Hybrid vehicle testing method and system

Abstract

The application provides a hybrid vehicle testing method and system, which comprises the following steps: collecting power signal samples generated by a hybrid system under different torque ratios in a full speed range, and extracting signal distortion characteristics generated in the power signal samples through wavelet transform; establishing a dynamic correlation model between torque fluctuation and signal distortion characteristics of the hybrid system, and storing standard signal characteristics generated under different speeds; after starting the test, calculating torque ratio fluctuation values of the engine and the battery pack, and generating distortion correction parameters adaptive to the torque ratio fluctuation values in real time through the dynamic correlation model; performing real-time correction processing on original power output signals generated by the hybrid system through the distortion correction parameters to generate corresponding corrected power signals, and completing power testing of the hybrid system according to the corrected power signals. The application can effectively improve the testing efficiency.

Description

Technical Field

[0001] This invention relates to the field of automotive technology, and in particular to a hybrid vehicle testing method and system. Background Technology

[0002] With the advancement of technology and the rapid development of productivity, automobiles have become widespread in people's daily lives and have become an indispensable means of transportation, greatly facilitating people's lives.

[0003] Hybrid vehicles, which combine battery packs and engines, have also developed rapidly. Due to their excellent driving performance, they have gradually gained recognition. Specifically, existing hybrid systems need to undergo corresponding performance tests before leaving the factory, and only those that pass the tests will be used in vehicle production.

[0004] Furthermore, most existing technologies use threshold segmentation calibration to test hybrid systems. However, while this testing method considers the diagnostic needs of the vehicle under different speed conditions, it ignores the influence of real-time torque on the power signal, which may cause distortion of the power signal output by the hybrid system, resulting in certain test errors and reducing the testing efficiency of the hybrid system. Summary of the Invention

[0005] Based on this, the purpose of this invention is to provide a hybrid vehicle testing method and system to solve the problem that the prior art ignores the influence of real-time torque on the power signal, which may cause distortion of the power signal output by the hybrid system, resulting in certain testing errors and a corresponding reduction in the testing efficiency of the hybrid system.

[0006] The first aspect of the present invention proposes:

[0007] A method for testing hybrid vehicles, wherein the method includes:

[0008] The system collects power signal samples generated under different torque ratios across the entire speed range, and extracts the corresponding signal distortion features from the power signal samples using wavelet transform.

[0009] Establish a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics, and store the standard signal characteristics generated at each speed.

[0010] After the test is started, the torque ratio fluctuation value between the engine and the battery pack is calculated, and distortion correction parameters adapted to the torque ratio fluctuation value are generated in real time through the dynamic correlation model.

[0011] The distortion correction parameters are used to correct the original power output signal generated by the hybrid system in real time to generate a corresponding corrected power signal, and the power test of the hybrid system is completed according to the corrected power signal.

[0012] The beneficial effects of this invention are: by collecting power signal samples of the hybrid system in real time, it is possible to obtain the signal distortion characteristics that occur in the actual test of the hybrid system. Based on this, a corresponding dynamic correlation model can be created in real time, and the corresponding distortion correction parameters can be output in real time. Based on this, the required corrected power signal can be generated, thereby effectively avoiding power signal distortion and improving test efficiency.

[0013] Furthermore, the step of establishing a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics includes:

[0014] The power signal samples within the full speed range are divided into grids according to a preset interval to form several speed nodes, and the kurtosis values ​​generated between the signal distortion features are calculated.

[0015] The signal distortion features are converted into corresponding feature vectors based on the kurtosis value, and a bidirectional LSTM network model based on the attention mechanism is constructed. The input layer receives the feature vectors, the hidden layer contains 64 LSTM units, and the attention layer is set with 8 attention heads.

[0016] The bidirectional LSTM network model is trained using several rotation speed nodes to generate the dynamic correlation model.

[0017] Furthermore, the step of training the bidirectional LSTM network model through several rotation speed nodes to generate the dynamic correlation model includes:

[0018] Each rotation speed node is converted into a corresponding node vector, and frequency domain decomposition is performed on each node vector to extract the corresponding low-frequency and high-frequency components.

[0019] Based on the low-frequency component and the high-frequency component, corresponding low-frequency feature matrices and high-frequency feature matrices are constructed respectively. The low-frequency feature matrix is ​​input into the forward propagation unit of the bidirectional LSTM network model, and the high-frequency feature matrix is ​​input into the backward propagation unit. Feature fusion is performed through the hidden layer.

[0020] Using the standard signal features as a supervision benchmark, when the Euclidean distance between the predicted distortion features output by the LSTM network model and the standard signal features is less than a preset distance threshold, the dynamic correlation model is generated.

[0021] Furthermore, the step of generating distortion correction parameters in real time through the dynamic correlation model that are adapted to the torque ratio fluctuation value includes:

[0022] The time-domain waveform of the torque ratio fluctuation value is obtained and a three-level wavelet packet decomposition is performed to extract the energy entropy value of each decomposition layer and form a fluctuation feature sequence.

[0023] The wave feature sequence is convolved with the standard signal feature to generate a corresponding feature response map, and the feature response map is input into the dynamic correlation model for feature mapping.

[0024] The mapping results output by the dynamic correlation model are corrected in real time according to the direction of torque fluctuation, so as to generate the distortion correction parameters accordingly based on the correction results.

[0025] Furthermore, the step of real-time segmenting and correcting the mapping result output by the dynamic correlation model according to the torque fluctuation direction, so as to generate the distortion correction parameters according to the correction result, includes:

[0026] By monitoring the sign change points of the torque fluctuation direction, several oscillation intervals in the same direction are obtained by dividing the intervals by adjacent sign change points;

[0027] The mapping results within each of the aforementioned unidirectional fluctuation intervals are subjected to polynomial fitting to extract the inflection point coordinates and slope change rate of each fitted curve.

[0028] A correction baseline is established based on the inflection point coordinates of each unidirectional fluctuation interval. The correction coefficient of each sampling point within the interval is calculated in combination with the slope change rate. The distortion correction parameter is generated by multiplying the correction coefficient with the mapping result.

[0029] Furthermore, the step of performing real-time correction processing on the original power output signal generated by the hybrid system using the distortion correction parameters to generate a corresponding corrected power signal includes:

[0030] The original power output signal is mapped in the time and frequency domains to obtain a two-dimensional signal matrix containing time and frequency axes.

[0031] The distortion correction coefficient is converted into a correction matrix with the same dimension as the two-dimensional signal matrix, wherein the element values ​​of the correction matrix increase linearly with the increase of the torque ratio fluctuation value;

[0032] The two-dimensional signal matrix and the correction matrix are multiplied element-wise, and the result is converted into a time-domain signal to generate the corrected dynamic signal.

[0033] Furthermore, the step of converting the distortion correction coefficients into a correction matrix of the same dimension as the two-dimensional signal matrix includes:

[0034] Extract the frequency distribution peak values ​​of the standard signal features to determine several characteristic frequency ranges;

[0035] The distortion correction coefficients are divided into corresponding segmented parameters according to the characteristic frequency range;

[0036] Using the time and frequency dimensions of the two-dimensional signal matrix as axes, the segmentation parameters are filled into the matrix elements of the characteristic frequency range to form a correction matrix with the same dimension as the two-dimensional signal matrix.

[0037] The second aspect of the present invention proposes:

[0038] A hybrid vehicle testing system, wherein the system comprises:

[0039] The acquisition module is used to acquire power signal samples generated by the hybrid system under different torque ratios across the entire speed range, and extract the corresponding signal distortion features from the power signal samples through wavelet transform.

[0040] The storage module is used to establish a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics, and to store the standard signal characteristics generated at each speed.

[0041] The calculation module is used to calculate the torque ratio fluctuation value between the engine and the battery pack after the test is started, and to generate distortion correction parameters that are adapted to the torque ratio fluctuation value in real time through the dynamic correlation model.

[0042] The correction module is used to perform real-time correction processing on the original power output signal generated by the hybrid system through the distortion correction parameters, so as to generate a corresponding corrected power signal, and to complete the power test of the hybrid system according to the corrected power signal.

[0043] Furthermore, the storage module is specifically used for:

[0044] The power signal samples within the full speed range are divided into grids according to a preset interval to form several speed nodes, and the kurtosis values ​​generated between the signal distortion features are calculated.

[0045] The signal distortion features are converted into corresponding feature vectors based on the kurtosis value, and a bidirectional LSTM network model based on the attention mechanism is constructed. The input layer receives the feature vectors, the hidden layer contains 64 LSTM units, and the attention layer is set with 8 attention heads.

[0046] The bidirectional LSTM network model is trained using several rotation speed nodes to generate the dynamic correlation model.

[0047] Furthermore, the storage module is specifically used for:

[0048] Each rotation speed node is converted into a corresponding node vector, and frequency domain decomposition is performed on each node vector to extract the corresponding low-frequency and high-frequency components.

[0049] Based on the low-frequency component and the high-frequency component, corresponding low-frequency feature matrices and high-frequency feature matrices are constructed respectively. The low-frequency feature matrix is ​​input into the forward propagation unit of the bidirectional LSTM network model, and the high-frequency feature matrix is ​​input into the backward propagation unit. Feature fusion is performed through the hidden layer.

[0050] Using the standard signal features as a supervision benchmark, when the Euclidean distance between the predicted distortion features output by the LSTM network model and the standard signal features is less than a preset distance threshold, the dynamic correlation model is generated.

[0051] Furthermore, the correction module is specifically used for:

[0052] The time-domain waveform of the torque ratio fluctuation value is obtained and a three-level wavelet packet decomposition is performed to extract the energy entropy value of each decomposition layer and form a fluctuation feature sequence.

[0053] The wave feature sequence is convolved with the standard signal feature to generate a corresponding feature response map, and the feature response map is input into the dynamic correlation model for feature mapping.

[0054] The mapping results output by the dynamic correlation model are corrected in real time according to the direction of torque fluctuation, so as to generate the distortion correction parameters accordingly based on the correction results.

[0055] Furthermore, the correction module is specifically used for:

[0056] By monitoring the sign change points of the torque fluctuation direction, several oscillation intervals in the same direction are obtained by dividing the intervals by adjacent sign change points;

[0057] The mapping results within each of the aforementioned unidirectional fluctuation intervals are subjected to polynomial fitting to extract the inflection point coordinates and slope change rate of each fitted curve.

[0058] A correction baseline is established based on the inflection point coordinates of each unidirectional fluctuation interval. The correction coefficient of each sampling point within the interval is calculated in combination with the slope change rate. The distortion correction parameter is generated by multiplying the correction coefficient with the mapping result.

[0059] Furthermore, the correction module is specifically used for:

[0060] The original power output signal is mapped in the time and frequency domains to obtain a two-dimensional signal matrix containing time and frequency axes.

[0061] The distortion correction coefficient is converted into a correction matrix with the same dimension as the two-dimensional signal matrix, wherein the element values ​​of the correction matrix increase linearly with the increase of the torque ratio fluctuation value;

[0062] The two-dimensional signal matrix and the correction matrix are multiplied element-wise, and the result is converted into a time-domain signal to generate the corrected dynamic signal.

[0063] Furthermore, the correction module is specifically used for:

[0064] Extract the frequency distribution peak values ​​of the standard signal features to determine several characteristic frequency ranges;

[0065] The distortion correction coefficients are divided into corresponding segmented parameters according to the characteristic frequency range;

[0066] Using the time and frequency dimensions of the two-dimensional signal matrix as axes, the segmentation parameters are filled into the matrix elements of the characteristic frequency range to form a correction matrix with the same dimension as the two-dimensional signal matrix.

[0067] The third aspect of the present invention proposes:

[0068] A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the hybrid vehicle testing method as described above.

[0069] The fourth aspect of the present invention proposes:

[0070] A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the hybrid vehicle testing method as described above.

[0071] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0072] Figure 1 A flowchart of the hybrid vehicle testing method provided in the first embodiment of the present invention;

[0073] Figure 2 This is a structural block diagram of a hybrid vehicle testing system provided in the third embodiment of the present invention.

[0074] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0075] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0076] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0077] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0078] Please see Figure 1 The figure shows a hybrid vehicle testing method provided in the first embodiment of the present invention. The hybrid vehicle testing method provided in this embodiment can avoid distorted power signals during the testing process, thereby improving the testing efficiency of hybrid vehicles.

[0079] Specifically, this embodiment provides:

[0080] A hybrid vehicle testing method specifically includes the following steps:

[0081] Step S10: Collect power signal samples generated by the hybrid system under different torque ratios across the entire speed range, and extract the corresponding signal distortion features from the power signal samples using wavelet transform;

[0082] It should be noted that the system first collects power signal samples (such as time-domain signals like speed, torque, and power) generated by the hybrid system across the entire speed range (e.g., from idle to maximum speed) under different engine and motor torque ratios (e.g., engine 30% + motor 70%). Then, wavelet transform (a time-frequency analysis method) is used to decompose the signals and extract signal distortion features. Specifically, these features include signal abrupt changes, harmonic interference, and abnormal amplitude fluctuations, reflecting the stability of the powertrain under different operating conditions. This step provides the foundational data for establishing the subsequent correlation model, facilitating subsequent processing.

[0083] Step S20: Establish a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics, and store the standard signal characteristics generated at each speed.

[0084] It's important to note that, based on the collected signal distortion characteristics, the system constructs a dynamic correlation model between torque fluctuations (the instantaneous deviation between engine and motor torque output) and signal distortion characteristics, quantifying the mapping relationship between the two (e.g., when torque fluctuations increase by 10%, signal distortion in a certain frequency band increases by 20%). Simultaneously, it stores standard signal characteristics at each speed (i.e., signal parameters without distortion under ideal operating conditions) as a benchmark for subsequent testing. This step achieves mathematical modeling from signal characteristics to system state, providing theoretical support for real-time correction and facilitating subsequent processing.

[0085] Step S30: After starting the test, calculate the torque ratio fluctuation value between the engine and the battery pack, and generate distortion correction parameters that are adapted to the torque ratio fluctuation value in real time through the dynamic correlation model.

[0086] It's important to note that after starting the test, the torque output of the engine and battery pack (drive motor) is monitored in real time, and the fluctuation value of their ratio (such as the difference between the actual ratio and the target ratio) is calculated. Through a dynamic correlation model, this fluctuation value is converted into corresponding distortion correction parameters. Specifically, these parameters are compensation coefficients used to offset signal distortion (such as amplitude correction and phase adjustment values), and their magnitudes are positively correlated with the degree of torque fluctuation. This step enables a rapid transition from system state to correction strategy, facilitating subsequent processing.

[0087] Step S40: The original power output signal generated by the hybrid system is corrected in real time using the distortion correction parameters to generate a corresponding corrected power signal, and the power test of the hybrid system is completed according to the corrected power signal.

[0088] It should be noted that the system applies distortion correction parameters to the original power output signal (the measured signal containing distortion interference) generated by the hybrid system. Through time-frequency domain adjustment, it eliminates or reduces distortion components, generating a corrected power signal (a signal close to the ideal state). Based on the corrected signal, accurate testing of the hybrid system's power performance (such as acceleration response, power smoothness, and energy efficiency ratio) is performed. This step eliminates testing errors through signal correction, ensuring the reliability of the test results and facilitating subsequent processing.

[0089] Second Embodiment

[0090] Furthermore, the step of establishing a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics includes:

[0091] The power signal samples within the full speed range are divided into grids according to a preset interval to form several speed nodes, and the kurtosis values ​​generated between the signal distortion features are calculated.

[0092] The signal distortion features are converted into corresponding feature vectors based on the kurtosis value, and a bidirectional LSTM network model based on the attention mechanism is constructed. The input layer receives the feature vectors, the hidden layer contains 64 LSTM units, and the attention layer is set with 8 attention heads.

[0093] The bidirectional LSTM network model is trained using several rotation speed nodes to generate the dynamic correlation model.

[0094] It should be noted that the system divides the entire speed range into several speed nodes at preset intervals (e.g., every 500 rpm), discretizing the continuous speed range into quantifiable analysis units. Simultaneously, it calculates the kurtosis values ​​between various signal distortion features. Specifically, kurtosis is a statistical measure of the steepness of the signal probability distribution; a larger value indicates that the signal contains more extreme values ​​(such as sudden distortions), which can be used to screen key features sensitive to torque fluctuations. Based on this, signal distortion features are converted into high-dimensional feature vectors (each vector representing a quantized value of a set of distortion features) using kurtosis values. A bidirectional Long Short-Term Memory (LSTM) network model based on an attention mechanism is then constructed. The input layer receives the feature vectors and converts them into tensors recognizable by the network. The hidden layer contains 64 LSTM units to capture the temporal dependencies of the feature vectors (e.g., the impact of distortion at a certain rotational speed on the next rotational speed). The attention layer uses eight attention heads to focus on the features that have the greatest impact on torque fluctuations (e.g., high-frequency vibration signals) through weighted calculations. Finally, using the feature vectors of each rotational speed node as training samples, the model iterates through forward propagation to predict the output and backpropagation to adjust the network parameters until convergence (e.g., prediction error below 5%). After training, the network model can output the corresponding signal distortion features in real time based on the input torque fluctuation, thus forming a "dynamic correlation model." This step achieves a precise mapping between torque fluctuations and signal distortion through a data-driven approach, facilitating subsequent processing.

[0095] Furthermore, the step of training the bidirectional LSTM network model through several rotation speed nodes to generate the dynamic correlation model includes:

[0096] Each rotation speed node is converted into a corresponding node vector, and frequency domain decomposition is performed on each node vector to extract the corresponding low-frequency and high-frequency components.

[0097] Based on the low-frequency component and the high-frequency component, corresponding low-frequency feature matrices and high-frequency feature matrices are constructed respectively. The low-frequency feature matrix is ​​input into the forward propagation unit of the bidirectional LSTM network model, and the high-frequency feature matrix is ​​input into the backward propagation unit. Feature fusion is performed through the hidden layer.

[0098] Using the standard signal features as a supervision benchmark, when the Euclidean distance between the predicted distortion features output by the LSTM network model and the standard signal features is less than a preset distance threshold, the dynamic correlation model is generated.

[0099] It should be noted that the feature vector of each speed node is converted into a corresponding node vector (integrating torque fluctuation and signal distortion information at that speed), and then frequency domain decomposition is performed through Fourier transform to separate low-frequency components (e.g., <10Hz, corresponding to engine cylinder vibration) and high-frequency components (e.g., >100Hz, corresponding to motor electromagnetic noise). This decomposition can specifically capture signal features in different frequency bands, avoiding high-frequency interference masking low-frequency trends. The low-frequency feature matrix (the set of low-frequency components for each speed node) is input into the forward propagation unit of the bidirectional LSTM to capture the feature change pattern from low speed to high speed (e.g., the attenuation trend of low-frequency vibration as speed increases); the high-frequency feature matrix (the set of high-frequency components for each speed node) is input into the backward propagation unit to capture the feature change pattern from high speed to low speed (e.g., the abrupt change characteristics of high-frequency noise in the high-speed range); the hidden layer fuses the output features of the bidirectional propagation through matrix operations to form a comprehensive feature containing time-series information across the entire speed range. Based on this, the stored standard signal features are used as a supervision benchmark, and the Euclidean distance (measuring the spatial difference between the two) between the predicted distortion features output by the model and the standard features is calculated. When the distance is less than a preset threshold (e.g., 0.01), it indicates that the model can accurately predict signal distortion. At this point, training stops, and the current model is designated as a "dynamic correlation model." This step uses supervised learning to ensure that the model output meets ideal operating conditions, facilitating subsequent processing.

[0100] Furthermore, the step of generating distortion correction parameters in real time through the dynamic correlation model that are adapted to the torque ratio fluctuation value includes:

[0101] The time-domain waveform of the torque ratio fluctuation value is obtained and a three-level wavelet packet decomposition is performed to extract the energy entropy value of each decomposition layer and form a fluctuation feature sequence.

[0102] The wave feature sequence is convolved with the standard signal feature to generate a corresponding feature response map, and the feature response map is input into the dynamic correlation model for feature mapping.

[0103] The mapping results output by the dynamic correlation model are corrected in real time according to the direction of torque fluctuation, so as to generate the distortion correction parameters accordingly based on the correction results.

[0104] It should be noted that the time-domain waveform of the torque ratio fluctuation value (such as a fluctuation curve changing over time) is obtained, and a three-level wavelet packet decomposition is performed (decomposing the signal into 8 frequency bands). The energy entropy value of each decomposition layer is calculated (measuring the degree of disorder in the signal energy distribution). These energy entropy values ​​form a fluctuation feature sequence, quantifying the frequency components and disorder of torque fluctuations (e.g., high entropy values ​​in the high-frequency band indicate severe fluctuations). Based on this, the fluctuation feature sequence is convolved with the standard signal features (weighted summation within a sliding window) to generate a feature response spectrum. Specifically, this spectrum visually shows the degree of matching between torque fluctuations and the standard signal (e.g., low response values ​​in a certain region indicate that the fluctuation in that frequency band deviates from the standard). The spectrum is input into a dynamic correlation model, and the model's feature mapping function (learned correspondence between torque fluctuations and distortion) outputs a mapping result matching the current fluctuation (theoretical distortion compensation). Finally, the mapping result is corrected in real time according to the direction of torque fluctuation (positive fluctuation / negative fluctuation): for example, positive fluctuation (actual torque is greater than the target value) corresponds to one set of correction coefficients, and negative fluctuation corresponds to another set of coefficients. By differentiating directions, the correction direction is ensured to be opposite to the distortion direction (e.g., applying an attenuation coefficient when the signal amplitude is too large), ultimately generating distortion correction parameters that can be directly applied. This step achieves dynamic adaptation of the correction parameters to the fluctuation characteristics, facilitating subsequent processing.

[0105] Furthermore, the step of real-time segmenting and correcting the mapping result output by the dynamic correlation model according to the torque fluctuation direction, so as to generate the distortion correction parameters according to the correction result, includes:

[0106] By monitoring the sign change points of the torque fluctuation direction, several oscillation intervals in the same direction are obtained by dividing the intervals by adjacent sign change points;

[0107] The mapping results within each of the aforementioned unidirectional fluctuation intervals are subjected to polynomial fitting to extract the inflection point coordinates and slope change rate of each fitted curve.

[0108] A correction baseline is established based on the inflection point coordinates of each unidirectional fluctuation interval. The correction coefficient of each sampling point within the interval is calculated in combination with the slope change rate. The distortion correction parameter is generated by multiplying the correction coefficient with the mapping result.

[0109] It should be noted that, by monitoring the signs of torque fluctuations in real time (e.g., the moment when they change from positive to negative), the fluctuation curve is divided into several unidirectional fluctuation intervals (e.g., continuous positive fluctuation segments, continuous negative fluctuation segments) using adjacent change points as boundaries. The fluctuation direction is consistent within each interval, facilitating the adoption of a unified correction strategy. Based on this, the mapping results (theoretical compensation amount) within each unidirectional fluctuation interval are subjected to polynomial fitting (e.g., cubic polynomial) to generate a smooth fitting curve. The inflection point coordinates (extreme points of the compensation amount's rate of change) and the slope change rate (reflecting the rate of change of the compensation amount over time) are extracted from the curve. These features reflect the dynamic changes in correction requirements within the interval (e.g., a large slope change rate within an interval indicates a need for rapid adjustment of the compensation amount). Finally, a correction baseline (e.g., a smooth curve passing through the inflection points) is established based on the inflection point coordinates of each interval. The correction coefficient for each sampling point within the interval is calculated using the slope change rate (the farther away from the baseline, the larger the coefficient). Multiplying the correction coefficient by the mapping result yields the final distortion correction parameter. Specifically, this parameter includes both the theoretical compensation predicted by the model and an interval dynamic adjustment factor, ensuring that the correction effect more closely matches the actual fluctuation characteristics, thus facilitating subsequent processing.

[0110] Furthermore, the step of performing real-time correction processing on the original power output signal generated by the hybrid system using the distortion correction parameters to generate a corresponding corrected power signal includes:

[0111] The original power output signal is mapped in the time and frequency domains to obtain a two-dimensional signal matrix containing time and frequency axes.

[0112] The distortion correction coefficient is converted into a correction matrix with the same dimension as the two-dimensional signal matrix, wherein the element values ​​of the correction matrix increase linearly with the increase of the torque ratio fluctuation value;

[0113] The two-dimensional signal matrix and the correction matrix are multiplied element-wise, and the result is converted into a time-domain signal to generate the corrected dynamic signal.

[0114] It should be noted that the original power signal (time-domain waveform) output by the hybrid system is converted into a two-dimensional signal matrix through a short-time Fourier transform. Specifically, the horizontal axis of the matrix represents time, the vertical axis represents frequency, and the element values ​​are the signal amplitudes at the corresponding time and frequency. This mapping can simultaneously reflect the signal's distribution characteristics in time and frequency (such as the high-frequency noise peak at a certain moment), providing a basis for targeted correction. Based on this, the distortion correction parameters are converted into a correction matrix with the same size as the two-dimensional signal matrix: the value of each element in the matrix is ​​positively correlated with the torque ratio fluctuation value at the corresponding position (e.g., the correction coefficient is larger in areas with large fluctuation values). For example, if the torque fluctuation in the high-frequency band is 5% at a certain moment, the corresponding correction matrix element value may be 1.05 (amplifying the signal) or 0.95 (attenuating the signal), depending on the distortion direction. Finally, the two-dimensional signal matrix and the correction matrix are multiplied element-wise (the signal amplitude at each position is multiplied by the corresponding correction coefficient) to achieve precise correction in the time and frequency domains (e.g., suppressing the amplitude of high-frequency noise regions and enhancing the effective low-frequency signal). After the calculation is complete, the corrected matrix is ​​converted back to a time-domain signal using an inverse short-time Fourier transform, thus generating the corrected dynamic signal. This step ensures that the correction process eliminates distortion while preserving the effective dynamic signal components, facilitating subsequent processing.

[0115] Furthermore, the step of converting the distortion correction coefficients into a correction matrix of the same dimension as the two-dimensional signal matrix includes:

[0116] Extract the frequency distribution peak values ​​of the standard signal features to determine several characteristic frequency ranges;

[0117] The distortion correction coefficients are divided into corresponding segmented parameters according to the characteristic frequency range;

[0118] Using the time and frequency dimensions of the two-dimensional signal matrix as axes, the segmentation parameters are filled into the matrix elements of the characteristic frequency range to form a correction matrix with the same dimension as the two-dimensional signal matrix.

[0119] It's important to note that the frequency distribution curve of the standard signal characteristics is analyzed, and the peak frequencies are extracted (e.g., 10Hz for engine idling and 500Hz for motor operation). These peaks are then used as the center to divide the signal into several characteristic frequency ranges (e.g., 5-15Hz, 450-550Hz). These ranges are the critical frequency bands for the power signal and have the greatest impact on the hybrid system's performance (e.g., low frequencies affect smoothness, while high frequencies affect noise levels). Based on this, the distortion correction parameters (overall correction strategy) are decomposed into segmented parameters according to the characteristic frequency ranges. Specifically, each range corresponds to a set of independent correction coefficients (e.g., small-amplitude linear correction for low-frequency ranges and large-amplitude nonlinear correction for high-frequency ranges). This decomposition allows for customized correction strategies based on the signal characteristics of different frequency bands, avoiding over-correction or under-correction caused by a "one-size-fits-all" approach. Finally, using the time axis (horizontal axis) and frequency axis (vertical axis) of the two-dimensional signal matrix as coordinates, the segmented parameters for each characteristic frequency interval are filled into the corresponding positions in the matrix: for example, the low-frequency segmented parameters are filled into the low-frequency region of the matrix (below the vertical axis), and the high-frequency parameters are filled into the high-frequency region (above the vertical axis). Non-characteristic frequency intervals can use the default correction coefficient (e.g., 1.0, no correction). This results in a correction matrix of the same dimension as the original signal matrix, ensuring spatial consistency of the correction operation for subsequent processing.

[0120] Please see Figure 2 The third embodiment of the present invention provides:

[0121] A hybrid vehicle testing system, wherein the system comprises:

[0122] The acquisition module is used to acquire power signal samples generated by the hybrid system under different torque ratios across the entire speed range, and extract the corresponding signal distortion features from the power signal samples through wavelet transform.

[0123] The storage module is used to establish a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics, and to store the standard signal characteristics generated at each speed.

[0124] The calculation module is used to calculate the torque ratio fluctuation value between the engine and the battery pack after the test is started, and to generate distortion correction parameters that are adapted to the torque ratio fluctuation value in real time through the dynamic correlation model.

[0125] The correction module is used to perform real-time correction processing on the original power output signal generated by the hybrid system through the distortion correction parameters, so as to generate a corresponding corrected power signal, and to complete the power test of the hybrid system according to the corrected power signal.

[0126] Furthermore, the storage module is specifically used for:

[0127] The power signal samples within the full speed range are divided into grids according to a preset interval to form several speed nodes, and the kurtosis values ​​generated between the signal distortion features are calculated.

[0128] The signal distortion features are converted into corresponding feature vectors based on the kurtosis value, and a bidirectional LSTM network model based on the attention mechanism is constructed. The input layer receives the feature vectors, the hidden layer contains 64 LSTM units, and the attention layer is set with 8 attention heads.

[0129] The bidirectional LSTM network model is trained using several rotation speed nodes to generate the dynamic correlation model.

[0130] Furthermore, the storage module is specifically used for:

[0131] Each rotation speed node is converted into a corresponding node vector, and frequency domain decomposition is performed on each node vector to extract the corresponding low-frequency and high-frequency components.

[0132] Based on the low-frequency component and the high-frequency component, corresponding low-frequency feature matrices and high-frequency feature matrices are constructed respectively. The low-frequency feature matrix is ​​input into the forward propagation unit of the bidirectional LSTM network model, and the high-frequency feature matrix is ​​input into the backward propagation unit. Feature fusion is performed through the hidden layer.

[0133] Using the standard signal features as a supervision benchmark, when the Euclidean distance between the predicted distortion features output by the LSTM network model and the standard signal features is less than a preset distance threshold, the dynamic correlation model is generated.

[0134] Furthermore, the correction module is specifically used for:

[0135] The time-domain waveform of the torque ratio fluctuation value is obtained and a three-level wavelet packet decomposition is performed to extract the energy entropy value of each decomposition layer and form a fluctuation feature sequence.

[0136] The wave feature sequence is convolved with the standard signal feature to generate a corresponding feature response map, and the feature response map is input into the dynamic correlation model for feature mapping.

[0137] The mapping results output by the dynamic correlation model are corrected in real time according to the direction of torque fluctuation, so as to generate the distortion correction parameters accordingly based on the correction results.

[0138] Furthermore, the correction module is specifically used for:

[0139] By monitoring the sign change points of the torque fluctuation direction, several oscillation intervals in the same direction are obtained by dividing the intervals by adjacent sign change points;

[0140] The mapping results within each of the aforementioned unidirectional fluctuation intervals are subjected to polynomial fitting to extract the inflection point coordinates and slope change rate of each fitted curve.

[0141] A correction baseline is established based on the inflection point coordinates of each unidirectional fluctuation interval. The correction coefficient of each sampling point within the interval is calculated in combination with the slope change rate. The distortion correction parameter is generated by multiplying the correction coefficient with the mapping result.

[0142] Furthermore, the correction module is specifically used for:

[0143] The original power output signal is mapped in the time and frequency domains to obtain a two-dimensional signal matrix containing time and frequency axes.

[0144] The distortion correction coefficient is converted into a correction matrix with the same dimension as the two-dimensional signal matrix, wherein the element values ​​of the correction matrix increase linearly with the increase of the torque ratio fluctuation value;

[0145] The two-dimensional signal matrix and the correction matrix are multiplied element-wise, and the result is converted into a time-domain signal to generate the corrected dynamic signal.

[0146] Furthermore, the correction module is specifically used for:

[0147] Extract the frequency distribution peak values ​​of the standard signal features to determine several characteristic frequency ranges;

[0148] The distortion correction coefficients are divided into corresponding segmented parameters according to the characteristic frequency range;

[0149] Using the time and frequency dimensions of the two-dimensional signal matrix as axes, the segmentation parameters are filled into the matrix elements of the characteristic frequency range to form a correction matrix with the same dimension as the two-dimensional signal matrix.

[0150] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the hybrid vehicle testing method as described above.

[0151] The fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the hybrid vehicle testing method as described above.

[0152] In summary, the hybrid vehicle testing method and system provided in the above embodiments of the present invention can avoid distorted power signals during the testing process of hybrid vehicles, thereby improving the testing efficiency of hybrid vehicle power performance.

[0153] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0154] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0155] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0156] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0157] In the description of this specification, 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 invention. In this specification, the illustrative expressions of the above terms do not necessarily refer 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.

[0158] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for testing hybrid vehicles, characterized in that, The method includes: The system collects power signal samples generated under different torque ratios across the entire speed range, and extracts the corresponding signal distortion features from the power signal samples using wavelet transform. Establish a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics, and store the standard signal characteristics generated at each speed. After the test is started, the torque ratio fluctuation value between the engine and the battery pack is calculated, and distortion correction parameters adapted to the torque ratio fluctuation value are generated in real time through the dynamic correlation model. The distortion correction parameters are used to correct the original power output signal generated by the hybrid system in real time to generate a corresponding corrected power signal, and the power test of the hybrid system is completed according to the corrected power signal. The step of performing real-time correction processing on the original power output signal generated by the hybrid system using the distortion correction parameters to generate a corresponding corrected power signal includes: The original power output signal is mapped in the time and frequency domains to obtain a two-dimensional signal matrix containing time and frequency axes. The distortion correction parameters are converted into a correction matrix with the same dimension as the two-dimensional signal matrix, wherein the element values ​​of the correction matrix increase linearly with the increase of the torque ratio fluctuation value. The two-dimensional signal matrix and the correction matrix are multiplied element-wise, and the result is converted into a time-domain signal to generate the corrected dynamic signal.

2. The hybrid vehicle testing method according to claim 1, characterized in that, The steps for establishing a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics include: The power signal samples within the full speed range are divided into grids according to a preset interval to form several speed nodes, and the kurtosis values ​​generated between the signal distortion features are calculated. The signal distortion features are converted into corresponding feature vectors based on the kurtosis value, and a bidirectional LSTM network model based on the attention mechanism is constructed. The input layer receives the feature vectors, the hidden layer contains 64 LSTM units, and the attention layer is set with 8 attention heads. The bidirectional LSTM network model is trained using several rotation speed nodes to generate the dynamic correlation model.

3. The hybrid vehicle testing method according to claim 2, characterized in that, The step of training the bidirectional LSTM network model through several rotation speed nodes to generate the dynamic correlation model includes: Each rotation speed node is converted into a corresponding node vector, and frequency domain decomposition is performed on each node vector to extract the corresponding low-frequency and high-frequency components. Based on the low-frequency component and the high-frequency component, corresponding low-frequency feature matrices and high-frequency feature matrices are constructed respectively. The low-frequency feature matrix is ​​input into the forward propagation unit of the bidirectional LSTM network model, and the high-frequency feature matrix is ​​input into the backward propagation unit. Feature fusion is performed through the hidden layer. Using the standard signal features as a supervision benchmark, when the Euclidean distance between the predicted distortion features output by the bidirectional LSTM network model and the standard signal features is less than a preset distance threshold, the dynamic correlation model is generated.

4. The hybrid vehicle testing method according to claim 1, characterized in that, The step of generating distortion correction parameters that are adapted to the torque ratio fluctuation value in real time through the dynamic correlation model includes: The time-domain waveform of the torque ratio fluctuation value is obtained and a three-level wavelet packet decomposition is performed to extract the energy entropy value of each decomposition layer and form a fluctuation feature sequence. The wave feature sequence is convolved with the standard signal feature to generate a corresponding feature response map, and the feature response map is input into the dynamic correlation model for feature mapping. The mapping results output by the dynamic correlation model are corrected in real time according to the direction of torque fluctuation, so as to generate the distortion correction parameters accordingly based on the correction results.

5. The hybrid vehicle testing method according to claim 4, characterized in that, The step of correcting the mapping result output by the dynamic correlation model in real time according to the torque fluctuation direction, and generating the distortion correction parameters according to the correction result, includes: By monitoring the sign change points of the torque fluctuation direction, several oscillation intervals in the same direction are obtained by dividing the intervals by adjacent sign change points; The mapping results within each of the aforementioned unidirectional fluctuation intervals are subjected to polynomial fitting to extract the inflection point coordinates and slope change rate of each fitted curve. A correction baseline is established based on the inflection point coordinates of each unidirectional fluctuation interval. The correction coefficient of each sampling point within the interval is calculated in combination with the slope change rate. The distortion correction parameter is generated by multiplying the correction coefficient with the mapping result.

6. The hybrid vehicle testing method according to claim 1, characterized in that, The step of converting the distortion correction parameters into a correction matrix of the same dimension as the two-dimensional signal matrix includes: Extract the frequency distribution peak values ​​of the standard signal features to determine several characteristic frequency ranges; The distortion correction parameters are divided into corresponding segmented parameters according to the characteristic frequency range; Using the time and frequency dimensions of the two-dimensional signal matrix as axes, the segmentation parameters are filled into the matrix elements of the characteristic frequency range to form a correction matrix with the same dimension as the two-dimensional signal matrix.

7. A hybrid vehicle testing system, characterized in that, The system for implementing the hybrid vehicle testing method as described in any one of claims 1 to 6 includes: The acquisition module is used to acquire power signal samples generated by the hybrid system under different torque ratios across the entire speed range, and extract the corresponding signal distortion features from the power signal samples through wavelet transform. The storage module is used to establish a dynamic correlation model between the torque fluctuation of the hybrid system and the signal distortion characteristics, and to store the standard signal characteristics generated at each speed. The calculation module is used to calculate the torque ratio fluctuation value between the engine and the battery pack after the test is started, and to generate distortion correction parameters that are adapted to the torque ratio fluctuation value in real time through the dynamic correlation model. The correction module is used to perform real-time correction processing on the original power output signal generated by the hybrid system through the distortion correction parameters, so as to generate a corresponding corrected power signal, and to complete the power test of the hybrid system according to the corrected power signal.

8. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the hybrid vehicle testing method as described in any one of claims 1 to 6.

9. A readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the hybrid vehicle testing method as described in any one of claims 1 to 6.

Images