Frequency modulation one-way grounding detection system based on frequency adjustment

Abstract

The invention relates to the technical field of grounding detection, and discloses a frequency modulation one-way grounding detection system based on frequency adjustment, which comprises a signal generator module used for generating frequency-adjustable sine wave or square wave detection signals, and the frequency adjustment range is 1 Hz to 1 kHz; the self-adaptive frequency modulation control module is internally provided with an LSTM neural network algorithm, collects power grid background noise in real time, generates a 0-500Hz noise spectrogram through fast Fourier transform, and dynamically outputs an interference avoidance frequency band and an optimal detection frequency set; a multi-frequency orthogonal coding module; a dynamic power distribution module; a signal acquisition and processing module; and a fault positioning module. Through a dynamic adaptive frequency modulation technology of an LSTM neural network, by analyzing a power grid noise spectrum in real time, an optimal detection set of a main frequency and an auxiliary frequency is dynamically generated, frequency decision logic is optimized by using closed-loop feedback, sudden interference frequency bands such as lightning stroke and load abrupt change are actively avoided, and thus the real-time performance and reliability of fault detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of grounding detection, and in particular to a frequency-modulated unidirectional grounding detection system based on frequency regulation. Background Art

[0002] As power systems expand and demand for intelligent systems increases, the number of nonlinear loads, distributed energy resources, and high-frequency power electronic equipment in the grid has increased dramatically, significantly increasing the complexity of the background noise spectrum. Traditional detection methods are unable to adapt to the dynamically changing electromagnetic environment. High-resistance ground faults, due to severe signal attenuation and weak signatures, are susceptible to multipath propagation and inter-frequency crosstalk in complex line topologies (such as multi-branch cables and hybrid lines). Traditional unidirectional ground fault detection methods often rely on fixed-frequency signal injection for detection.

[0003] However, in current technologies, detection methods based on fixed frequency bands are easily affected by grid harmonics, equipment noise, and environmental interference, resulting in insufficient sensitivity in high-resistance ground fault detection and a high false alarm rate. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a frequency-modulated unidirectional grounding detection system based on frequency regulation to solve the problem that the fixed frequency band is easily affected by grid harmonics, equipment noise and environmental interference, resulting in insufficient sensitivity and high false alarm rate in high-resistance grounding fault detection.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a frequency modulation unidirectional grounding detection system based on frequency regulation, comprising:

[0006] A signal generator module is used to generate a frequency-adjustable sine wave or square wave detection signal with a frequency adjustment range of 1Hz to 1kHz;

[0007] The adaptive frequency modulation control module has a built-in LSTM neural network algorithm, which collects power grid background noise in real time and generates a 0-500Hz noise spectrum through fast Fourier transform, dynamically outputting the interference avoidance frequency band and the optimal detection frequency set;

[0008] The multi-frequency orthogonal encoding module uses orthogonal frequency division multiplexing technology to convert the received frequency instructions into multiple orthogonal subcarrier signals. Each subcarrier is loaded with a unique Gold code sequence and the frequency band is isolated by a digital bandpass filter;

[0009] A dynamic power allocation module is configured to control a digitally controlled gain amplifier based on frequency priority, allocating no less than 60% of the total power to the primary frequency signal and allocating the power to the secondary frequency signal in descending order of interference level;

[0010] The signal acquisition and processing module, synchronous detection unit, analog-to-digital converter and filter are used to extract the amplitude attenuation rate and phase offset of the response signal;

[0011] The fault location module calculates the fault distance based on the phase difference and frequency difference, and outputs the position correction value in combination with the pre-trained random forest model.

[0012] Preferably, the adaptive frequency modulation control module includes:

[0013] The noise spectrum sensing unit obtains the frequency domain distribution of the power grid background noise in real time through 1024-point FFT;

[0014] The frequency decision unit uses an LSTM neural network to predict the interference avoidance frequency band and generate an optimal detection frequency set consisting of a primary frequency and at least two auxiliary frequencies;

[0015] The feedback optimization unit dynamically updates the neural network weight parameters according to the signal-to-noise ratio of the multi-frequency response data.

[0016] Preferably, including:

[0017] The code allocation unit allocates a 31-bit Gold code sequence to each subcarrier frequency through a pseudo-random sequence generator;

[0018] The frequency band isolation unit performs frequency band filtering on each orthogonal coded signal through a digital bandpass filter to isolate subcarrier signals of different frequencies;

[0019] The redundancy check unit verifies the integrity of the main frequency band data based on the correlation between the coding sequence of the auxiliary frequency signal and the main frequency signal.

[0020] Preferably, the dynamic power allocation module includes:

[0021] The power priority mapping unit maps the frequency priority to the gain coefficient of the voltage-controlled amplifier through a table lookup method. The mapping relationship satisfies that the gain of the main frequency signal is not less than 3 times that of the auxiliary frequency.

[0022] Digitally controlled gain amplifier, which adjusts the gain value of each frequency band signal in real time according to the gain coefficient;

[0023] The power feedback unit monitors the transmit signal strength through the RMS detection circuit and corrects the gain parameters in a closed loop.

[0024] Preferably, the signal acquisition and processing module operates as follows:

[0025] Synchronous detection: The digital phase-locked loop (PLL) is used to track the detection signal frequency and the quadrature demodulator is used to generate baseband I / Q components.

[0026] Analog-to-digital conversion: I / Q signals are sampled by a 24-bit Σ-Δ analog-to-digital converter with a 128x oversampling rate, and quantization noise ≤-120dB;

[0027] Feature extraction: Calculate the amplitude attenuation rate and phase offset of the response signal and compensate for the phase error using temperature sensor data.

[0028] Preferably, the fault location module includes:

[0029] A phase difference positioning unit is configured to calculate the initial fault distance based on the phase difference and frequency difference between the main frequency signal and the auxiliary frequency signal by a linear proportional relationship between the time difference and the distance;

[0030] A random forest verification unit is configured to input the initial fault distance, amplitude decay rate, and ambient temperature into a pre-trained random forest model and output a corrected fault distance and confidence probability;

[0031] The alarm trigger unit triggers the adaptive frequency modulation control module to reselect the detection frequency and start secondary detection when the confidence probability is lower than 90%.

[0032] A frequency modulation unidirectional grounding detection method based on frequency regulation, the method comprising the following steps:

[0033] S1, dynamic frequency modulation decision, uses LSTM neural network to analyze the power grid background noise spectrum in real time, generates an optimal detection frequency set containing a main frequency and at least two auxiliary frequencies, and prioritizes the frequency band with a signal-to-noise ratio ≥ 20dB as the main frequency. The frequency set is dynamically updated based on the signal-to-noise ratio of the multi-frequency response data;

[0034] S2. Multi-frequency signal generation and power allocation: A frequency-adjustable sine wave or square wave detection signal is generated through a signal generator module. Orthogonal frequency division multiplexing technology is used to generate multi-frequency detection signals. Each subcarrier is loaded with a unique Gold code sequence. The power of the primary frequency signal accounts for ≥ 60%, and the power of the auxiliary frequency signal is allocated in a step-by-step manner according to the interference level.

[0035] S3, signal injection and synchronous acquisition: inject a detection signal into the power grid, track the signal frequency through a digital phase-locked loop, and demodulate the baseband I / Q components of the response current. A 24-bit Σ-Δ analog-to-digital converter is used to sample the signal at a 128x oversampling rate, with a quantization noise of ≤-120dB. Phase error is compensated using temperature sensor data.

[0036] S4, fault location and closed-loop verification, calculate the initial fault distance based on the phase difference and frequency difference between the main and auxiliary frequencies, input the initial distance, amplitude attenuation rate, and temperature into the pre-trained random forest model to output the corrected fault distance and confidence probability. When the confidence probability is lower than 90%, the system is triggered to reselect the detection frequency and start secondary detection until the confidence probability is ≥90%.

[0037] Preferably, the power allocation in step S2 includes:

[0038] The frequency priority is mapped to the gain coefficient through the table lookup method. The gain of the main frequency signal is +6dB, and the gain of the auxiliary frequency signal decreases by 3dB according to the interference level.

[0039] The power distribution deviation is monitored in real time. If the deviation exceeds ±5%, the gain coefficient is adjusted through closed-loop feedback.

[0040] Preferably, the synchronous acquisition in step S3 includes:

[0041] Dynamically compensate the phase offset using temperature sensor data with a compensation coefficient of 0.1 degrees per degree Celsius;

[0042] The demodulated I / Q components are oversampled by 128 times to ensure that the quantization noise is ≤-120dB.

[0043] Preferably, the random forest model pre-trained in step S4 is trained by inputting features including the main and auxiliary frequency phase difference, the amplitude attenuation rate, the line impedance parameter and the ambient temperature, and the output label is the true distance.

[0044] The present invention provides a frequency modulation unidirectional grounding detection system based on frequency regulation. It has the following beneficial effects:

[0045] 1. The present invention uses the dynamic adaptive frequency modulation technology of the LSTM neural network to analyze the power grid noise spectrum in real time, dynamically generate the optimal detection set of the main frequency and auxiliary frequency, use closed-loop feedback to optimize the frequency decision logic, and actively avoid sudden interference frequency bands such as lightning strikes and load mutations, thereby improving the real-time and reliability of fault detection.

[0046] 2. The present invention generates multi-frequency orthogonal subcarrier signals, loads a unique Gold code sequence on each subcarrier, and combines digital bandpass filters to achieve frequency band isolation and multipath signal separation, suppressing inter-frequency interference and grid topology crosstalk, providing high-purity frequency domain data for fault location, and improving positioning accuracy and adaptability to complex lines.

[0047] 3. The present invention calculates the initial fault distance through the phase difference and frequency difference of the main and auxiliary frequencies, combines the random forest model with multi-dimensional features to perform nonlinear correction on the result, and outputs a high-confidence probability positioning result, and automatically triggers secondary detection in low-confidence scenarios, solving the error problem of a single positioning model in complex lines and high-resistance grounding scenarios, and realizing accurate fault positioning and stable detection in dynamic interference environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is an architecture diagram of a frequency modulation unidirectional grounding detection system based on frequency regulation of the present invention;

[0049] Figure 2The present invention is a flow chart of a frequency modulation unidirectional grounding detection method based on frequency regulation. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all 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.

[0051] Please see the attached Figure 1 The embodiment of the present invention provides a frequency modulation unidirectional grounding detection system based on frequency regulation, comprising:

[0052] A signal generator module is used to generate a frequency-adjustable sine wave or square wave detection signal with a frequency adjustment range of 1Hz to 1kHz;

[0053] The adaptive frequency modulation control module has a built-in LSTM neural network algorithm, which collects power grid background noise in real time and generates a 0-500Hz noise spectrum through fast Fourier transform, dynamically outputting the interference avoidance frequency band and the optimal detection frequency set;

[0054] The multi-frequency orthogonal encoding module uses orthogonal frequency division multiplexing technology to convert the received frequency instructions into multiple orthogonal subcarrier signals. Each subcarrier is loaded with a unique Gold code sequence and the frequency band is isolated by a digital bandpass filter;

[0055] A dynamic power allocation module is configured to control a digitally controlled gain amplifier based on frequency priority, allocating no less than 60% of the total power to the primary frequency signal and allocating the power to the secondary frequency signal in descending order of interference level;

[0056] The signal acquisition and processing module, synchronous detection unit, analog-to-digital converter and filter are used to extract the amplitude attenuation rate and phase offset of the response signal;

[0057] The fault location module calculates the fault distance based on the phase difference and frequency difference, and outputs the position correction value in combination with the pre-trained random forest model.

[0058] Specifically, the signal generator module generates a frequency-adjustable sine wave or square wave detection signal (1Hz-1kHz), injecting a recognizable characteristic signal into the power grid to stimulate the response characteristics of the ground fault point. It dynamically adjusts the signal frequency and waveform type (sine waves are suitable for low-noise environments, and the steep edges of square waves are conducive to high-frequency interference detection) to ensure that the detection signal effectively avoids interference from the power grid background noise and adapts to different line impedance characteristics.

[0059] The adaptive frequency modulation control module uses the LSTM neural network algorithm to analyze the power grid background noise spectrum (0-500Hz) in real time, dynamically identify and avoid high-interference frequency bands, and output the optimal detection frequency set (primary frequency + auxiliary frequency). This ensures that the detection signal always avoids strong noise areas, significantly improving the signal-to-noise ratio and anti-interference capabilities. At the same time, it optimizes the frequency decision logic through closed-loop feedback to ensure real-time and reliable fault detection under complex working conditions (such as lightning strikes and sudden load changes), and solves the problem that traditional fixed-frequency detection methods are susceptible to harmonic interference and have a high false alarm rate.

[0060] The multi-frequency orthogonal encoding module uses orthogonal frequency division multiplexing technology to decompose the received frequency instructions into multiple orthogonal subcarrier signals. Each subcarrier is loaded with a unique Gold code sequence to achieve signal identification. The target frequency band is accurately isolated through a digital bandpass filter, effectively eliminating inter-frequency interference and multipath effects, and achieving efficient parallel transmission and separation of multi-frequency signals, thereby improving the system's anti-interference capability and multi-path signal separation accuracy. This ensures stable extraction of fault characteristics in complex electromagnetic environments and provides a crosstalk-free multi-frequency response data foundation for high-precision positioning.

[0061] The dynamic power allocation module dynamically allocates transmission power according to frequency priority through a digitally controlled gain amplifier, allocating no less than 60% of the total power to the main frequency signal to ensure a high signal-to-noise ratio and anti-interference capability in the core frequency band. The auxiliary frequency signal is allocated exponentially decreasingly according to the interference level. Combined with closed-loop feedback, the gain error is calibrated in real time to achieve optimal utilization of power resources, significantly enhance the penetration of high-frequency signals and the sensitivity of weak feature extraction, and ensure stable transmission and accurate analysis of multi-frequency detection signals in complex electromagnetic environments.

[0062] The signal acquisition and processing module uses a synchronous detection unit to track and demodulate the frequency characteristics of the detection signal in real time. It then uses an analog-to-digital converter to perform high-precision digital sampling of the response signal. It then uses a filter to suppress background noise and harmonic interference. Ultimately, it accurately extracts the amplitude attenuation rate and phase offset of the response signal, providing key characteristic parameters for fault location, enabling reliable detection and high-precision location of ground faults.

[0063] The fault location module analyzes the phase and frequency differences between the main and auxiliary frequency signals, calculates the initial fault distance based on the signal propagation speed, and uses a pre-trained random forest model to verify and correct the initial results using multi-dimensional data (amplitude attenuation, ambient temperature and humidity, and line parameters). This effectively solves the error problems of traditional single positioning methods in complex scenarios such as high-resistance grounding and multi-branch lines, and enables rapid and accurate positioning of power grounding faults.

[0064] The adaptive frequency modulation control module includes:

[0065] The noise spectrum sensing unit obtains the frequency domain distribution of the power grid background noise in real time through 1024-point FFT;

[0066] The frequency decision unit uses an LSTM neural network to predict the interference avoidance frequency band and generate an optimal detection frequency set consisting of a primary frequency and at least two auxiliary frequencies;

[0067] The feedback optimization unit dynamically updates the neural network weight parameters according to the signal-to-noise ratio of the multi-frequency response data.

[0068] Specifically, the noise spectrum sensing unit uses a 1024-point fast Fourier transform (FFT) to capture the frequency domain distribution (0-500Hz) of the power grid background noise in real time, accurately identifying the location and intensity of interference sources such as power frequency harmonics and equipment switching noise. It provides a spectrum diagram with a resolution of 0.5Hz and updates the data every 10ms, ensuring real-time perception of dynamic changes in power grid noise. This provides high-precision input to the frequency decision unit and prevents overlap between the detection signal and the interference frequency band.

[0069] The frequency decision unit uses an LSTM neural network to analyze noise spectra and historical response data, predict interference avoidance strategies within the next 10ms, and dynamically generate an optimal frequency set consisting of a primary frequency (signal-to-noise ratio ≥ 20dB) and at least two auxiliary frequencies. This allows for proactive avoidance of sudden interference, supports multi-frequency collaborative detection, and adapts to complex power grid topologies.

[0070] The feedback optimization unit dynamically updates the LSTM weight parameters based on the signal-to-noise ratio of multi-frequency response data, optimizes the frequency selection logic in a closed-loop manner, automatically enhances the high-frequency signal weight for high-attenuation lines, improves detection sensitivity by 30%, and offsets the impact of temperature and humidity changes.

[0071] Multi-frequency orthogonal coding module, including:

[0072] The code allocation unit allocates a 31-bit Gold code sequence to each subcarrier frequency through a pseudo-random sequence generator;

[0073] The frequency band isolation unit performs frequency band filtering on each orthogonal coded signal through a digital bandpass filter to isolate subcarrier signals of different frequencies;

[0074] The redundancy check unit verifies the integrity of the main frequency band data based on the correlation between the coding sequence of the auxiliary frequency signal and the main frequency signal.

[0075] Specifically, the code allocation unit assigns a unique Gold code sequence to each subcarrier frequency through a pseudo-random sequence generator, giving identifiable code identifiers to detection signals of different frequencies. This ensures that the receiver accurately separates signals of each frequency band through correlation operations, avoids mutual crosstalk between multi-frequency signals, and achieves efficient analysis and parallel processing of multipath signals, providing an independent and complete frequency domain data source for subsequent feature extraction.

[0076] The frequency band isolation unit uses a digital bandpass filter to perform frequency band filtering on the orthogonal coded signal. Through precise frequency band selection and isolation technology, it eliminates the spectrum overlap interference between adjacent subcarriers, ensuring that each subcarrier signal is transmitted independently within the preset frequency band. This guarantees the purity and orthogonality of the multi-frequency signal and provides cross-interference-free response data input for the fault location module.

[0077] The redundancy check unit verifies the integrity and consistency of the main frequency band data based on the correlation between the coding sequence of the auxiliary frequency signal and the main frequency signal. It identifies anomalies or distortions in the signal transmission process through matching analysis of the coding sequence, actively eliminates low-confidence data fragments, and enhances the system's fault tolerance in complex electromagnetic environments. It ensures the reliability and stability of fault feature data and avoids misjudgment or missed detection.

[0078] The dynamic power allocation module includes:

[0079] The power priority mapping unit maps the frequency priority to the gain coefficient of the voltage-controlled amplifier through a table lookup method. The mapping relationship satisfies that the gain of the main frequency signal is not less than 3 times that of the auxiliary frequency.

[0080] Digitally controlled gain amplifier, which adjusts the gain value of each frequency band signal in real time according to the gain coefficient;

[0081] The power feedback unit monitors the transmit signal strength through the RMS detection circuit and corrects the gain parameters in a closed loop.

[0082] Specifically, the power priority mapping unit maps the frequency priority to the gain coefficient of the voltage-controlled amplifier through a table lookup method, and dynamically allocates the amplification factor of each frequency band signal according to preset rules, ensuring that the main frequency signal occupies a dominant position in power distribution and providing stable signal strength guarantee for the core detection frequency band. At the same time, it optimizes the gain ratio in real time according to changes in the power grid environment and adapts to different line impedance characteristics.

[0083] The digitally controlled gain amplifier adjusts the amplification factor of each frequency band signal in real time based on the received gain coefficient, and realizes precise power distribution through voltage control, ensuring that the power gradient between the main frequency signal and the auxiliary frequency signal conforms to the preset strategy, suppressing signal crosstalk and maintaining the overall stability of multi-band signals, and adapting to the dynamic power requirements under complex working conditions.

[0084] The power feedback unit monitors the actual strength of the transmitted signal in real time through an RMS detection circuit, generates an error signal based on the preset power target value, and uses a closed-loop control algorithm to dynamically correct the gain parameters to compensate for power deviations caused by line attenuation or environmental interference, ensuring the accurate execution of the power allocation strategy and improving the reliability of the system in long-distance transmission or high-noise environments.

[0085] The signal acquisition and processing module operation includes:

[0086] Synchronous detection: The digital phase-locked loop (PLL) is used to track the detection signal frequency and the quadrature demodulator is used to generate baseband I / Q components.

[0087] Analog-to-digital conversion: I / Q signals are sampled by a 24-bit Σ-Δ analog-to-digital converter with a 128x oversampling rate, and quantization noise ≤-120dB;

[0088] Feature extraction: Calculate the amplitude attenuation rate and phase offset of the response signal and compensate for the phase error using temperature sensor data.

[0089] Specifically, the signal acquisition and processing module uses synchronous detection technology to accurately track the detection signal frequency and demodulate the baseband orthogonal components, ensuring the synchronous analysis of the signal phase and amplitude. The analog-to-digital conversion link uses a high-resolution converter and oversampling strategy to convert the analog signal into a high-fidelity digital signal, effectively suppressing quantization noise and harmonic interference to ensure the accuracy of data acquisition. It also calculates the amplitude attenuation rate and phase offset based on the demodulated signal, and simultaneously integrates temperature sensor data to dynamically compensate for phase deviations introduced by environmental factors, eliminating the impact of external interference on characteristic parameters. Thus, through multi-link collaborative processing, high-precision extraction and noise suppression of key features of the response signal are achieved, providing a stable and reliable data foundation for fault location.

[0090] The fault location module includes:

[0091] A phase difference positioning unit is configured to calculate the initial fault distance based on the phase difference and frequency difference between the main frequency signal and the auxiliary frequency signal by a linear proportional relationship between the time difference and the distance;

[0092] A random forest verification unit is configured to input the initial fault distance, amplitude decay rate, and ambient temperature into a pre-trained random forest model and output a corrected fault distance and confidence probability;

[0093] The alarm trigger unit triggers the adaptive frequency modulation control module to reselect the detection frequency and start secondary detection when the confidence probability is lower than 90%.

[0094] Specifically, the phase difference positioning unit calculates the initial fault distance based on the phase and frequency differences of the main and auxiliary frequency signals, combined with the linear relationship of signal propagation speed. It offsets the environmental interference error of single-frequency measurement through multi-band delay differences, providing a preliminary positioning benchmark for subsequent verification and ensuring basic positioning reliability in complex line scenarios.

[0095] The random forest verification unit inputs the initial fault distance, amplitude attenuation rate, and ambient temperature multi-dimensional data into a pre-trained model. Through feature fusion and nonlinear relationship modeling, it outputs the corrected fault location and confidence probability. This solves the misjudgment problem of traditional linear models in complex scenarios such as high-resistance grounding and multi-branch lines, and improves the credibility and adaptability of positioning results.

[0096] The alarm trigger unit makes dynamic decisions based on the confidence probability threshold output by the random forest. When the result is not credible enough, it triggers the adaptive frequency modulation module to reselect the detection frequency and start secondary detection, forming a closed-loop optimization mechanism of "detection-verification-rechecking". This ensures the stability of the system results in dynamic interference or signal attenuation scenarios, and avoids false positives and missed negatives of low-quality data.

[0097] Please see the attached Figure 2 A frequency modulation unidirectional grounding detection method based on frequency regulation comprises the following steps:

[0098] S1, dynamic frequency modulation decision, uses LSTM neural network to analyze the power grid background noise spectrum in real time, generates an optimal detection frequency set containing a main frequency and at least two auxiliary frequencies, and prioritizes the frequency band with a signal-to-noise ratio ≥ 20dB as the main frequency. The frequency set is dynamically updated based on the signal-to-noise ratio of the multi-frequency response data;

[0099] S2. Multi-frequency signal generation and power allocation: A frequency-adjustable sine wave or square wave detection signal is generated through a signal generator module. Orthogonal frequency division multiplexing technology is used to generate multi-frequency detection signals. Each subcarrier is loaded with a unique Gold code sequence. The power of the primary frequency signal accounts for ≥ 60%, and the power of the auxiliary frequency signal is allocated in a step-by-step manner according to the interference level.

[0100] S3, signal injection and synchronous acquisition: inject a detection signal into the power grid, track the signal frequency through a digital phase-locked loop, and demodulate the baseband I / Q components of the response current. A 24-bit Σ-Δ analog-to-digital converter is used to sample the signal at a 128x oversampling rate, with a quantization noise of ≤-120dB. Phase error is compensated using temperature sensor data.

[0101] S4, fault location and closed-loop verification, calculate the initial fault distance based on the phase difference and frequency difference between the main and auxiliary frequencies, input the initial distance, amplitude attenuation rate, and temperature into the pre-trained random forest model to output the corrected fault distance and confidence probability. When the confidence probability is lower than 90%, the system is triggered to reselect the detection frequency and start secondary detection until the confidence probability is ≥90%.

[0102] Specifically, S1 uses an LSTM neural network to analyze the power grid noise spectrum in real time, dynamically generates a primary and secondary frequency detection set, actively avoids interference frequency bands and prioritizes high signal-to-noise ratio frequency bands, ensuring that the detection signal always adapts to changes in the power grid environment and improves anti-interference capabilities;

[0103] S2 generates multi-frequency detection signals based on orthogonal frequency division multiplexing technology, identifies frequency bands and isolates interference through Gold codes, allocates power to primary and secondary frequencies according to priority, optimizes signal strength distribution, and ensures high-frequency penetration and weak signal detectability.

[0104] After the S3 injects a detection signal into the power grid, it extracts the amplitude and phase characteristics of the response signal through synchronous detection and high-precision analog-to-digital conversion. Combined with temperature compensation, it eliminates measurement errors introduced by environmental factors to ensure the stability and consistency of data acquisition.

[0105] S4 calculates the initial distance based on the delay difference of multi-frequency signals, uses the random forest model to fuse multi-dimensional data for correction and confidence evaluation, and triggers frequency reselection and secondary detection when the confidence level is low, forming a closed-loop optimization mechanism to improve positioning reliability and adaptability to complex scenarios.

[0106] The power allocation in step S2 includes:

[0107] The frequency priority is mapped to the gain coefficient through the table lookup method. The gain of the main frequency signal is +6dB, and the gain of the auxiliary frequency signal decreases by 3dB according to the interference level.

[0108] The power distribution deviation is monitored in real time. If the deviation exceeds ±5%, the gain coefficient is adjusted through closed-loop feedback.

[0109] Specifically, the power allocation in S2 maps the frequency priority to the gain coefficient through a table lookup method, ensuring that the strength of the main frequency signal is prioritized, and the auxiliary frequency signal gradually reduces the gain according to the interference level. Combined with closed-loop feedback, it monitors and calibrates the power deviation in real time, dynamically suppresses signal crosstalk in the interference frequency band, optimizes the overall power distribution of multi-frequency signals, and improves the anti-attenuation capability of high-frequency signals.

[0110] The synchronous acquisition in step S3 includes:

[0111] Dynamically compensate the phase offset using temperature sensor data with a compensation coefficient of 0.1 degrees per degree Celsius;

[0112] The demodulated I / Q components are oversampled by 128 times to ensure that the quantization noise is ≤-120dB.

[0113] Specifically, synchronous acquisition uses temperature sensor data to dynamically compensate for phase shifts caused by ambient temperature changes. High-rate oversampling technology is used to improve signal digitization accuracy, suppress quantization noise and power frequency harmonic interference, and ensure that the amplitude and phase characteristics of the demodulated baseband signal are highly consistent, providing stable, low-error input data for the fault location module.

[0114] The pre-trained random forest model in step S4 is trained by inputting features including the phase difference between the main and auxiliary frequencies, the amplitude attenuation rate, the line impedance parameters, and the ambient temperature, and the output label is the true distance.

[0115] Specifically, the pre-trained random forest model integrates the multi-dimensional characteristics of the main and auxiliary frequency phase difference, amplitude attenuation rate, line impedance parameters and ambient temperature to model and correct the nonlinear relationship of the initial fault distance, thereby solving the positioning deviation problem of the traditional single physical model in complex scenarios such as high-resistance grounding and multi-branch lines, and significantly improving the accuracy of fault location in complex power grid environments.

[0116] While embodiments of the present invention have been shown and described, it will be appreciated 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 invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A frequency modulation unidirectional grounding detection system based on frequency regulation, characterized in that: include: A signal generator module is used to generate a frequency-adjustable sine wave or square wave detection signal with a frequency adjustment range of 1Hz to 1kHz; The adaptive frequency modulation control module has a built-in LSTM neural network algorithm, which collects power grid background noise in real time and generates a 0-500Hz noise spectrum through fast Fourier transform, dynamically outputting the interference avoidance frequency band and the optimal detection frequency set; The multi-frequency orthogonal encoding module uses orthogonal frequency division multiplexing technology to convert the received frequency instructions into multiple orthogonal subcarrier signals. Each subcarrier is loaded with a unique Gold code sequence and the frequency band is isolated by a digital bandpass filter; A dynamic power allocation module is configured to control a digitally controlled gain amplifier based on frequency priority, allocating no less than 60% of the total power to the primary frequency signal and allocating the power to the secondary frequency signal in descending order of interference level; The signal acquisition and processing module, synchronous detection unit, analog-to-digital converter and filter are used to extract the amplitude attenuation rate and phase offset of the response signal; The fault location module calculates the fault distance based on the phase difference and frequency difference, and outputs the position correction value in combination with the pre-trained random forest model.

2. The frequency modulation unidirectional grounding detection system based on frequency regulation according to claim 1, characterized in that: The adaptive frequency modulation control module includes: The noise spectrum sensing unit obtains the frequency domain distribution of the power grid background noise in real time through 1024-point FFT; The frequency decision unit uses an LSTM neural network to predict the interference avoidance frequency band and generate an optimal detection frequency set consisting of a primary frequency and at least two auxiliary frequencies; The feedback optimization unit dynamically updates the neural network weight parameters according to the signal-to-noise ratio of the multi-frequency response data.

3. The frequency modulation unidirectional grounding detection system based on frequency regulation according to claim 1, characterized in that: The multi-frequency orthogonal encoding module includes: The code allocation unit allocates a 31-bit Gold code sequence to each subcarrier frequency through a pseudo-random sequence generator; The frequency band isolation unit performs frequency band filtering on each orthogonal coded signal through a digital bandpass filter to isolate subcarrier signals of different frequencies; The redundancy check unit verifies the integrity of the main frequency band data based on the correlation between the coding sequence of the auxiliary frequency signal and the main frequency signal.

4. The frequency modulation unidirectional grounding detection system based on frequency regulation according to claim 1, characterized in that: The dynamic power allocation module includes: The power priority mapping unit maps the frequency priority to the gain coefficient of the voltage-controlled amplifier through a table lookup method. The mapping relationship satisfies that the gain of the main frequency signal is not less than 3 times that of the auxiliary frequency. Digitally controlled gain amplifier, which adjusts the gain value of each frequency band signal in real time according to the gain coefficient; The power feedback unit monitors the transmit signal strength through the RMS detection circuit and corrects the gain parameters in a closed loop.

5. The frequency modulation unidirectional grounding detection system based on frequency regulation according to claim 1, characterized in that: The signal acquisition and processing module operation includes: Synchronous detection: The digital phase-locked loop (PLL) is used to track the detection signal frequency and the quadrature demodulator is used to generate baseband I / Q components. Analog-to-digital conversion: I / Q signals are sampled by a 24-bit Σ-Δ analog-to-digital converter with a 128x oversampling rate, and quantization noise ≤-120dB; Feature extraction: Calculate the amplitude attenuation rate and phase offset of the response signal and compensate for the phase error using temperature sensor data.

6. The frequency modulation unidirectional grounding detection system based on frequency regulation according to claim 1, characterized in that: The fault location module includes: A phase difference positioning unit is configured to calculate the initial fault distance based on the phase difference and frequency difference between the main frequency signal and the auxiliary frequency signal by a linear proportional relationship between the time difference and the distance; A random forest verification unit is configured to input the initial fault distance, amplitude decay rate, and ambient temperature into a pre-trained random forest model and output a corrected fault distance and confidence probability; The alarm trigger unit triggers the adaptive frequency modulation control module to reselect the detection frequency and start secondary detection when the confidence probability is lower than 90%.

7. A frequency modulation unidirectional grounding detection method based on frequency regulation, characterized in that: A frequency modulation unidirectional grounding detection system based on frequency regulation according to any one of claims 1 to 6, the method comprising the following steps: S1, dynamic frequency modulation decision, uses LSTM neural network to analyze the power grid background noise spectrum in real time, generates an optimal detection frequency set containing a main frequency and at least two auxiliary frequencies, and prioritizes the frequency band with a signal-to-noise ratio ≥ 20dB as the main frequency. The frequency set is dynamically updated based on the signal-to-noise ratio of the multi-frequency response data; S2. Multi-frequency signal generation and power allocation: A frequency-adjustable sine wave or square wave detection signal is generated through a signal generator module. Orthogonal frequency division multiplexing technology is used to generate multi-frequency detection signals. Each subcarrier is loaded with a unique Gold code sequence. The power of the primary frequency signal accounts for ≥ 60%, and the power of the auxiliary frequency signal is allocated in a step-by-step manner according to the interference level. S3, signal injection and synchronous acquisition: inject a detection signal into the power grid, track the signal frequency through a digital phase-locked loop, and demodulate the baseband I / Q components of the response current. A 24-bit Σ-Δ analog-to-digital converter is used to sample the signal at a 128x oversampling rate, with a quantization noise of ≤-120dB. Phase error is compensated using temperature sensor data. S4, fault location and closed-loop verification, calculate the initial fault distance based on the phase difference and frequency difference between the main and auxiliary frequencies, input the initial distance, amplitude attenuation rate, and temperature into the pre-trained random forest model to output the corrected fault distance and confidence probability. When the confidence probability is lower than 90%, the system is triggered to reselect the detection frequency and start secondary detection until the confidence probability is ≥90%.

8. The frequency modulation unidirectional grounding detection method based on frequency regulation according to claim 7, characterized in that: The power allocation in step S2 includes: The frequency priority is mapped to the gain coefficient through the table lookup method. The gain of the main frequency signal is +6dB, and the gain of the auxiliary frequency signal decreases by 3dB according to the interference level. The power distribution deviation is monitored in real time. If the deviation exceeds ±5%, the gain coefficient is adjusted through closed-loop feedback.

9. The frequency modulation unidirectional grounding detection method based on frequency regulation according to claim 7, characterized in that: The synchronous acquisition in step S3 includes: Dynamically compensate the phase offset using temperature sensor data with a compensation coefficient of 0.1 degrees per degree Celsius; The demodulated I / Q components are oversampled by 128 times to ensure that the quantization noise is ≤-120dB.

10. The frequency modulation unidirectional grounding detection method based on frequency regulation according to claim 7, characterized in that: The random forest model pre-trained in step S4 is trained by inputting features including the main and auxiliary frequency phase difference, the amplitude attenuation rate, the line impedance parameter and the ambient temperature, and the output label is the true distance.

Images