Lithium ion battery health state estimation method based on ultrasonic characteristics

Abstract

The invention relates to a lithium ion battery health state estimation method based on ultrasonic characteristics, and belongs to the technical field of battery detection and diagnosis. The method comprises the following steps: applying pulse ultrasonic excitation to a to-be-detected battery and collecting an ultrasonic echo signal formed by multi-interface reflection; carrying out a standardized accelerated aging test on the battery under the conditions of constant temperature and fixed multiplying power, obtaining current, voltage and capacity data, and calculating the SOH of the battery at different aging stages; preprocessing the ultrasonic echo signals, extracting ultrasonic time of flight (TOF) features corresponding to main echoes by adopting a cross-correlation analysis method, and performing time synchronization; and constructing an ultrasonic health factor based on the TOF features and the SOC, and introducing a BP neural network model to realize online estimation of the SOH of the battery. According to the method, through fusion of ultrasonic nondestructive testing and an intelligent algorithm, rapid and accurate evaluation of the health state of the lithium ion battery under non-intrusive and non-charging and non-discharging working conditions is realized, and the method has the advantages of high detection speed, high applicability and high engineering popularization value.

Description

Technical Field

[0001] This invention belongs to the field of battery testing and diagnostic technology, specifically relating to a method for estimating the health status of lithium-ion batteries based on ultrasonic features. Background Technology

[0002] Lithium-ion batteries, due to their high energy density, long cycle life, and low self-discharge rate, have been widely used in new energy vehicles, energy storage power stations, communication base stations, and high-end equipment. However, in actual use, under the influence of long-term cycling, rate shocks, and temperature fluctuations, the internal electrode structure, electrolyte performance, and interface state of lithium-ion batteries gradually deteriorate, leading to a decrease in usable capacity, an increase in internal resistance, and an increase in safety risks. If the battery health status cannot be accurately assessed, it may lead to safety hazards such as overcharging, over-discharging, and thermal runaway, seriously affecting system reliability and operational safety.

[0003] Currently, methods for estimating the state of harmonics (SOH) of batteries can be mainly classified into the following categories:

[0004] (1) The capacity calibration method calculates the SOH by measuring the ratio of the capacity to the nominal capacity through a complete charge-discharge cycle. This method has high accuracy, but the test cycle is long and it is difficult to meet the needs of online applications.

[0005] (2) Methods based on internal resistance or impedance characteristics, such as DC internal resistance method and electrochemical impedance spectroscopy, are highly dependent on test conditions and equipment, and their applicability in the field is limited.

[0006] (3) Model-based methods estimate SOH by identifying parameters of equivalent circuit models or electrochemical models. However, these methods are highly sensitive to model structure and parameter accuracy and lack robustness.

[0007] (4) Data-driven methods use operating data such as voltage and current to train machine learning models to estimate SOH. Although this type of method has a certain generalization ability, it is highly dependent on the completeness of the data conditions and the number of samples. In actual operating conditions, it is often difficult to obtain high-quality full data.

[0008] The methods described above generally suffer from drawbacks such as reliance on complete charge-discharge cycles, strong indirectness of electrical characteristics, and insufficient reflection of changes in the internal structure of the battery. In contrast, ultrasonic testing technology can directly reflect changes in the internal dielectric and structural state of the battery, and has advantages such as non-destructiveness, repeatability, and high sensitivity. However, current quantitative modeling methods for SOH based on ultrasonic characteristics are still immature and urgently require systematic research. Summary of the Invention

[0009] The purpose of this invention is to address the problems of existing battery health state estimation methods, such as heavy reliance on complete charge-discharge cycles, difficulty in directly characterizing the aging features of the battery's internal structure, and insufficient stability in online applications. This invention provides a lithium-ion battery health state estimation method based on ultrasonic features. By constructing a joint database of standardized battery aging tests and ultrasonic detection, a cross-correlation algorithm is used to accurately extract time-of-flight features from ultrasonic echoes. Furthermore, a backpropagation (BP) neural network algorithm is introduced to establish a nonlinear mapping relationship between ultrasonic health factors and battery state of health (SOH), achieving high-precision, non-invasive online estimation of lithium-ion battery health. The proposed method can estimate SOH without requiring a complete charge-discharge cycle, making it suitable for battery health state assessment under complex operating conditions and showing promising engineering application prospects.

[0010] To achieve the above objectives, the technical solution of the present invention is: a method for estimating the health status of a lithium-ion battery based on ultrasonic features, comprising the following steps:

[0011] S1. Under constant temperature conditions, an ultrasonic flaw detector is used to apply pulsed ultrasonic excitation to the soft-pack lithium-ion battery under test, and ultrasonic echo signals formed by reflections from multiple interfaces such as the battery electrode layer, separator layer, electrolyte layer and encapsulation layer are collected. Multiple repeated measurements are performed under the same SOH condition, and the echo signals are superimposed and averaged to construct a stable and reliable ultrasonic echo feature database.

[0012] S2. Under constant temperature conditions, the battery is subjected to accelerated aging cycle test using 2C rate constant current constant voltage charging and 2C rate constant current discharging. The actual discharge capacity of the battery is measured at each aging cycle node, and the ratio of this capacity to the initial nominal capacity is used as the SOH value under the corresponding cycle. The SOH value is matched one-to-one with the synchronously acquired ultrasonic echo data to construct a standardized calibration database between battery aging state and SOH.

[0013] S3. Zero-point calibration, baseline correction, and time window segmentation are performed on the acquired raw ultrasonic echo signals. The main echo signal within each echo time window is selected as the analysis object. The initial measurement echo is used as the reference signal. The normalized cross-correlation algorithm is used to calculate the relative time delay of each measurement echo relative to the reference echo. The ultrasonic time of flight (TOF) is determined based on the hysteresis corresponding to the maximum correlation coefficient. The TOF sequence is subjected to moving average filtering and time synchronization processing to obtain stable TOF characteristic quantities.

[0014] S4. Normalize the TOF features obtained in step S3, construct an ultrasound health factor, and introduce the battery state of charge (SOC) as an auxiliary input feature to establish a training sample dataset with ultrasound health factor and SOC as input and SOH as output; construct a three-layer feedforward BP neural network model, train the sample data, and obtain a nonlinear mapping model between ultrasound features and SOH; during the actual operation of the battery, collect ultrasound echoes in real time and extract TOF features, input them into the trained neural network model, and output the corresponding SOH estimate.

[0015] The present invention also provides an electronic device, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0016] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0017] Compared with existing technologies, this invention has the following advantages: This invention proposes a lithium-ion battery health state estimation method based on the fusion of ultrasonic features and neural networks. By establishing a correspondence between ultrasonic Time-of-Flight (TOF) features and the state of health (SOH) of the battery during aging, it achieves high-precision non-destructive detection of the battery's health state. This invention has three significant advantages: First, this method, based on ultrasonic detection, can estimate SOH without relying on a complete charge-discharge cycle, avoiding the stringent requirements of traditional capacity and electrochemical methods regarding operating conditions. Second, by using a cross-correlation algorithm to extract TOF from the ultrasonic echoes and combining it with sliding filtering, the influence of environmental noise and probe coupling fluctuations on the measurement results is effectively suppressed, significantly improving the stability and repeatability of the ultrasonic features. Third, by using a BP neural network to establish a nonlinear mapping relationship between ultrasonic health factors and SOH, it can achieve unified SOH estimation under multiple aging stages and operating conditions, exhibiting good adaptability to SOC changes and operating condition fluctuations, and possessing strong engineering application value. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention;

[0019] Figure 2 This is a schematic diagram of ultrasonic feature detection.

[0020] Figure 3 The capacity degradation curve of the NCM battery;

[0021] Figure 4 For NCM battery charge-discharge cycles;

[0022] Figure 5 The relative waveforms of each ultrasonic echo signal. Detailed Implementation

[0023] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0024] like Figure 1 As shown, the present invention provides a method for estimating the health status of lithium-ion batteries based on ultrasonic features, comprising the following steps:

[0025] S1. The probe of an ultrasonic flaw detector is used to perform ultrasonic excitation and echo acquisition on the soft-pack lithium-ion battery under test, and the ultrasonic echo characteristic signal formed by multi-interface reflection inside the battery is obtained.

[0026] S101. The ambient temperature fluctuation range, ultrasonic signal acquisition accuracy, time synchronization accuracy and excitation voltage stability of the ultrasonic testing experiment shall be performed in accordance with national or industry standards. The soft-pack lithium-ion battery to be tested shall be placed in a constant temperature environment, with the ambient temperature controlled at 25℃±1℃, and left to stand for no less than 30 minutes without charging or discharging to allow the internal temperature field of the battery to reach a thermally stable state.

[0027] S102. Clean the surface of the soft-pack battery to be tested, remove dust and oil, and uniformly coat the contact area between the battery surface and the ultrasonic probe with ultrasonic coupling agent to reduce the influence of air gap on ultrasonic wave propagation and improve ultrasonic energy coupling efficiency and echo signal-to-noise ratio.

[0028] S103. Based on the structural dimensions, number of electrode layers, and packaging material characteristics of the soft-pack lithium-ion battery under test, adjust the transmission frequency, excitation voltage, pulse width, and repetition frequency parameters of the ultrasonic flaw detector. For the lithium-ion soft-pack battery under test in this example, the transmission frequency is set to 200Hz, the pulse width is set to 30 ns, and the transmission intensity is set to 500V. Perform time base calibration and gain calibration on the ultrasonic flaw detector.

[0029] S104. Place the ultrasonic probe perpendicularly to the surface of the soft-pack lithium-ion battery. Emitterly transmit pulsed ultrasonic signals into the battery in a longitudinal wave pattern. While maintaining the probe position, acquire multiple ultrasonic echo signals formed by reflections from various interfaces including the battery electrode layer, separator layer, electrolyte layer, and encapsulation layer. The arrangement is as follows: Figure 2 As shown;

[0030] S105. The acquired ultrasonic echo signals are synchronously acquired and stored. At least 10 ultrasonic excitations and echo acquisitions are repeated under the same SOH state. The multiple echo signals are superimposed and averaged to obtain stable and reliable ultrasonic echo characteristic data under the SOH state, and an ultrasonic echo characteristic database corresponding to the SOH state is established.

[0031] S2. Conduct standardized aging tests on the batteries under test at constant temperature and fixed rate, collect current and voltage data during the aging process, and obtain battery samples at different aging stages.

[0032] S201. Place the soft-pack lithium-ion battery to be tested in a constant temperature test environment, with the ambient temperature controlled at 25℃±2℃. Leave the battery in an open circuit state for no less than 10 minutes without charging or discharging, so that the active materials, electrolyte and electrode structure inside the battery reach thermal equilibrium with the external environment, and eliminate the influence of temperature gradient on the subsequent high-rate charging and discharging process and aging consistency.

[0033] S202. The battery is charged using a constant current constant voltage (CCCV) method. First, in the constant current stage, the battery is charged at a 2C rate until the battery terminal voltage reaches the upper limit charging voltage of 4.2V. Then, the constant voltage charging stage is entered, and the terminal voltage is kept constant at 4.2V until the charging current naturally decays to the set cutoff current of 0.05C. The charging process is then considered complete.

[0034] S203. After charging is complete, the battery is left to stand in open circuit condition for 30 minutes to eliminate the ohmic polarization, electrochemical polarization and concentration polarization effects caused by high-rate charging, so that the battery terminal voltage gradually returns to a stable open circuit state, thereby avoiding the polarization effect from interfering with the subsequent discharge capacity test and aging characteristic calibration results.

[0035] S204. Under constant temperature conditions, the battery is discharged using a 2C rate constant current method. During the discharge process, the battery terminal voltage, current, and discharge time data are collected in real time until the battery terminal voltage drops to the lower cutoff voltage of 2.5V, at which point the discharge process is terminated. The actual discharge capacity for this cycle is calculated based on the current and time data during the discharge process. Figure 3 As shown.

[0036] S205. Steps S201–S204 constitute a complete high-rate accelerated aging cycle, and this is recorded as one aging cycle. The above cycle process is repeated continuously until the actual discharge capacity of the battery decays to 70% of its initial nominal capacity. At each cycle or after each preset number of cycles, the battery health state (SOH) is calculated through a capacity test, and is defined as follows:

[0037]

[0038] The obtained SOH values ​​are then stored in a one-to-one correspondence with the ultrasonic detection characteristic data corresponding to that cycle, ultimately constructing a standardized calibration database between battery aging state and SOH, such as... Figure 4As shown, this is used to establish a subsequent quantitative estimation model for ultrasound features and SOH.

[0039] S3. Preprocess the acquired ultrasonic echo signals, extract the main echo signals, and use cross-correlation analysis to calculate the corresponding time of flight (TOF).

[0040] S301. Zero-point calibration and baseline correction of ultrasonic echo signals. Let the original discrete ultrasonic echo signal acquired in the i-th acquisition be: ,in , This represents the number of sampling points.

[0041] To eliminate zero-point bias in the analog-to-digital converter (ADC), zero-point calibration is performed on the signal:

[0042]

[0043] in, The system has a fixed zero-point offset, which is set to 256 here.

[0044] S302, such as Figure 5 As shown, multiple echo time windows are pre-set based on the distribution characteristics of different echoes on the time axis:

[0045]

[0046] Within the j-th echo analysis window, extract the echo sub-signal of the i-th measurement:

[0047]

[0048] Furthermore, the echo signal within the corresponding window from the first measurement data is selected as the reference echo signal:

[0049]

[0050] S303, The echo signal in the i-th measurement With reference echo signal Perform normalized cross-correlation calculation:

[0051]

[0052] Where: k is the discrete hysteresis; the search range is limited to: ;

[0053] Subsequently, in the cross-correlation function Within the range of values, determine the lag point corresponding to the maximum correlation coefficient:

[0054]

[0055] The ultrasound sampling time interval is set as follows:

[0056]

[0057] The time of flight of the ultrasound for the i-th measurement and the j-th echo is:

[0058]

[0059] Subsequently, to suppress probe coupling fluctuations and transient electromagnetic interference, the TOF sequence was subjected to moving average filtering:

[0060]

[0061] The length of the sliding window is taken as: The system also converts the "day-hour-minute-second" time information output by the ultrasound acquisition system into a unified second-level time axis.

[0062]

[0063] in For date, For hours, For minutes, The time is in seconds and is synchronized with the battery data time.

[0064] S4. A neural network estimation method for SOH based on ultrasound TOF.

[0065] S401, TOF Feature Standardization and Health Factor Construction: The feature time-of-flight (TOF) obtained in step S3 using the cross-correlation algorithm at the k-th iteration is... Normalization was performed, and ultrasound health factors were constructed. :

[0066]

[0067] in: The baseline flight time is the initial state of the new battery. This is the time-of-flight characteristic after cross-correlation and sliding filtering in S3.

[0068] Synchronously introduce the battery's state of charge As auxiliary input features, they constitute a multidimensional input vector:

[0069]

[0070] Subsequently, the SOH of the battery obtained in step S2 of the aging test is used as the output feature to construct training sample data pairs:

[0071]

[0072] S402. Construct a three-layer feedforward BP neural network, including:

[0073] Input layer: 2 groups of neurons, corresponding to , ;

[0074] Hidden layer: m neurons;

[0075] Output layer: 1 neuron, corresponding to the estimated SOH value. .

[0076] The network forward propagation relationship is as follows:

[0077]

[0078]

[0079] in For weights; For threshold; The hidden layer activation function is either Sigmoid or ReLU. It is a linear function of the output layer.

[0080] S403. Define the network training error function as follows:

[0081]

[0082] Weight adjustment is performed using gradient descent.

[0083]

[0084]

[0085] in, For learning rate, For network layer indexes.

[0086] S404. During battery operation, ultrasonic signals are acquired in real time, and TOF features are extracted according to step S3 to construct the input vector. The input is fed into the trained BP neural network, which outputs a real-time SOH estimate:

[0087] .

[0088] In summary, this invention proposes a method for estimating the state of health (SOH) of lithium-ion batteries based on ultrasonic features. By applying pulsed ultrasonic excitation to a pouch lithium-ion battery and collecting multi-interface reflected echo signals, a cross-correlation algorithm is used to accurately extract ultrasonic time-of-flight (TOF) features. Furthermore, a standardized aging test is combined to construct the correspondence between ultrasonic health factors and SOH. A BP neural network model is introduced to achieve high-precision online estimation of the battery's SOH. This invention effectively overcomes the dependence of traditional methods on complete charge-discharge data and complex electrochemical models, achieving rapid and reliable assessment of battery health under non-invasive and non-destructive conditions. It has good engineering applicability and promotional value.

[0089] The present invention also provides an electronic device, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0090] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, it can implement the steps of the method described above.

[0091] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for state of health estimation of lithium-ion batteries based on ultrasonic signatures, characterized in that, The method comprises the following steps: S1, applying pulse ultrasonic excitation to the soft pack lithium ion battery to be tested and collecting ultrasonic echo signals formed by multi-interface reflection in the battery to obtain ultrasonic echo characteristic data under different battery health states SOH; S2, performing a standardized accelerated aging test on the battery under constant temperature and fixed rate conditions, collecting current, voltage and capacity data, and calculating the battery health state SOH at different aging stages to construct a battery aging calibration database; S3, preprocessing the collected ultrasonic echo signals, extracting the ultrasonic time of flight TOF characteristics corresponding to the main echo signals by using cross-correlation analysis method, and completing time synchronization; S4, constructing an ultrasonic health factor based on the TOF characteristics and the battery state of charge SOC, introducing a BP neural network to establish a mapping relationship between the TOF characteristics and the SOH, and realizing online estimation of the lithium ion battery SOH. 2.The method of claim 1, wherein, Step S1 includes the following sub-steps: S101, place the soft pack lithium ion battery to be tested in a constant temperature environment, control the environment temperature to be 25℃±1℃, and stand still for not less than 30 minutes under no charging and discharging conditions to make the battery reach a thermal stable state; S102, clean the surface of the battery and uniformly coat ultrasonic coupling agent on the contact area between the probe and the battery to improve the ultrasonic energy coupling efficiency; S103, set the ultrasonic excitation parameters according to the battery structure size and material characteristics, including the transmission frequency, pulse width, excitation voltage and repetition frequency, wherein the transmission frequency is 200Hz, the pulse width is 30ns, the excitation voltage is 500V, and the flaw detector time base and gain calibration are completed; S104, vertically contact the ultrasonic probe with the surface of the battery, transmit pulse ultrasonic signals to the inside of the battery, and collect multiple ultrasonic echo signals formed by multi-interface reflection; S105, repeat the ultrasonic excitation and echo collection for not less than 10 times under the same SOH state, perform superposition and average processing on the multiple ultrasonic echo signals, and construct an ultrasonic echo characteristic database corresponding to the SOH state.

3. The method of claim 1, wherein, Step S2 includes the following sub-steps: S201, place the battery in a constant temperature environment of 25℃±2℃ to stand still and make it reach a thermal equilibrium state; S202, charge the battery at a 2C rate by using a constant current and constant voltage method, change to constant voltage charging after constant current charging to the upper limit voltage 4.2V, and continue until the current decays to 0.05C; S203, continue to stand still for 30 minutes after charging is completed to eliminate polarization effect; S204, discharge the battery to the lower limit cutoff voltage 2.5V by using a 2C rate constant current method, and calculate the actual discharge capacity of a single cycle; S205, repeat S201-S204 to form an accelerated aging cycle, until the battery capacity decays to 70% of the initial nominal capacity, and calculate the corresponding SOH value at each cycle to construct a calibration database of aging state and SOH. 4.The method of claim 1, wherein, Step S3 includes the following sub-steps: S301, perform zero point calibration and baseline correction on the collected ultrasonic echo discrete signals to eliminate the zero point bias and baseline drift of the analog-to-digital converter; S302, a plurality of echo time windows are set according to the distribution characteristics of the ultrasonic echo on the time axis, echo sub-signals are extracted in each time window, and initial measurement data are selected as reference echo signals; S303, normalized cross-correlation operation is performed on the i-th measurement echo signal and the reference echo signal, and a lag corresponding to a maximum correlation coefficient is searched; S304, the ultrasonic time of flight (TOF) feature under each measurement is calculated according to the lag and the ultrasonic sampling time interval; S305, a sliding average filtering process is performed on the multiple TOF sequences, and the ultrasonic acquisition system time information is converted into a unified second-level time axis, so that time synchronization with the battery operation data is realized.

5. The method of claim 1, wherein, Step S4 includes the following sub-steps: S401, the extracted TOF feature is normalized, and a reference time of flight in the initial state of the new battery is taken as a reference to construct an ultrasonic health factor; S402, the state of charge (SOC) of the battery is synchronously introduced as an auxiliary input feature, a training sample set taking the ultrasonic health factor and the SOC as inputs and the SOH as output is constructed; S403, a three-layer BP neural network model including an input layer, a hidden layer and an output layer is constructed, and a nonlinear mapping relationship between the ultrasonic feature and the SOH is established; S404, the BP neural network model is trained by using a gradient descent algorithm, and a model after convergence of an error function is taken as an SOH estimation model; S405, in the actual operation process of the battery, ultrasonic signals are collected in real time and TOF features are extracted, which are input into the trained SOH estimation model, and an SOH estimation value of the battery at the corresponding time is output.

6. The method of claim 4, wherein, The lag of the cross-correlation operation is limited according to the echo time window width, and the accurate time of flight of the main echo is determined by the lag corresponding to the maximum correlation coefficient.

7. The method of claim 5, wherein, The number of neurons in the hidden layer of the BP neural network model is adaptively set according to the convergence of the training error, the Sigmoid function or the ReLU function is used as the activation function of the hidden layer, and the linear activation function is used for the output layer.

8. The method of claim 1, wherein, The method is suitable for online health state rapid detection and evaluation of soft package lithium ion batteries under non-charging and discharging conditions.

9. An electronic device, comprising: The computer program instructions stored on the memory and capable of being executed by the processor can realize the steps of the method according to any one of claims 1-8 when the processor executes the computer program instructions.

10. A computer readable storage medium having stored thereon computer program instructions capable of being executed by a processor, which can realize the steps of the method according to any one of claims 1-8 when the processor executes the computer program instructions.

Images