A method and system for online measurement and estimation of electrochemical impedance spectroscopy of lithium batteries based on charging piles.

Abstract

This invention relates to an online measurement and estimation method and system for the electrochemical impedance spectroscopy of lithium batteries based on charging piles, belonging to battery testing technology. The system includes a measurement module, an estimation module, and a result output module. The measurement module includes a signal generation submodule, a waveform amplification submodule, and a response processing submodule. During charging, the signal generation submodule inserts a pulse control signal, which is amplified by the waveform amplification submodule into a measurement excitation signal applied to the battery. The response processing submodule samples the response signal and calculates the impedance spectrum. The estimation module includes a neural network submodule and an impedance spectrum estimation submodule. The neural network submodule uses the impedance response to train a physical information neural network, outputting the parameters of a Randle circuit model to the estimation submodule to predict changes in the electrochemical impedance spectrum. The result output module displays and saves the measured and estimated impedance spectrum values. This invention can measure and estimate the impedance values ​​of the battery pack under test online at multiple frequencies, and the impedance spectrum prediction has physical interpretability.

Description

Technical Field

[0001] This invention belongs to the field of battery testing technology, and in particular relates to a method and system for online measurement and estimation of electrochemical impedance spectroscopy of lithium batteries based on charging piles. Background Technology

[0002] Lithium-ion batteries possess advantages such as high energy density, good cycle performance, and low self-discharge rate, making them widely used as the primary energy storage component in electric vehicles. To meet the ever-increasing charging demands of on-board lithium-ion batteries, charging stations must not only pursue fast charging but also consider the real-time state of health (SOH) of the lithium-ion batteries. Electrochemical impedance spectroscopy (EIS) is an important tool for analyzing the internal electrochemical reaction states of lithium-ion batteries. With its wide frequency domain detection (typically from kilohertz to millihertz), EIS can obtain coupled kinetic information on mass transport, ion diffusion, and charge transfer during battery charging.

[0003] Traditional EIS measurements require the lithium battery to be in equilibrium. A small voltage or current signal (to avoid affecting the battery's state of charge during measurement) is applied to the battery using an expensive waveform generator. The power amplifier circuit in the electrochemical workstation amplifies the measurement signal into an excitation signal, which is then applied to the battery terminals. The voltage and current responses of the lithium battery under the measurement excitation are sampled, and a discrete Fourier transform is used to convert the signals to the time and frequency domains. The impedance response at different frequencies is measured based on the amplitude ratio and phase difference of the response signal relative to the excitation signal, and then an impedance spectrum curve is plotted. Traditional EIS measurements are not only inefficient but also only suitable for measuring battery impedance response data offline.

[0004] Estimating the electrochemical impedance spectroscopy (EIS) of lithium-ion batteries typically involves equivalent model analysis, referred to here as model-based analysis methods. These include equivalent circuit models and electrochemical impedance models. Equivalent circuit models characterize the battery's impedance response through circuit elements, featuring a simple circuit structure and a smaller number of parameters. Electrochemical impedance models describe the battery's intrinsic mechanisms, such as electrode process kinetics and ion diffusion processes.

[0005] Model-based methods are typically only suitable for offline analysis of electrochemical impedance spectroscopy (EIS) of lithium-ion batteries. While data-based methods avoid the complexity of analyzing circuit structure and chemical mechanisms, they lack interpretability in terms of physical principles. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a method and system for online measurement and estimation of electrochemical impedance spectroscopy (EIS) of lithium batteries based on charging piles. This method utilizes the existing control circuits and power conversion circuits of the charging piles for impedance measurement; it measures the impedance values ​​of the battery pack under test at multiple frequencies simultaneously, improving the efficiency of impedance measurement and avoiding the generation of impedance values ​​for only one measurement frequency at a time; it can perform fast and accurate online prediction of EIS based on offline impedance data provided by the battery management system and the measurement excitation signal of the input battery under test; and it can simultaneously provide physical interpretability for the online prediction of EIS using a physical information neural network and Randle circuits.

[0007] To achieve the above objectives, the system of this invention is implemented through the following technical solution: an online measurement and estimation system for electrochemical impedance spectroscopy of lithium batteries based on charging piles, including an impedance spectrum measurement module, an impedance spectrum estimation module, and a result output module;

[0008] The impedance spectrum measurement module includes the following sub-modules:

[0009] The signal generation submodule modifies the charging strategy of the battery under test at the charging pile end, inserts a small-amplitude pulse control signal required for measuring impedance during the charging process of the battery under test, and transmits it to the waveform amplification submodule.

[0010] The waveform amplification submodule amplifies the small-amplitude pulse control signal into a wide-bandwidth, large-amplitude measurement excitation signal through the power conversion circuit of the charging pile, and applies it to both ends of the battery under test.

[0011] The response processing submodule is used to sample the voltage and current responses of the battery under test under the measurement excitation signal, compensate for the time delay and voltage drop of the current and voltage responses, perform data processing, and calculate the impedance response of the battery under test.

[0012] The impedance spectrum prediction module includes the following sub-modules:

[0013] The neural network submodule is used to retrieve historical charging and discharging data, extract impedance parameter vectors, train the physical information neural network model offline, output the Randles circuit model parameters, and pass them to the impedance spectrum estimation submodule.

[0014] The impedance spectrum estimation submodule, by combining the parameters of the Randle circuit model and the current measurement excitation response signal, performs online estimation of the electrochemical impedance spectrum of the battery under test and predicts the future electrochemical impedance spectrum curve.

[0015] The results output module is used to display and save the measured and estimated values ​​of the impedance spectrum.

[0016] On the other hand, the method of this invention is implemented through the following technical solution: an online measurement and estimation method for the electrochemical impedance spectroscopy of a lithium battery based on a charging pile, implemented based on the online measurement and estimation system, the online measurement and estimation method including the following steps:

[0017] The signal generation submodule modifies the charging strategy of the battery under test at the charging pile end, inserts the pulse control signal required for measuring impedance during the charging process of the battery under test, and transmits the pulse control signal to the waveform amplification submodule.

[0018] After receiving the pulse control signal from the signal generation submodule, the waveform amplification submodule amplifies the pulse control signal using the power conversion circuit of the charging pile to obtain the measurement excitation signal, which is then applied to both ends of the battery under test.

[0019] The response processing submodule is connected to the battery under test, samples the voltage and current responses of the battery under test under the measurement excitation signal, compensates for the time delay and voltage drop of the current and voltage responses, and calculates the impedance response data of the battery under test under the measurement excitation signal in parallel.

[0020] The neural network submodule retrieves the neural network model trained using offline impedance data and outputs the resistance and capacitance parameters of the Randles circuit model.

[0021] The impedance spectrum estimation submodule performs online prediction of the electrochemical impedance spectrum of the battery under test based on the resistance and capacitance parameters of the Randle circuit model and the measurement excitation signal.

[0022] Compared with the prior art, the beneficial effects achieved by the present invention include:

[0023] 1. This invention can monitor the health status of lithium batteries and provide a basis for formulating lithium battery charging strategies by measuring and predicting the current and future values ​​of the electrochemical impedance spectrum of lithium batteries online while charging them at charging piles.

[0024] 2. This invention utilizes the existing control circuit of the charging pile to generate a programmable composite wave pulse signal required for impedance measurement. This signal is easier to generate than a conventional periodic sine wave signal and does not require a traditional, complex, and expensive waveform generator. The programmable composite wave pulse signal makes it possible to measure the impedance value of the battery pack under test at multiple frequencies at one time, avoiding the generation of impedance value for a single measurement frequency and improving the efficiency of impedance measurement.

[0025] 3. This invention utilizes the existing power conversion circuit of the charging pile to amplify the power of the composite wave pulse signal, avoiding the use of expensive dedicated power amplifier circuits in traditional measuring instruments.

[0026] 4. This invention can perform fast and accurate online prediction of electrochemical impedance spectroscopy based on offline impedance data provided by the battery management system and the measurement excitation signal of the battery under test, which can efficiently obtain the health status information of the battery under test. At the same time, the physical information neural network and Randle circuit provide physical interpretability for online prediction of electrochemical impedance spectroscopy, avoiding the problems of model inaccuracy in model-based prediction and lack of physical interpretability in data-based prediction, thus improving the accuracy and efficiency of online prediction. Attached Figure Description

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0028] Figure 1 This is a schematic diagram of the structure of an online measurement and estimation system for the electrochemical impedance spectroscopy of a lithium battery based on a charging pile, as proposed in an embodiment of the present invention.

[0029] Figure 2 This is a flowchart illustrating an online measurement and estimation method for the electrochemical impedance spectroscopy of lithium batteries based on charging piles, proposed in this embodiment of the invention, used to predict the electrochemical impedance spectroscopy of the battery under test.

[0030] Figure 3 The waveforms of the programmable composite wave pulse and the measurement excitation signal and their relaxation time are shown in the embodiments of the present invention.

[0031] Figure 4 This is a comparison chart of the online estimated impedance spectrum and the measured impedance spectrum in a certain embodiment of the present invention.

[0032] Figure 5 This is a schematic diagram of the physical information neural network structure in an embodiment of the present invention.

[0033] Figure 6 This is a schematic diagram of the Randles circuit model structure in an embodiment of the present invention. Detailed Implementation

[0034] An embodiment of the present invention is described in detail below to make the technical solution of the present invention clearer. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0035] In the description of this invention, unless otherwise expressly defined, the technical and scientific terms used have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The technical and scientific terms used in this invention are merely details describing specific embodiments and should not be construed as limiting the invention.

[0036] This invention addresses the problems existing in the prior art by using a physical information neural network to perform online measurement and prediction of the electrochemical impedance spectroscopy of lithium batteries in charging piles. This approach not only utilizes a physical model to provide interpretability for solving the problem but also leverages the advantages of data to simplify the analysis complexity.

[0037] The following describes, with reference to the accompanying drawings, a method and system for online measurement and estimation of electrochemical impedance spectroscopy of lithium batteries based on charging piles.

[0038] like Figure 1 As shown in the figure, an online measurement and estimation system for the electrochemical impedance spectroscopy of lithium batteries based on charging piles is proposed in this embodiment of the invention, including an impedance spectroscopy measurement module, an impedance spectroscopy prediction module, and a result output module.

[0039] The impedance spectrum measurement module includes the following sub-modules:

[0040] The signal generation submodule, connected to the battery management system, modifies the charging strategy of the battery under test at the charging pile. During the charging process of the battery under test, it inserts a small-amplitude programmable composite wave pulse required for impedance measurement as a pulse control signal. For example, under pulse charging conditions, the programmable composite wave pulse is applied during the relaxation time of pulse charging to avoid direct interference with pulse charging; while under constant current charging conditions, the programmable composite wave pulse is a small-amplitude short current pulse. The pulse control signal contains multiple components with different amplitudes and frequencies, and transmits the pulse control signal to the waveform amplification submodule.

[0041] The waveform amplification submodule, influenced by the programmable composite wave pulse, amplifies the programmable composite wave pulse using the power conversion circuit of the charging pile during the charging process of the battery under test. The small-amplitude programmable composite wave pulse is amplified into a wide-bandwidth large-amplitude measurement excitation signal, which is then applied to both ends of the battery under test, i.e., input to the battery under test.

[0042] The response processing submodule is used to sample the voltage and current responses of the battery under test under the measurement excitation signal, perform data processing, and calculate the impedance response of the battery under test.

[0043] The impedance spectrum prediction module includes the following sub-modules:

[0044] The neural network submodule is used to retrieve the Randles circuit model parameters output by the neural network model trained from offline impedance data based on the impedance response of the battery under test, and pass them to the impedance spectrum estimation submodule.

[0045] The impedance spectrum estimation submodule is used to combine the Randles circuit model parameters output by the neural network submodule with the measurement excitation signal to perform online prediction of the electrochemical impedance spectrum of the battery under test and obtain the future electrochemical impedance spectrum curve.

[0046] The results output module is used to display and save the measured and estimated values ​​of the impedance spectrum.

[0047] The signal generation submodule inserts programmable composite wave pulses during the charging process based on the charging strategy characteristics of the battery under test. These programmable composite wave pulses are easy to create and contain rich harmonic components. They can be generated directly by the charging pile and amplified into a non-periodic measurement excitation signal by a power conversion circuit, eliminating the need for traditional, complex, and expensive waveform generators to input sinusoidal excitation signals.

[0048] The measurement excitation signal of the battery under test input to the waveform amplification submodule can be expressed in the following form:

[0049]

[0050] Where A is the amplitude of the programmable composite pulse, amplified by a power conversion circuit; f is the frequency of the programmable composite pulse; and n is the number of harmonics. In terms of timing, relaxation time is set before and after each programmable composite pulse (measurement excitation signal) to eliminate the influence of the programmable composite pulse (measurement excitation signal) on the state of charge of the battery under test.

[0051] The response processing submodule includes a Chebyshev filter and an analog-to-digital converter (ADC). These filters and samples the measured excitation signal and the voltage and current response signals of the battery under test (BUT) in real time. Multiple measurements are used to compensate for the time delay and voltage drop in the current and voltage responses. Data processing is then performed, including but not limited to Fourier decomposition to obtain the amplitude components of the measured excitation signal and response signal at the same frequency, and further calculation of the impedance response of the BUT. The cutoff frequency of the Chebyshev filter is set to 15kHz to ensure complete coverage of the frequency range of the measured excitation signal (mHZ to kiloHZ), fully extracting effective information in the high-frequency range, while also possessing strong attenuation characteristics at the cutoff frequency to avoid the introduction of high-frequency noise. The ADC samples and discretizes the programmable composite pulse and the voltage and current response signals of the BUT in continuous time. The sampling frequency of the ADC follows the Shannon sampling theorem, meaning that to avoid signal distortion, the sampling frequency should be greater than twice the frequency being sampled. Simultaneously, to achieve fast and efficient real-time measurement, the computational burden caused by high-frequency sampling should be avoided. In this embodiment, the sampling frequency of the analog-to-digital converter is set to 5 times the maximum value of the frequency f of the measured excitation signal.

[0052] The analog-to-digital converter (ADC) is installed at the charging station, independent of the battery management system (BMS), to avoid the sampling signal being filtered out by the low-pass filtering characteristics of the Delta-Sigma ADC. Furthermore, to improve quantization accuracy while avoiding the storage and processing burden of high resolution, the ADC has a resolution of 16 bits or higher.

[0053] Furthermore, the response processing submodule converts the sampled voltage and current response signals from the time domain to the frequency domain using an analog-to-digital converter. It then analyzes the frequency response using Discrete Fourier Transform (DFT) to calculate the impedance response of the battery under test (BUT) at multiple measurement frequencies. The impedance response data of the BUT is then transmitted to the impedance spectrum prediction module, which further calculates the impedance value of the BUT. Finally, the impedance value of the BUT is transmitted to the result output module for display and storage. In this embodiment, the formula for calculating the impedance value of the BUT is:

[0054]

[0055] In the formula, Z represents the impedance value of the battery under test in the frequency domain, U represents the voltage response of the battery under test in the frequency domain, and I represents the current response of the battery under test in the frequency domain. The response processing submodule calculates the real and imaginary parts of the impedance response, obtains the electrochemical impedance spectrum using the Nyquist plot as a standard, and then transmits it to the impedance spectrum prediction module.

[0056] The impedance spectrum prediction module includes the following sub-modules:

[0057] The neural network submodule, whose framework is Physical-Informed Neural Networks (PINN), fits and trains the offline impedance data provided by the battery management system, and outputs the resistance and capacitance parameters of the Randles circuit model.

[0058] The impedance spectrum estimation submodule is used to perform online prediction of the electrochemical impedance spectrum of the battery under test by combining the resistance and capacitance parameters of the Randle circuit model output by the neural network submodule with the measurement excitation signal.

[0059] In this embodiment, charging data can be obtained from the battery management system, battery data cloud platform, charging pile, etc.; offline impedance data includes characteristic information of ohmic impedance, SEI impedance, SEI capacitance, charge transfer impedance, double layer capacitance, and diffusion impedance extracted from the impedance spectrum curve, as well as the frequency and amplitude of the measured excitation signal.

[0060] refer to Figure 2 The present invention proposes an online measurement and estimation method for the electrochemical impedance spectroscopy of lithium batteries based on charging piles, comprising the following steps:

[0061] After receiving the information from the battery management system, the signal generation submodule will identify the battery under test; generate a programmable composite wave pulse based on the charging strategy characteristics of the battery under test, and transmit the programmable composite wave pulse to the waveform amplification submodule.

[0062] After receiving the programmable composite wave pulse input from the signal generation submodule, the waveform amplification submodule uses the power conversion circuit of the charging pile to amplify the small-amplitude programmable composite wave pulse into a wide-band large-amplitude measurement excitation signal including multiple different amplitudes A and different frequencies f, which is then applied to the two ends of the battery under test.

[0063] Figure 3 The image shows waveforms of the programmable composite wave pulse, the measurement excitation signal, and their relaxation times in an embodiment of the present invention. (Reference) Figure 3 In terms of timing, a certain relaxation time is set after each programmable composite pulse (measurement excitation signal). The relaxation time is designed to eliminate the influence of the programmable composite pulse (measurement excitation signal) on the state of charge (SOC) of the battery under test, so as to obtain an accurate impedance response under that state of charge.

[0064] The response processing submodule is connected to the battery under test and calculates the impedance response data of the battery under test under the measurement excitation signal. The specific steps are as follows:

[0065] Step 1.1: When the battery under test is under stable DC polarization conditions (equilibrium conditions), input the measurement excitation signal to the battery under test;

[0066] Step 1.2: Sample the measurement excitation signal and the current and voltage responses of the battery under test;

[0067] Step 1.3: Compensate for the time delay and voltage drop of the current and voltage responses to obtain the compensation values;

[0068] Step 1.4: Calculate the impedance response data of the battery under test using discrete Fourier transform, including amplitude and frequency, calculate the real and imaginary parts of the corresponding Nyquist plot, and plot the impedance spectrum curve.

[0069] The time delay and voltage drop in the current and voltage responses originate from the parasitic capacitance and contact resistance of the cables between the battery under test and the charging pile. The compensation value is obtained by comparing the actual current and voltage responses measured by the offline comparison response processing submodule and the electrochemical workstation under the action of the measurement excitation signal; and compensation is applied to the time delay and voltage drop when measuring the current and voltage responses of the battery under test.

[0070] The neural network submodule retrieves historical charge and discharge measurement data provided by the battery management system and extracts impedance parameter vectors from the impedance spectrum curve. The amplitude and frequency of the measured excitation signal δ, as well as the impedance parameter vectors extracted from the impedance spectrum curve, including ohmic impedance, SEI impedance, SEI capacitance, charge transfer impedance, double layer capacitance, and diffusion impedance, are used as offline impedance data.

[0071] The frequency range of offline impedance data includes kilohertz to millihertz; the sampling frequency is selected to be the same as the sampling frequency of the battery under test: following Shannon's sampling theorem, and to avoid the complexity caused by high-frequency calculations, the sampling frequency is selected to be 5 times the frequency of the measurement excitation signal.

[0072] The extraction of impedance spectrum feature information specifically involves: extracting ohmic impedance feature information r in the frequency band above kilohertz. ohm Extracting SEI impedance feature information from the frequency band from 100 Hz to 1 kilohertz. SEI Characteristic information of SEI capacitors cS EI Extracting characteristic information of charge transfer impedance r in the frequency band from Hertz to 100 Hertz ct and the characteristic information of double-layer capacitance c dl Extracting characteristic information of diffusion impedance in the millihertz to hertz frequency range z w .

[0073] Furthermore, offline impedance data is input into the physical information neural network of the neural network submodule to train the physical information neural network and output the parameters of the Randle circuit model. This embodiment extracts different feature information from low to high frequencies, which is sufficient to reflect the multi-level information within the lithium battery, from slow material diffusion to rapid charge transfer. In the impedance spectrum curve, ohmic impedance, double-layer capacitance, diffusion impedance, and other feature parameters are coupled together and exhibit dynamic changes from low to high frequencies. Therefore, the feature parameters extracted from the impedance spectrum curve can comprehensively reflect the health status of the battery under test, providing sufficient features for training the physical information neural network.

[0074] In this embodiment, the training and parameter output process of the physical information neural network includes the following specific steps:

[0075] Step 3.1: Construct a 7-dimensional offline impedance data vector x(δ, r) based on the offline impedance data. ohm ,r SEI ,c SEI ,r ct ,c dl ,z w The input data is then preprocessed. Preprocessing includes replacing missing and erroneous offline impedance data using a sliding time window to ensure that the input data provides valid historical information for model training.

[0076] Step 3.2: Construct a physical information neural network and train the constructed physical information neural network by inputting the offline impedance data vector.

[0077] The constructed physical information neural network comprises an input layer, hidden layers, and an output layer. The input layer receives the offline impedance data, including characteristic information on the measured excitation signal, ohmic impedance, SEI impedance, SEI capacitance, charge transfer impedance, double-layer capacitance, and diffusion impedance, and inputs it as a 7-dimensional vector. The hidden layer contains three fully connected hidden layers, each with 30 neurons. Under the action of the activation function σ, the hidden layers iteratively update the weight matrix w and bias vector b to obtain the output. The weight matrix w and bias vector b are obtained during training. In this invention, the activation function σ is the tanh function. The output layer outputs a 4-dimensional vector u(R). s ,R ct C dl Z w ), representing the parameters of the Randle circuit model, including the ohmic impedance R. s Charge transfer impedance R ct Double-layer capacitance C dl Diffusion impedance Z w .

[0078] Step 3.3: Iteratively update the weight matrix w and bias vector b of the physical information neural network to minimize the loss function L of the physical information neural network. The formula for calculating the loss function L is as follows:

[0079] L=λ1L EIS +λ2L p

[0080]

[0081] In the formula, L represents the total loss function of the physical information neural network, and λ1 represents the loss value L of the impedance spectrum. EIS The weights, λ2 represents the loss value L of the output Randle circuit model parameters. p The weights are set as follows: in this embodiment, λ1 = 0.9 and λ2 = 0.1. n represents the total number of samples, m is the total number of data points in the impedance spectrum, and q is the number of Randle circuit parameters, which is q = 4 in this embodiment. It is the estimated impedance response of the k-th sample at the j-th impedance spectrum data point. It is the impedance response measurement of the k-th sample at the j-th impedance spectrum data point. It is the estimated value of the parameters of the a-th Randles circuit model for the k-th sample. It is the true value of the parameter of the a-th Randles circuit model of the k-th sample.

[0082] The total loss function L characterizes the degree of matching between the parameters of the Randle circuit model predicted by the Physical Information Neural Network and the impedance spectrum. The goal of the Physical Information Neural Network is to find an optimal set of network parameters (w, b) to minimize the total loss function L.

[0083] Step 3.4: Utilize the physical information neural network to output the resistance and capacitance parameters of the Randle circuit model, and perform online prediction of the electrochemical impedance spectroscopy of the battery under test based on the resistance and capacitance parameters.

[0084] The output Randle circuit model parameters will be used by the impedance spectrum estimation submodule to estimate the impedance response Z under a measurement excitation signal with an input frequency of f. The calculation formula is as follows:

[0085]

[0086] The calculated impedance response Z(f) is plotted as an electrochemical impedance spectrum using the Nyquist plot as the standard, thus obtaining the online estimated value of the electrochemical impedance spectrum of the battery under test.

[0087] The results output module displays and saves the measured and estimated values ​​of the impedance spectrum.

[0088] This invention utilizes a charging pile to generate a measurement excitation signal with rich harmonic components, enabling simultaneous measurement of the impedance values ​​of the battery under test at multiple frequencies, thus improving the speed of impedance measurement. Furthermore, this invention employs a physical information neural network to perform online prediction of the impedance spectrum, forecasting future changes in the electrochemical impedance spectrum, which is of great significance for monitoring the health status of lithium batteries and formulating charging strategies.

[0089] The above embodiments are one of the implementation methods of the present invention. The implementation methods of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A system for online measurement and estimation of electrochemical impedance spectroscopy of lithium batteries based on charging piles, characterized in that, It includes an impedance spectrum measurement module, an impedance spectrum prediction module, and a result output module; The impedance spectrum measurement module includes the following sub-modules: The signal generation submodule modifies the charging strategy of the battery under test at the charging pile end, inserts a small-amplitude pulse control signal required for measuring impedance during the charging process of the battery under test, and transmits it to the waveform amplification submodule. The waveform amplification submodule amplifies the small-amplitude pulse control signal into a wide-bandwidth, large-amplitude measurement excitation signal through the power conversion circuit of the charging pile, and applies it to both ends of the battery under test. The response processing submodule is used to sample the voltage and current responses of the battery under test under the measurement excitation signal, perform data processing, and calculate the impedance response of the battery under test. The impedance spectrum prediction module includes the following sub-modules: The neural network submodule is used to retrieve historical charging and discharging data, extract impedance parameter vectors, train the physical information neural network model offline, output the Randles circuit model parameters, and pass them to the impedance spectrum estimation submodule. The impedance spectrum estimation submodule, by combining the parameters of the Randle circuit model and the current measurement excitation response signal, performs online estimation of the electrochemical impedance spectrum of the battery under test and predicts the future electrochemical impedance spectrum curve. The result output module is used to display and save the measured and estimated values ​​of the impedance spectrum; The pulse control signal inserted by the signal generation submodule during the charging process is a programmable composite wave pulse, which includes multiple signal components with different amplitudes A and different frequencies f, as shown below: Where A is the amplitude of the programmable composite pulse, f is the frequency of the programmable composite pulse, and n is the number of harmonics; In terms of timing, relaxation times are set before and after each programmable composite wave pulse.

2. The online measurement and estimation system according to claim 1, characterized in that, The programmable composite wave pulse output by the signal generation submodule will be directly amplified by the power conversion circuit of the charging pile in the waveform amplification submodule.

3. The online measurement and estimation system according to claim 1, characterized in that, The response processing submodule samples the voltage and current response signals through an analog-to-digital converter; compensates for the time delay and voltage drop of the current and voltage responses; performs time-domain to frequency-domain conversion using discrete Fourier transform; calculates the impedance response of the battery under test at multiple measurement frequencies in parallel; obtains the real and imaginary parts of the corresponding Nyquist plot; and plots the impedance spectrum curve.

4. The online measurement and estimation system according to claim 1, characterized in that, The neural network submodule in the impedance spectrum prediction module is based on a physical information neural network. It fits and trains offline impedance data to output the resistance and capacitance parameters of the Randle circuit model. The impedance spectrum estimation submodule performs online prediction of the electrochemical impedance spectrum of the battery under test based on the resistance and capacitance parameters of the Randle circuit model and the measurement excitation signal.

5. The online measurement and estimation system according to claim 1 or 4, characterized in that, The training and parameter output process of a physical information neural network includes the following steps: Historical charging and discharging data are retrieved, and impedance parameter vectors are extracted. The impedance parameter vectors include characteristic information of ohmic impedance, SEI impedance, SEI capacitance, charge transfer impedance, double layer capacitance, and diffusion impedance extracted from the impedance spectrum curve. The extracted impedance parameter vectors, along with the frequency and amplitude of the measured excitation signal, are used as offline impedance data. Preprocess the offline impedance data; A physical information neural network is constructed, and the offline impedance data is input into the physical information neural network for training. Iteratively update the weight matrix and bias vector of the physical information neural network to minimize the loss function of the physical information neural network; The resistance and capacitance parameters of the Randle circuit model are output using a physical information neural network.

6. The online measurement and estimation system according to claim 5, characterized in that, The frequency range of offline impedance data includes kilohertz to millihertz; The process of extracting the impedance parameter vector includes: extracting the characteristic information r of ohmic impedance in the frequency band above kilohertz. ohm Extracting SEI impedance feature information from the frequency band from 100 Hz to 1 kilohertz. SEI and the characteristic information of SEI capacitor c SEI Extracting characteristic information of charge transfer impedance r in the frequency band from Hertz to 100 Hertz ct and the characteristic information of double-layer capacitance c dl Extracting characteristic information of diffusion impedance in the millihertz to hertz frequency range z w Construct a 7-dimensional offline impedance data vector based on the offline impedance data. The input is preprocessed and then fed into the physical information neural network, where... To measure the excitation signal.

7. A method for online measurement and estimation of electrochemical impedance spectroscopy of lithium batteries based on charging piles, characterized in that, Based on the online measurement and estimation system according to any one of claims 1-6, the online measurement and estimation method includes the following steps: The signal generation submodule modifies the charging strategy of the battery under test at the charging pile end, inserts the pulse control signal required for measuring impedance during the charging process of the battery under test, and transmits the pulse control signal to the waveform amplification submodule. After receiving the pulse control signal from the signal generation submodule, the waveform amplification submodule amplifies the pulse control signal using the power conversion circuit of the charging pile to obtain the measurement excitation signal, which is then applied to both ends of the battery under test. The response processing submodule is connected to the battery under test, samples the voltage and current responses of the battery under test under the measurement excitation signal, compensates for the time delay and voltage drop of the current and voltage responses, performs time-frequency domain conversion through Fourier transform, and then calculates the impedance response data of the battery under test under the measurement excitation signal in parallel. The neural network submodule retrieves the neural network model trained using offline impedance data and outputs the resistance and capacitance parameters of the Randles circuit model. The impedance spectrum estimation submodule performs online prediction of the electrochemical impedance spectrum of the battery under test based on the resistance and capacitance parameters of the Randle circuit model and the measurement excitation signal.

Images