Lithium ion battery expansion force estimation method based on ultrasonic signals

Abstract

The invention relates to a lithium ion battery expansion force estimation method based on an ultrasonic signal, and belongs to the technical field of battery monitoring. The method comprises the following steps: S1, fixing two ultrasonic sensors on the same side of a battery at an interval; s2, selecting a plurality of lithium ion batteries integrated with ultrasonic sensors, applying a preload to the batteries by using a programmable dynamic load clamp, collecting expansive force data in real time through a force sensor, synchronously connecting a battery test system, designing a constant-current charge-discharge and NEDC dynamic working condition cycle experiment, and recording voltage, current, temperature and internal resistance data; s3, data batch processing and signal feature extraction are carried out on the collected waveform original data, and battery current data and expansive force data in the circulation process are extracted; s4, constructing a random forest model, and training the model by taking the ultrasonic signal feature data and the current data as model input and taking the expansive force data as model output; and S5, carrying out lightweight processing on the trained model.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery monitoring, and relates to a method for estimating the swelling force of a lithium-ion battery based on ultrasonic signals. Background Art

[0002] In recent years, with the rapid development of the new energy industry, as a core energy storage device, the application field of lithium-ion batteries has expanded from consumer electronics to key fields such as electric vehicles, aerospace, and smart grids. However, with the technological evolution of increasing energy density, the volume expansion phenomenon generated by the battery system during the charging and discharging process has become increasingly prominent, and has become a key technical bottleneck restricting the development of high specific energy batteries.

[0003] The swelling behavior of lithium-ion batteries is essentially the result of the coupling of multiple physical fields. At the microscopic level, the lattice parameters of the positive electrode material (such as NCM, NCA) change during the insertion and extraction of lithium ions, resulting in volume strain; the negative electrode material (such as graphite, silicon-carbon composite material) has a higher volume expansion rate during the lithiation process. At the mesoscopic level, the decomposition reaction of the electrolyte in a high-pressure environment (>4.3V) generates gas (CO2, CH4, etc.), resulting in an increase in the internal pressure of the battery cell. Macroscopically, the growth of lithium dendrites not only causes the loss of active lithium, but also poses a risk of diaphragm puncture. The mechanical stress generated by the superposition of these multi-scale swelling effects can increase the thickness of the battery cell, leading to systematic failures such as deformation of the battery module structure and breakage of the tab.

[0004] Traditional swelling force monitoring technologies have significant limitations: the pressure sensor (such as MEMS piezoresistive) will exhibit thermal drift in a high-temperature environment, resulting in large measurement errors; the contact displacement sensor (LVDT type) needs to be in direct contact with the surface of the battery cell, and its installation gap requirement of 0.5-1mm significantly changes the module stacking design. Although the fiber Bragg grating sensor (FBG) can achieve distributed strain measurement, the single-point cost is high, and it is necessary to pre-embed the sensing optical fiber during the battery cell manufacturing stage. More notably, the existing technologies generally have the problem of mismatched time-frequency characteristics - the pressure sensor (response frequency <1kHz) is difficult to capture high-frequency events such as electrolyte boiling, while the acceleration sensor (>10kHz) lacks sensitivity to slow creep (<0.1Hz).

[0005] Acoustic sensing technology provides a new technical path to solve the above problems. Research shows that the internal state change of a lithium-ion battery will significantly change its acoustic characteristic parameters: the change in electrolyte wettability can cause fluctuations in the ultrasonic attenuation coefficient; the change in material stiffness caused by lithium dendrite growth can shift the longitudinal wave speed; there is a definite functional relationship between the cavity resonance frequency generated during the gas production process and the gas volume. Compared with traditional expansion force monitoring technology, acoustic sensors have significant advantages. They can achieve non-contact monitoring through the characteristics of sound wave propagation, avoiding direct interference with the battery structure. At the same time, acoustic sensors show good adaptability under different environmental conditions and can reflect the internal state changes of the battery in real time. This highly sensitive feedback mechanism enables acoustic sensors to provide more accurate and comprehensive data support when evaluating the expansion characteristics of lithium-ion batteries, thus providing a reliable basis for battery design and safety improvement. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to provide a method for estimating the expansion force of a lithium-ion battery based on ultrasonic signals, realizing non-destructive online detection of the expansion force of the lithium-ion battery while reducing the detection cost.

[0007] To achieve the above purpose, the present invention provides the following technical solutions:

[0008] A method for estimating the expansion force of a lithium-ion battery based on ultrasonic signals specifically includes the following steps:

[0009] S1: Select a lithium-ion soft-pack battery, fix two ultrasonic sensors at intervals on the same side of the battery. Among them, the ultrasonic signal transmitting sensor is connected to a signal generator, and the ultrasonic signal receiving sensor is connected to an oscilloscope;

[0010] S2: Select several lithium-ion batteries integrated with ultrasonic sensors, apply stepped or sinusoidal fluctuation preloads of 0 MPa, 1 MPa, 2 MPa, 3 MPa, and 4 MPa to the batteries using a programmable dynamic load fixture, and collect expansion force data in real time through a high-precision force sensor. Synchronously connect to the battery test system, and design 0.5C, 1C, 2C constant current charge and discharge and NEDC dynamic working condition cycle experiments, and record voltage, current, temperature, and internal resistance data every 10 seconds;

[0011] S3: Use data processing software to batch process the collected waveform raw data and extract signal characteristics (including the signal amplitude of specific wave peaks and signal flight time), and extract the battery current data and expansion force data during the cycle. The processed ultrasonic signal characteristic data, as well as the collected current data and the expansion force data of the battery, form a data set;

[0012] S4: Construct a random forest model, and use the ultrasonic signal feature data and current data as the model inputs and the swelling force data as the model output to train the random forest model;

[0013] S5: Perform lightweight processing on the trained random forest model. Prune the decision trees with a contribution degree < 1% according to the Gini index for the random forest, deploy it to the Raspberry Pi 4B platform, and test the time and power consumption for a single prediction;

[0014] S6: Compare the prediction accuracies (MSE, RMSE, MAE) of the traditional BP, single random forest, and voltage-SOC linear models, simulate abnormal conditions such as overcharge (>4.3V) and over-discharge (<2.8V), verify the model warning mechanism and dynamic parameter adjustment ability, and output the standardized data set, lightweight model file, and visualization analysis report.

[0015] Further, in step S1, the ultrasonic sensor includes an ultrasonic signal transmitting sensor and an ultrasonic signal receiving sensor. The ultrasonic signal transmitting sensor is connected to a signal generator, and the signal generator controls the signal transmitting sensor to emit a five-cycle sine pulse wave with a center frequency of 80 kHz, emitting a pulse every 1 ms. The ultrasonic signal receiving sensor is connected to an oscilloscope, the oscilloscope is connected to a computer, and the computer controls the oscilloscope to collect the ultrasonic signal received by the ultrasonic signal receiving sensor every 3 s through data acquisition software.

[0016] Further, in step S2, the charge and discharge cycle process under different conditions is repeated for lithium-ion batteries under different loads, using constant current and constant voltage charging at different rates, and then constant current discharging.

[0017] Further, in step S4, the random forest model realizes the training and optimization of the model by constructing multiple decision trees and combining their outputs, specifically including the following steps:

[0018] S41: Model construction: Use multiple decision trees to construct a random forest model; each decision tree is generated from a different subset of the training set (by the bootstrap sampling method), and node splitting is performed using randomly selected features;

[0019] S42: Tree training: Each decision tree is independently trained using the corresponding subset to generate a prediction output for the input data;

[0020] S43: Integrated prediction process: Integrate all the trained decision trees, and generate the final prediction result, that is, the estimated value of the swelling force, by voting (in classification tasks) or taking the average (in regression tasks);

[0021] S44: Model evaluation: Calculate the performance metrics (such as mean squared error) of the model according to the prediction results of the random forest model and the actual labels to evaluate the effectiveness of the model;

[0022] S45: Model training and optimization: Repeatedly train and evaluate under different parameter conditions, and adjust model parameters (such as the number and depth of trees, etc.) to improve the generalization ability and prediction accuracy of the model.

[0023] Furthermore, in step S5, lightweight processing is performed on the trained random forest model. Specifically, the random forest prunes the decision trees with a contribution degree < 1% according to the Gini index, and deploys it to the Raspberry Pi 4B platform to test the time consumption and power consumption of a single prediction.

[0024] Furthermore, compare the prediction accuracies (MSE, RMSE, MAE) of traditional BP, single random forest, and voltage-SOC linear models, simulate abnormal conditions such as overcharge (>4.3V) and over-discharge (<2.8V), verify the model warning mechanism and dynamic parameter adjustment ability, and output a standardized data set, lightweight model file, and visualization analysis report.

[0025] The beneficial effects of the present invention are as follows:

[0026] (1) The sensor integration advantage is significant: Using miniaturized and low-cost ultrasonic sensors, it is easy to integrate into the battery management system to realize synchronous monitoring of multiple parameters (expansion force, SOC, temperature, etc.), breaking through the limitations of traditional single-function sensors.

[0027] (2) The cost-benefit is prominent: Through the multi-functional reuse of ultrasonic signals, the number of sensors is reduced, the hardware cost and system complexity are lowered, and at the same time, the data acquisition process is simplified.

[0028] (3) High-precision dynamic prediction ability: Based on the random forest model, fusing ultrasonic features (amplitude, flight time) and current data to achieve high-precision estimation of the expansion force (optimization of MSE, RMSE, MAE indicators), especially showing robustness in dynamic conditions (such as NEDC cycle).

[0029] (4) Real-time and lightweight deployment: The model is lightweight processed by pruning (Gini index screening), and low-time-consumption and low-power-consumption prediction are realized on the Raspberry Pi 4B platform, meeting the real-time requirements of the embedded system.

[0030] (5) Abnormal warning and adaptive adjustment: The model can trigger the warning mechanism under abnormal conditions such as overcharge / over-discharge, and supports dynamic parameter adjustment, improving the safety and reliability of the battery management system.

[0031] (6) Expandability and universality: Provide a standardized data set and visualization analysis tool, which is convenient for model iteration optimization and adaptation to different battery types or working conditions, and promotes the wide application of ultrasonic technology in battery health management.

[0032] The present invention not only has obvious advantages in terms of sensors, but also provides innovative solutions for improving the performance and saving costs of lithium-ion battery management systems, and has high practical and market application values.

[0033] Other advantages, objectives, and features of the present invention will to some extent be elaborated in the subsequent description, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following description. Brief Description of the Drawings

[0034] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0035] Figure 1 is a flowchart of the method for estimating the swelling force of a lithium-ion battery based on ultrasonic signals according to the present invention;

[0036] Figure 2 is a schematic diagram of the experimental platform framework in an embodiment of the present invention;

[0037] Figure 3 is a waveform diagram of ultrasonic signals in an embodiment of the present invention;

[0038] Figure 4 is a schematic diagram of the structure of a random forest model in an embodiment of the present invention;

[0039] Figure 5 is the prediction result of the random forest model in an embodiment of the present invention for the 0.5C constant current charge and discharge of a lithium-ion battery under a preload of 0.25 MPa;

[0040] Reference numerals: 1 - signal generator, 2 - fixture, 3 - ultrasonic signal transmitting sensor, 4 - ultrasonic signal receiving sensor, 5 - lithium-ion soft-pack battery, 6 - oscilloscope, 7 - data acquisition card, 8 - power sensor, 9 - battery test system. Detailed Embodiments

[0041] The following specific examples illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention schematically, and the following embodiments and the features in the embodiments can be combined with each other without conflict.

[0042] Among them, the attached drawings are only for illustrative purposes, showing only schematic diagrams rather than physical diagrams, and should not be construed as limiting the present invention; in order to better illustrate the embodiments of the present invention, some components in the attached drawings will be omitted, enlarged or reduced, which does not represent the size of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the attached drawings may be omitted.

[0043] In the attached drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the attached drawings. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the attached drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0044] Please refer to Figures 1 to 5 , the present invention provides a method for estimating the swelling force of a lithium-ion battery based on ultrasonic signals, including the following steps:

[0045] Step 1: Select a lithium-ion soft-pack battery, fix two ultrasonic sensors on the same side of the battery at a certain distance apart. The signal transmitting sensor is connected to a signal generator, and the signal receiving sensor is connected to an oscilloscope. The ultrasonic sensors include an ultrasonic signal transmitting sensor and an ultrasonic signal receiving sensor. The ultrasonic signal transmitting sensor is connected to the signal generator, and the signal generator controls the signal transmitting sensor to emit a five-cycle sine pulse wave with a center frequency of 80 kHz, and emits a pulse every 1 ms. The ultrasonic signal receiving sensor is connected to the oscilloscope, the oscilloscope is connected to a computer, and the computer controls the oscilloscope to collect the ultrasonic signals received by the ultrasonic signal receiving sensor every 3 s through data acquisition software.

[0046] Step 2: Select several lithium-ion batteries integrated with ultrasonic sensors as in Step 1, apply step or sine wave preloads of 0 MPa, 1 MPa, 2 MPa, 3 MPa, and 4 MPa to the lithium-ion batteries using a programmable dynamic load fixture, and collect the swelling force data in real time through a high-precision force sensor. Synchronously connect to the battery test system, and design 0.5C, 1C, 2C constant current charge and discharge and NEDC dynamic working condition cycle experiments, and conduct charge and discharge experiments under the following three working conditions under these five loads respectively:

[0047] Working condition 1: Constant current and constant voltage charge at 0.5C to 100% SOC, and constant current discharge at 0.5C to 0% SOC;

[0048] Condition 2: Constant current and constant voltage charging at 1C to 100% SOC, constant current discharging at 1C to 0% SOC;

[0049] Condition 3: Constant current and constant voltage charging at 2C to 100% SOC, constant current discharging at 2C to 0% SOC.

[0050] Step 3: Use data processing software to batch process the collected original waveform data and extract the signal amplitudes, signal flight times, and the expansion force of the battery during the cycle for the two peaks (the two peaks boxed in Figure 4 ). The collected signal feature data and the battery expansion force data form a data set. The ultrasonic signal data and current data are used as inputs for the expansion force estimation.

[0051] Step 4: Based on the collected ultrasonic signal feature data and load data, establish a random forest model. The random forest model is trained and optimized by constructing multiple decision trees and combining the outputs of each tree. The input layer of the model inputs a data set obtained by normalizing the collected ultrasonic signal feature data and load data, and the output layer outputs the expansion force of the lithium-ion battery. Specifically, it includes the following steps:

[0052] (1) Model construction: Use 100 decision trees to construct a random forest model. Each decision tree is generated from a different subset of the training set (by the bootstrap sampling method) and is split using randomly selected features to reduce the risk of overfitting.

[0053] (2) Tree training: Each decision tree is independently trained using the training set to generate a predicted output for the input data. The maximum depth of the tree is set to be unlimited to ensure that the model can capture the complex relationships in the data.

[0054] (3) Ensemble prediction process: Integrate the prediction results of all decision trees and use the average method to generate the final prediction result. In a regression task, calculate the average of the outputs of all decision trees as the predicted value of the expansion force of the lithium-ion battery.

[0055] (4) Model evaluation and optimization: Calculate the performance metrics (such as mean squared error (MSE)) of the model based on the prediction results of the random forest model and the actual labels, use the cross-validation method to evaluate the generalization ability of the model, and continuously adjust the model parameters (such as the number of trees, minimum sample split number, and maximum number of features, etc.) to improve the prediction accuracy of the model.

[0056] Step 5: Normalize the collected ultrasonic signal feature data and load data and divide them into a training set and a test set. Use 70% of the data as the training set and 30% of the data as the test set. Use the training set to train the random forest model and predict the expansion force of the lithium-ion battery based on this model.

[0057] Step 6: Compare the prediction accuracies (MSE, RMSE, MAE) of the traditional BP, single random forest, and voltage-SOC linear models, simulate abnormal conditions such as overcharging (>4.3V) and over-discharging (<2.8V), verify the model warning mechanism and dynamic parameter adjustment ability, and output the standardized data set, lightweight model file, and visualization analysis report.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for estimating the swelling force of a lithium-ion battery based on ultrasonic signals, characterized in that, The method specifically includes the following steps: S1: Select a lithium-ion soft-pack battery, and fixedly install two ultrasonic sensors at the same side of the battery at intervals. Among them, the ultrasonic signal transmitting sensor is connected to a signal generator, and the ultrasonic signal receiving sensor is connected to an oscilloscope; S2: Select several lithium-ion batteries integrated with ultrasonic sensors, apply a preload to the batteries using a programmable dynamic load fixture, and collect expansion force data in real time through a force sensor. Synchronously connect to a battery test system, and design constant current charge and discharge and NEDC dynamic working condition cycle experiments, and record voltage, current, temperature and internal resistance data; S3: Use data processing software to batch process the collected waveform raw data and extract signal features, and extract the battery current data and expansion force data during the cycle, the processed ultrasonic signal feature data, as well as the collected current data and the expansion force data of the battery to form a data set; S4: Construct a random forest model, and use the ultrasonic signal feature data and current data as model inputs, and the expansion force data as the model output to train the random forest model; S5: Perform lightweight processing on the trained random forest model, and use the lightweight processed model to estimate the expansion force of the lithium-ion battery.

2. The method for estimating the swelling force of a lithium-ion battery according to claim 1, wherein In step S1, the ultrasonic sensor includes an ultrasonic signal transmitting sensor and an ultrasonic signal receiving sensor. The ultrasonic signal transmitting sensor is connected to a signal generator, and the signal generator controls the signal transmitting sensor to emit a five-cycle sine pulse wave with a center frequency of 80 kHz, and emits a pulse every 1 ms; the ultrasonic signal receiving sensor is connected to an oscilloscope, the oscilloscope is connected to a computer, and the computer controls the oscilloscope to collect the ultrasonic signal received by the ultrasonic signal receiving sensor every 3 s through data acquisition software.

3. The method for estimating the swelling force of a lithium-ion battery according to claim 1, wherein In step S2, a stepped or sinusoidal fluctuating preload is applied to the battery.

4. The method for estimating the swelling force of a lithium-ion battery according to claim 1 or 3, characterized in that In step S2, the charge and discharge cycle process of different working conditions is repeated for lithium-ion batteries subjected to different magnitudes of loads, and constant current constant voltage charging at different rates is used, followed by constant current discharging.

5. The method for estimating the swelling force of a lithium-ion battery according to claim 1, wherein In step S3, the extracted ultrasonic signal features include the signal amplitude of a specific wave peak and the signal flight time.

6. The method for estimating the swelling force of a lithium-ion battery according to claim 1, wherein In step S4, the random forest model realizes the training and optimization of the model by constructing multiple decision trees and combining their outputs, specifically including the following steps: S41: Model construction: Use multiple decision trees to construct a random forest model; each decision tree is generated from a different subset of the training set, and node splitting is performed using randomly selected features; S42: Tree training: Each decision tree is independently trained using the corresponding subset to generate a predicted output for the input data; S43: Integrated prediction process: Integrate all trained decision trees, and generate a final prediction result, that is, an estimated value of the expansion force, by voting or taking an average; S44: Model evaluation: Calculate the performance index of the model according to the prediction result of the random forest model and the actual label to evaluate the effectiveness of the model; S45: Model training and optimization: Repeat training and evaluation under different parameter conditions, and adjust the model parameters to improve the generalization ability and prediction accuracy of the model.

7. The method for estimating the swelling force of a lithium-ion battery according to claim 1, wherein In step S5, the trained random forest model is lightweight processed. Specifically, the random forest prunes the decision trees with a contribution degree of <1% according to the Gini index, deploys them to the Raspberry Pi 4B platform, and tests the time and power consumption of a single prediction.

Images