GIS partial discharge on-line monitoring method

Abstract

The invention discloses a GIS partial discharge on-line monitoring method, and relates to the technical field of power equipment monitoring and fault diagnosis, and the method comprises the following steps: a detection system carries out the real-time collection of an electrical signal caused by a partial discharge phenomenon in the operation process of GIS equipment through a high-frequency sensor disposed on the GIS equipment, therefore, a data basis for subsequent distortion analysis and positioning is obtained. Partial discharge signals are collected in real time through the high-frequency sensor, signal distortion is evaluated in combination with a machine learning model, the time domain and the frequency domain of the signals are accurately corrected by adopting a dynamic time warping algorithm and a frequency compensation means, and the real characteristics of the signals are recovered. According to the method, the signal positioning precision is improved, the GIS equipment fault detection capability is enhanced, potential risks are warned in advance, equipment faults and safety accidents are reduced, and the safety and stability of a power system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring and fault diagnosis, and particularly to an on-line monitoring method for partial discharge in GIS. Background Art

[0002] On-line monitoring of partial discharge in GIS (Gas Insulated Switchgear) refers to the technology of real-time monitoring and analyzing whether partial discharge phenomena occur inside GIS equipment. Partial discharge means that in high-voltage electrical equipment, due to local deterioration or damage of the insulating material, when the electric field strength exceeds the insulation strength, discharge phenomena occur in local areas. Partial discharge will cause the gradual degradation of the insulating material of the equipment over time, and may eventually lead to equipment failure or even outage. On-line monitoring methods usually rely on sensors and data acquisition systems to detect partial discharge signals (such as voltage pulses) in real time, and judge the health status of the equipment by analyzing these signals. Through this monitoring, early warnings can be issued before potential equipment failures occur, thus providing a decision-making basis for the maintenance and management of power equipment and ensuring the stable operation of the power system.

[0003] The prior art has the following deficiencies: In the prior art, partial discharge signals usually propagate inside GIS (Gas Insulated Switchgear) equipment. However, in some complex equipment structures, due to the multi-level structure of the equipment and the presence of multiple media, partial discharge signals often propagate through multiple paths, such as metal enclosures, insulating media, gases, etc. This multi-path propagation phenomenon may cause signal distortion during propagation, thereby increasing the difficulty of signal positioning and affecting the accurate detection and positioning of faults.

[0004] Especially in the case of large equipment size or the presence of multiple layers of insulating materials, the propagation of discharge signals will show more complexity. Specifically, phenomena such as signal reflection, refraction or attenuation may occur. These physical effects may cause changes in the intensity and propagation path of the discharge signal, further exacerbating the complexity of signal processing. Therefore, the discharge signals in the prior art often cannot truly and accurately reflect the actual position of partial discharge, resulting in possible errors in the positioning algorithm when identifying the discharge source.

[0005] If the partial discharge source cannot be accurately located, the following serious consequences may occur: Maintenance personnel may overlook or misjudge the fault point, and cannot carry out targeted repairs or replacements in time, resulting in the equipment remaining in a potential fault state. Prolonged undetected partial discharge may cause further degradation of the equipment's insulating material, and may even lead to more serious equipment damage. In addition, if the partial discharge problem cannot be detected and accurately located in time, in extreme cases, it may lead to safety accidents such as electrical fires and explosions, causing significant economic losses and safety hazards.

[0006] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure. Therefore, it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The object of the present invention is to provide an on-line monitoring method for partial discharge in GIS. By using high-frequency sensors to collect partial discharge signals in real time, combined with a machine learning model to evaluate signal distortion, and adopting the dynamic time warping (DTW) algorithm and frequency compensation means to accurately correct the time domain and frequency domain of the signal, the true characteristics of the signal are restored. This method improves the signal positioning accuracy, enhances the fault detection ability of GIS equipment, warns of potential risks in advance, reduces equipment failures and safety accidents, and significantly improves the safety and stability of the power system to solve the problems in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions: An on-line monitoring method for partial discharge in GIS, including the following steps: The detection system uses high-frequency sensors deployed on GIS equipment to collect in real time the electrical signals generated by partial discharge phenomena during the operation of GIS equipment, so as to obtain the data basis for subsequent distortion analysis and positioning; Preprocess the collected original signals, and organize and store the preprocessed signals according to rules and standards to form a data set; Extract the key indicators reflecting the distortion of partial discharge signals from the data set, and comprehensively analyze the extracted key indicators to quantify the influence of each indicator on the distortion degree of partial discharge signals; Input the key indicators after comprehensive analysis into a machine learning model pre-trained based on historical data. The model intelligently evaluates the electrical signals generated by partial discharge of GIS equipment according to the decision rules learned internally, and judges whether there is a distortion phenomenon in the electrical signals according to the matching result between the characteristic indicators and the model decision boundary; When it is recognized that the discharge signal is distorted, by comparing the reference signal with the actually collected signal, the dynamic time warping algorithm is used to correct the signal waveform distortion caused by multi-path propagation, calculate the optimal matching path between the signals, dynamically adjust the time axis of the signal to achieve time sequence alignment, and combined with frequency compensation means, further improve the time-frequency consistency of the signal.

[0009] Preferably, the specific steps of the detection system using high-frequency sensors deployed on GIS equipment to collect in real time the electrical signals generated by partial discharge phenomena during the operation of GIS equipment are as follows: First, use high-frequency sensors deployed on GIS equipment to monitor and obtain the electrical signals inside the equipment in real time; Subsequently, the acquired signals are transmitted to the central control unit in real time through a high-speed data transmission system to ensure the timely synchronization and accurate transmission of the signals; Finally, all the collected signal data are stored in the database and stored in a structured manner, providing a basis for subsequent signal analysis, distortion detection, and equipment fault location.

[0010] Preferably, key indicators reflecting the distortion of partial discharge signals are extracted from the data set. The extracted indicators include the change in the number of zero crossings per unit time of the signal and the trend of signal complexity over time. Under the detection window, the change in the number of zero crossings per unit time of the signal and the trend of signal complexity over time are comprehensively analyzed, and a zero-crossing density reference value and a signal complexity growth reference value are generated respectively. The degree of distortion of partial discharge signals is quantified through the zero-crossing density reference value and the signal complexity growth reference value.

[0011] Preferably, the specific steps for comprehensively analyzing the change in the number of zero crossings per unit time of the signal to generate a zero-crossing density reference value under the detection window are as follows: Identify the zero-crossing points of the collected partial discharge signals and establish a set , , where is the position of the th zero-crossing point, is the total number of zero-crossing points within the detection window. By dividing the zero-crossing points into multiple sub-segments, calculate the change amplitude index of the zero-crossing interval within each sub-segment. The calculation formula for the change amplitude index of the zero-crossing interval is as follows: , where in the formula, is the change amplitude index of the zero-crossing interval, is the sampling point distance between adjacent zero-crossing points, representing the sampling point distance between the th zero-crossing point and the th zero-crossing point, , is the th zero-crossing point and the th zero-crossing point, is the change amplification coefficient, is the maximum change amplitude of the zero-crossing interval within the sub-segment, is the hyperbolic tangent suppression function; After extracting the change amplitude index of each sub-segment, further integrate the index values of all sub-segments to generate a zero-crossing density reference value. The generation formula is as follows: , where in the formula, is the zero-crossing density reference value, is the total number of sub-segments, is the zero-crossing interval change amplitude index of the th sub-segment, and is the first trend amplification factor.

[0012] Preferably, the specific steps for comprehensively analyzing the trend of signal complexity increasing over time under the detection window to generate a signal complexity growth reference value are as follows: Divide the obtained partial discharge electrical signal into sub-intervals. For each sub-interval, extract the number of local extrema as a complexity characterization index, and calculate the local extremum density volatility. The calculation expression is as follows: , where in the formula, is the local extremum density volatility, that is, the local extremum density volatility of the th sub-segment, is the th number of local maxima that appear on the signal curve within the th sub-segment, is the th number of local minima that appear on the signal curve within the th sub-segment, is the second trend amplification factor, and is the hyperbolic tangent suppression function; , where in the formula, is the signal complexity growth reference value, is the maximum value of the local extremum density volatility, is the minimum value of the local extremum density volatility, is the th local extremum density volatility of the sub-segment, that is, the local extremum density volatility of the previous sub-segment.

[0013] Preferably, input the zero-crossing density reference value and the signal complexity growth reference value after comprehensive analysis into a machine learning model pre-trained based on historical data. Generate a signal distortion risk coefficient through the model, conduct an intelligent evaluation of the electrical signal generated by partial discharge of the GIS device through the signal distortion risk coefficient, and judge whether there is a signal distortion phenomenon according to the matching result between the characteristic index and the model decision boundary.

[0014] Preferably, compare and analyze the signal distortion risk coefficient generated when conducting an intelligent evaluation of the electrical signal generated by partial discharge of the GIS device through a machine learning model pre-trained based on historical data with a pre-set signal distortion risk coefficient reference threshold to judge whether there is a signal distortion phenomenon. The judgment logic is as follows: If the signal distortion risk coefficient is greater than the preset signal distortion risk coefficient, it is determined that the acquired electrical signal is distorted; if the signal distortion risk coefficient is less than or equal to the preset signal distortion risk coefficient, it is determined that the acquired electrical signal is not distorted.

[0015] Preferably, when it is recognized that the discharge signal is distorted, by comparing the reference signal with the actually acquired signal, the dynamic time warping algorithm is used to correct the signal waveform distortion caused by multipath propagation, calculate the optimal matching path between the signals, dynamically adjust the time axis of the signal to achieve timing alignment, and combine frequency compensation means to further improve the time-frequency consistency of the signal. The specific steps are as follows: After it is confirmed that the electrical signal is distorted, that is, when the signal distortion risk coefficient satisfies: where, is the signal distortion risk coefficient, is the reference threshold of the signal distortion risk coefficient, immediately start the DTW algorithm for signal correction, and the calculation formula of the DTW optimal matching path is as follows: , where, is the overall loss value of the optimal matching path of the signal calculated by the DTW algorithm, is the set of matching paths, is the amplitude of the th data point of the actually acquired signal, is the amplitude of the th data point of the reference signal, is the amplitude difference weight coefficient, is the path distance penalty coefficient, is the natural base; Through the DTW optimal matching path, dynamically correct the time axis offset of the actual signal to obtain the signal after timing alignment. The calculation formula is as follows: , where, is the actual signal amplitude of the th data point after dynamic adjustment of the time axis, is the time axis offset of the th data point with respect to the reference signal, is the adaptive time correction sensitivity coefficient, is the overall loss value of the optimal matching path with respect to the data point local change rate, that is, "local path gradient" is the rounding function; After completing the timing alignment, to further improve the time-frequency consistency of the signal, perform dynamic compensation processing in the frequency domain on the signal after time domain correction. The formula is as follows: , where is the actual signal amplitude of the -th data point after dynamic adjustment through the time axis, is the time axis offset of the -th data point with respect to the reference signal, is the adaptive time correction sensitivity coefficient, is the overall loss value of the optimal matching path with respect to the local change rate of the data point , i.e., the "local path gradient" is the rounding function; After completing the timing alignment, to further improve the signal time-frequency consistency, perform dynamic compensation processing in the frequency domain on the signal after time domain correction. The formula is as follows: , where is the frequency domain signal obtained by performing Fourier transform on , i.e., the amplitude of the signal after timing alignment in the frequency domain, is the dynamic frequency compensation intensity coefficient, is the energy distribution of the reference signal at frequency , representing the ideal spectral characteristics, is the spectral energy distribution of the actual signal after timing correction, representing the actual spectral state of the signal, is a very small positive number used to avoid a division-by-zero error in the denominator , is the hyperbolic tangent suppression function, is the signal amplitude after frequency compensation.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are as follows: The present invention collects signals in real time through high-frequency sensors and combines a machine learning model to intelligently evaluate the signal distortion degree, ensuring the accurate identification and positioning of partial discharge signals. The combination of the dynamic time warping (DTW) algorithm and frequency compensation means can accurately correct the time domain and frequency domain of the signal, thereby restoring the true characteristics of the signal and significantly improving the accuracy and reliability of partial discharge signal positioning. This method not only improves the fault detection ability of GIS equipment but also can early warn of potential risks, reduce equipment failures and safety accidents, and greatly ensure the safety and stability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0018] Figure 1 This is the method flow chart of the on-line partial discharge monitoring method for GIS of the present invention. Detailed implementation manners

[0019] Now, the exemplary embodiments will be described more comprehensively with reference to the drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that the present disclosure will be more complete and comprehensive, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0020] The present invention provides an on-line partial discharge monitoring method for GIS as shown in Figure 1 and includes the following steps: The detection system uses high-frequency sensors deployed on GIS equipment to collect electrical signals caused by partial discharge phenomena during the operation of the GIS equipment in real time, so as to obtain the data basis for subsequent distortion analysis and positioning. During the operation of GIS (Gas Insulated Switchgear), the detection system uses high-frequency sensors deployed inside or outside the equipment to collect electrical signals caused by partial discharge phenomena in real time. Partial discharge is usually caused by factors such as insulation aging, defects, or moisture contamination, and will release high-frequency electromagnetic waves. These signals can reflect potential insulation abnormalities inside the equipment. High-frequency sensors mainly include ultra-high frequency (UHF) sensors, transient earth voltage (TEV) sensors, high-frequency current sensors (HFCT), etc. These sensors have good sensitivity and response speed to high-frequency pulse signals generated by partial discharge. Real-time collection means that the signal is captured and recorded immediately at the moment of discharge, ensuring the timeliness and integrity of the data. These electrical signals will be used as the basic data for subsequent analysis to identify whether the signal is distorted due to factors such as multipath propagation, and further perform distortion correction and discharge source positioning. Through this process, potential insulation hazards inside the GIS can be detected in a timely manner, realizing real-time monitoring and intelligent evaluation of the equipment status, and effectively improving the operation safety and fault warning ability of the GIS equipment.

[0021] The specific steps for the detection system to use high-frequency sensors deployed on GIS equipment to collect electrical signals caused by partial discharge phenomena during the operation of the GIS equipment in real time are as follows: First, high-frequency sensors deployed on GIS devices are used to monitor and acquire the electrical signals inside the devices in real time, especially the high-frequency signals caused by partial discharges. These sensors can accurately capture the electromagnetic waves caused by the discharge phenomenon. Subsequently, the acquired signals are transmitted to the central control unit in real time through a high-speed data transmission system to ensure the timely synchronization and accurate transmission of the signals. Finally, all the collected signal data are stored in a database and stored in a structured manner, providing the necessary data basis for subsequent signal analysis, distortion detection, and equipment fault location. Through these steps, it is ensured that the partial discharge signals can be collected and processed quickly and accurately, providing reliable data support for subsequent analysis.

[0022] Preprocess the collected original signals, and organize and store the preprocessed signals according to rules and standards to form a data set. The collected signals will be preprocessed, mainly including denoising, filtering, signal normalization, and time window framing. Denoising is to eliminate the interference of environmental noise on the signals, and filtering removes high-frequency noise and retains the meaningful part of the signals. Normalization helps to unify the measurement standards of signals from different sources and ensures the stability of subsequent processing. Time window framing divides the long signal into multiple time windows, which helps to analyze the signal characteristics within a local time period and facilitates the accurate extraction of key information. The preprocessing steps ensure the quality and consistency of the original data and prepare for subsequent signal analysis and feature extraction.

[0023] The preprocessed signals are organized into a data set, forming a representative and uniformly formatted data set. The data set contains signals collected by multiple sensors at multiple times, covering different possible discharge situations and background noises. The data set includes not only the original signals but also various statistical features of the signals (such as mean, variance, peak value, etc.), as well as the signal patterns generated by the preprocessing steps. The role of this step is to convert the signals into structured data that is convenient for subsequent analysis and input into machine learning models.

[0024] Extract the key indicators reflecting the distortion of the partial discharge signals from the data set, and conduct a comprehensive analysis of the extracted key indicators to quantify the influence of each indicator on the distortion degree of the partial discharge signals. Extract the key indicators reflecting the distortion of the partial discharge signals from the data set. The extracted indicators include the change in the number of zero crossings per unit time of the signal and the trend of signal complexity increasing with time. Under the detection window, conduct a comprehensive analysis of the change in the number of zero crossings per unit time of the signal and the trend of signal complexity increasing with time, and generate the zero crossing density reference value and the signal complexity growth reference value respectively. Quantify the distortion degree of the partial discharge signals through the zero crossing density reference value and the signal complexity growth reference value.

[0025] An abnormal increase in the number of zero crossings of a signal per unit time usually indicates that the acquired partial discharge signal has been distorted. From the perspective of the characteristics of partial discharge signals, normal discharge signals are mainly clear and sharp pulses, with a limited number of zero crossings and a stable distribution pattern; once distortion occurs, especially due to multipath propagation, noise superposition, or reflection interference, the signal waveform becomes more complex, generating a large number of high-frequency jitters and abnormally small oscillations. These oscillations cause the signal to cross the zero point frequently in a short time, resulting in a significant increase in the number of zero crossings per unit time. The abnormal change in the number of zero crossings actually reflects the structural damage and energy distribution disorder of the signal, meaning that the original single and clear partial discharge pulse has been severely interfered with and lost its due simplicity. Therefore, the abnormal zero crossing density is a sensitive and intuitive dynamic indicator of partial discharge signal distortion. By monitoring the change in the number of zero crossings, it is possible to effectively capture the early signs of signal damage, distortion, or external interference, providing an important basis for subsequent signal correction and distortion identification.

[0026] The specific steps for comprehensively analyzing the change in the number of zero crossings of a signal per unit time under a detection window to generate a zero crossing density reference value are as follows: Identify the zero crossing points of the acquired partial discharge signal. The zero crossing point is defined as the position where the signal crosses the zero point from positive to negative or from negative to positive, and establish a set , , where is the position of the th zero crossing point, is the total number of zero crossing points within the detection window. By dividing the zero crossing points into multiple sub-segments, calculate the change amplitude index of the zero crossing interval within each sub-segment. The calculation formula for the change amplitude index of the zero crossing interval is as follows: , where in the formula, is the change amplitude index of the zero crossing interval, is the sampling point distance between adjacent zero crossing points, representing the sampling point distance between the th zero crossing point and the th zero crossing point, , is the sampling point distance between the th zero crossing point and the th zero crossing point, is the change amplification coefficient, which is used to amplify the influence of the zero crossing interval change amplitude on the final index , and its value range is , and it is flexibly set according to the detection sensitivity requirements. Usually, it is most commonly taken between 2.0 and 3.0, taking into account both sensitivity and system stability, is the maximum change amplitude of the zero-crossing intervals within a sub-segment, which means finding the difference between two consecutive zero-crossing intervals within the sub-segment and taking the one with the largest change amplitude. is the hyperbolic tangent suppression function, and its value range is ; The role of the hyperbolic tangent suppression function ( function) in the process of generating the zero-crossing density index is to perform non-linear compression on the maximum zero-crossing interval difference after being amplified by variation, and suppress the influence of extreme outliers on the final index result. Since the local zero-crossing interval changes may show sudden extremely large or small offsets when the partial discharge signal is affected by factors such as multi-path propagation and noise superposition, if the un-suppressed change amplitude is directly used, it is easy to cause the zero-crossing density index to be too sensitive to individual outliers, resulting in a misjudgment phenomenon of distortion amplification. By introducing the hyperbolic tangent function, the large-amplitude changes can be naturally and smoothly restricted within the finite interval of [-1,1], realizing the soft limit processing of large change amplitudes, while maintaining a high response sensitivity to small-amplitude normal changes, so as to ensure that the zero-crossing density index can not only sensitively reflect the signal distortion trend, but also have good stability and anti-interference ability, avoid the system from over-amplifying local extreme jitters, and improve the overall monitoring accuracy and robustness.

[0027] This step calculates the amplification of the maximum zero-crossing interval change, which can highlight the sudden changes inside the signal and reflect the degree of signal distortion.

[0028] After extracting the zero-crossing interval change amplitude index of each sub-segment further integrate the index values of all sub-segments to generate the zero-crossing density reference value, and the generation formula is as follows: , where, is the zero-crossing density reference value, is the total number of sub-segments, is the th sub-segment's zero-crossing interval change amplitude index, is the first trend amplification factor, which is used to weight the zero-crossing interval change amplitude index of each sub-segment, and is usually set to a value greater than 1, such as 1.2, 1.5.

[0029] This step can amplify the contribution of the local distortion area to the zero-crossing density reference value by performing exponential weighted accumulation on the zero-crossing interval change amplitude index of each sub-segment, thereby improving the sensitivity to the distortion of the partial discharge signal.

[0030] After comprehensively analyzing the change in the number of zero crossings of a signal within a unit time under a detection window, a zero-crossing density reference value is generated. When the partial discharge signal is distorted, it is usually accompanied by high-frequency noise, waveform distortion, or multipath reflection effects, causing a large number of abnormal oscillations in the signal within a short period, resulting in a significant increase in the number of zero crossings and ultimately manifested as an increase in the value of the zero-crossing density reference value. Therefore, the larger the zero-crossing density reference value, the more it usually indicates that the continuity and stability within the signal are disrupted, reflecting an obvious distortion phenomenon in the partial discharge signal; conversely, when the zero-crossing density reference value remains at a low and stable level, it indicates that the signal waveform is regular, the pulse characteristics are clear, and it has not been significantly interfered with or distorted, and it can be considered that the obtained partial discharge signal is in a normal state.

[0031] A sharp increase in signal complexity over time usually indicates that the obtained partial discharge signal has been distorted. From the perspective of partial discharge signal distortion, a normal partial discharge signal has relatively stable and regular pulse characteristics, a simple signal form, concentrated spectral energy, and the change in the time series also has a certain degree of predictability. However, when the partial discharge signal is affected by multipath effects, medium inhomogeneity, noise interference, or other abnormal factors during propagation, the original clear pulse waveform will exhibit abnormal superposition, distortion, reflection, and frequency drift, resulting in a more complex change pattern of the signal in the time domain and frequency domain, and a frequent increase in local detail fluctuations. At this time, the complexity of the signal (which can be measured by entropy, fractal dimension, or other complex metric indicators) will increase significantly and at a relatively fast rate, reflecting a rapid increase in the disorder and uncertainty within the signal. Therefore, the sharp growth of signal complexity directly indicates that the signal has deviated from the characteristics of normal partial discharge behavior and has become one of the important dynamic characteristics of a distorted signal, which can be used as an important basis for distortion detection and location analysis.

[0032] The specific steps for comprehensively analyzing the growth trend of signal complexity over time under a detection window to generate a signal complexity growth reference value are as follows: Divide the obtained partial discharge electrical signal into sub-intervals. For each sub-interval, extract the number of local extrema as a complexity characterization index, and calculate the local extreme density volatility. The calculation formula is as follows: , where is the local extreme density volatility, that is, the local extreme density volatility of the th sub-interval, is the number of local maxima (peaks) that appear on the signal curve within the th sub-interval, is the number of local minima (valleys) that appear on the signal curve within the th sub-interval, is the second trend amplification factor, which is used to control the sensitivity of the change in the number of local extrema to the final volatility output. is the hyperbolic tangent suppression function, and its value range is ; In the analysis of the complexity of partial discharge signals, introducing the hyperbolic tangent suppression function mainly serves to perform non-linear compression processing on the total number of extrema, so as to prevent the complexity index from being severely amplified and the numerical value getting out of control due to the abnormal increase in the number of local maxima and minima. When the partial discharge signal is in a distorted state, there are often a large number of short-term fine fluctuations, which causes the number of local extrema to increase sharply; if the number of extrema is directly linearly accumulated, it may lead to too large a jump in the complexity curve, and the system cannot distinguish the responses to slight distortion and severe distortion. By using function, it can maintain high sensitivity in a small range of changes (that is, even a slight increase in extrema can be recognized), while when the number of extrema increases sharply, the output value tends to saturation (close to 1), achieving natural suppression and smoothing of abnormal extreme situations. Therefore, the hyperbolic tangent function plays a dual role of dynamically enhancing the perception of small changes and preventing extreme anomalies from getting out of control, making the signal complexity volatility have both high dynamic response and ensuring the stability and comparability of the overall curve, and improving the accuracy and robustness of partial discharge signal distortion detection.

[0033] Through the above steps, the non-linear compression characterization of the complexity changes of the signal in each sub-interval within the detection window can be carried out, avoiding the influence of extreme abnormal values on the overall evaluation result, and effectively reflecting the dynamic change characteristics of the partial discharge signal waveform structure.

[0034] Based on all the obtained local extreme density volatilities , calculate the signal complexity growth reference value to quantify the growth trend of the overall complexity over time. The calculation expression is as follows: , where, is the signal complexity growth reference value, is the maximum value of the local extreme density volatility, indicating the maximum value of the local extreme density volatility in all sub-intervals within the detection window, is the minimum value of the local extreme density volatility, indicating the minimum value of the local extreme density volatility in all sub-intervals within the detection window, is the local extreme density volatility of the th sub-interval, that is, the local extreme density volatility of the previous sub-interval.

[0035] Through the above steps, while highlighting the drastic changes in local extreme value fluctuations, it can be corrected by combining the overall change density, thereby enhancing the sensitive perception ability of the dynamic characteristics of signal distortion. When a significant distortion phenomenon occurs in the partial discharge signal, the complexity growth reference value will rise sharply, reflecting the abnormal evolution process of the signal structure; conversely, if the signal structure is stable, it will show a low level, indicating that the signal is in a normal state.

[0036] The growth trend of the signal complexity over time is comprehensively analyzed under the detection window to generate a signal complexity growth reference value. Under normal circumstances, the waveform change of the partial discharge signal has good regularity and predictability, and the signal complexity grows slowly and changes steadily within the monitoring window. Therefore, the signal complexity growth reference value shows a relatively low level. When the partial discharge signal is affected by factors such as multipath propagation, noise interference, and medium anomalies, the waveform structure of the signal will become disordered and fluctuate frequently, resulting in a sharp increase in the signal complexity within a short period of time, and the growth trend of the complexity becomes significantly steeper, thus making the complexity growth reference value increase significantly. Therefore, the larger the performance value of the signal complexity growth reference value, the more serious the distortion degree of the collected partial discharge signal; conversely, when the growth reference value remains at a relatively low level, it indicates that the signal waveform structure is stable and the partial discharge signal belongs to the normal acquisition state.

[0037] The key indicators after comprehensive analysis are input into a machine learning model pre-trained based on historical data. The model intelligently evaluates the electrical signals generated by partial discharge of GIS equipment according to the decision rules learned internally, and judges whether there is a distortion phenomenon in the electrical signals based on the matching results between the characteristic indicators and the model decision boundary; The zero-crossing density reference value and the signal complexity growth reference value after comprehensive analysis are input into a machine learning model pre-trained based on historical data. The signal distortion risk coefficient is generated through the model, and the electrical signals generated by partial discharge of GIS equipment are intelligently evaluated through the signal distortion risk coefficient, and it is judged whether there is a distortion phenomenon in the electrical signals based on the matching results between the characteristic indicators and the model decision boundary.

[0038] A machine learning model pre-trained based on historical data refers to, before officially engaging in the task of partial discharge signal distortion detection, systematically training a specific machine learning algorithm using a large amount of historical partial discharge signal data that has been manually labeled or verified by reliable sources, enabling it to autonomously learn the distortion laws and characteristic patterns of partial discharge signals. Specifically, these historical datasets usually contain a large number of partial discharge signal samples with different types, degrees of distortion and non-distortion. Each sample includes not only basic time-domain and frequency-domain signal characteristics (such as zero-crossing density, signal complexity growth, etc.), but also an explicit classification label (such as "distorted" or "undistorted"). By using these data to train the machine learning model, the model can automatically identify the internal relationship between key features (such as the magnitude of index changes, waveform morphological complexity, etc.) and the signal distortion state, and form a set of decision rules that can adaptively judge whether a new signal is distorted by continuously iteratively optimizing the prediction error in the training process. This pre-training process usually adopts supervised learning methods such as support vector machine (SVM), random forest (RF), gradient boosting decision tree (GBDT) or deep neural network (DNN), etc., aiming to enable the model to accurately give a risk judgment of signal distortion according to the input features when new signals that have not been seen arrive.

[0039] Furthermore, in this solution, the zero-crossing density reference value and the signal complexity growth reference value after comprehensive analysis are used as input features and fed into this pre-trained model for intelligent evaluation. After receiving these characteristic indicators, the model will calculate a quantified signal distortion risk coefficient based on the decision boundary system established inside it (that is, the best discrimination boundary between "distorted" and "undistorted" samples learned by the model during the training stage). This risk coefficient represents the likelihood of the current signal being distorted. When the risk coefficient exceeds the set judgment threshold, the system can determine that the partial discharge signal is distorted. On the contrary, if the risk coefficient is low, it is considered that the signal is not distorted. In this way, the system can not only achieve real-time and automatic signal quality evaluation, but also because the model is trained based on a large number of historical samples, it has a certain generalization ability and fault tolerance, and can adapt to the changes of partial discharge signals in different discharge environments, different equipment models or different operating states, greatly improving the accuracy and reliability of signal distortion recognition in on-line monitoring of partial discharges in GIS equipment.

[0040] The machine learning model is not limited here, and any machine learning model that can comprehensively analyze the zero-crossing density reference value and the signal complexity growth reference value and generate a signal distortion risk coefficient Signal distortion risk coefficient The generation formula is as follows: , where and are respectively the preset proportionality coefficients of the zero-crossing density reference value and the signal complexity growth reference value , and and are both greater than 0.

[0041] The preset proportionality coefficients ( and ) refer to the coefficients used to adjust and balance the relative influence degrees of the zero-crossing density reference value and the signal complexity growth reference value when generating the signal distortion risk coefficient . Specifically, and are respectively multiplied by the zero-crossing density reference value and the signal complexity growth reference value in the formula, indicating the contribution ratios of these two parameters to the final risk coefficient.

[0042] and 's functions: These two coefficients allow for flexible adjustment of the weights of the zero-crossing density reference value and the signal complexity growth reference value in the risk assessment according to different actual situations, thereby affecting the calculation result of the signal distortion risk coefficient . By adjusting the values of and , the accuracy of the distortion risk assessment model can be optimized according to experimental data or actual application requirements.

[0043] Setting of the preset proportionality coefficients: These proportionality coefficients are usually determined during the model training process. Through data fitting or empirical adjustment, the model can more accurately reflect the distortion risk. In practice, and 's values will vary according to different devices, environmental conditions, and signal characteristics, ensuring that the model can provide reasonable risk assessments under various conditions.

[0044] Therefore, the preset proportionality coefficients are mainly used to control the influence weights of different signal characteristics (such as the zero-crossing density reference value and the signal complexity growth reference value ) on the final signal distortion risk assessment result, ensuring the reasonable calculation and accuracy of the risk coefficient.

[0045] From the signal distortion risk coefficient, it can be seen that the larger the zero-crossing density reference value generated by comprehensively analyzing the change in the number of zero-crossings of the signal per unit time under the detection window, and the larger the complexity growth reference value of the signal generated by comprehensively analyzing the trend of signal complexity over time under the detection window, the greater the signal distortion risk coefficient generated when the electrical signal generated by partial discharge of GIS equipment is intelligently evaluated by a machine learning model pre-trained based on historical data, indicating that the probability of distortion of the electrical signal generated by partial discharge of the GIS equipment obtained is greater. On the contrary, it indicates that the probability of distortion of the electrical signal generated by partial discharge of the GIS equipment obtained is smaller.

[0046] Compare and analyze the signal distortion risk coefficient generated when the electrical signal generated by partial discharge of GIS equipment is intelligently evaluated by a machine learning model pre-trained based on historical data with the pre-set signal distortion risk coefficient reference threshold to determine whether there is a distortion phenomenon in the electrical signal. The judgment logic is as follows: If the signal distortion risk coefficient is greater than the pre-set signal distortion risk coefficient, it is judged that the obtained electrical signal is distorted; if the signal distortion risk coefficient is less than or equal to the pre-set signal distortion risk coefficient, it is judged that the obtained electrical signal is not distorted.

[0047] When it is identified that the discharge signal is distorted, by comparing the reference signal with the actually collected signal, the dynamic time warping (DTW) algorithm is used to correct the signal waveform distortion caused by multi-path propagation, calculate the optimal matching path between the signals, dynamically adjust the time axis of the signal to achieve time series alignment, and combine frequency compensation means to further improve the time-frequency consistency of the signal; When it is identified that the discharge signal is distorted, the core function of this step is to correct the signal waveform distortion caused by factors such as multi-path propagation by comparing the reference signal with the actually collected signal and using the dynamic time warping (DTW) algorithm, so as to restore the time series and frequency consistency of the signal. The DTW algorithm can effectively solve the non-linear deformation of the signal in the time domain, such as stretching, compression or time delay of the signal, by calculating the optimal matching path between the reference signal and the actual signal. By calculating the minimum distance between different paths point by point, the algorithm dynamically adjusts the time axis of the signal, so that the corresponding relationship between the actual signal and the reference signal is accurately aligned, thus eliminating the time delay error and waveform deviation caused by multi-path propagation.

[0048] In addition, to further optimize the frequency response of the signal, this step also introduces a frequency compensation method. This compensation method can adjust the frequency characteristics according to the frequency change of the signal, making the signal closer to the true partial discharge signal in the frequency domain. Since multipath propagation not only affects the time axis of the signal but also may cause attenuation or offset of the frequency components, frequency compensation is crucial for restoring the true characteristics of the signal. Through frequency compensation, the frequency distortion of the signal can be effectively repaired, making it consistent in the time-frequency domain, thereby improving the overall quality and accuracy of the signal.

[0049] Finally, through the combination of timing alignment and frequency compensation, this step can not only restore the original form of the distorted signal but also provide a more accurate signal input for subsequent fault location, anomaly detection, and equipment health assessment, improving the overall accuracy and reliability of the partial discharge monitoring system.

[0050] When it is identified that the discharge signal is distorted, by comparing the reference signal with the actually collected signal, the dynamic time warping (DTW) algorithm is used to correct the waveform distortion of the signal caused by multipath propagation, calculate the optimal matching path between the signals, dynamically adjust the time axis of the signal to achieve timing alignment, and combined with the frequency compensation method, the specific steps to further improve the time-frequency consistency of the signal are as follows: After confirming that the electrical signal is distorted, that is, when the signal distortion risk coefficient satisfies: where, is the signal distortion risk coefficient, is the reference threshold of the signal distortion risk coefficient, immediately start the DTW algorithm for signal correction. The calculation formula for the optimal matching path of DTW is as follows: , where, is the overall loss value of the optimal matching path of the signal calculated by the DTW algorithm, representing the actual collected partial discharge signal and the reference signal After being optimized by the DTW (dynamic time warping) algorithm, the minimum cumulative cost value along the optimal matching path, is the set of matching paths, including all possible corresponding relationships (matching paths) between the collected partial discharge signal and the reference signal . The DTW algorithm finds an optimal path from it to make the total cumulative cost the smallest, is the amplitude of the th data point of the actually collected signal, is the amplitude of the th data point of the reference signal, is the amplitude difference weight coefficient, a parameter used to adjust the non-linear amplification or compression of the amplitude error term. When , large differences are amplified, emphasizing more the regions where the signal is severely distorted; when , small differences are amplified, paying more attention to minor but potentially important distortions; when , it is an ordinary linear amplitude difference measure. is the path distance penalty coefficient, a parameter used to adjust the weight affected by the path distance (time difference). When , the greater the difference between and on the path, the greater the penalty term , encouraging "short-distance correspondence" (smooth alignment). When , the path length is not particularly emphasized, allowing a greater degree of stretching of the time axis. By penalizing long-time misalignments, the probability of unreasonable matches is reduced. is the natural base; Precisely and dynamically calculates the optimal matching path between the actually collected signal and the reference signal through the above steps. By weighing the signal amplitude difference and path distance factors, it realizes the accurate positioning and quantitative characterization of the distortion characteristics caused by multi-path propagation.

[0051] Dynamically corrects the time axis offset of the actual signal through the DTW optimal matching path to obtain the signal after time series alignment. The calculation formula is as follows: , where in the formula, is the actual signal amplitude of the th data point after dynamic adjustment of the time axis, is the time axis offset of the th data point with respect to the reference signal (in units of data points), is the adaptive time correction sensitivity coefficient, a parameter used to adjust the correction offset amplitude, generally a decimal less than 1 or a value close to 1. is the overall loss value of the optimal matching path with respect to the local change rate of the data point , that is, the "local path gradient". is the rounding function, ensuring that the time axis correction operation conforms to the discrete characteristics of digital sampled signals in practical applications; This step dynamically adjusts the time series offset of each data point through the adaptive time correction sensitivity coefficient and the path sensitivity term, realizing a more refined dynamic correction, ensuring a high degree of restoration of the signal waveform details, and achieving a high-precision time series alignment effect of the signal.

[0052] After completing the timing alignment, to further improve the signal's time-frequency consistency, perform dynamic compensation processing in the frequency domain on the signal after time-domain correction. The formula is as follows: , where is the frequency-domain signal obtained by performing a Fourier Transform on , that is, the amplitude of the signal after timing alignment in the frequency domain. It represents the actual signal amplitude at the frequency point after correcting the time axis through the Dynamic Time Warping (DTW) algorithm. is the dynamic frequency compensation intensity coefficient, a coefficient that controls the strength of frequency compensation and correction. is the energy distribution of the reference signal at the frequency , representing the ideal spectral characteristics. is the spectral energy distribution of the actual signal after timing correction, representing the actual spectral state of the signal. is a very small positive number (such as or smaller), used to avoid a division-by-zero error in the denominator . is the hyperbolic tangent suppression function, with a value range of is the amplitude of the signal after frequency compensation, referring to the actual signal amplitude after compensation processing at the frequency point.

[0053] The role of the hyperbolic tangent suppression function (tanh) is to smooth and control the frequency compensation amplitude of the signal through its non-linear characteristics. When there is a large energy difference between the partial discharge signal and the reference signal, direct compensation may lead to over-adjustment, generating new distortion or unstable frequency components. Using the tanh function can suppress and smooth these excessive compensation amplitudes, ensuring that the spectral correction process does not overreact. Specifically, the tanh function limits the deviation of the signal within the range of [-1, 1], avoiding the unlimited expansion of the compensation amount when there is a large deviation, thus making the compensation process more stable and natural, and effectively maintaining the time-frequency consistency of the signal, avoiding unnecessary oscillations or distortions in the signal.

[0054] The role of the above steps is to dynamically adjust the frequency response of the actual signal based on the spectral difference between the reference signal and the actual signal, further precisely eliminating the frequency-domain distortion caused by multi-path propagation, realizing the high-quality restoration of the signal, and significantly improving the time-frequency consistency of the overall signal.

[0055] The present invention collects signals in real time through high-frequency sensors and combines machine learning models to intelligently evaluate the degree of signal distortion, ensuring the accurate identification and positioning of partial discharge signals. The combination of the Dynamic Time Warping (DTW) algorithm and frequency compensation means can accurately correct the time domain and frequency domain of signals, thereby restoring the true characteristics of the signals and significantly improving the accuracy and reliability of partial discharge signal positioning. This method not only enhances the fault detection ability of GIS equipment but also can early warn of potential risks, reduce the occurrence of equipment failures and safety accidents, and greatly ensure the safety and stability of the power system.

[0056] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0057] Only some exemplary embodiments of the present invention have been described by way of illustration above. Undoubtedly, for those of ordinary skill in the art, without departing from the spirit and scope of the present invention, the described embodiments can be modified in various different ways. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0058] It should be noted that in this text, if there are relationship terms such as first and second, they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.

[0059] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution is prior or subsequent. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0060] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0061] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0062] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0063] In addition, the functional units in various embodiments of this application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0064] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0065] Only some exemplary embodiments of the present invention have been described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, without departing from the spirit and scope of the present invention, the described embodiments can be modified in various different ways. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. On-line monitoring method for partial discharge of GIS, characterized in that, It includes the following steps: The detection system uses high-frequency sensors deployed on GIS devices to collect in real time the electrical signals caused by partial discharge phenomena during the operation of GIS devices, so as to obtain the data basis for subsequent distortion analysis and positioning; Preprocess the collected original signals, and organize and store the preprocessed signals according to rules and standards to form a data set; Extract the key indicators reflecting the distortion of partial discharge signals from the data set, and comprehensively analyze the extracted key indicators to quantify the influence of each indicator on the distortion degree of partial discharge signals; Input the key indicators after comprehensive analysis into a machine learning model that has been pre-trained based on historical data. The model intelligently evaluates the electrical signals generated by partial discharge of GIS devices according to the decision rules learned internally, and judges whether there is a distortion phenomenon in the electrical signals according to the matching result between the characteristic indicators and the model decision boundary; When it is recognized that the discharge signal is distorted, by comparing the reference signal with the actually collected signal, use the dynamic time warping algorithm to correct the signal waveform distortion caused by multi-path propagation, calculate the optimal matching path between signals, dynamically adjust the time axis of the signal to achieve timing alignment, and combine frequency compensation means to further improve the time-frequency consistency of the signal.

2. The on-line monitoring method for partial discharge of GIS according to claim 1, characterized in that The specific steps of the detection system using high-frequency sensors deployed on GIS devices to collect in real time the electrical signals caused by partial discharge phenomena during the operation of GIS devices are as follows: First, use high-frequency sensors deployed on GIS devices to monitor and obtain the electrical signals inside the devices in real time; Subsequently, transmit the obtained signals to the central control unit in real time through a high-speed data transmission system to ensure the timely synchronization and accurate transmission of the signals; Finally, all the collected signal data are stored in the database and stored in a structured manner to provide a data basis for subsequent signal analysis, distortion detection and equipment fault positioning.

3. The on-line partial discharge monitoring method for GIS according to claim 1, characterized in that Extract the key indicators reflecting the distortion of partial discharge signals from the data set. The extracted indicators include the change in the number of zero crossings per unit time of the signal and the trend of signal complexity increasing with time. Comprehensively analyze the change in the number of zero crossings per unit time of the signal and the trend of signal complexity increasing with time under the detection window, and generate a zero-crossing density reference value and a signal complexity growth reference value respectively. Quantify the distortion degree of partial discharge signals through the zero-crossing density reference value and the signal complexity growth reference value.

4. The on-line partial discharge monitoring method for GIS according to claim 3, characterized in that The specific steps of comprehensively analyzing the change in the number of zero crossings per unit time of the signal under the detection window to generate a zero-crossing density reference value are as follows: Identify the zero-crossing points of the collected partial discharge signals and establish a set , , where is the position of the th zero-crossing point, and is the total number of zero-crossing points within the detection window. By dividing the zero-crossing points into multiple sub-segments, calculate the change amplitude index of the zero-crossing interval within each sub-segment. The formula for the change amplitude index of the zero-crossing interval is as follows: , where is the zero-crossing interval variation amplitude index, is the sampling point distance between adjacent zero-crossing points, representing the sampling point distance between the th zero-crossing point and the th zero-crossing point, , is the sampling point distance between the th zero-crossing point and the th zero-crossing point, is the variation amplification coefficient, is the maximum variation amplitude of the zero-crossing interval within the sub-segment, is the hyperbolic tangent suppression function; After extracting the change amplitude index of the zero-crossing interval for each sub-segment then, further integrate the index values of all sub-segments to generate a zero-crossing density reference value. The generation formula is as follows: , where is the zero-crossing density reference value, is the total number of sub-segments, is the amplitude change index of the zero-crossing interval of the -th sub-segment, is the first trend amplification factor.

5. The on-line monitoring method for partial discharge of GIS according to claim 3, characterized in that, The specific steps of comprehensively analyzing the trend of signal complexity increasing with time under the detection window to generate a signal complexity growth reference value are as follows: Divide the obtained partial discharge electrical signals into sub-intervals. For each sub-interval, extract the number of local extrema as the complexity characterization index, and calculate the local extremum density volatility. The calculation formula is as follows: , where is the local extreme density volatility, i.e., the local extreme density volatility of the -th subinterval, is the -th subinterval, the number of local maxima that appear on the signal curve, is the -th subinterval, the number of local minima that appear on the signal curve, is the second trend amplification factor, is the hyperbolic tangent suppression function; Based on all the obtained local extreme density volatilities , calculate the reference value of signal complexity growth to quantify the growth trend of the overall complexity over time. The calculation expression is as follows: , where is the reference value for the growth of signal complexity, is the maximum value of the volatility of the local extreme density, is the minimum value of the volatility of the local extreme density, is the volatility of the local extreme density of the th sub - interval, that is, the volatility of the local extreme density of the previous sub - interval.

6. The on-line partial discharge monitoring method for GIS according to claim 3, characterized in that, The zero-crossing density reference value and the signal complexity growth reference value after comprehensive analysis are input into a machine learning model pre-trained based on historical data. The model generates a signal distortion risk coefficient, which is used to intelligently evaluate the electrical signals generated by partial discharge of GIS equipment. According to the matching result between the characteristic index and the model decision boundary, it is judged whether there is a distortion phenomenon in the electrical signals.

7. The on-line partial discharge monitoring method for GIS according to claim 1, characterized in that The signal distortion risk coefficient generated when the electrical signals generated by partial discharge of GIS equipment are intelligently evaluated by a machine learning model pre-trained based on historical data is compared and analyzed with a pre-set signal distortion risk coefficient reference threshold to judge whether there is a distortion phenomenon in the electrical signals. The judgment logic is as follows: If the signal distortion risk coefficient is greater than the pre-set signal distortion risk coefficient, it is judged that the obtained electrical signal is distorted; if the signal distortion risk coefficient is less than or equal to the pre-set signal distortion risk coefficient, it is judged that the obtained electrical signal is not distorted.

8. The on-line partial discharge monitoring method for GIS according to claim 7, characterized in that, When it is identified that there is distortion in the discharge signal, by comparing the reference signal with the actually collected signal, the dynamic time warping algorithm is used to correct the signal waveform distortion caused by multipath propagation, calculate the optimal matching path between the signals, dynamically adjust the time axis of the signal to achieve timing alignment, and combined with frequency compensation means, the specific steps to further improve the time-frequency consistency of the signal are as follows: After it is confirmed that the electrical signal is distorted, that is, when the signal distortion risk coefficient satisfies: wherein, is the signal distortion risk coefficient, is the reference threshold of the signal distortion risk coefficient, the DTW algorithm is immediately started for signal correction, and the calculation formula of the DTW optimal matching path is as follows: , where is the overall loss value of the optimal matching path of the signal calculated by the DTW algorithm, is the set of matching paths, is the amplitude of the th data point of the actually collected signal, is the amplitude of the th data point of the reference signal, is the amplitude difference weight coefficient, is the path distance penalty coefficient, is the natural base; Through the DTW optimal matching path, the time axis offset of the actual signal is dynamically corrected to obtain the signal after timing alignment. The calculation formula is as follows: , where is the actual signal amplitude of the -th data point after dynamic adjustment through the time axis, is the time axis offset of the -th data point with respect to the reference signal, is the adaptive time correction sensitivity coefficient, is the overall loss value of the optimal matching path with respect to the local change rate of the data point , i.e., the "local path gradient" is the rounding function; After completing the timing alignment, to further improve the time-frequency consistency of the signal, dynamic compensation processing in the frequency domain is performed on the signal after time-domain correction. The formula is as follows: , where is the frequency-domain signal obtained by performing a Fourier transform on , i.e., the amplitude in the frequency domain of the signal after time-series alignment, is the dynamic frequency compensation intensity coefficient, is the energy distribution of the reference signal at frequency , representing the ideal spectral characteristics, The spectral energy distribution of the actual signal after time-series correction, representing the actual spectral state of the signal, is a very small positive number used to avoid a division-by-zero error in the denominator , is the hyperbolic tangent suppression function, is the amplitude of the signal after frequency compensation.

Images