratio SNR in step S5 is:

[0051] ;

[0052] in is the average power of the partial discharge pulse, is the noise power in the pulse-free section.

[0053] Preferably, the system comprises:

[0054] Signal acquisition and processing module, dynamic frequency band optimization module, and intelligent identification and analysis module;

[0055] The signal acquisition and processing module includes:

[0056] Ultra-high frequency sensor array: arranged at the basin insulator of GIS equipment to collect 900MHz electromagnetic signals;

[0057] Signal conditioning unit: includes preamplifier and 24-bit ADC converter;

[0058] The dynamic frequency band optimization module includes:

[0059] Programmable multi-bandpass filter bank: hardware implementation of four parallel filtering chains;

[0060] Band switching controller: FPGA-based real-time SNR evaluation and switching logic circuit;

[0061] The intelligent recognition and analysis module includes:

[0062] Feature extraction engine: performs time domain, frequency domain, and statistical feature calculations;

[0063] Neural network processor: deploying CNN-LSTM classification model;

[0064] Online learning unit: supports incremental model updates.

[0065] Preferably, the programmable multi-bandpass filter group adopts a microstrip line structure design, comprising:

[0066] Low-frequency channel: center frequency 600MHz, bandwidth ±200MHz;

[0067] High frequency channel: center frequency 1250MHz, bandwidth ±250MHz;

[0068] Dual-frequency combination channel: switching frequency band combination mode through PIN diode.

[0069] The present invention has the following beneficial effects:

[0070] 1. In the present invention, by applying a dynamic frequency band selection mechanism, when conducting real-time monitoring of electrical equipment discharge phenomena, the signal-to-noise ratio and environmental interference level of different monitoring frequency bands are intelligently evaluated, and the optimal monitoring frequency band selection strategy is set for specific discharge types. This ensures the clarity and pertinence of the different discharge signal acquisition processes. At the same time, the discharge signal is captured within the selected frequency band and analyzed in real time. It is possible to determine in real time whether the selected frequency band can still effectively capture the target discharge characteristics, avoid signal distortion or omission problems caused by interference in the fixed frequency band, ensure the accuracy and integrity of discharge signal acquisition, and reduce the risk of monitoring blind spots and false alarms.

[0071] 2. In the present invention, by integrating the network recognition algorithm, when performing discharge signal feature extraction and pattern discrimination, the pattern category to which the discharge signal belongs is identified in real time by calculating the distinction between the discharge pulse waveform and the background noise and the degree of aggregation in the feature space, so that the system can reduce the situation where noise or non-target signals are misjudged as target discharges. Moreover, when it is found that the signal characteristics deviate significantly from the preset pattern during the monitoring process, the preset discharge feature map library can be used to assist or correct the recognition results in real time, so that when the discharge pattern discrimination is abnormal, it can be quickly prompted or corrected, thereby ensuring the accuracy and reliability of discharge type identification.

[0072] 3. In the present invention, by constructing an intelligent alarm grading terminal, when conducting a comprehensive assessment and early warning of the status of electrical equipment, after identifying a valid discharge signal, the intensity, frequency, development trend and pattern evolution of the discharge signal are automatically analyzed and evaluated in multiple dimensions and stages, and the hazard level deviation value of the current discharge activity is quantified in real time. According to the hazard level deviation values ​​of different dimensions, the health risks of the equipment insulation condition are comprehensively evaluated and graded early warnings are given in real time, so that the system can implement differentiated and stepped alarm management for different discharge types and severities, reduce the situation where the monitoring system generates a large number of invalid alarms or misses serious status reports due to a single threshold trigger, and further improve the effectiveness of discharge monitoring, the accuracy of operation and maintenance guidance and the overall monitoring effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 This is a flow chart of the discharge monitoring system and method based on dynamic frequency band selection and network identification of the present invention;

[0074] Figure 2 This is a flow chart of the discharge monitoring system and method based on dynamic frequency band selection and network identification of the present invention. DETAILED DESCRIPTION

[0075] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0076] See also Figure 1-Figure 2 The discharge monitoring method based on dynamic frequency band selection and network identification includes the following steps:

[0077] Step S1: collecting original electromagnetic signals of power equipment through a UHF sensor array to generate an initial UHF signal data set;

[0078] Step S2: inputting the initial UHF signal data set into a multi-band filtering module for parallel frequency band processing to generate a low-frequency bandpass signal, a high-frequency bandpass signal, a full-frequency direct communication signal, and a selective dual-band combination signal;

[0079] Step S3, performing time domain feature extraction, frequency domain feature extraction, and statistical feature extraction on the signals of each frequency band outputted in step S2 to generate a multi-dimensional feature vector set;

[0080] Step S4: inputting the multidimensional feature vector set into a pre-trained neural network classification model to verify signal validity, outputting a partial discharge signal probability value and a noise probability value, and generating signal classification identification data;

[0081] Step S5: Calculate the real-time signal-to-noise ratio (SNR) of each frequency band based on the signal classification identification data, and generate a frequency band performance evaluation matrix in combination with a preset pulse density threshold;

[0082] Step S6: Execute dynamic frequency band switching decision according to the frequency band performance evaluation matrix;

[0083] Step S7: adjusting the frequency band polling period according to the switching mode parameters configured by the user, generating an optimized partial discharge monitoring result and outputting an alarm instruction;

[0084] Step S2 includes:

[0085] Step S21: Configure a four-way parallel filtering chain through a programmable filter module:

[0086] Link 1: Use a 550MHz bandpass filter to generate a low-frequency bandpass signal;

[0087] Link 2: Use a 1300MHz bandpass filter to generate a high-frequency bandpass signal;

[0088] Link 3: Uses a 900MHz all-pass filter to generate a full-band direct-through signal;

[0089] Link 4: Dynamically selects two non-overlapping frequency bands for combined filtering to generate a dual-band combined signal;

[0090] Step S22: controlling the signal output timing of each filter link through a multi-way selection switch;

[0091] The specific operation of frequency domain feature extraction in step S3 is:

[0092] Step S31: performing discrete wavelet transform on each frequency band signal to decompose it into 8 sub-bands;

[0093] Step S32: Calculate the energy proportion of each sub-band:

[0094] ;

[0095] in, is the energy proportion of the kth sub-band, is the wavelet coefficient of the kth sub-band, is the sampling point index, N is the number of sampling points, is the sub-band index;

[0096] Step S33: using the energy proportion distribution vector as a frequency domain feature;

[0097] The neural network classification model construction in step S4 includes:

[0098] Step S41: The model structure adopts a CNN-LSTM hybrid architecture:

[0099] Input layer: receives a feature vector of dimension 12;

[0100] Convolutional layer: 3 layers of 1D-CNN with a convolution kernel size of 3×1;

[0101] LSTM layer: bidirectional LSTM unit number 64;

[0102] Output layer: Softmax activation function outputs partial discharge / noise probability;

[0103] Step S42: The training data includes:

[0104] Positive samples: pulse waveforms of corona discharge and suspension discharge;

[0105] Negative samples: mobile phone signals, arc noise, and switch operation interference;

[0106] The logic of dynamic frequency band switching decision in step S6 is:

[0107] Step S61: Initialize the polling cycle T and activate each frequency band in sequence to collect signals;

[0108] Step S62: Calculate the comprehensive performance score of each frequency band:

[0109] ;

[0110] in, is the comprehensive performance score of the i-th frequency band, is the current signal-to-noise ratio, The switching time is is the weight coefficient;

[0111] Step S63: Select The highest frequency band serves as the main monitoring channel;

[0112] The switching mode parameter configuration in step S7 includes:

[0113] The system includes: Emergency mode: polling period T = 1s, SNR drop threshold = 15%;

[0114] Steady-state mode: polling period T = 15 minutes, SNR drop threshold = 25%;

[0115] Step S71: When unidentified signal types appear continuously, trigger the data enhancement module to generate synthetic samples;

[0116] Step S72: Update the neural network classification model using an incremental learning algorithm:

[0117] ;

[0118] in, are the updated model parameters, are the model parameters before updating, is the learning rate, is the gradient of the loss function L with respect to the parameter θ, Adding samples to the model The predicted output of is the true label of the newly added sample;

[0119] The calculation formula of the signal-to-noise ratio SNR in step S5 is:

[0120] ;

[0121] in is the average power of the partial discharge pulse, is the noise power in the pulse-free section;

[0122] The system includes:

[0123] Signal acquisition and processing module, dynamic frequency band optimization module, and intelligent identification and analysis module;

[0124] The signal acquisition and processing module includes:

[0125] Ultra-high frequency sensor array: arranged at the basin insulator of GIS equipment to collect 900MHz electromagnetic signals;

[0126] Signal conditioning unit: includes preamplifier and 24-bit ADC converter;

[0127] The dynamic frequency band optimization module includes:

[0128] Programmable multi-bandpass filter bank: hardware implementation of four parallel filtering chains;

[0129] Band switching controller: FPGA-based real-time SNR evaluation and switching logic circuit;

[0130] Intelligent recognition and analysis modules include:

[0131] Feature extraction engine: performs time domain, frequency domain, and statistical feature calculations;

[0132] Neural network processor: deploying CNN-LSTM classification model;

[0133] Online learning unit: supports incremental model updates;

[0134] The programmable multi-bandpass filter bank adopts a microstrip line structure design and includes:

[0135] Low-frequency channel: center frequency 600MHz, bandwidth ±200MHz;

[0136] High frequency channel: center frequency 1250MHz, bandwidth ±250MHz;

[0137] Dual-frequency combination channel: switching frequency band combination mode through PIN diode.

[0138] Example 1: GIS equipment insulation fault monitoring

[0139] 1. Signal acquisition and preprocessing

[0140] The 900MHz original electromagnetic signal is collected by a UHF sensor array arranged in a ring on the basin insulator. After conditioning by a 40dB gain low-noise amplifier, the 500MS / s sampling rate ADC module completes the digital conversion.

[0141] 2. Dynamic frequency band optimization

[0142] Initialize four filter channels: low-frequency bandpass, high-frequency bandpass, full-frequency passthrough, and dual-frequency combination;

[0143] Real-time analysis of the ambient noise spectrum, automatically shutting down the high-frequency channel when 1.8GHz communication interference is detected;

[0144] Calculate the signal-to-noise ratio of the remaining channels and lock the low-frequency channel as the main monitoring path.

[0145] 3. Feature fusion and recognition

[0146] Extract the time domain features and frequency domain features of low-frequency channel signals;

[0147] Input 12-dimensional feature vector to the pre-trained CNN-LSTM model;

[0148] The model outputs a suspended discharge probability greater than the threshold, generating an insulation defect alarm.

[0149] 4. Online learning trigger

[0150] When unrecognized pulses appear continuously at the same position:

[0151] Take new samples and add time-shift dithering;

[0152] Start incremental training to update the weights of the fully connected layer;

[0153] After the accuracy of the validation set reaches the standard, the classification model is hot-replaced.

[0154] Example 2: Transformer monitoring in plateau substations

[0155] 1. Anti-interference initialization

[0156] Low-temperature-resistant sensors are deployed at sites at an altitude of 3,000 meters to automatically detect the corona noise background upon startup and establish a baseline noise library.

[0157] 2. Response to severe working conditions

[0158] Blizzard weather triggers emergency mode, and the frequency band polling cycle is shortened to milliseconds;

[0159] The dual-band combined channel tracks the discharge pulse in real time;

[0160] Filter ice drop interference through phase distribution statistics.

[0161] 3. Multidimensional diagnostic decision-making

[0162] Fusion of oil chromatography data and discharge pulse density;

[0163] Identify the development trajectory of inter-turn discharge in windings;

[0164] Output insulation aging level assessment report.

[0165] Example 3: Coastal Wind Power Converter Monitoring

[0166] 1. Salt spray environment adaptation

[0167] During the initialization phase, the characteristics of sea breeze salt spray corona are collected;

[0168] Dynamically shield 1.3GHz high-interference frequency band;

[0169] Enable full frequency pass-through mode to capture discharges with a wide frequency spectrum.

[0170] 2. Mechanical vibration decoupling

[0171] Install vibration accelerometer for synchronous data acquisition;

[0172] Freeze pulse count when vibration frequency is greater than 100Hz;

[0173] 3. Lifespan prediction linkage

[0174] Correlate converter operating time and discharge intensity trends;

[0175] Establish the remaining life model of IGBT module;

[0176] Provide 8 weeks advance warning of the power device replacement window.

[0177] Example 4: Partial discharge monitoring of large generators in hydropower stations

[0178] 1. Anti-interference deployment in strong electromagnetic environment

[0179] High-temperature resistant fiber optic composite sensors are embedded in the stator slots of the generator and sealed with fluororubber to ensure stable insulation performance in an environment with 95% humidity. The sensor array has 24 measuring points equidistantly distributed along the circumference. Fiber Bragg gratings are used to achieve synchronous acquisition of temperature and discharge dual parameters. A baseline calibration is performed at system startup to collect the harmonic interference characteristics of the excitation system under no-load conditions, and a background noise model that includes the rotor rotation modulation effect is established.

[0180] 2. Wideband dynamic perception and optimization

[0181] Scan the 1.5GHz spectrum in real time. When spark interference from the excitation carbon brush is detected:

[0182] Activate high-frequency monitoring channel

[0183] Inject the inverted noise cancellation signal:

[0184] ;

[0185] in, is the generated denoised signal, A is the interference amplitude, is the central time point of the interference signal, is the Gaussian envelope width parameter, is the interference center frequency;

[0186] Use wavelet packet reconstruction technology to enhance the effective frequency band:

[0187] Decompose the original signal into 8 layers;

[0188] Select energy-focused subbands to reconstruct signals;

[0189] The signal-to-noise ratio is improved by more than 22dB.

[0190] 3. Decoupling of multi-source discharge characteristics

[0191] Establish a discharge pulse fingerprint library:

[0192] Stator bar corona discharge: pulse rise time greater than 15ns, phase dispersion;

[0193] Wedge loosening discharge: pulse cluster period 30ms, phase concentrated;

[0194] Cooling water pipe creepage: accompanied by a temperature gradient greater than 3°C / cm;

[0195] Developing spatiotemporal joint analysis algorithms:

[0196] Spatial domain: Calculate the coordinates of the released power source based on the measurement point position matrix;

[0197] ;

[0198] in, is the spatial coordinate of the discharge source to be solved, is the coordinate of the i-th sensor, is the propagation speed of electromagnetic waves, is the time delay of the discharge pulse reaching the i-th sensor;

[0199] Time domain: extract the phase relationship between the discharge pulse and the power frequency voltage;

[0200] When slot discharge signature is detected:

[0201] Correlate vibration monitoring data;

[0202] Output the wedge tightening operation instruction.

[0203] 4. Special response to hydraulic environment

[0204] High humidity environment:

[0205] Start the microwave drying module to maintain the dew point of the sensor cavity below -15°C;

[0206] Use dielectric constant compensation algorithm to correct pulse amplitude:

[0207] ;

[0208] Water quality deterioration during flood season:

[0209] Monitor changes in cooling water conductivity;

[0210] When creeping discharge is detected:

[0211] If the cooling water conductivity is greater than 10μS / cm, it is determined that the insulator surface is contaminated with discharge;

[0212] If the conductivity is normal but the discharge increases sharply, it is a warning of water leakage in the stator bar;

[0213] Low temperature freezing conditions:

[0214] The heating module maintains the sensor temperature above 5°C;

[0215] Pulse width identification technology is used to eliminate interference from ice crystal breakage.

[0216] Example 5: UHV Converter Station Valve Tower Monitoring

[0217] 1. Multi-layer electromagnetic shielding deployment

[0218] In the ±800kV valve hall, a carbon fiber composite shielding cover is used to encapsulate the ultra-high frequency sensor, and the signal line is led out through a waveguide penetrator. The sensor array is distributed according to the valve tower level, with three groups of orthogonal polarization antennas arranged on each level to cover the omnidirectional radiation field of the thyristor assembly. Baseline calibration is performed at startup to collect background noise without working conditions, and establish an interference feature library including cooling fan harmonics and thyristor switching spikes.

[0219] 2. Dynamic Spectrum Sensing and Avoidance

[0220] Scan the 1.8GHz spectrum in real time and detect the communication frequency of the DC protection system;

[0221] Activate dual-band combination mode;

[0222] Adaptive notch filter is used to deeply suppress the 2.4GHz±50MHz frequency band;

[0223] The calculated signal-to-noise ratio gain within the effective monitoring bandwidth is greater than 18dB.

[0224] 3. Multi-physics field fusion diagnosis

[0225] Synchronously collect valve tower cooling water flow and busbar temperature data;

[0226] When an abnormal discharge pulse is detected, the fault matrix is ​​established by correlating the cooling parameters:

[0227] If the flow rate decreases and the discharge phase is concentrated, it is determined that the water channel is blocked and local overheating discharge occurs;

[0228] If the temperature is normal but the pulse repetition rate increases suddenly, it will warn that the insulation of the thyristor trigger unit is deteriorating;

[0229] The discharge source is mapped to specific layer coordinates through a three-dimensional positioning algorithm.

[0230] 4. Emergency Mechanism for Extreme Working Conditions

[0231] When the grounding grid potential rises due to lightning strike, the optical fiber transmission channel is automatically switched;

[0232] Triggering emergency monitoring mode:

[0233] The polling cycle is compressed to 500ms;

[0234] Enable full-band scanning to capture transient discharges;

[0235] The neural network model is switched to a lightweight version;

[0236] Record the electromagnetic transient process and discharge correlation characteristics to generate the insulation strength decay curve.

[0237] Example 6: Offshore Platform Power System Monitoring

[0238] 1. Adaptation to high salt fog environment

[0239] Adopt helium sealed sensor, with polytetrafluoroethylene anti-corrosion layer sprayed on the surface;

[0240] The sea fog corona noise characteristics are collected during the initialization phase;

[0241] Dynamically shield high-interference frequency bands and lock the optimal monitoring window.

[0242] 2. Mechanical vibration decoupling analysis

[0243] Install a three-axis accelerometer to synchronously collect the platform vibration spectrum;

[0244] Developing a vibration-discharge coupling separation algorithm:

[0245] ;

[0246] in, is the spectrum of the discharge signal after separation, is the original mixed signal spectrum, is the vibration signal spectrum, is the transfer function, is the coupling coefficient;

[0247] The effective pulse capture rate remained above 92% during the typhoon's passage.

[0248] 3. Collaborative monitoring of multiple energy systems

[0249] Associated gas turbine generator sets, photovoltaic inverters, and energy storage converter operating status;

[0250] When abnormal discharge is detected:

[0251] If the discharge occurs on the gas turbine busbar side and is accompanied by an increase in CO concentration, an early warning of thermal decomposition of the insulation material will be issued;

[0252] If the discharge is synchronized with the PV inverter switching frequency, the DC bus support capacitor is diagnosed as failing;

[0253] Establish equipment health scoring models to drive preventive maintenance decisions.

[0254] Example 7: Monitoring underground substations in urban core areas

[0255] 1. Ultra-dense interference suppression

[0256] Constructing a panoramic map of the electromagnetic environment:

[0257] 5G base station downlink frequency band

[0258] Subway train-to-ground communication frequency band

[0259] Scattered radio frequency points of IoT devices

[0260] Develop joint frequency and spatial domain filtering:

[0261] Spatial domain: beamforming technology is used to enhance the directional gain of the transformer;

[0262] Frequency domain: Design a comb-type stopband filter to suppress the communication carrier;

[0263] Extract discharge signals greater than 5pC under -85dBm background noise.

[0264] 2. Distributed diagnosis of cable trenches

[0265] 32 sensor nodes are deployed along the 10kV cable trench, with a time base synchronization accuracy of less than 1ns;

[0266] Using the pulse arrival time difference positioning method:

[0267] ;

[0268] in, is the time difference between the pulse reaching sensors i and j, is the discharge source coordinate, is the propagation speed of electromagnetic waves, and are the coordinates of sensors i and j;

[0269] Draw a three-dimensional discharge hotspot cloud map.

[0270] 3. Digital twin linkage

[0271] Mapping real-time monitoring data to BIM models;

[0272] When a discharge is detected at the cable intermediate joint:

[0273] Retrieve historical load curves to evaluate insulation aging rate;

[0274] Combined with ground current data to diagnose the degree of sheath damage;

[0275] Predict remaining life and optimize maintenance scheduling;

[0276] Generate a health record for the entire life cycle of the equipment.

[0277] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A discharge monitoring method based on dynamic frequency band selection and network identification, characterized in that: The method comprises the following steps: Step S1: collecting original electromagnetic signals of power equipment through a UHF sensor array to generate an initial UHF signal data set; Step S2: inputting the initial UHF signal data set into a multi-band filtering module for parallel frequency band processing to generate a low-frequency bandpass signal, a high-frequency bandpass signal, a full-frequency direct-pass signal, and a selective dual-band combination signal; Step S3, performing time domain feature extraction, frequency domain feature extraction, and statistical feature extraction on the signals of each frequency band outputted in step S2 to generate a multi-dimensional feature vector set; Step S4: inputting the multidimensional feature vector set into a pre-trained neural network classification model to verify signal validity, outputting a partial discharge signal probability value and a noise probability value, and generating signal classification identification data; Step S5: Calculate the real-time signal-to-noise ratio (SNR) of each frequency band based on the signal classification identification data, and generate a frequency band performance evaluation matrix in combination with a preset pulse density threshold; Step S6: Execute dynamic frequency band switching decision according to the frequency band performance evaluation matrix; Step S7: Adjust the frequency band polling period according to the switching mode parameters configured by the user, generate optimized partial discharge monitoring results and output an alarm instruction.

2. The discharge monitoring method based on dynamic frequency band selection and network identification according to claim 1, characterized in that: The step S2 comprises: Step S21: Configure a four-way parallel filtering chain through a programmable filter module: Link 1: Use a 550MHz bandpass filter to generate a low-frequency bandpass signal; Link 2: Use a 1300MHz bandpass filter to generate a high-frequency bandpass signal; Link 3: Uses a 900MHz all-pass filter to generate a full-band direct-through signal; Link 4: Dynamically selects two non-overlapping frequency bands for combined filtering to generate a dual-band combined signal; Step S22: Control the signal output timing of each filter chain through a multi-way selection switch.

3. The discharge monitoring method based on dynamic frequency band selection and network identification according to claim 1, characterized in that: The operation of frequency domain feature extraction in step S3 is specifically as follows: Step S31: performing discrete wavelet transform on each frequency band signal to decompose it into 8 sub-bands; Step S32: Calculate the energy proportion of each sub-band: Among them, E k is the energy proportion of the kth sub-band, D k (i) is the wavelet coefficient of the kth sub-band, i is the sampling point index, N is the number of sampling points, and k is the sub-band index; Step S33: Use the energy proportion distribution vector as a frequency domain feature.

4. The discharge monitoring method based on dynamic frequency band selection and network identification according to claim 1, characterized in that: The neural network classification model construction in step S4 includes: Step S41: The model structure adopts a CNN-LSTM hybrid architecture: Input layer: receives a feature vector of dimension 12; Convolutional layer: 3 layers of 1D-CNN with a convolution kernel size of 3×1; LSTM layer: bidirectional LSTM unit number 64; Output layer: Softmax activation function outputs partial discharge / noise probability; Step S42: The training data includes: Positive samples: pulse waveforms of corona discharge and suspension discharge; Negative samples: mobile phone signals, arc noise, and switch operation interference.

5. The discharge monitoring method based on dynamic frequency band selection and network identification according to claim 1, characterized in that: The logic of the dynamic frequency band switching decision in step S6 is: Step S61: Initialize the polling cycle T and activate each frequency band in sequence to collect signals; Step S62: Calculate the comprehensive performance score of each frequency band: Score i =w1×P PD +w2×SNR i -w3×Δt; Among them, Score i is the comprehensive performance score of the i-th frequency band, P PD is the probability of partial discharge, SNR i is the current signal-to-noise ratio, Δt is the switching time, w1, w2, w3 are weight coefficients; Step S63: Select Score i The highest frequency band serves as the main monitoring channel.

6. The discharge monitoring method based on dynamic frequency band selection and network identification according to claim 1, characterized in that: The switching mode parameter configuration in step S7 includes: Emergency mode: polling period T = 1s, SNR drop threshold = 15%; Steady-state mode: polling period T = 15 minutes, SNR drop threshold = 25%.

7. The discharge monitoring method based on dynamic frequency band selection and network identification according to claim 1, characterized in that: The method also includes an online learning mechanism: Step S71: When unidentified signal types appear continuously, trigger the data enhancement module to generate synthetic samples; Step S72: Update the neural network classification model using an incremental learning algorithm: Among them, θ new is the updated model parameter, θ old is the model parameter before updating, η is the learning rate, is the gradient of the loss function L with respect to the parameter θ, f θ (x new ) is the new sample x for the model new The predicted output, y new is the true label of the newly added sample.

8. The discharge monitoring method based on dynamic frequency band selection and network identification according to any one of claims 1 to 7, characterized in that: The calculation formula of the signal-to-noise ratio SNR in step S5 is: Among them, P signal is the average power of the partial discharge pulse, P noise is the noise power in the pulse-free section.

9. A discharge monitoring system based on dynamic frequency band selection and network identification, used to implement the method according to any one of claims 1 to 8, characterized in that: The system comprises: Signal acquisition and processing module, dynamic frequency band optimization module, and intelligent identification and analysis module; The signal acquisition and processing module includes: Ultra-high frequency sensor array: arranged at the basin insulator of GIS equipment to collect 900MHz electromagnetic signals; Signal conditioning unit: includes preamplifier and 24-bit ADC converter; The dynamic frequency band optimization module includes: Programmable multi-bandpass filter bank: hardware implementation of four parallel filtering chains; Band switching controller: FPGA-based real-time SNR evaluation and switching logic circuit; The intelligent recognition and analysis module includes: Feature extraction engine: performs time domain, frequency domain, and statistical feature calculations; Neural network processor: deploying CNN-LSTM classification model; Online learning unit: supports incremental model updates.

10. The discharge monitoring system based on dynamic frequency band selection and network identification according to claim 9, characterized in that: The programmable multi-bandpass filter group adopts a microstrip line structure design and includes: Low-frequency channel: center frequency 600MHz, bandwidth ±200MHz; High frequency channel: center frequency 1250MHz, bandwidth ±250MHz; Dual-frequency combination channel: switching frequency band combination mode through PIN diode.

Images