A simulink-based partial discharge signal processing link simulation optimization method

Abstract

The application discloses a partial discharge signal processing link simulation optimization method based on Simulink, which comprises the following steps: generating a first digital signal based on a model of a partial discharge signal, and superimposing random noise on the first digital signal to generate a second digital signal; performing frequency domain analysis on the second digital signal to extract frequency domain features, and inputting the frequency domain features and preset environmental factor parameters into a pre-trained parameter prediction model as inputs, and outputting a processing structure type and corresponding structure parameters for constructing a simulation model; constructing a simulation model of a signal processing link according to the processing structure type and the structure parameters in a Simulink simulation environment, inputting the second digital signal into the simulation model and running simulation to obtain a third digital signal; performing complete set empirical mode decomposition processing on the third digital signal to obtain a fourth digital signal; and calculating a similarity index between the fourth digital signal and the first digital signal.

Description

Technical Field

[0001] This invention relates to the field of simulation technology, and in particular to a simulation optimization method for partial discharge signal processing links based on Simulink. Background Technology

[0002] Partial discharge (PD) signals are electrical signals generated by minute discharges occurring in certain areas within an electrical insulation system. Partial discharge occurs when defects, impurities, or other factors within or on the surface of the electrical insulation system cause the electric field strength in a localized area to exceed the breakdown strength of the insulating material, resulting in localized dielectric breakdown or ionization in that tiny area. In current technology, partial discharge signals are generally used to diagnose faults in electrical systems.

[0003] In engineering applications of partial discharge signals, the signal amplitude is extremely small and is often interfered with by complex environmental noise. Therefore, noise suppression and signal enhancement are achieved through filtering and amplification circuits. However, the traditional development process of filtering and amplification circuits often relies on repeated board-brushing and experimental debugging of actual circuit prototypes. This approach has a long development cycle and high trial-and-error costs. In addition, different operating conditions and environmental parameters (such as temperature and humidity) have a significant impact on circuit performance, and actual hardware cannot cover all possible scenarios in a short period of time, leading to potential reliability and adaptability issues.

[0004] Therefore, the industry is increasingly adopting software simulation methods to model the partial discharge signal and circuit system throughout the entire process; however, existing simulation methods have two main technical shortcomings: on the one hand, the complexity and time-varying nature of real environmental noise are difficult to fully replicate with traditional simulation methods, resulting in a gap between the simulated input signal and the actual operation; on the other hand, the modeling of the filter amplifier circuit still has idealized simplifications, and does not take into account the non-ideal characteristics of components and environmental influences, resulting in some simulation results deviating from actual engineering performance.

[0005] More importantly, even if the optimal filter circuit parameters are derived through simulation, inaccurate simulation configurations, such as improper simulation step size settings or insufficient solver type precision, will significantly interfere with the obtained performance indicators. This means that the optimal parameters may not truly meet engineering requirements in actual printed hardware applications. This problem is often overlooked in existing technologies, leading to inconsistencies between subsequent hardware design and actual operating conditions, and increasing the risk of engineering failure.

[0006] The purpose of this invention is to design a Simulink-based partial discharge signal processing link simulation optimization method to address the problems existing in the prior art. Summary of the Invention

[0007] In view of this, the purpose of this invention is to propose a Simulink-based simulation optimization method for partial discharge signal processing links that can solve the above-mentioned problems.

[0008] This invention provides a Simulink-based simulation optimization method for partial discharge signal processing links, comprising:

[0009] S1 generates a first digital signal based on a mathematical model of the partial discharge signal, and then superimposes random noise onto the first digital signal to generate a second digital signal.

[0010] S2 performs frequency domain analysis on the second digital signal to extract frequency domain features, and takes the frequency domain features and preset environmental factor parameters as inputs to a pre-trained parameter prediction model to output the processing structure type and corresponding structure parameters used to build the simulation model.

[0011] S3 constructs a simulation model of the signal processing link in the Simulink simulation environment based on the processing structure type and structural parameters, inputs the second digital signal into the simulation model and runs the simulation to obtain the third digital signal;

[0012] S4 performs complete set empirical mode decomposition on the third digital signal to obtain the fourth digital signal;

[0013] S5 calculates the similarity index between the fourth digital signal and the first digital signal, and iteratively optimizes the structural parameters of the simulation model based on the similarity index to update the structural parameters.

[0014] The beneficial effects of this invention are:

[0015] First, by simulating partial discharge signal data models with different pulse widths, amplitudes, and oscillation periods, and combining various typical noises and unequal signal-to-noise ratios, it can realistically reflect the diverse environments and characteristics of multiple batches of discharge signals in actual electrical sites, ensuring the full realism and coverage of subsequent simulation inputs.

[0016] Secondly, by predicting circuit parameters under different temperatures and humidity through circuit parameter models, it is possible to ensure that the filter circuit operates in a real environment, making the filtered partial discharge signal more consistent with the actual working scenario. At the same time, the model can also select the corresponding type of filter circuit based on the frequency domain characteristics of the input signal. The final filter circuit parameters obtained through the model can be applied to the simulation to ensure the realism and accuracy of the simulation.

[0017] Third, we use simulation component models that can reflect the non-ideal characteristics of components and fully recreate the filtering and amplification links in Simulink to ensure that the simulation conditions are mapped one-to-one with the actual hardware conditions, laying a realistic foundation for subsequent signal optimization and comparative analysis.

[0018] Fourth, through a complete empirical mode decomposition algorithm, based on the adaptive decomposition mode number and correlation coefficient threshold dynamic selection, combined with multi-component collaboration, it can efficiently separate complex noise and signals, maximize the clarity and accuracy of signals, and provide solid data support for backend fault detection and analysis.

[0019] Fifth, a multi-dimensional evaluation system is established with dynamic time warping, frequency domain KL divergence, and energy fidelity as core indicators. When all indicators fail to meet the standards, the noise decomposition algorithm and simulation environment parameters are automatically optimized step by step to achieve full-process adaptive intelligent optimization, ensuring that the simulation effect evolves self-evolves in a high-precision closed loop. The final results can directly guide the actual hardware filter settings and deployment, significantly reducing the cost of repeated manual parameter tuning and hardware trial and error. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the method in this embodiment.

[0022] Figure 2 This is a schematic diagram of the filtering and amplification model in this embodiment.

[0023] Figure 3 This is a comparison diagram of the filtered and amplified waveforms in this embodiment. Detailed Implementation

[0024] To facilitate understanding by those skilled in the art, the structure of the present invention will now be described in further detail with reference to the accompanying drawings. It should be understood that, unless otherwise specified, the order of the steps mentioned in this embodiment can be adjusted according to actual needs, and they can even be executed simultaneously or partially simultaneously.

[0025] like Figure 1 As shown, this embodiment of the invention provides a Simulink-based simulation optimization method for partial discharge signal processing links, including:

[0026] S1 generates a first digital signal based on a mathematical model of the partial discharge signal, and then superimposes random noise onto the first digital signal to generate a second digital signal.

[0027] S101 changes the amplitude, signal pulse width, and oscillation period of the mathematical model to simulate and generate the first digital signal, which is then imported into Simulink;

[0028] S102 randomly adds noise signals with different signal-to-noise ratios to the first digital signal, the noise signals including one or more of Gaussian noise, white noise, and periodic noise, to obtain the second digital signal.

[0029] The mathematical model includes any one of the following: single exponential decay model, double exponential decay model, single exponential decay oscillation model, and double exponential decay oscillation model.

[0030] In this step, the mathematical model can be compiled using MATLAB. The formula of the mathematical model is as follows:

[0031] Single exponential decay model:

[0032] Double exponential decay model:

[0033] Single exponential decaying oscillation model:

[0034] Double exponential decay oscillation model:

[0035] in, It is the time when the signal occurs. A It is the signal amplitude. The attenuation coefficient is... It is the oscillation frequency.

[0036] By adjusting the parameters in the formula, the amplitude and pulse width of the partial discharge signal can be changed. Based on this, multiple partial discharges can be simulated over a period of time through periodic extension. Finally, one or more of three typical noise signals (Gaussian noise, white noise, and periodic noise) are added. By adjusting the signal-to-noise ratio of the noise, the partial discharge signal in the real environment can be simulated as the signal input source for the simulation.

[0037] S2 performs frequency domain analysis on the second digital signal to extract frequency domain features, and takes the frequency domain features and preset environmental factor parameters as inputs to a pre-trained parameter prediction model to output the processing structure type and corresponding structure parameters used to build the simulation model.

[0038] S201 performs a Fourier transform on the second digital signal to extract the frequency domain features of the second digital signal. The frequency domain features include: frequency range, spectral amplitude, signal-to-noise ratio, power spectral density, and bandwidth.

[0039] Based on actual application scenarios and known experience, S202 sets the range of environmental parameter variation, and samples temperature and humidity evenly distributed within the set range to obtain a set of environmental factor characteristics.

[0040] S203 inputs the frequency domain characteristics and environmental factor characteristics of the second digital signal into the pre-trained parameter prediction model to obtain the filter circuit type and its resistance parameter value, capacitance parameter value, and inductance parameter value.

[0041] Specifically, the pre-trained parameter prediction model is trained through the following steps:

[0042] By actually building or simulating various filter circuits under different temperature and humidity conditions, their frequency domain characteristics and component parameters are recorded.

[0043] Frequency domain features, temperature data, and humidity data are used as inputs to the parameter prediction model, while the corresponding filter circuit type and circuit component parameters are used as outputs to construct training samples.

[0044] A parameter prediction model is established by a multi-layer fully connected neural network. Its input layer receives frequency domain features, temperature and humidity data, and the middle layer of the network sets an appropriate number of hidden units to learn the complex correlation between input features. Its output layer outputs the filter circuit type and circuit component parameters respectively.

[0045] We design a joint loss function, using cross-entropy loss for classification output and mean squared error for regression output, and then weighted summation to optimize the overall model performance.

[0046] In this step, traditional software often uses ideal models (linear RLC components) for co-simulation of filters. Non-ideal effects (such as changes in component capacitance and quality factor with temperature and humidity, aging, etc.) are highly dependent on manual settings, requiring experts to input complex physical parameters, making it difficult to comprehensively and accurately cover all actual operating conditions. Many simulation parameters are assumed to be constant, ideal, and without dynamic drift, lacking flexible modeling of environmental factors such as temperature and humidity changes, resulting in significant errors between actual measurements and simulations. During engineering implementation, it is common to encounter situations where the simulation appears fine, but parameter mismatch and significantly deteriorated filtering performance occur when implemented on the actual board.

[0047] Environmental parameters (such as temperature and humidity) are explicitly defined as input variables and input into the model along with frequency domain requirements. The model automatically learns the influence of each environmental component (even nonlinear coupling) on ​​the filter type and the parameters of components such as R / L / C. During inference, the parameters are automatically adjusted to achieve self-compensation design that adapts to the environment.

[0048] Some signals and environmental characteristics are better suited to low-pass circuits, while others are better suited to band-pass circuits. The model can automatically decide on the structure, eliminating the tedious process of manual diagram lookup and decision-making. At the same time, each structure has a different parameter space, and the model can conditionally output the most suitable parameters, performing end-to-end automatic inference in one go, which is more efficient and accurate.

[0049] S3 constructs a simulation model of the signal processing link in the Simulink simulation environment based on the processing structure type and structural parameters, inputs the second digital signal into the simulation model and runs the simulation to obtain the third digital signal;

[0050] Based on the optimized circuit component parameters and filter circuit type, S301 builds the corresponding filter circuit in the Simulink platform and selects a component model that can describe the non-ideal characteristics of the components.

[0051] In this step, the filtering circuit includes any one of a low-pass filter circuit, a high-pass filter circuit, and a band-pass filter circuit. The non-ideal characteristics include, but are not limited to, component temperature coefficients, device tolerances, and parasitic parameters (such as winding inductance, parasitic capacitance, and equivalent series / parallel impedance), which are implemented by setting relevant parameters or calling custom / advanced library models. The temperature and humidity parameters in the simulation environment are set to be the same as in stage S2 to reflect the impact of environmental changes on device performance in real time.

[0052] S302 cascades the filter circuit and the amplifier circuit in sequence to form a complete filter amplifier circuit model;

[0053] S303 configures simulation parameters according to the actual needs of the model, including step size and solver type.

[0054] In this step, as shown, the four partial discharge signal models are then imported into Simulink, and Simulink tools are used to build the following... Figure 2 A filtering and amplification circuit mainly consists of two parts: a filter circuit and an amplifier. The filter circuit can employ low-pass filtering, high-pass filtering, band-pass filtering, etc. Low-pass filtering only allows signals below a certain frequency band to pass, high-pass filtering only allows signals above a certain frequency band to pass, and band-pass filtering only allows signals within a specific frequency band to pass. The parameter design of the filter circuit considers the frequency domain characteristics of the partial discharge signal. For example, the design of the band-pass filter circuit considers the frequency range of the partial discharge signal, filtering out signals with frequencies below 3MHz and above 100MHz to reduce noise interference. The amplifier then amplifies the filtered signal appropriately according to actual needs to enhance signal strength and stability. Through simulation and debugging, the parameters of the filter circuit can be optimized to ensure the reliability and stability of the system under different operating conditions. The simulation effect of filtering and amplification is shown below. Figure 3 As shown, from top to bottom, they are the input signal, the filtered signal, and the amplified signal.

[0055] S4 performs complete set empirical mode decomposition on the third digital signal to obtain the fourth digital signal;

[0056] S401 decomposes the third digital signal using a complete empirical mode decomposition algorithm according to an adaptively set number of decomposition modes to obtain several intrinsic mode functions (IMFs), and averages the IMFs with the same index to improve decomposition stability.

[0057] In this step, the original signal is adaptively decomposed into a set of IMFs according to the set decomposition mode number. The mode number parameter is determined adaptively by the system and can be automatically adjusted based on subsequent feedback. Noise is added to the signal for multiple decompositions, and the corresponding mode components are averaged to improve the decomposition stability and robustness of the results.

[0058] S402 identifies and analyzes Intrinsic Mode Functions (IMFs) based on an adaptively set decomposition correlation coefficient threshold. IMFs with correlation coefficients higher than the threshold are identified as signal components, while those with correlation coefficients lower than the threshold are identified as noise components.

[0059] In this step, the correlation coefficient threshold is used to measure the correlation between each IMF and the original signal. Components with a correlation coefficient higher than the threshold are considered signal components, while those with a correlation coefficient lower than the threshold are considered noise.

[0060] The S403 removes noise components, reconstructs signal components, and outputs a fourth digital signal.

[0061] S5 calculates the similarity index between the fourth digital signal and the first digital signal, and iteratively optimizes the structural parameters of the simulation model based on the similarity index to update the structural parameters.

[0062] S501 calculates the similarity between the fourth digital signal and the first digital signal using dynamic time warping, frequency domain KL divergence, and energy fidelity.

[0063] In this step, the partial discharge signal exhibits strong non-stationarity, containing rich time-domain details (pulse sequence), frequency-domain structure (energy distribution in specific frequency bands), and overall signal energy characteristics. A single indicator cannot simultaneously reflect changes in all dimensions. Dynamic Time Warping (DTW) distance primarily measures the similarity in the time-domain structure, i.e., the overall shape and positional relationship of the waveform, and is most sensitive to alignment of peaks, edges, and main pulses; KL divergence measures the fidelity of the spectral distribution, detecting distortions where the waveforms appear similar but the frequency bands have been abnormally modified, or where energy is lost or enhanced at certain frequencies; Energy fidelity reflects whether the overall or sub-frequency band energy distribution is reasonable, ensuring that characteristic energy is not lost. The above calculation formulas are existing technologies, and the actual implementation can be achieved by directly calling functions.

[0064] If all indicators reach the preset threshold, the simulation result is judged to be excellent; otherwise, it is recorded as substandard and the parameters of the complete empirical mode decomposition algorithm are adaptively adjusted.

[0065] If the dynamic time warping distance exceeds its preset threshold by 20%, S5021 will automatically increase the number of decomposed modes until the maximum limit is reached.

[0066] In this step, Dynamic Time Warping (DTW) is commonly used to calculate the similarity between two one-dimensional time-domain signals (which may have different lengths). It is particularly suitable for handling non-stationary pulse signals such as those involved in partial discharge, and can resist small changes in signal length or phase drift. In simulation evaluation, DTW reflects the high degree of similarity in the waveform's time-domain structure. In practical applications, the function can be directly called for calculation. The specific calculation steps are as follows:

[0067] Reference signal: Simulated signal: ;

[0068] Construct the distance matrix Each element ;

[0069] The formula for calculating the cumulative distance (cumulative cost matrix) C is as follows:

[0070]

[0071] Optimal registration path from Backtracking The registration path is obtained. ;

[0072] The final dynamic time-warped distance is defined as:

[0073]

[0074] Where L is the registration path length.

[0075] In partial discharge signal analysis, due to the short duration and concentrated energy of pulse signals, time-to-wave (DTW) is extremely sensitive to signal shape; even small time drifts or amplitude disturbances can cause DTW changes. If the threshold is set too low, even unrelated noise or sampling errors will frequently trigger parameter adjustments, leading to system oversensitivity, "over-tuning," or even optimization loops, resulting in extremely low simulation efficiency. When the DTW error exceeds the preset threshold by 20%, the corresponding signal peak and shape often show significant misalignment or distortion, which will have a substantial negative impact on the accuracy of subsequent fault location and monitoring.

[0076] When the evaluation finds that the Dynamic Time Warping (DTW) distance is greater than its threshold of 20%, it indicates that the decomposition does not adequately restore the time-domain characteristics of the original signal. At this point, the target modal number parameter for modal decomposition is automatically incremented by 1 (or increased by a specified step size), and the new modal number is used for decomposition in the next S4 decomposition stage. The decomposition, identification, reconstruction, and evaluation process is repeated until the DTW index meets the target or is increased to the maximum modal number (e.g., 15 or 20), at which point the process stops.

[0077] Increasing the number of decomposed modes allows for finer-grained IMF decomposition of the signal, preserving or effectively separating more time-domain details. This helps to reconstruct the pulse structure of the original signal and reduce DTW distance. However, the number of modes should not be increased indefinitely, otherwise mode confusion and the introduction of high-frequency noise can easily occur.

[0078] If the energy fidelity is lower than its preset threshold, S5022 automatically lowers the decomposition correlation coefficient threshold of the mode function and relaxes the mode function retention standard until the minimum limit is reached;

[0079] In this step, energy fidelity reflects whether the overall energy of the simulated signal is consistent with that of the original signal. In practical applications, the function can be directly called for calculation. The specific calculation steps are as follows:

[0080] Common signals Norm energy definition, energy fidelity E This can be represented as normalized similarity:

[0081] ,

[0082] in, , X is the reference signal, Y is the simulation signal, and the value range is: E≤1. The closer E is to 1, the higher the energy reduction degree.

[0083] If the assessment finds that the energy fidelity (such as the wavelet energy distribution correlation coefficient) is lower than the standard, it indicates that the energy spectrum of the reconstructed signal differs significantly from the original signal, and too many signal components have been incorrectly discarded. The IMF correlation coefficient screening threshold is automatically lowered, allowing some IMFs that were previously judged as noise due to insufficient correlation to participate in signal reconstruction.

[0084] If the frequency domain KL divergence of S5023 exceeds its preset threshold, it will automatically increase the number of decomposed modes until the maximum limit is reached.

[0085] In this step, the frequency domain KL divergence directly reflects the deviation between the two signals in terms of frequency structure / power spectral density. In practical applications, the function can be directly called for calculation. The specific calculation steps are as follows:

[0086] Performing a Fourier transform on X and Y yields the amplitude spectrum. and ;

[0087] Power spectral density normalization (as a probability distribution):

[0088] ,

[0089] ,

[0090] in, Frequency The power spectrum at that location;

[0091] KL divergence definition:

[0092] ,

[0093] in, It is a very small positive number to prevent the denominator from being zero. The smaller the KL divergence, the closer the spectral distribution of the simulated signal is to the original signal.

[0094] If the KL divergence exceeds the preset threshold, it indicates a significant difference in the spectral structure between the signal and the original signal, resulting in distorted distribution of frequency components in the decomposed or reconstructed signal. First, automatically increase the number of decomposed modes (steps are the same as S5021) to refine the spectral decomposition. It is recommended to have an upper limit on the number of times each decomposition algorithm is switched and parameters are tried to prevent system malfunction.

[0095] Furthermore, after all S4 parameters (number of decomposed modes, screening threshold, decomposition algorithm) have been adaptively traversed, if the various indicators of the simulated signal (DTW, KL divergence, energy fidelity, etc.) still cannot meet the standards, or if the evaluation results show that the original signal itself has systematic defects, then it is necessary to feed back to the front-end stages such as S1–S3 to adjust the signal generation, acquisition, or preprocessing settings.

[0096] If, after several adaptive adjustments to the parameters of the complete empirical mode decomposition algorithm, the parameters of the S503 filter amplifier circuit model still fail to reach the preset threshold, the simulation parameters of the filter amplifier circuit model will be adaptively adjusted.

[0097] S5031 If the energy fidelity does not reach the preset threshold, the solver type will be switched first.

[0098] If the energy fidelity meets the standard but the dynamic time warping distance exceeds the preset threshold, the S5032 will prioritize reducing the simulation step size.

[0099] S5033 If the energy fidelity and dynamic time warp distance both meet the standards, but the frequency domain KL divergence exceeds the preset threshold, the solver type will be switched first.

[0100] If multiple indicators of S5034 exceed the limit at the same time, the simulation parameters shall be adjusted in the order of priority: energy fidelity, dynamic time warping distance, and frequency domain KL divergence.

[0101] After each round of parameter adjustments, the S5035 re-evaluates the three indicators in real time until all indicators meet the standards or the simulation parameters reach their limits.

[0102] In this step, different types of errors originate from different sources in actual simulations. Globally indiscriminate adjustments will lead to low parameter tuning efficiency and high computational consumption. Energy deviation will cause a "basic loss" of all downstream indicators and must be prioritized. When energy fidelity fails to meet the standard, the simulation solver should be switched first to correct the overall energy conservation. For abnormal Dynamic Time Warping Distance (DTW), the step size should be reduced first to improve time-domain fitting accuracy. For abnormal KL divergence, the solver should be adjusted first to refine the spectral recovery. When multiple indicators exceed the standard simultaneously, simulation parameters are adjusted in the order of energy fidelity, DTW, and KL divergence to ensure a clear optimization direction and stable convergence. The three indicators are monitored in real time after each round of parameter tuning to ensure a closed-loop process, ultimately improving the accuracy and optimization efficiency of the simulation results.

[0103] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0104] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0105] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0106] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0107] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0108] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0109] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0110] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0111] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms should not be construed as necessarily referring to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

Claims

1. A Simulink-based simulation optimization method for partial discharge signal processing links, characterized in that, include: S1 generates a first digital signal based on a mathematical model of the partial discharge signal, and then superimposes random noise onto the first digital signal to generate a second digital signal. S2 performs frequency domain analysis on the second digital signal to extract frequency domain features, and takes the frequency domain features and preset environmental factor parameters as inputs to a pre-trained parameter prediction model to output the filter circuit type and corresponding structural parameters used to build the simulation model. S3 constructs a simulation model of the signal processing link in the Simulink simulation environment based on the filter circuit type and structural parameters, inputs the second digital signal into the simulation model and runs the simulation to obtain the third digital signal; S4 performs complete set empirical mode decomposition on the third digital signal to obtain the fourth digital signal; S5 calculates the similarity index between the fourth digital signal and the first digital signal, and iteratively optimizes the structural parameters of the simulation model based on the similarity index to update the structural parameters.

2. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 1, characterized in that, The mathematical model includes any one of the following: single exponential decay model, double exponential decay model, single exponential decay oscillation model, and double exponential decay oscillation model.

3. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 1, characterized in that, The process of generating a first digital signal using a mathematical model based on partial discharge signals, and then superimposing random noise onto the first digital signal to generate a second digital signal, includes: S101 changes the amplitude, signal pulse width, and oscillation period of the mathematical model to simulate and generate the first digital signal, which is then imported into Simulink; S102 randomly adds noise signals with different signal-to-noise ratios to the first digital signal, the noise signals including one or more of Gaussian noise, white noise, and periodic noise, to obtain the second digital signal.

4. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 1, characterized in that, The step of performing frequency domain analysis on the second digital signal to extract frequency domain features, and then inputting the frequency domain features and preset environmental factor parameters into a pre-trained parameter prediction model to output the filter circuit type and corresponding structural parameters used to construct the simulation model, includes: S201 performs a Fourier transform on the second digital signal to extract the frequency domain features of the second digital signal. The frequency domain features include: frequency range, spectral amplitude, signal-to-noise ratio, power spectral density, and bandwidth. Based on actual application scenarios and known experience, S202 sets the range of environmental parameter variation, and samples temperature and humidity evenly distributed within the set range to obtain a set of environmental factor characteristics. S203 inputs the frequency domain characteristics and environmental factor characteristics of the second digital signal into the pre-trained parameter prediction model to obtain the filter circuit type and its resistance parameter value, capacitance parameter value, and inductance parameter value.

5. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 1, characterized in that, The pre-trained parameter prediction model is trained through the following steps: By actually building or simulating various filter circuits under different temperature and humidity conditions, their frequency domain characteristics and component parameters are recorded. Frequency domain features, temperature data, and humidity data are used as inputs to the parameter prediction model, while the corresponding filter circuit type and circuit component parameters are used as outputs to construct training samples. A circuit parameter model is established by a multi-layer fully connected neural network. Its input layer receives frequency domain features, temperature and humidity data, and the middle layer of the network sets an appropriate number of hidden units to learn the complex correlation between input features. Its output layer outputs the filter circuit type and circuit component parameters respectively. We design a joint loss function, using cross-entropy loss for classification output and mean squared error for regression output, and then weighted summation to optimize the overall model performance.

6. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 1, characterized in that, The simulation model of the signal processing link constructed in the Simulink simulation environment based on the filter circuit type and structural parameters includes: Based on the optimized circuit component parameters and filter circuit type, S301 builds the corresponding filter circuit in the Simulink platform and selects a component model that can describe the non-ideal characteristics of the components. S302 cascades the filter circuit and the amplifier circuit in sequence to form a complete filter amplifier circuit model; S303 configures simulation parameters according to the actual needs of the model, including step size and solver type.

7. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 1, characterized in that, The process of performing complete set empirical mode decomposition on the third digital signal to obtain the fourth digital signal includes: S401 decomposes the third digital signal using a complete empirical mode decomposition algorithm according to an adaptively set number of decomposition modes to obtain several intrinsic mode functions (IMFs), and averages the IMFs with the same index to improve decomposition stability. S402 identifies and analyzes Intrinsic Mode Functions (IMFs) based on an adaptively set decomposition correlation coefficient threshold. IMFs with correlation coefficients higher than the threshold are identified as signal components, while those with correlation coefficients lower than the threshold are identified as noise components. The S403 removes noise components, reconstructs signal components, and outputs a fourth digital signal.

8. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 1, characterized in that, The calculation of the similarity index between the fourth digital signal and the first digital signal, and the iterative optimization of the structural parameters of the simulation model based on the similarity index to update the structural parameters, includes: S501 calculates the similarity between the fourth digital signal and the first digital signal using dynamic time warping, frequency domain KL divergence, and energy fidelity. If all indicators reach the preset threshold, the simulation result is judged to be excellent; otherwise, it is recorded as substandard and the parameters of the complete empirical mode decomposition algorithm are adaptively adjusted. If, after several adaptive adjustments to the parameters of the complete empirical mode decomposition algorithm, the parameters of the S503 filter amplifier circuit model still fail to reach the preset threshold, the simulation parameters of the filter amplifier circuit model will be adaptively adjusted.

9. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 8, characterized in that, If all indicators reach the preset threshold, the simulation result is judged to be excellent; otherwise, it is recorded as substandard. The adaptive adjustment of the parameters of the complete empirical mode decomposition algorithm includes: If the dynamic time warping distance error exceeds its preset threshold of 20%, the S5021 will automatically increase the number of decomposed modes until the maximum limit is reached. If the energy fidelity is lower than its preset threshold, S5022 automatically lowers the decomposition correlation coefficient threshold of the mode function and relaxes the mode function retention standard until the minimum limit is reached; If the frequency domain KL divergence of S5023 exceeds its preset threshold, it will automatically increase the number of decomposed modes until the maximum limit is reached.

10. The Simulink-based partial discharge signal processing link simulation optimization method according to claim 8, characterized in that, If, after adaptively adjusting the parameters of the complete empirical mode decomposition algorithm several times, the index still fails to reach the preset threshold, the simulation parameters of the filter amplifier circuit model will be adaptively adjusted as follows: If the energy fidelity does not reach the preset threshold, S5031 will prioritize switching the solver type; If the energy fidelity meets the standard but the dynamic time warping distance exceeds the preset threshold, the S5032 will prioritize reducing the simulation step size. S5033 If the energy fidelity and dynamic time warp distance both meet the standards, but the frequency domain KL divergence exceeds the preset threshold, the solver type will be switched first. If multiple indicators of S5034 exceed the limit at the same time, the simulation parameters shall be adjusted in the order of priority: energy fidelity, dynamic time warping distance, and frequency domain KL divergence. After each round of parameter adjustments, the S5035 re-evaluates three indicators in real time until all indicators meet the standards or the simulation parameters reach their limits. The simulation parameters include step size and solver type.

Images