Method, device and equipment for measuring internal resistance of storage battery based on improved characteristic mode decomposition

Abstract

The invention discloses a storage battery internal resistance measurement method, device and equipment for improving characteristic mode decomposition, and relates to the technical field of storage battery internal resistance measurement. The storage battery internal resistance measuring method comprises the following steps: performing pulse direct current discharge on a storage battery to obtain voltage and current sequence data of the storage battery; and according to the voltage and current sequence data, searching optimal parameters of characteristic mode decomposition in a parameter space through a Newton-quasi-Newton hybrid optimization algorithm. And according to the optimal parameters, carrying out dynamic frequency band scaling, correlation merging and zero-phase filtering to obtain a denoising mode and a reconstruction signal. And according to the de-noising mode and the reconstruction signal, carrying out feature evaluation and fusion screening through the spectrum peak matching degree, the square envelope spectrum negentropy and the maximum cross correlation, and obtaining de-noised voltage and current sequence data. And acquiring a voltage difference value and a current difference value according to the denoised voltage and current sequence data. And obtaining the internal resistance of the storage battery according to the voltage difference and the current difference.

Description

Technical Field

[0001] This invention relates to the field of battery internal resistance measurement technology, and more specifically, to an improved characteristic mode decomposition method, apparatus, and device for measuring battery internal resistance. Background Technology

[0002] Batteries play a crucial role in energy storage, serving critical functions from grid peak shaving and electric vehicle backup to communication backup. Their health directly determines the reliability and security of the entire system's power supply. Operational experience shows that internal resistance values ​​can sensitively reflect battery aging, capacity decay, and potential faults. Therefore, the industry generally regards online internal resistance monitoring as a core parameter that battery management systems must continuously acquire.

[0003] Currently, DC pulse discharge and AC injection methods are commonly used in engineering to extract internal resistance. The DC pulse discharge method captures voltage steps through instantaneous constant current discharge and calculates the internal resistance using Ohm's law. The AC injection method couples a small-amplitude, high-frequency current to the battery and obtains the impedance using synchronous detection. To suppress electromagnetic, polarization, and harmonic interference in the field, subsequent signal processing typically introduces empirical mode decomposition, variational mode decomposition, or wavelet thresholding for denoising. The sampled sequence is then filtered, reconstructed, and the high-low level difference is calculated.

[0004] The above methods exhibit significant drawbacks in field applications: the voltage and current waveforms obtained by the DC pulse discharge method show quasi-square wave characteristics, with severe overlap between effective harmonics and broadband noise in the spectrum; traditional mode decomposition algorithms based on the periodic impact assumption cannot align transient edges, resulting in weakened effective amplitude. The AC injection method is limited by frequency selection, has complex measurement circuits, and its weak response signals are easily submerged by environmental electromagnetic noise. Empirical mode decomposition suffers from mode aliasing and endpoint effects, while variational mode decomposition and wavelet transform are highly sensitive to the number of modes, penalty factors, or basis functions. When parameters are mismatched, residual noise is directly converted into internal resistance calculation errors, causing the repeated measurement results of the same battery to be dispersed, with the maximum absolute error exceeding 0.01 mΩ, making it difficult to meet the repeatability accuracy requirements of high-rate energy storage systems below 0.3%. Summary of the Invention

[0005] The present invention provides an improved method, apparatus, device and medium for measuring the internal resistance of a battery by characteristic mode decomposition, in order to improve at least one of the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides an improved method for measuring the internal resistance of a storage battery based on characteristic mode decomposition, comprising S1 to S6.

[0007] S1. Perform pulsed DC discharge on the battery to obtain the battery's voltage and current sequence data.

[0008] S2. Based on the voltage and current sequence data, the optimal parameters for eigenmode decomposition are searched in the parameter space using a Newton-quasi-Newton hybrid optimization algorithm.

[0009] S3. Based on the optimal parameters, perform dynamic bandwidth scaling, correlation merging, and zero-phase filtering to obtain the denoised mode and reconstructed signal.

[0010] S4. Based on the denoised mode and reconstructed signal, feature evaluation and fusion screening are performed by peak matching degree, square envelope spectrum negative entropy and maximum cross-correlation to obtain denoised voltage and current sequence data.

[0011] S5. Based on the denoised voltage and current sequence data, obtain the voltage difference and current difference.

[0012] S6. Obtain the internal resistance of the battery based on the voltage difference and the current difference.

[0013] Secondly, the present invention provides an improved characteristic mode decomposition battery internal resistance measurement device, which includes a raw data module, a parameter search module, a signal reconstruction module, a noise reduction module, a voltage and current module, and an internal resistance module.

[0014] The raw data module is used to perform pulsed DC discharge on the battery to obtain voltage and current sequence data of the battery.

[0015] The parameter search module is used to search for the optimal parameters of the characteristic mode decomposition in the parameter space based on the voltage and current sequence data using a Newton-quasi-Newton hybrid optimization algorithm.

[0016] The signal reconstruction module is used to perform dynamic bandwidth scaling, correlation merging, and zero-phase filtering based on the optimal parameters to obtain the denoised mode and the reconstructed signal.

[0017] The denoising module is used to evaluate and fuse features based on the denoised mode and reconstructed signal through spectral peak matching degree, squared envelope spectrum negative entropy and maximum cross-correlation, to obtain denoised voltage and current sequence data.

[0018] The voltage and current module is used to obtain the voltage difference and current difference based on the denoised voltage and current sequence data.

[0019] An internal resistance module is used to obtain the internal resistance of the battery based on the voltage difference and the current difference.

[0020] Thirdly, the present invention provides an improved characteristic mode decomposition (EMD) battery internal resistance measurement device, comprising a processor, a memory, and a computer program stored in the memory. The computer program can be executed by the processor to implement an improved EMD battery internal resistance measurement method as described in any paragraph of the first aspect.

[0021] By adopting the above technical solution, the present invention can achieve the following technical effects: The improved characteristic mode decomposition method for measuring battery internal resistance of the present invention can suppress noise and preserve the characteristics of voltage and current signals, so that the measured battery internal resistance can be accurately measured. Attached Figure Description

[0022] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the specific embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some specific embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is the voltage waveform diagram of the original voltage.

[0024] Figure 2 This is the waveform of the original current.

[0025] Figure 3 This represents the energy harmonic distribution of the original voltage.

[0026] Figure 4 This represents the energy harmonic distribution of the original current.

[0027] Figure 5 This represents the broadband noise characteristics of the original voltage.

[0028] Figure 6 This represents the broadband noise characteristics of the original current.

[0029] Figure 7 This is an equivalent model diagram for measuring the internal resistance of a storage battery.

[0030] Figure 8 The flowchart shows the parameter adaptive optimization process driven by the Newton-quasi-Newton hybrid optimization algorithm.

[0031] Figure 9 These are the voltage fluctuation signals before and after Newton-FMD denoising.

[0032] Figure 10 These are the current fluctuation signals before and after Newton-FMD denoising.

[0033] Figure 11 The voltage fluctuation signals are before and after denoising using four different methods.

[0034] Figure 12 The current fluctuation signals are before and after denoising using four different algorithms. Detailed Implementation

[0035] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention.

[0036] Example 1: The main purpose of this example is to address the limitations of current signal processing methods for battery internal resistance detection, such as parameter sensitivity and unsuitable objective function. A battery internal resistance measurement method based on Newton-FMD optimization is designed to overcome the inaccuracies in current technologies caused by noise. Newton-FMD is a signal processing method that combines the Newton-Raphson optimization algorithm (NRBO or Newton's method) with Feature Mode Decomposition (FMD).

[0037] Please see Figures 1 to 12 The first embodiment of the present invention provides a battery internal resistance measurement method with improved eigenmode decomposition, which can be executed by a battery internal resistance measurement device with improved eigenmode decomposition (hereinafter referred to as: battery internal resistance measurement device). In particular, it is executed by one or more processors in the battery internal resistance measurement device to implement steps S1 to S6.

[0038] S1. Perform pulsed DC discharge on the battery to obtain the battery's voltage and current sequence data.

[0039] Specifically, the main control module of the battery internal resistance measurement drive circuit controls the switching on to control the conduction of the MOSFET, thereby performing pulsed DC discharge on the battery. The voltage signal U and current signal I across the battery terminals are simultaneously acquired by the internal resistance measurement circuit. Then, the main control chip processes the voltage and current signals to obtain voltage and current data sequences. The Newton-FMD algorithm is then used to adaptively decompose and denoise the voltage and current data, extracting the effective signal components. The battery internal resistance measurement drive circuit and the internal resistance measurement circuit are existing technologies and will not be described further in this invention.

[0040] S2. Based on the voltage and current sequence data, the optimal parameters for eigenmode decomposition are searched in the parameter space using a Newton-quasi-Newton hybrid optimization algorithm.

[0041] like Figure 8As shown, this embodiment uses a hybrid optimization algorithm (BFGS) to drive adaptive parameter optimization. The BFGS algorithm is a hybrid optimization algorithm combining Newton's method and the quasi-Newton method (Broyden–Fletcher–Goldfarb–Shanno algorithm). By using the BFGS algorithm to globally optimize the three core parameters of the FMD (Focus Device Demodulation)—number of modes, filter length, and number of filters—the optimal denoising effect of the FMD on voltage and current signals is achieved. The specific process includes: parameter space modeling, fitness function design, direction update and step size search, line search and parameter update, convergence judgment, and termination.

[0042] Based on the above embodiments, in an optional embodiment of the present invention, such as Figure 8 As shown, step S2 specifically includes steps S21 to S25.

[0043] S21. Parameter Space Modeling: Define the vector of parameters to be optimized and its feasible region. : .

[0044] .

[0045] In the formula, For the parameter vector to be optimized, For filter length, For the lower limit of filter length, Upper limit of filter length, For the number of filters, Lower limit of the number of filters Maximum number of filters For the number of modes, The lower limit of the number of modes, The upper limit of the number of modes, For integer fields, This means taking the smaller of the two values.

[0046] In this embodiment, the median value of the parameter range is used as the starting point for Newton optimization.

[0047] S22. Construct a fitness function that integrates residual autocorrelation and spectral flatness.

[0048] .

[0049] In the formula, For fitness, For residual autocorrelation, for Weighting coefficients For spectral flatness, for The weighting coefficients.

[0050] This embodiment , These correspond to the priorities of periodic noise suppression and broadband noise suppression, respectively.

[0051] It can quantify the periodic components remaining in the residual signal after denoising. The smaller the value, the more thoroughly the periodic noise is suppressed. Its calculation is based on the normalized autocorrelation function.

[0052] .

[0053] .

[0054] .

[0055] In the formula, This means taking the larger of the two values. The autocorrelation function of the residual signal, For delay, The zero-delay autocorrelation value, i.e., the total signal energy, The upper limit of latency, For sampling frequency, For expected value, For residual signals, The mean of the residual signal, The variance of the residual signal, For the original noisy signal, For the denoised signal, For time.

[0056] It can characterize the uniformity of energy distribution of the power spectrum of the denoised signal. The value range of this index is [0,1]. The lower the value, the more prominent the signal components are and the better the broadband background noise suppression effect.

[0057] .

[0058] In the formula, For the natural base, For frequency points, This represents the number of iterations. The power spectral density of the denoised signal.

[0059] Estimated using windowing (such as the Kaiser window) and the Welch method. N is typically taken as 256 points.

[0060] S23. Determine the iterative search direction of parameters based on the approximate Hessian matrix using the quasi-Newton method and gradient information. .

[0061] .

[0062] In the formula, For iterative search direction, For approximate matrices, For pseudo gradient, For fitness function, For the first The parameter vector at the next iteration.

[0063] The derivation of the iterative search direction for the parameters is as follows.

[0064] fitness function in We use a second-order Taylor expansion to perform a second-order approximation, and obtain an approximate model of the fitness function.

[0065] .

[0066] In the formula, For fitness function, For the first The parameter vector at the next iteration For iterative search direction, for gradient, Indicates transpose, This is the Hessian matrix.

[0067] fitness function approximation model Differentiate to obtain the parameter update direction.

[0068] .

[0069] In the discrete parameter space, the gradient needs to be redefined using the discrete coordinate transformation method: .

[0070] In the formula, parameter vector The Middle Discrete gradient direction of each parameter For the modified parameter vector, For the first parameter in the parameter vector to be optimized One parameter, For preset change amount, This indicates that only the first one is changed. One parameter.

[0071] The direction of the discrete gradient reflects the local descent trend within the discrete neighborhood.

[0072] To be compatible with the continuous optimization framework of the quasi-Newton method, the discrete gradient direction is mapped to a pseudo gradient.

[0073] .

[0074] .

[0075] In the formula, For pseudo gradient, For discrete gradient direction, For element-wise multiplication, For changes, For the filter step size, The step size of the number of filters, The step size for the number of modes, Indicates transpose, The discrete gradient direction of the filter length, Discrete gradient directions of the number of filters, The discrete gradient direction represents the number of modes.

[0076] Pseudo-gradients preserve discrete directional information, but through the amount of change... Normalization scaling, at which point the parameter update direction is corrected to: .

[0077] In the formula, For iterative search direction, For the Hessian matrix, For pseudo gradient, For fitness function, For the first The parameter vector at the next iteration.

[0078] Each iteration of Newton's method requires calculating the first derivative and the Hessian inverse matrix, which is computationally intensive and time-consuming, and the Hessian matrix may not be positive definite. Therefore, this embodiment introduces the quasi-Newton method (BFGS). The goal of the BFGS method is to find an approximate matrix that satisfies the secant condition. At the same time, make it approximate with the matrix from the previous step. The differences should be minimized. This objective is achieved by minimizing the weighted Frobenius norm of the matrix. The approximation matrix is ​​constructed using a quasi-Newton method to approximate the Hessian matrix. The weight matrix is ​​typically chosen as... The inverse or other symmetric positive definite matrix.

[0079] .

[0080] .

[0081] In the formula, For norm, For the weight matrix, For objects used to calculate the Frobenius norm, Let be the trace of the matrix.

[0082] Construct the Lagrangian function and combine it with the secant condition. ,get: .

[0083] In the formula, For Lagrange function, For Lagrange multipliers, For gradient change, This represents the change in parameters.

[0084] right Taking the derivative and setting it to zero, we get: .

[0085] Assumption It is a symmetric matrix, and is updated to its rank-2 correct form: .

[0086] In the formula, For the first undetermined scalar coefficient, For scaling factor, For the second undetermined scalar coefficient, This is an auxiliary vector.

[0087] Substituting the secant condition, we get: .

[0088] After unfolding: .

[0089] To simplify calculations, a specific vector is chosen, let... , Substituting into the equation expanded in the previous step, we get: .

[0090] By comparing the coefficients, we can obtain: , .

[0091] Therefore, the update formula for the quasi-Newton method is: .

[0092] By forcibly satisfying the curvature condition ,make sure Always positive definite, avoiding iterative divergence. Therefore, the parameter update direction, determined by the product of the approximate Hessian inverse matrix and the gradient vector, can be obtained: .

[0093] S24. After determining the update direction, determine the step size through line search and complete the parameter iterative update, as well as record the historical state.

[0094] Specifically, after determining the update direction, the step size is determined through line search. The parameters are iteratively updated, and historical states are recorded to support subsequent convergence analysis. To balance convergence speed and stability, the Armijo criterion is used to find the maximum step size that allows the fitness function to decrease sufficiently.

[0095] The condition for finding the maximum step size is: .

[0096] In the formula, For the decrease coefficient, For the first The step size of the secondary search attempt.

[0097] Initial step size set to This inherits the second-order convergence advantage of Newton's method. (Descent coefficient) The degree of looseness used to control the magnitude of the decline.

[0098] If the candidate step size does not meet the maximum step size search condition, then update the step size according to the geometric shrinkage rule: .

[0099] In the formula, For the new step size after contraction, It is a contraction factor.

[0100] contraction factor =0.5 ensures the step size decays rapidly to avoid invalid searches. For example, the initial step size... If the condition is not met, then shrink sequentially to , This continues until the maximum step size search condition is met or the minimum step size threshold is reached. .

[0101] After determining the feasible step size, update the parameters and force them to fall within the feasible region.

[0102] If the value exceeds the feasible region, it is set as the feasible region boundary, and the gradient norm is recorded. Rate of change of objective function Parameter sequence and step size sequence .in, For the first The parameter vector at the next iteration For the first Filter length at the next iteration For the first Number of filters in the next iteration For the first The number of modes in the next iteration.

[0103] S25. Evaluate the convergence of the current iteration based on the historical state and obtain the output parameters. The convergence evaluation is based on the gradient norm and the transformation rate criterion.

[0104] Specifically, based on the historical states recorded in step S24, the convergence of the current iteration is quantitatively evaluated. The convergence criterion is: first, the gradient norm. This indicates that the objective function tends to flatten at the current point, reaching a local extremum. Second, the transformation rate criterion: .in, For gradient, L2 norm, For changes, For fitness function, This is the parameter vector to be optimized.

[0105] If the gradient condition is met, directly output the optimal parameters. . .

[0106] If only the rate of change criterion is met, continue monitoring for the next 3 iterations to avoid false convergence.

[0107] If the number of iterations The process terminates and displays a warning of non-convergence. This represents the maximum number of iterations.

[0108] In addition, an exception handling mechanism is introduced: First, if an exception occurs in 5 consecutive iterations... The oscillation was determined, and the BFGS matrix was reset. Second, if the gradient norm suddenly increases or the fitness function becomes abnormal, roll back to the most recent stable state. Re-iterate. Among them, It is the identity matrix or the initial matrix. This is the parameter vector from the previous iteration.

[0109] S3. Based on the voltage and current sequence data and the optimal parameters, perform dynamic bandwidth scaling, correlation merging and zero-phase filtering to obtain the denoised mode and reconstructed signal.

[0110] Specifically, after global optimization is completed using the Newton-BFGS hybrid optimization algorithm, the optimal parameters are transmitted to the Feature Mode Decomposition (FMD). Simultaneously, the FMD is improved by proposing a Dynamic Band Adaptive and Correlation-Constrained Feature Mode Decomposition (FMD) method to adapt to the broadband and non-periodic noise characteristics of quasi-square wave signals such as battery voltage and current during unsteady-state processes.

[0111] This invention employs a DC pulse discharge method to measure the internal resistance of a battery. The discharge current used in this method is a small current of approximately 1A, which will not damage the battery and is suitable for batteries of various capacities. Addressing the problem of battery polarization effects, this invention proposes a dynamic band adaptive and correlation-constrained FMD method, comprising steps S31 to S34.

[0112] S31. Construct a dynamic frequency band scaling mechanism. Specifically, the lower limit of the first frequency band is forced to a minimum value (value is...). The upper limit is expanded to To avoid low-frequency leakage, subsequent frequency bands will be based on the theoretical center frequency. Based on the baseline, expand left and right And set up protective boundaries.

[0113] .

[0114] .

[0115] In the formula, For the number of modes, For the number of filters, To protect the lower boundary, To protect the upper boundary, This means taking the larger of the two values. This means taking the smaller of two values. For the offset of the frequency band extension, To prevent division by zero, the value is a very small value.

[0116] When there is a frequency band boundary conflict, the offset is automatically increased (the value is...). This ensures the feasibility of the filter.

[0117] S32, using the intermodal time-domain correlation coefficient matrix Quantize waveform similarity, remove autocorrelation terms, and define a dynamic merging threshold.

[0118] .

[0119] In the formula, For the maximum correlation coefficient threshold, This represents the current mode number.

[0120] S33, For the maximum correlation coefficient exceeding The modes are fused by weight to preserve the energy of the dominant mode and suppress redundant oscillations.

[0121] .

[0122] In the formula, For the fused threshold, For the first Modal component signal, For modality With mode The time-domain correlation coefficient (i.e., weight) between them.

[0123] S34. Dynamically optimize the filter using principal component analysis of the covariance matrix to extract the maximum energy frequency band: .

[0124] In the formula, For the energy distribution of the corresponding dominant frequency band of the signal, This means finding the input value that makes a function reach its maximum value. For the frequency components of the signal, Indicates transpose, For the time delay matrix, The eigenvalues ​​of the covariance matrix, For identity matrix or initial matrix, It is an L2 norm.

[0125] After frequency domain peak normalization, zero-phase filtering is used to maintain the transient alignment of the quasi-square wave, which significantly improves harmonic alignment accuracy and noise robustness.

[0126] S4. Based on the denoised mode and reconstructed signal, feature evaluation and fusion screening are performed by peak matching degree, square envelope spectrum negative entropy and maximum cross-correlation to obtain denoised voltage and current sequence data.

[0127] Despite the progress made in improving FMD in terms of parameter optimization and noise suppression, its mode screening mechanism is still limited to a single-dimensional feature evaluation, making it difficult to accurately distinguish effective harmonics, transient impulses and noise components in quasi-square wave signals.

[0128] To further distinguish between effective harmonics, transient impulses, and noise components in quasi-square wave signals, this invention constructs a three-dimensional feature system (i.e., multi-mode screening index) based on peak matching degree (SPM), squared envelope spectral negative entropy, and maximum cross-correlation (MCC). This index is used to select the optimal mode for signal reconstruction, thereby achieving the best noise reduction effect. Specific indices are as follows: The expression for peak matching degree is: .

[0129] In the formula, No. Peak matching degree of each mode, For the first The dominant frequency of the power spectrum of each mode, Indicates extracting the median, The set of peak frequencies of the original noise signal To prevent the minimum value from being divided by zero.

[0130] Extracted from the set of peak frequencies in the spectrum of the noise signal, it is used to dynamically estimate the actual fundamental frequency, replacing the theoretical harmonic fundamental frequency, in order to be compatible with the frequency shift and noise interference of the signal in practical applications.

[0131] Specifically, the energy of a quasi-square wave signal is concentrated in the fundamental frequency and its odd harmonics, while the traditional energy proportion method is prone to mistakenly eliminating high-frequency effective components. SPM (spectral peak matching degree) accurately locks the discrete harmonic frequency band and suppresses broadband noise interference by quantifying the matching degree between the modal spectrum and the theoretical harmonic distribution.

[0132] The expression for the negative entropy of the squared envelope spectrum is: .

[0133] .

[0134] .

[0135] In the formula, For the negative entropy of the squared envelope spectrum, The reference entropy of the Gaussian distribution, For the Shannon entropy of the envelope spectrum, For the natural base, The variance of the squared envelope signal, Indicates taking the average value. This represents the frequency domain form of the square envelope signal.

[0136] During battery charging and discharging, polarization effects cause drastic fluctuations in the voltage signal at the moment of switching between charging and discharging. The negative entropy of the square envelope spectrum can balance transient characteristics and prevent excessive smoothing from causing loss of information about internal resistance jumps. The larger the negative entropy value of the square envelope spectrum, the more significant the transient characteristics. However, excessively high values ​​may introduce noise and misjudgment, so it is necessary to weigh its corresponding weighting coefficients.

[0137] The expression for the negative entropy of the squared envelope spectrum is:

[0138] In the formula, For the negative entropy of the squared envelope spectrum, The reference entropy of the Gaussian distribution, is the Shannon entropy of the envelope spectrum.

[0139] The specific calculation of the negative entropy of the squared envelope spectrum is divided into three steps.

[0140] First, the square envelope is extracted, that is, for a length of... signal conduct The analytic signal is obtained by transformation, and then its squared envelope is taken to obtain the final signal. .

[0141] .

[0142] In the formula, For square envelope signal, For imaginary units, for Transformation.

[0143] Secondly, envelope spectrum calculation is performed, that is, for Perform a discrete Fourier transform to convert the squared envelope to the frequency domain to obtain... .

[0144] .

[0145] In the formula, The frequency domain morphology of the square envelope signal, This is the Discrete Fourier Transform.

[0146] Finally, the entropy value is calculated, and the result is substituted into the expression for the negative entropy of the square envelope spectrum to obtain the corresponding negative entropy of the square envelope spectrum.

[0147] .

[0148] .

[0149] In the formula, Indicates taking the average value. The frequency domain morphology of the square envelope signal, For the natural base, Let V be the variance of the squared envelope signal.

[0150] The expression for maximum cross-correlation is:

[0151] In the formula, For maximum cross-correlation, To calculate the normalized cross-correlation function, For the first Modal component signal, For noisy raw signals, for Standard deviation for The standard deviation.

[0152] Maximum cross-correlation (MCC) solves the phase alignment problem of transient edges of quasi-square wave signals in battery internal resistance detection by quantizing the similarity of the time-domain waveform of the mode with the original noisy signal: it prioritizes the selection of modes with strict time-domain waveform matching with a weight of 0.6, suppresses Gibbs oscillations and high-frequency pseudo-components caused by noise, avoids the loss of internal resistance jump information caused by waveform distortion at the moment of charge-discharge switching, and ensures time-domain fidelity under dynamic operating conditions.

[0153] Feature evaluation and fusion screening are performed using spectral peak matching degree, squared envelope spectral negative entropy, and maximum cross-correlation, specifically as follows: A comprehensive evaluation index is constructed using weighted linear combination spectral peak matching degree, squared envelope spectral negative entropy, and maximum cross-correlation: In the formula, For comprehensive evaluation indicators, Spectral peak matching degree weighting coefficient Square envelope spectrum negative entropy weighting coefficient, Maximum cross-correlation weighting coefficient.

[0154] The selection criteria consist of three layers: First, modes with a DC energy percentage >40% are forcibly retained. Second, the top 70% of components are selected in descending order of scores. Third, in anomaly handling, the maximum number of modes is limited to ≤5, and the validity of the index is ensured.

[0155] In this embodiment, , , This mechanism precisely locks the harmonic frequency band through spectral peak matching, enhances transient impact retention through the negative entropy of the square envelope spectrum, and suppresses noise interference through maximum cross-correlation, significantly improving the noise reduction effect.

[0156] S5. Obtain the voltage difference and current difference based on the denoised voltage and current sequence data. Preferably, step S5 specifically includes steps S51 to S52.

[0157] S51. Calculate the average value of the high level of the denoised voltage signal. and the average value of low level Obtain the voltage difference . S52. Calculate the average high level of the denoised current signal. and the average value of low level Obtain the current difference .

[0158] Specifically, a high level refers to the battery's voltage state during pulse discharge (when the MOSFET is on). A low level refers to the battery's voltage state during the end of pulse discharge (when the MOSFET is off).

[0159] S6. Obtain the internal resistance of the battery based on the voltage difference and the current difference.

[0160] Based on the equivalent model diagram of battery internal resistance measurement ( Figure 7 Substitute the following formula for calculating the internal resistance of a battery: The internal resistance of the battery can then be obtained. In the formula, This indicates the internal resistance of the battery.

[0161] This invention combines the advantages of both AC and DC methods, employing a DC pulse discharge method for resistance measurement. This method discharges the battery with DC pulses of a constant frequency and collects the voltage fluctuation signals across the battery terminals. Due to the battery polarization effect, the voltage signal exhibits a quasi-square wave with steep rises and falls during charge-discharge switching. Meanwhile, the current signal produces brief spikes during the initial stage of battery discharge, such as... Figure 1 and Figure 2 As shown. This non-steady-state process leads to two types of characteristics in the signal spectrum: first, the effective signal energy is concentrated at the fundamental frequency (consistent with the pulse frequency), such as... Figure 3 and Figure 4 As shown. Secondly, the broadband interference generated by the polarization effect, combined with environmental and hardware noise, forms a continuous spectrum, such as... Figure 5 and Figure 6 As shown, this differs fundamentally from the equally spaced harmonic clusters induced by periodic impacts in mechanical vibration signals. Given the complexity of these signal characteristics, especially the significant overlap between broadband interference and fundamental harmonics, traditional signal decomposition methods based on periodic assumptions face severe challenges in processing voltage and current signals under such unsteady, broadband noise backgrounds. Therefore, selecting a suitable decomposition method with strong noise suppression capabilities is crucial for improving the accuracy of battery internal resistance measurements.

[0162] This invention uses Newton-FMD to suppress noise and preserve the characteristics of voltage and current signals, thereby enabling accurate measurement of the battery's internal resistance.

[0163] To verify the effect of Newton-FMD on improving accuracy in battery internal resistance measurement, the present invention will be further described below through specific embodiments.

[0164] Experimental tests and analyses were conducted at room temperature. A 12V / 100Ah valve-regulated sealed lead-acid battery was used, and an online battery monitoring system was built to better monitor the battery's signal data.

[0165] This invention uses a constant frequency of 1kHz for DC pulse discharge. By collecting the high and low levels of fluctuating voltage and current, the average value of the fluctuation is obtained. Substituting this value into the battery internal resistance calculation formula yields the battery internal resistance. However, it is worth noting that the voltage fluctuation signal is too small, so the voltage is amplified in the hardware design. Therefore, the final calculation still needs to be divided by the voltage gain of the amplification circuit, which is 39.22.

[0166] Figure 9 These are the voltage fluctuation signals before and after Newton-FMD denoising. Figure 10 The images show the current fluctuation signals before and after Newton-FMD denoising. Compared to the signal before denoising, the signal after denoising has smaller fluctuations and a smoother curve, demonstrating the excellent noise reduction performance of Newton-FMD.

[0167] Figure 11 The voltage fluctuation signals are before and after denoising using four different methods. Figure 12 The figures show the current fluctuation signals before and after denoising using four different algorithms. As can be seen from the figures, the waveform after denoising using the Newton-FMD method is smoother.

[0168] Example 2: The present invention provides an improved characteristic mode decomposition battery internal resistance measurement device, which includes a raw data module, a parameter search module, a signal reconstruction module, a noise reduction module, a voltage and current module, and an internal resistance module.

[0169] The raw data module is used to perform pulsed DC discharge on the battery to obtain voltage and current sequence data of the battery.

[0170] The parameter search module is used to search for the optimal parameters of the characteristic mode decomposition in the parameter space based on the voltage and current sequence data using a Newton-quasi-Newton hybrid optimization algorithm.

[0171] The signal reconstruction module is used to perform dynamic bandwidth scaling, correlation merging, and zero-phase filtering based on the optimal parameters to obtain the denoised mode and the reconstructed signal.

[0172] The denoising module is used to evaluate and fuse features based on the denoised mode and reconstructed signal through spectral peak matching degree, squared envelope spectrum negative entropy and maximum cross-correlation, to obtain denoised voltage and current sequence data.

[0173] The voltage and current module is used to obtain the voltage difference and current difference based on the denoised voltage and current sequence data.

[0174] An internal resistance module is used to obtain the internal resistance of the battery based on the voltage difference and the current difference.

[0175] In embodiment three, the present invention provides an improved characteristic mode decomposition (EMD) battery internal resistance measurement device, which includes a processor, a memory, and a computer program stored in the memory. The computer program can be executed by the processor to implement an improved EMD battery internal resistance measurement method as described in any paragraph of embodiment one.

[0176] In a fourth aspect, the present invention provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform an improved eigenmode decomposition method for measuring the internal resistance of a battery as described in any paragraph of Embodiment 1.

[0177] Obviously, the embodiments described above are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0178] In the several embodiments provided in this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0179] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0180] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0181] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention are also intended to include the plural forms unless the context clearly indicates otherwise.

[0182] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0183] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0184] The terms "first" and "second" used in the embodiments are merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" can be interchanged in a specific order or sequence where permitted. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0185] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An improved method for measuring the internal resistance of a storage battery based on characteristic mode decomposition, characterized in that, Include: S1. Perform pulsed DC discharge on the battery to obtain the battery's voltage and current sequence data; S2. Based on the voltage and current sequence data, search for the optimal parameters of the characteristic mode decomposition in the parameter space using a Newton-quasi-Newton hybrid optimization algorithm; S3. Based on the optimal parameters, perform dynamic bandwidth scaling, correlation merging, and zero-phase filtering to obtain the denoised mode and reconstructed signal; S4. Based on the denoised mode and reconstructed signal, feature evaluation and fusion screening are performed by spectral peak matching degree, squared envelope spectrum negative entropy and maximum cross-correlation to obtain denoised voltage and current sequence data. S5. Based on the denoised voltage and current sequence data, obtain the voltage difference and current difference; S6. Obtain the internal resistance of the battery based on the voltage difference and the current difference.

2. The improved characteristic mode decomposition method for measuring the internal resistance of a battery according to claim 1, characterized in that, The optimal parameters for eigenmode decomposition are searched in the parameter space using a Newton-quasi-Newton hybrid optimization algorithm, specifically including: Parameter space modeling: Define the vector of parameters to be optimized and its feasible region. : ； ； In the formula, For the parameter vector to be optimized, For filter length, For the lower limit of filter length, Upper limit of filter length, For the number of filters, Lower limit of the number of filters Maximum number of filters For the number of modes, The lower limit of the number of modes, The upper limit of the number of modes, For integer fields, This indicates taking the smaller of the two values; Construct a fitness function that integrates residual autocorrelation and spectral flatness; ； In the formula, For fitness, For residual autocorrelation, for Weighting coefficients For spectral flatness, for Weighting coefficients; Based on the approximate Hessian matrix using the quasi-Newton method and gradient information, the iterative search direction of the parameters is determined. ； In the formula, For iterative search direction, For approximate matrices, For pseudo gradient, For fitness function, For the first The parameter vector at the next iteration; After determining the update direction, the step size is determined through line search, the parameter iterative update is completed, and the historical state is recorded; The convergence of the current iteration is evaluated based on the historical state, and the output parameters are obtained.

3. The improved characteristic mode decomposition method for measuring the internal resistance of a battery according to claim 2, characterized in that, The criteria for convergence evaluation include: gradient norm. ; conversion rate ;in, For gradient, L2 norm, The change; The convergence of the current iteration is evaluated based on historical states, and the output parameters are obtained, including: If the gradient condition is met, the optimal parameters are output directly. If only the rate of change criterion is met, continue monitoring for the next 3 iterations to avoid false convergence; If the number of iterations reaches the maximum number of iterations, the process terminates and a non-convergence warning is displayed. If in 5 consecutive iterations The oscillation was determined, and the quasi-Newton matrix was reset. ; It is the identity matrix or the initial matrix; If the gradient norm suddenly increases or the fitness function becomes abnormal, roll back to the most recent stable state. Re-iterate; This is the parameter vector from the previous iteration.

4. The improved characteristic mode decomposition method for measuring the internal resistance of a battery according to claim 1, characterized in that, Based on the optimal parameters, dynamic bandwidth scaling, correlation merging, and zero-phase filtering are performed to obtain the denoised mode and reconstructed signal, specifically including: Construct a dynamic frequency band scaling mechanism; wherein, the lower limit of the first frequency band is forced to the minimum value, and the upper limit is expanded to... Subsequent frequency bands are based on the theoretical center frequency. Based on the baseline, expand left and right And set up protective boundaries; ； ； In the formula, For the number of modes, For the number of filters, To protect the lower boundary, To protect the upper boundary, This means taking the larger of the two values. This means taking the smaller of two values. For the offset of the frequency band extension, To prevent division by zero; Through the intermodal time-domain correlation coefficient matrix Quantize waveform similarity, remove autocorrelation terms, and define a dynamic merging threshold: In the formula, For the maximum correlation coefficient threshold, This represents the current mode number; For the maximum correlation coefficient exceeding The modalities are fused according to their weights; In the formula, For the fused threshold, For the first Modal component signal, For modality With mode The temporal correlation coefficient between them; The maximum energy frequency band is extracted by dynamically optimizing the filter through principal component analysis of the covariance matrix. In the formula, For the energy distribution of the corresponding dominant frequency band of the signal, This means finding the input value that makes a function reach its maximum value. For the frequency components of the signal, Indicates transpose, For the time delay matrix, The eigenvalues ​​of the covariance matrix, For identity matrix or initial matrix, It is an L2 norm.

5. The improved characteristic mode decomposition method for measuring the internal resistance of a battery according to claim 1, characterized in that, The expression for peak matching degree is: ； In the formula, No. Peak matching degree of each mode, For the first The dominant frequency of the power spectrum of each mode, Indicates extracting the median, The set of peak frequencies of the original noise signal To prevent the minimum value of division by zero; The expression for the negative entropy of the squared envelope spectrum is: ； ； ； In the formula, For the negative entropy of the squared envelope spectrum, The reference entropy of the Gaussian distribution, For the Shannon entropy of the envelope spectrum, For the natural base, The variance of the squared envelope signal, Indicates taking the average value. The frequency domain form of the square envelope signal; The expression for maximum cross-correlation is: ； In the formula, For maximum cross-correlation, To calculate the normalized cross-correlation function, For the first Modal component signal, For noisy raw signals, for Standard deviation for Standard deviation; Feature evaluation and fusion screening are performed using spectral peak matching degree, squared envelope spectral negative entropy, and maximum cross-correlation, specifically as follows: A comprehensive evaluation index is constructed using weighted linear combination spectral peak matching degree, squared envelope spectral negative entropy, and maximum cross-correlation: In the formula, For comprehensive evaluation indicators, Spectral peak matching degree weighting coefficient Square envelope spectrum negative entropy weighting coefficient, Maximum cross-correlation weighting coefficient; The screening rules consist of three layers: First, modes with a DC energy ratio >40% are forcibly retained; second, the top 70% of components are selected in descending order of scores; and third, the maximum number of modes is limited to ≤5 in anomaly handling while ensuring the validity of the index.

6. A method for measuring the internal resistance of a storage battery using improved characteristic mode decomposition according to any one of claims 1 to 5, characterized in that, Based on the denoised voltage and current sequence data, the voltage difference and current difference are obtained, specifically including: Calculate the average high level of the denoised voltage signal and the average value of low level Obtain the voltage difference ; Calculate the average high level of the denoised current signal and the average value of low level Obtain the current difference .

7. The improved characteristic mode decomposition method for measuring the internal resistance of a battery according to claim 6, characterized in that, The battery internal resistance is obtained based on the voltage difference and the current difference, specifically including: Substitute the voltage difference and the current difference into the battery internal resistance calculation formula. To obtain the internal resistance of the battery; where, This indicates the internal resistance of the battery.

8. A battery internal resistance measurement device with improved characteristic mode decomposition, characterized in that, Include: The raw data module is used to perform pulsed DC discharge on the battery to obtain the battery's voltage and current sequence data; The parameter search module is used to search for the optimal parameters of the characteristic mode decomposition in the parameter space based on the voltage and current sequence data using a Newton-quasi-Newton hybrid optimization algorithm. The signal reconstruction module is used to perform dynamic bandwidth scaling, correlation merging, and zero-phase filtering based on the optimal parameters to obtain the denoised mode and the reconstructed signal. The denoising module is used to evaluate and fuse features based on the denoised mode and reconstructed signal through spectral peak matching degree, squared envelope spectrum negative entropy and maximum cross-correlation, and obtain denoised voltage and current sequence data. The voltage and current module is used to obtain the voltage difference and current difference based on the denoised voltage and current sequence data; An internal resistance module is used to obtain the internal resistance of the battery based on the voltage difference and the current difference.

9. An improved characteristic mode decomposition method for measuring the internal resistance of a storage battery, characterized in that, It includes a processor, a memory, and a computer program stored in the memory; the computer program can be executed by the processor to implement an improved characteristic mode decomposition method for measuring the internal resistance of a battery as described in any one of claims 1 to 7.

Images