A power frequency signal depth analysis method and system

Abstract

This invention discloses a method and system for deep analysis of power frequency signals, relating to the field of power signal processing technology. It includes: a high-speed acquisition step, using a 16-bit or higher ADC to acquire signals at a sampling rate of 200 kS / s; a data preprocessing step, compressing data using the adjacent averaging method; a frequency domain analysis step, performing DFT transformation on the sequence and extracting the fundamental and harmonic frequencies using spectral correction; a time domain integration step, calculating the true RMS value through digital integration; a time domain differentiation step, obtaining the differential sequence using the central difference method and identifying waveform distortion using a threshold of K times the RMS value; and a result output step, fusing and outputting the spectrum, harmonic amplitude, RMS value, and distortion characteristics. The system includes analog acquisition, digital processing, storage, and output units. This invention synergistically utilizes frequency domain analysis and time domain differentiation / integration techniques to achieve comprehensive perception of the steady-state spectrum and transient behavior of power frequency signals, with harmonic measurement error less than 0.5% and transient event identification accuracy reaching 92%, making it suitable for real-time monitoring on edge computing terminals.

Description

Technical Field

[0001] This invention relates to the field of power signal processing technology, and in particular to a method and system for deep analysis of power frequency signals. Background Technology

[0002] The acquisition and analysis of power frequency signals (50Hz or 60Hz voltage / current signals) are fundamental tasks for power system operation and maintenance, power quality monitoring, and equipment condition assessment. Accurately obtaining the amplitude, frequency, harmonic components, and transient characteristics of power frequency signals is of great significance for ensuring the safe and stable operation of the power system.

[0003] Currently, commercially available power frequency testing instruments (such as digital multimeters, clamp meters, power quality analyzers, etc.) mainly adopt the following technical solutions: Zero-crossing detection method: This method uses a hardware comparator to detect the moment when the signal waveform crosses a zero, and calculates the time interval between adjacent zero-crossings to obtain the frequency. While simple to implement, this method is extremely sensitive to harmonic interference and noise. When the signal is distorted or has multiple zero-crossings, it is prone to significant frequency measurement errors.

[0004] Fourier Transform Method: This method uses Discrete Fourier Transform (DFT) or Fast Fourier Transform (FFT) to convert the time-domain signal to the frequency domain and extract the fundamental and harmonic components. However, traditional FFT methods require the sampling frequency to be strictly synchronized with the signal's fundamental frequency; otherwise, spectral leakage and the picket fence effect will reduce the accuracy of the analysis. In actual power grids, there are frequency fluctuations ranging from ±0.2Hz to ±0.5Hz. Existing instruments often use fixed sampling rates or simple windowing, making it difficult to completely eliminate synchronization errors, thus limiting the accuracy of harmonic amplitude measurement.

[0005] True RMS conversion method: This method uses a dedicated true RMS converter chip (such as AD636, AD737, etc.) to convert AC signals into DC levels, and then obtains the RMS value through ADC acquisition. This method has high accuracy under sinusoidal wave conditions, but for non-sinusoidal signals containing high-order harmonics, the bandwidth limitation and crest factor adaptation capability of the chip will affect the measurement accuracy, and it cannot provide specific information on harmonic components.

[0006] Comprehensive parameter analysis method: Some high-end power quality analyzers employ a fusion strategy of multiple algorithms to simultaneously calculate parameters such as RMS value, peak value, and waveform distortion rate. However, these devices tend to focus on the statistics of steady-state parameters and have limited ability to capture transient events (such as voltage sags, pulse interference, and partial discharge), making it difficult to accurately locate waveform distortion points and quantitatively assess the degree of distortion.

[0007] The limitations of existing technologies are mainly reflected in the following aspects: The disconnect between frequency domain and time domain analysis: Most instruments focus either on frequency domain harmonic analysis or on time domain RMS value calculation, lacking in-depth integrated analysis of signal frequency domain characteristics and time domain transient behavior, making it difficult to fully characterize signal characteristics under complex operating conditions.

[0008] Insufficient transient event detection capability: Event detection methods based on changes in effective values ​​have slow response speeds and cannot capture transient interference at the microsecond or millisecond level; while methods based on direct waveform storage have excessively high requirements for storage resources and transmission bandwidth, making it difficult to achieve real-time continuous monitoring in embedded devices.

[0009] The disconnect between laboratory calibration and field application: Current metrology laboratories typically use standard sine waves to calibrate instruments, which perform well under pure signal conditions. However, field signals often contain rich harmonics, noise, and transient interference, leading to significant differences in measurement results for the same complex signal from instruments from different manufacturers. This seriously affects the accuracy of field testing personnel's judgment of equipment status.

[0010] The contradiction between algorithm complexity and real-time performance: High-precision spectrum analysis and transient detection algorithms often involve large amounts of computation, making it difficult to run in real time on resource-constrained edge computing terminals, which limits the widespread application of deep analysis technology in distributed monitoring scenarios.

[0011] In summary, existing power frequency signal analysis technologies are insufficient in terms of frequency domain accuracy, time domain sensitivity, transient event capture capability, and real-time edge processing capability, making it difficult to meet the needs of refined power quality management and intelligent equipment status diagnosis. Summary of the Invention

[0012] This invention aims to provide a method and system for in-depth analysis of power frequency signals, addressing the problems of fragmented frequency and time domain analysis, low sensitivity in transient event detection, and insufficient accuracy in harmonic analysis under asynchronous sampling in existing technologies. By synergistically utilizing DFT frequency domain analysis, digital differentiation, and digital integration techniques, it achieves comprehensive perception of the steady-state spectrum and transient behavior of power frequency signals, improving the accuracy of high-order harmonic extraction, RMS value calculation, and waveform distortion detection, thus providing a scientific assessment tool for power operation and maintenance.

[0013] To achieve the above-mentioned objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for deep analysis of power frequency signals, comprising the following steps: High-speed acquisition steps: Perform high-speed analog-to-digital conversion on the power frequency analog signal to obtain the original digital sequence. ,in This represents the number of sampling points; Data preprocessing step: The original digital sequence is compressed to reduce the data volume, resulting in a preprocessed digital sequence. ,in ; Frequency domain analysis steps: Perform discrete Fourier transform on the preprocessed digital sequence to obtain the frequency domain representation, and extract the amplitude, frequency and phase information of the fundamental wave and each harmonic; Time-domain integration step: Perform digital integration on the preprocessed digital sequence to calculate the energy value of the signal, and then obtain the true effective value of the signal; Time-domain differentiation step: Perform digital differentiation operation on the preprocessed digital sequence to obtain the differential sequence, and identify signal waveform distortion points and transient interference events according to the dynamic threshold; Result output steps: Integrate the analysis results from the frequency domain analysis step, time domain integration step, and time domain differentiation step to output multi-dimensional in-depth analysis information of the power frequency signal.

[0014] Furthermore, in the high-speed acquisition step, an ADC chip with a resolution of no less than 16 bits is used, the sampling rate is set to more than 4000 times the power frequency signal frequency, and the acquisition duration covers at least 5 complete power frequency cycles to ensure that high-frequency harmonic components are acquired without distortion.

[0015] Furthermore, in the data preprocessing step, data compression is performed using the adjacent data averaging method, specifically as follows: ;in, This method also has anti-aliasing filtering effect.

[0016] Furthermore, the frequency domain analysis step further includes a spectrum correction sub-step, which uses interpolation or phase difference correction to correct the spectrum leakage and picket fence effect errors caused by power grid frequency fluctuations, and controls the relative error of harmonic amplitude measurement to within 0.1%.

[0017] Furthermore, in the time-domain integration step, the energy of the signal is calculated using a composite trapezoidal integral algorithm or a Simpson integral algorithm, and the formula for calculating the true RMS value of the signal is: ,in M is the length of the preprocessed digital sequence.

[0018] Furthermore, in the time-domain differentiation step, the central difference method is used for digital differentiation operations, and the specific calculation formula is as follows: Where Δt is the sampling time interval, This method has a higher signal-to-noise ratio compared to forward difference.

[0019] Furthermore, the dynamic threshold is set to K times the effective value obtained in the time-domain integration step, where K ranges from 1.2 to 2.5; when the differential value When the threshold is exceeded, the corresponding sampling point is marked as a waveform distortion point or a transient interference point.

[0020] Furthermore, the time-domain differentiation step further includes a distortion feature extraction sub-step, which counts the number, distribution density, and maximum steepness of differentiation points exceeding the dynamic threshold, in order to quantify the degree of waveform distortion and generate a distortion intensity index.

[0021] Furthermore, in the result output step, the output depth analysis information includes at least the following three types: spectrum diagram of the fundamental wave and each harmonic, harmonic amplitude histogram, harmonic phase diagram, signal true RMS value, waveform distortion rate (THD), waveform distortion occurrence time point, distortion intensity, differential waveform diagram, and transient event record.

[0022] Secondly, the present invention provides a power frequency signal deep analysis system for implementing the above-mentioned method, comprising: Analog signal acquisition unit: includes analog front-end conditioning circuit and high-speed ADC module with a resolution of not less than 16 bits, used to filter, amplify and convert the power frequency analog signal to analog-to-digital, and output the original digital sequence; Digital signal processing unit: includes a data preprocessing module, a DFT frequency domain analysis module, a digital integration module, a digital differentiation module, and a result fusion module. The digital signal processing unit is built into an FPGA, DSP, or high-performance MCU to realize real-time in-depth analysis of power frequency signals. Storage unit: used to cache the original digital sequence, preprocessed sequence, and intermediate calculation results; Output and communication unit: including display interface and wired / wireless communication module, used to output in-depth analysis results and interact with host computer, cloud platform or mobile terminal; Power Management Unit: Provides stable, low-noise operating power to each unit, supporting both battery power and external power supply modes.

[0023] Compared with the prior art, the present invention has the following advantages and beneficial effects: I. Deep Fusion Analysis of Frequency and Time Domains This invention achieves comprehensive perception of the steady-state spectrum and transient behavior of power frequency signals through the coordinated operation of three major modules: DFT frequency domain analysis, digital integration, and digital differentiation. As described in Examples 1 and 9, the system can simultaneously output spectrum diagrams, harmonic histograms, true RMS values, waveform distortion rates, distortion point locations, and differential waveform diagrams, overcoming the shortcomings of traditional instrument frequency domain and time domain analysis that are fragmented, and significantly improving the ability to perceive situations under complex operating conditions.

[0024] II. High-precision spectrum analysis with strong resistance to frequency fluctuations. A 16-bit or higher ADC is used, with a sampling rate set at least 4000 times the power frequency (preferably 200 kS / s) to ensure distortion-free acquisition of harmonics up to the 50th order. See Example 2 and... Figure 2As shown, the measurement error of the 25th harmonic amplitude can be controlled within 0.8%. After introducing the three-point interpolation spectrum correction technique, when the grid frequency fluctuates within the range of 49.8-50.2Hz, the measurement error of the fundamental and harmonic amplitudes decreases from 2.3% to within 0.1% (Example 4 and...). Figure 4 This solves the accuracy bottleneck under asynchronous sampling.

[0025] III. High sensitivity in transient event detection By replacing the traditional forward difference method with the central difference method, the signal-to-noise ratio of the differential sequence is increased from 25dB to 38dB, significantly enhancing the detection capability of weak transient signals (Example 6 and...). Figure 6 Employing a dynamic threshold mechanism based on effective values ​​(K=1.2-2.5), the system can adapt to different signal amplitude levels. As described in Example 8, by extracting features such as the number of points exceeding the threshold, distribution density, and maximum steepness, the system achieves a 92% accuracy rate in identifying three types of events: voltage sag, pulse interference, and harmonic surge, filling the technical gap of "only alarming, not classifying".

[0026] IV. Scientific and accurate calculation of effective values The true effective value is calculated using a digital integration method based on the definition of energy, strictly following... The physical definition. See Example 5 and... Figure 5 As shown, the composite trapezoidal integral algorithm has an effective value calculation error of only 0.3% when the harmonic distortion rate reaches 14.6%, which is far better than the 5.2% error of the peak detection method, ensuring the measurement accuracy under non-sinusoidal waveforms.

[0027] V. Edge computing friendly and real-time responsive The algorithm design balances accuracy and computational efficiency, with a processing time of only 65ms per analysis cycle (100ms of data) (Example 10). It can be easily deployed on resource-constrained edge computing terminals to achieve real-time online monitoring of power frequency signals. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the power frequency signal deep analysis system of the present invention; Figure 2 A comparison chart of measurement errors for the amplitude of the 25th harmonic at different sampling rates; Figure 3 A comparison of the spectral performance of the adjacent average compression method and the direct extraction method; Figure 4 A comparison chart of the measurement errors of fundamental and harmonic amplitudes before and after spectrum correction; Figure 5 A comparison chart of the calculation errors for the effective values ​​of different integration algorithms; Figure 6 This is a comparison of the responses of the forward difference method and the central difference method to transient pulses. Detailed Implementation

[0029] The present invention will be further described in detail below with reference to experimental examples and specific embodiments. However, this should not be construed as limiting the scope of the above-mentioned subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0030] Example 1 This embodiment provides a method for in-depth analysis of power frequency signals, including the following steps: High-speed acquisition steps: Perform high-speed analog-to-digital conversion on the power frequency analog signal to obtain the original digital sequence. ,in This represents the number of sampling points; Data preprocessing step: The original digital sequence is compressed to reduce the data volume, resulting in a preprocessed digital sequence. ,in ; Frequency domain analysis steps: Perform discrete Fourier transform on the preprocessed digital sequence to obtain the frequency domain representation, and extract the amplitude, frequency and phase information of the fundamental wave and each harmonic; Time-domain integration step: Perform digital integration on the preprocessed digital sequence to calculate the energy value of the signal, and then obtain the true effective value of the signal; Time-domain differentiation step: Perform digital differentiation operation on the preprocessed digital sequence to obtain the differential sequence, and identify signal waveform distortion points and transient interference events according to the dynamic threshold; Result output steps: By fusing the analysis results from the frequency domain analysis step, time domain integration step, and time domain differentiation step, multi-dimensional in-depth analysis information of the power frequency signal is output. Specifically: This embodiment provides a complete implementation flow of a power frequency signal deep analysis method, referring to... Figure 1 The system architecture shown.

[0031] Taking the voltage signal monitoring of a 10kV distribution line as an example, the signal is a 50Hz power frequency voltage, which includes the 5th harmonic (amplitude 8%), the 7th harmonic (amplitude 5%), and random noise.

[0032] Step 101 (High-speed acquisition): A 16-bit ADC chip is used to sample the conditioned power frequency analog signal. The sampling rate is set to 200 kS / s, and the acquisition time is 100 ms (i.e., 5 power frequency cycles) to obtain the original digital sequence. .

[0033] Step 102 (Data Preprocessing): For The sequence is compressed by averaging the values ​​of adjacent numbers. Obtain the preprocessed sequence This reduces the amount of subsequent calculations.

[0034] Step 103 (Frequency Domain Analysis): The preprocessed sequence... Input the DFT frequency domain analysis module to perform a discrete Fourier transform and obtain the frequency domain spectrum. In the spectrum, the fundamental wave (50Hz) amplitude is identified as 100.2%, the 5th harmonic (250Hz) amplitude as 7.98%, and the 7th harmonic (350Hz) amplitude as 4.95%, with a frequency identification error of less than 0.01Hz.

[0035] Step 104 (Time Domain Integration): Integrate the preprocessed sequence The input digital integrator module uses a composite trapezoidal integration algorithm to calculate the signal energy, and then calculates the true RMS value: (Per unit value) accurately reflects the contribution of harmonic energy.

[0036] Step 105 (Time Domain Differentiation): The preprocessed sequence... Input the digital differential module and calculate the differential sequence using the central difference method: Using 1.5 times the effective value obtained in step 104 as the dynamic threshold, three points exceeding the threshold were detected, corresponding to tiny glitch noise in the waveform.

[0037] Step 106 (Result Output): The result fusion output module receives the frequency domain analysis, integration, and differentiation results, and simultaneously outputs the spectrum, harmonic histogram, RMS value 1.015, waveform distortion rate THD=9.8%, distortion point location, and differential waveform diagram.

[0038] Example 2: Furthermore, in the high-speed acquisition step, an ADC chip with a resolution of no less than 16 bits is used, the sampling rate is set to more than 4000 times the power frequency signal frequency, and the acquisition duration covers at least 5 complete power frequency cycles to ensure that high-frequency harmonic components are acquired without distortion.

[0039] This embodiment verifies the impact of different sampling parameters on the analysis accuracy. Figure 1 The specific parameter selection for module 101 and the experimental results are as follows. Figure 2 As shown.

[0040] Set up a comparative experiment: Use a 12-bit ADC and a 16-bit ADC to collect the same power frequency signal containing the 25th harmonic at sampling rates of 50kS / s, 100kS / s, and 200kS / s, respectively.

[0041] Experimental conditions: The signal source outputs a 50Hz fundamental frequency, superimposed with 3rd, 5th, 7th, 15th, and 25th harmonics, with amplitudes of 10%, 8%, 6%, 4%, and 2% of the fundamental frequency, respectively. The acquisition time is 100ms (5 cycles), and the data is input into the DFT module for harmonic analysis.

[0042] Experimental results are as follows Figure 2 As shown: Curve A (12-bit ADC, 50kS / s): The amplitude measurement error of the 25th harmonic (1250Hz) reaches 25%, which is caused by high-frequency aliasing due to insufficient sampling rate; Curve B (16-bit ADC, 50kS / s): 25th harmonic amplitude error 8%, due to improved resolution but still insufficient sampling rate; Curve C (16-bit ADC, 100kS / s): 25th harmonic amplitude error 3.2%; Curve D (16-bit ADC, 200kS / s): 25th harmonic amplitude error 0.8%, all harmonics can be accurately identified.

[0043] from Figure 2 It can be clearly seen that when the sampling rate reaches 200kS / s (4000 times the power frequency), the amplitude measurement error of harmonics within the 25th order is controlled within 1%.

[0044] Conclusion: By using a resolution of 16 bits or higher, a sampling rate of 200 kS / s or higher (i.e., more than 4000 times the power frequency), and acquiring more than 5 Hz waves, it is possible to ensure distortion-free acquisition of harmonics up to the 50th order.

[0045] Example 3: Furthermore, in the data preprocessing step, data compression is performed using the adjacent data averaging method, specifically as follows: ;in, This method also has anti-aliasing filtering effect.

[0046] This embodiment verifies the effectiveness of the adjacent data averaging compression method, corresponding to Figure 1 The specific implementation of module 102, and its spectrum effect are as follows: Figure 3 As shown.

[0047] Using the original sequence (Sampling at 200kS / s, data in 100ms), the following steps were performed: Method A: Direct sampling (take 1 point out of every 2 points) Method B: Adjacent average compression ; Test signal: 50Hz fundamental frequency superimposed with 15th harmonic (amplitude 5%) and white noise (signal-to-noise ratio 40dB). The compressed sequence was input into the DFT module for spectrum analysis.

[0048] The results are as follows Figure 3 As shown: Figure 3(a) The spectrum after processing by method A: There is a significant noise increase near the 15th harmonic (750Hz), the noise floor is raised by about 8dB, and the signal-to-noise ratio drops to 32dB; Figure 3 (b) is the spectrum after processing by method B: the 15th harmonic is clearly distinguishable, the noise floor remains at a low level, the signal-to-noise ratio remains at 38dB, and high-frequency noise aliasing is effectively suppressed.

[0049] from Figure 3 It can be clearly seen that the adjacent average compression method effectively suppresses high-frequency noise aliasing while reducing the amount of data.

[0050] Conclusion: The adjacent average compression method has the dual effects of data compression and anti-aliasing filtering.

[0051] Example 4: Furthermore, the frequency domain analysis step further includes a spectrum correction sub-step, which uses interpolation or phase difference correction to correct the spectrum leakage and picket fence effect errors caused by power grid frequency fluctuations, and controls the relative error of harmonic amplitude measurement to within 0.1%.

[0052] This embodiment verifies the adaptability of spectrum correction technology to frequency fluctuations, corresponding to Figure 1 The extended functions of module 103 provide correction effects such as... Figure 4 As shown.

[0053] The power grid frequency was set to fluctuate slowly between 49.8Hz and 50.2Hz, and five cycles of data were collected using a fixed sampling rate of 200kS / s. The two processing methods were compared: Method A: Direct DFT analysis (Module 103 Basic Functions) Method B: Spectral correction is performed using three-point interpolation after DFT (Module 103 extended function). The results are as follows Figure 4 As shown: Figure 4 (a) The fundamental amplitude measurement error of method A: the error reaches 2.3% at a frequency of 49.8 Hz and 2.1% at a frequency of 50.2 Hz. Figure 4 (b) The fundamental amplitude measurement error of method B: the error is controlled within 0.1% throughout the entire frequency fluctuation range.

[0054] Example 5: Furthermore, in the time-domain integration step, the energy of the signal is calculated using a composite trapezoidal integral algorithm or a Simpson integral algorithm, and the formula for calculating the true RMS value of the signal is: ,in M is the length of the preprocessed digital sequence.

[0055] This embodiment verifies the impact of different integration algorithms on the accuracy of effective value calculation. Figure 1 The specific implementation of module 104, and the error comparison, are as follows. Figure 5 As shown.

[0056] Test signal: 50Hz fundamental frequency superimposed with higher harmonics (3rd harmonic 10%, 5th harmonic 8%, 7th harmonic 6%, 9th harmonic 4%), waveform distortion rate THD=14.6%.

[0057] Four algorithms were used to calculate the valid values ​​in module 104, and the results are as follows: Figure 5 As shown: Algorithm A: Peak detection method (RMS = peak value / √2), error 5.2% Algorithm B: Rectangular integration method, error 1.8% Algorithm C: Composite trapezoidal integral method, error 0.3% Algorithm D: Simpson's integral method, error 0.2% from Figure 5 The error comparison bar chart clearly shows that the accuracy of the composite trapezoidal integral and Simpson integral algorithms is much higher than that of the peak detection method and the rectangular integral method.

[0058] Conclusion: The composite trapezoidal integral and Simpson's integral algorithms have the highest accuracy. Module 104 uses the composite trapezoidal integral algorithm by default to balance accuracy and computational cost.

[0059] Example 6: Furthermore, in the time-domain differentiation step, the central difference method is used for digital differentiation operations, and the specific calculation formula is as follows: Where Δt is the sampling time interval, This method has a higher signal-to-noise ratio compared to forward difference.

[0060] This embodiment verifies the advantages of the central difference method compared to the forward difference method, corresponding to Figure 1 The specific implementation of module 105, and the response object, for example Figure 6 As shown.

[0061] Test signal: 50Hz fundamental frequency, with a spike pulse of 0.1ms width and 3 times the amplitude of the fundamental frequency superimposed at 50ms.

[0062] In module 105, two difference algorithms are implemented and compared: Forward difference: ; Central difference: ; The results are as follows Figure 6 As shown: Figure 6(a) shows the original signal waveform, with the spike pulse at 50ms clearly visible; Figure 6 (b) Forward difference results: The noise of the differential sequence is significantly amplified. Although the pulse position can be identified, the background noise is high, and the signal-to-noise ratio is 25dB. Figure 6 (c) Central difference results: The noise level of the differential sequence is significantly reduced, the pulse position is prominent, the signal-to-noise ratio is 38dB, and the pulse width and amplitude characteristics are well preserved.

[0063] from Figure 6 It can be clearly seen that the central difference method effectively suppresses noise amplification while maintaining the sensitivity of transient event detection.

[0064] Conclusion: The central difference method has a higher signal-to-noise ratio, and module 105 uses the central difference method by default.

[0065] Example 7: Furthermore, the dynamic threshold is set to K times the effective value obtained in the time-domain integration step, where K ranges from 1.2 to 2.5; when the differential value When the threshold is exceeded, the corresponding sampling point is marked as a waveform distortion point or a transient interference point.

[0066] This embodiment verifies the basis for selecting the dynamic threshold K value, corresponding to Figure 1 Threshold setting mechanism of module 105.

[0067] 100 sets of actual power frequency signals from the field are collected, including normal waveforms, harmonic distortion waveforms, and transient pulse waveforms. Module 105 obtains the current effective value in real time from module 104, and sets the dynamic threshold = K × effective value. Statistical analysis of the differential sequence distribution characteristics is performed. Experimental results: Under normal waveform conditions, the maximum derivative value shall not exceed 1.2 times the effective value; Under waveforms containing harmonic distortion, the differential value can reach 1.3-1.8 times the effective value; Under transient pulse waveforms, the differential value can reach 2.5-5.0 times the effective value.

[0068] Threshold setting recommendations: K=1.2: Sensitive to weak distortions, but may falsely report noise; K=1.5: Balances sensitivity and specificity, suitable for general monitoring; K=2.0: Only captures significant transient events, suitable for event logging.

[0069] Conclusion: A K value ranging from 1.2 to 2.5 can cover the needs of different application scenarios. Module 105 supports parameterized configuration of the K value.

[0070] Example 8: Furthermore, the time-domain differentiation step further includes a distortion feature extraction sub-step, which counts the number, distribution density, and maximum steepness of differentiation points exceeding the dynamic threshold, in order to quantify the degree of waveform distortion and generate a distortion intensity index.

[0071] This embodiment verifies the ability of the distortion feature extraction method to distinguish event types. Figure 1 The extended functions of module 105 and the fusion processing of module 106.

[0072] Collect 20 sets of each of the three types of typical events: Class A: Voltage sag (amplitude decreases by 30%, lasting for 3 cycles) Class B: Impulse interference (width 0.1ms, amplitude 3 times) Class C: Harmonic surge (5th harmonic surge from 2% to 15%) Module 105 Feature Extraction: Counting the Number of Points Exceeding the Threshold Distribution density Maximum steepness Module 106 receives these features and, in conjunction with the harmonic background information provided by module 103, performs event classification.

[0073] result: Category A: A relatively large number (approximately 150 points). medium, medium; Category B: Very few (1-3 o'clock). Extremely high great; Class C: A relatively large number (approximately 200 points). High, Above average.

[0074] Classification accuracy: Based on these three features, module 106 uses a simple decision tree algorithm to achieve an event type recognition accuracy of 92%.

[0075] Conclusion: Distortion feature extraction effectively quantifies the degree of waveform distortion and supports event classification.

[0076] Example 9: Furthermore, in the result output step, the output depth analysis information includes at least the following three types: spectrum diagram of the fundamental wave and each harmonic, harmonic amplitude histogram, harmonic phase diagram, signal true RMS value, waveform distortion rate (THD), waveform distortion occurrence time point, distortion intensity, differential waveform diagram, and transient event record.

[0077] This embodiment verifies the practicality of multi-dimensional result output, corresponding to Figure 1 Output function of module 106.

[0078] A week-long continuous monitoring was conducted on the 10kV busbar of a certain substation. The output of module 106 included: Regular output (updated every 10 seconds): Fundamental amplitude and frequency Bar chart of amplitude values ​​of harmonics 2-25 True RMS, THD Event-triggered output (when module 105 detects distortion): Time of distortion occurrence (accurate to 0.1ms) Distortion intensity index (0-100) Differential waveforms of one cycle before and after distortion Preliminary identification of transient event type (sag / impulse / harmonic surge) Application Results: By analyzing event logs, maintenance personnel successfully located a pulse interference caused by partial discharge of a cable, verifying the practical value of multi-dimensional output for on-site fault diagnosis.

[0079] Example 10: This embodiment provides a power frequency signal deep analysis system for implementing the above method, including: Analog signal acquisition unit: includes analog front-end conditioning circuit and high-speed ADC module with a resolution of not less than 16 bits, used to filter, amplify and convert the power frequency analog signal to analog-to-digital, and output the original digital sequence; Digital signal processing unit: includes a data preprocessing module, a DFT frequency domain analysis module, a digital integration module, a digital differentiation module, and a result fusion module. The digital signal processing unit is built into an FPGA, DSP, or high-performance MCU to realize real-time in-depth analysis of power frequency signals. Storage unit: used to cache the original digital sequence, preprocessed sequence, and intermediate calculation results; Output and communication unit: including display interface and wired / wireless communication module, used to output in-depth analysis results and interact with host computer, cloud platform or mobile terminal; Power Management Unit: Provides stable, low-noise operating power to each unit, supporting both battery power and external power supply modes. Details are as follows: This embodiment provides a complete implementation of a deep analysis system for power frequency signals, comprehensively corresponding to... Figure 1 The various modules shown are integrated. Figures 2 to 6 Optimization of the verification algorithm.

[0080] Hardware platform (corresponding) Figure 1(Physical implementation of each module) Module 101 Analog Signal Acquisition Unit: Analog front-end (differential input, programmable gain amplifier, anti-aliasing filter) + 16-bit ADC (ADS8363), sampling rate 200kS / s (based on...) Figure 2 (Verification result selected) Modules 102-105 Digital Signal Processing Unit: STM32H743 (Cortex-M7, 400MHz) handles data preprocessing, integration, and differentiation algorithms; FPGA (XC7S25) hardware-accelerated DFT calculations and spectrum correction (based on...) Figure 4 (Verification results achieved) Storage unit: 32MB SDRAM cache for raw data and intermediate results Module 106 Output and Communication Unit: 4.3-inch LCD display, Ethernet port, 4G module, RS485 interface Power Management Unit: Wide voltage input (9-36V DC), battery backup, low-noise LDO, providing stable power to all modules. Software implementation and module correspondence: Module 102 (Data Preprocessing): MCU implements adjacent average compression (based on...) Figure 3 (Validation results optimization) Module 103 (DFT Frequency Domain Analysis): FPGA implements parallel FFT calculations, and MCU implements three-point interpolation spectrum correction (based on...). Figure 4 (Verification results) Module 104 (Digital Integrator): MCU implements composite trapezoidal integrals (based on...) Figure 5 (Verification results) Module 105 (Digital Differentiation and Distortion Detection): MCU implements center differential (based on...) Figure 6 Validation results), dynamic threshold comparison (K=1.5 configurable), feature extraction Module 106 (Result Fusion Output): MCU implements result fusion, display driver, and communication protocol stack.

[0081] Performance testing: Processing time for a single analysis cycle (100ms of data): 65ms (including data transmission between FPGA and MCU) It ran continuously for 72 hours without crashing or losing data. Power consumption: 3.5W on average; battery power allows for up to 4 hours of continuous operation. Deployment and Application: This system, deployed as an edge computing terminal in a company's power distribution room, monitors the power frequency signal of the incoming cabinet in real time. During operation, the system... Figures 2-6 The optimized algorithm, after verification, successfully issued warnings for three voltage sag events, accurately identifying the event types and validating its effectiveness. Figure 1The reliability and practicality of the system shown.

[0082] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for deep analysis of power frequency signals, characterized in that, Includes the following steps: High-speed acquisition steps: Perform high-speed analog-to-digital conversion on the power frequency analog signal to obtain the original digital sequence. ,in This represents the number of sampling points; Data preprocessing step: The original digital sequence is compressed to reduce the data volume, resulting in a preprocessed digital sequence. ,in ; Frequency domain analysis steps: Perform discrete Fourier transform on the preprocessed digital sequence to obtain the frequency domain representation, and extract the amplitude, frequency and phase information of the fundamental wave and each harmonic; Time-domain integration step: Perform digital integration on the preprocessed digital sequence to calculate the energy value of the signal, and then obtain the true effective value of the signal; Time-domain differentiation step: Perform digital differentiation operation on the preprocessed digital sequence to obtain the differential sequence, and identify signal waveform distortion points and transient interference events according to the dynamic threshold; Result output steps: Integrate the analysis results from the frequency domain analysis step, time domain integration step, and time domain differentiation step to output multi-dimensional in-depth analysis information of the power frequency signal.

2. The power frequency signal deep analysis method according to claim 1, characterized in that, In the high-speed acquisition step, an ADC chip with a resolution of no less than 16 bits is used, the sampling rate is set to more than 4000 times the power frequency signal frequency, and the acquisition time covers at least 5 complete power frequency cycles to ensure that high-frequency harmonic components are acquired without distortion.

3. The power frequency signal deep analysis method according to claim 1, characterized in that, In the data preprocessing step, data compression is performed using the adjacent data averaging method, specifically as follows: ;in, This method also has anti-aliasing filtering effect.

4. The power frequency signal deep analysis method according to claim 1, characterized in that, The frequency domain analysis step further includes a spectrum correction sub-step, which uses interpolation or phase difference correction to correct the spectrum leakage and picket fence effect errors caused by power grid frequency fluctuations, and controls the relative error of harmonic amplitude measurement to within 0.5%.

5. The power frequency signal deep analysis method according to claim 1, characterized in that, In the time-domain integration step, the energy of the signal is calculated using either the composite trapezoidal integral algorithm or the Simpson integral algorithm. The formula for calculating the true RMS value of the signal is as follows: ,in M is the length of the preprocessed digital sequence.

6. The power frequency signal deep analysis method according to claim 1, characterized in that, In the time-domain differentiation step, the central difference method is used for digital differentiation, and the specific calculation formula is as follows: Where Δt is the sampling time interval, This method has a higher signal-to-noise ratio compared to forward difference.

7. The power frequency signal deep analysis method according to claim 6, characterized in that, The dynamic threshold is set to K times the effective value obtained in the time-domain integration step, where K ranges from 1.2 to 2.5; when the differential value When the threshold is exceeded, the corresponding sampling point is marked as a waveform distortion point or a transient interference point.

8. The power frequency signal deep analysis method according to claim 7, characterized in that, The time-domain differentiation step further includes a distortion feature extraction sub-step, which counts the number, distribution density, and maximum steepness of differentiation points exceeding the dynamic threshold, to quantify the degree of waveform distortion and generate a distortion intensity index.

9. The power frequency signal deep analysis method according to claim 1, characterized in that, In the result output step, the output depth analysis information includes at least the following three types: spectrum diagram of fundamental wave and each harmonic, harmonic amplitude histogram, harmonic phase diagram, signal true RMS value, waveform distortion rate (THD), waveform distortion occurrence time point, distortion intensity, differential waveform diagram, and transient event record.

10. A power frequency signal deep analysis system, used to implement the method according to any one of claims 1 to 9, characterized in that, include: Analog signal acquisition unit: includes analog front-end conditioning circuit and high-speed ADC module with a resolution of not less than 16 bits, used to filter, amplify and convert the power frequency analog signal to analog-to-digital, and output the original digital sequence; Digital signal processing unit: includes a data preprocessing module, a DFT frequency domain analysis module, a digital integration module, a digital differentiation module, and a result fusion module. The digital signal processing unit is built into an FPGA, DSP, or high-performance MCU to realize real-time in-depth analysis of power frequency signals. Storage unit: used to cache the original digital sequence, preprocessed sequence, and intermediate calculation results; Output and communication unit: including display interface and wired / wireless communication module, used to output in-depth analysis results and interact with host computer, cloud platform or mobile terminal; Power Management Unit: Provides stable, low-noise operating power to each unit, supporting both battery power and external power supply modes.

Images