A Sensorless Temperature Estimation Method and Device for Lithium-Ion Batteries

Abstract

The present invention discloses a method and device for sensorless temperature estimation of a lithium-ion battery. Considering the state of charge, temperature, and current, the method describes the non-linear dynamic time-varying characteristics of the lithium-ion battery through a state-dependent model; then optimizes the parameters of the state-dependent model and the lumped mass thermal model offline according to experimental data; finally, based on the state-dependent lithium-ion battery model, through the measurement feedback of the terminal voltage, the sensorless temperature estimation of the lithium-ion battery is realized by combining the unscented Kalman filter. This method enables the large-scale energy storage system to monitor the temperature of each battery cell in real time without using temperature sensors, facilitating the battery management system to perform thermal management on the battery.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery management, and in particular relates to a sensorless temperature estimation method and device for a lithium-ion battery. Background Art

[0002] Energy storage systems are currently widely used in areas such as transportation electrification and smart grids. Lithium-ion batteries, due to their advantages such as long cycle life, high energy density, and low self-discharge, are the preferred energy storage unit for energy storage systems. In the application of these energy storage systems, temperature is a critical state variable affecting the safety and economic viability of lithium-ion batteries. For example, lithium-ion batteries must operate at an appropriate temperature to ensure optimal performance. However, under certain high-current operating conditions, lithium-ion batteries can heat up rapidly, leading to excessively high real-time temperatures. Cooling measures are required to prevent thermal runaway and heat spread. Therefore, real-time and accurate temperature monitoring of the lithium-ion batteries that comprise the energy storage system is necessary to ensure safe and stable system operation. Existing large-scale energy storage systems consist of thousands of battery cells connected in series and parallel. For cost and safety reasons, temperature sensors can only be installed at key locations to obtain temperature information at a limited number of points. This makes it impossible for battery management systems to accurately monitor the temperature of each battery cell. Furthermore, if a single battery cell experiences thermal runaway, it can cause similar thermal runaway in adjacent cells, triggering a chain reaction with potentially catastrophic consequences. Therefore, estimating the temperature of each battery cell without the aid of sensors is an urgent problem that needs to be solved in the thermal management technology of energy storage systems. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a sensorless temperature estimation method and device for lithium-ion batteries, which uses collected battery data to estimate the temperature of lithium-ion battery cells in an energy storage system without installing a temperature sensor, thereby achieving real-time and effective battery temperature estimation.

[0004] In order to solve the above technical problems, the present invention discloses a sensorless temperature estimation method for a lithium-ion battery, comprising:

[0005] Obtain the data set of the battery to be tested and input it into the ampere-hour integration formula and the parameterized lumped mass thermal model to obtain the predicted state of charge and predicted temperature value;

[0006] Inputting the predicted state of charge and the predicted temperature value into a trained terminal voltage model to obtain a predicted terminal voltage;

[0007] Combining the measured terminal voltage of the battery to be tested and the predicted terminal voltage, an unscented Kalman filter is used to perform feedback correction to output a temperature estimate of the battery to be tested;

[0008] The terminal voltage model is constructed based on the state-dependent model.

[0009] Furthermore, the terminal voltage model is as follows:

[0010]

[0011] In the formula, U t is the terminal voltage, I is the current, p and q are the regression orders of the terminal voltage U t and the current I respectively, ζ0, ζ Ut,i , ζ I,i are functional coefficients, w(k) is the state variable, and e is Gaussian white noise.

[0012] Furthermore, the state variable w(k) = [SOC(k) I(k) T(k)];

[0013] In the formula, SOC is the state of charge, I is the current, and T is the temperature of the lithium-ion battery.

[0014] Furthermore, the functional coefficients ζ0, ζ Ut,i , ζ I,i are determined by a radial basis function neural network:

[0015]

[0016] In the formula, m is the number of neurons in the radial basis function neural network, and are the linear weights of the radial basis function neural network, is the center of the radial basis function neural network, is the scaling factor of the radial basis function neural network.

[0017] Furthermore, the lumped mass heat model is as follows:

[0018]

[0019] In the formula, T is the temperature of the lithium-ion battery, h is the equivalent convective heat transfer coefficient, A is the battery surface area, m q is the battery mass, c is the battery specific heat capacity, Δt is the sampling period, is the heat generation rate, T a is the ambient temperature.

[0020] Furthermore, the heat generation rate is calculated by the heat generation model ; in the formula, OCV is the open circuit voltage, and the open circuit voltage OCV is a function of the state of charge SOC.

[0021] Further, the open circuit voltage OCV is obtained by calculating through the open circuit voltage - state of charge curve; the open circuit voltage - state of charge curve is fitted by the following steps:

[0022] 001. Perform small - current charge - discharge tests on the lithium - ion battery, collect the small - current charge - discharge data of the lithium - ion battery, and the small - current charge - discharge data includes the terminal voltage and the state of charge;

[0023] 002. Take 1% of the state of charge as the interval as the state - of - charge points, extract the terminal voltage corresponding to the state - of - charge points as the open - circuit voltage points, and fit the state - of - charge points and the open - circuit voltage points into the open circuit voltage - state of charge curve through a polynomial function.

[0024] Further, the equivalent convective heat transfer coefficient h and the specific heat capacity c are optimized for parameters by using the particle swarm optimization algorithm.

[0025] Further, the ampere - hour integration formula is:

[0026]

[0027] In the formula, C b is the capacity of the lithium - ion battery.

[0028] Based on the same inventive concept, another aspect of the present invention further provides a lithium - ion battery sensorless temperature estimation device, and the device includes:

[0029] A lithium - ion battery sensorless temperature prediction module, which obtains the data set of the battery to be measured, inputs it into the ampere - hour integration formula and the parameterized lumped mass thermal model for processing to obtain the predicted state of charge and the predicted temperature value; inputs the predicted state of charge and the predicted temperature value into the trained terminal voltage model to obtain the predicted terminal voltage;

[0030] A lithium - ion battery sensorless temperature correction and output module, which combines the measured terminal voltage of the battery to be measured and the predicted terminal voltage, and uses the unscented Kalman filter for feedback correction to output the temperature estimation value of the battery to be measured.

[0031] The present invention has the following beneficial effects:

[0032] The present invention provides a method for estimating the temperature of a lithium-ion battery without a sensor. Considering the state of charge, temperature, and current, the method realizes the modeling of the terminal voltage of the lithium-ion battery through a state-dependent model. The established state-dependent terminal voltage model has the ability to reflect the nonlinearity and time-variation of the lithium-ion battery in different temperature ranges and operating conditions. A dynamic experiment is designed to optimize the parameters of the state-dependent model and the lumped mass thermal model offline according to the experimental data. Finally, based on the state-dependent model, through the measurement feedback of the terminal voltage and combined with the unscented Kalman filter, accurate temperature estimation of the lithium-ion battery is achieved. This method can monitor the temperature of a single battery in real time without using a temperature sensor, facilitating the thermal management of the battery by the battery management system. It is particularly suitable for energy storage systems containing a large number of batteries, reducing the cost of installing temperature sensors in the energy storage system and providing thermal safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is one of the flowcharts of the method for estimating the temperature of a lithium-ion battery without a sensor provided by some embodiments of the present invention.

[0034] Figure 2 is the second flowchart of the method for estimating the temperature of a lithium-ion battery without a sensor provided by some embodiments of the present invention.

[0035] Figure 3 is a schematic diagram of the data collected by some embodiments of the present invention, including the terminal voltage, current diagram, state of charge SOC, surface temperature, and ambient temperature.

[0036] Figure 4 is a schematic diagram comparing the true temperature and estimated temperature of the battery provided by an embodiment of the present invention.

[0037] Figure 5 is a schematic diagram of the error of the estimation result provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] In order to describe the technical solution of the present invention more clearly and completely, the present invention will be further described in detail below through specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Various changes can be made within the scope defined by the rights of the present invention.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0040] As Figure 1As shown in the figure, the present invention provides a method for estimating the temperature of a lithium-ion battery without a sensor, comprising the following steps:

[0041] S1. Obtain the data set of the battery to be measured, input it into the ampere-hour integration formula and the parameterized lumped mass thermal model for processing, and obtain the predicted state of charge data and the predicted temperature value;

[0042] S2. Input the predicted state of charge data and the predicted temperature value into the trained terminal voltage model to obtain the predicted terminal voltage;

[0043] S3. Combine the measured terminal voltage of the battery to be measured and the predicted terminal voltage, and use the unscented Kalman filter for feedback correction to output the temperature estimation value of the battery to be measured;

[0044] Among them, the terminal voltage model is constructed based on the state-dependent model.

[0045] In some embodiments provided by the present invention, the terminal voltage model is:

[0046]

[0047] In the formula, U t is the model output, that is, the terminal voltage, I is the model input, that is, the current, p and q are the regression orders of the output-terminal voltage U t and the input-current I respectively, ζ0, ζ Ut,i , ζ I,i are functional coefficients, w(k) is the state quantity, and e is Gaussian white noise.

[0048] In some embodiments provided by the present invention, according to the differences in the active materials, sizes, core forms, and external shapes of the lithium-ion batteries to be measured, batteries with the same model and specifications as the lithium-ion batteries to be measured are selected as the modeling and acquisition objects, and dynamic working condition tests are performed on them. The terminal voltage, current, temperature, ambient temperature, and SOC data of the lithium-ion batteries are collected, and then the temperature data in the dynamic working condition test data is subjected to Gaussian average filtering to reduce noise.

[0049] In some embodiments provided by the present invention, considering the significant influence of current, SOC, and the temperature of the lithium-ion battery on the terminal voltage of the lithium-ion battery, the state quantity w(k) in the terminal voltage model constructed by the state-dependent model is determined, w(k)=[SOC(k) I(k) T(k)]. In the formula, SOC is the state of charge, I is the current, and T is the temperature of the lithium-ion battery.

[0050] In some embodiments provided by the present invention, the functional coefficients ζ0, ζ Ut,i , ζ I,i in the terminal voltage model constructed according to the state-dependent model are determined by a radial basis function neural network:

[0051]

[0052] Wherein, m is the number of neurons of the radial basis function neural network, and are the linear weights of the radial basis function neural network, is the center of the radial basis function neural network, is the scaling factor of the radial basis function neural network. All parameters of the state-dependent model include linear weights, centers, and scaling factors.

[0053] According to the processed dynamic condition data described above, all parameters in the state-dependent model are optimized by the gradient descent method, and the terminal voltage model (1-1) established based on the state-dependent model is stored as the observation equation.

[0054]

[0055] In some embodiments provided by the present invention, the lumped mass heat model is:

[0056]

[0057] Wherein, T is the temperature of the lithium-ion battery, h is the equivalent convective heat transfer coefficient, A is the battery surface area, m q is the battery mass, c is the battery specific heat capacity, Δt is the sampling period, [[ID=—33]]is the heat generation rate, T a is the ambient temperature.

[0058] In some embodiments provided by the present invention, the heat generation rate is calculated by the heat generation model ; wherein, OCV is the open circuit voltage, and the open circuit voltage OCV is a function of the state of charge SOC.

[0059] In some embodiments provided by the present invention, the open circuit voltage OCV is obtained by calculating the open circuit voltage-state of charge curve; the open circuit voltage-state of charge curve is fitted by the following steps:

[0060] 001. Perform small current charge and discharge tests on the lithium-ion battery. For example, the lithium-ion battery is charged and discharged at a small current of 1 / 20C, and the small current charge and discharge data of the lithium-ion battery are collected. The small current charge and discharge data include the terminal voltage and the state of charge;

[0061] 002. Using 1% of the state of charge as an interval to obtain state of charge points, extracting the corresponding terminal voltage at the state of charge points as open circuit voltage points, and fitting the state of charge points and the open circuit voltage points into an open circuit voltage - state of charge curve through a polynomial function.

[0062] In some embodiments provided by the present invention, according to the processed dynamic operating condition data, the particle swarm optimization algorithm is used to optimize the parameters of the equivalent convective heat transfer coefficient h and the specific heat capacity c, and the optimized lumped mass heat model is stored:

[0063]

[0064] In some embodiments provided by the present invention, the state of charge SOC of the lithium - ion battery is obtained by the ampere - hour integration formula:

[0065]

[0066] where C b is the capacity of the lithium - ion battery, which is obtained from the battery manual.

[0067] Taking (1 - 2) and (1 - 3) as state equations; the state equations combined with the observation equation can predict the terminal voltage, temperature and state of charge of the lithium - ion battery to be measured and combine the feedback correction of the unscented Kalman filter to output the estimated value of the lithium - ion temperature, so that the temperature of the lithium - ion battery can be obtained without a temperature sensor. As Figure 2 shown, the specific process is as follows: First, input the measured ambient temperature, current and the calculated heat generation rate into the state equation to obtain the state of charge and temperature predicted by the model. Then input the state of charge and temperature into the observation equation to obtain the terminal voltage predicted by the model. There is a difference between the actually measured terminal voltage of the lithium - ion battery and the terminal voltage predicted by the model. Through the error value between the two and combining the feedback correction ability of the unscented Kalman filter, the predicted temperature is corrected, and the corrected temperature, that is, the estimated temperature of the lithium - ion battery, can be output.

[0068] The following will illustrate the lithium - ion battery sensor - less temperature estimation method provided by the present invention with specific embodiments:

[0069] Embodiment

[0070] This embodiment provides a lithium - ion battery sensor - less temperature estimation method, including two stages: offline establishment and online application.

[0071] Offline establishment stage:

[0072] According to the differences in the active materials, sizes, core forms and outer shapes of the lithium - ion batteries to be measured, batteries with the same model and specifications as the lithium - ion batteries to be measured are selected as the acquisition objects.

[0073] (1.1) Perform small current (1 / 20C) charge and discharge tests on the lithium-ion battery to be collected, and collect the terminal voltage and state of charge (SOC) data of the lithium-ion battery.

[0074] (1.2) Extract the terminal voltage of the lithium-ion battery at intervals of 1% SOC from the small current charge and discharge data collected in (1.1). Since the current is very small, the polarization effect can be ignored, and the terminal voltage can be considered as the open circuit voltage (OCV). Thus, the open circuit voltage OCV corresponding to each SOC point can be obtained. Finally, these points are fitted into an OCV-SOC curve through a polynomial function.

[0075] (1.3) Conduct dynamic operating condition tests on the lithium-ion battery, and collect the terminal voltage, current, temperature, ambient temperature, and state of charge SOC data of the lithium-ion battery, as Figure 3 shown. Then, perform Gaussian average filtering on the temperature data in the dynamic operating condition test data to reduce noise.

[0076] (1.4) Construct a terminal voltage model of the lithium-ion battery through a state-dependent model:

[0077]

[0078] where U t is the model output, i.e., the terminal voltage; I is the model input, i.e., the current; p and q are the regression orders of the output and input respectively; e is Gaussian white noise. ζ0, ζ Ut,i , ζ I,i are the functional coefficients of the model; w(k) is the state quantity of the model;

[0079] (1.5) Considering the influence of current, SOC, and temperature on the terminal voltage of the lithium-ion battery, determine the state quantity in the state-dependent model (1.4):

[0080] w(k) = [SOC(k) I(k) T(k)]

[0081] where SOC is the SOC of the lithium-ion battery, I is the current, and T is the temperature of the lithium-ion battery.

[0082] (1.6) Determine the functional coefficients through a radial basis function neural network with local approximation ability:

[0083]

[0084] where m is the number of neurons in the radial basis function neural network; and is the linear weight of the radial basis function neural network; is the center of the radial basis function neural network; is the scaling factor of the radial basis function neural network. All parameters of the state-dependent model include the linear weight, center, and scaling factor.

[0085] (1.7) Further obtains the following complete state-dependent model.

[0086]

[0087] (1.8) According to the processed dynamic condition data in (1.3), optimize all parameters in the state-dependent model by the gradient descent method and store the model.

[0088] (1.9) According to the collected dynamic condition data in (1.3), use the particle swarm algorithm to optimize the parameters of the specific heat capacity c and the equivalent convective heat transfer coefficient h of the following lumped mass heat model of the battery, and store the lumped mass heat model.

[0089]

[0090] Where A is the surface area; m q is the mass, and these two parameters can be obtained from the battery manual. c is the specific heat capacity of the lithium-ion battery; h is the equivalent convective heat transfer coefficient, and these two parameters need to be optimized. Δt is the sampling period. T a is the ambient temperature. is the heat generation rate, and the heat generation rate is calculated by the following heat generation model.

[0091]

[0092] Where OCV is the open circuit voltage of the lithium-ion battery OCV, which is a function of the state of charge SOC of the lithium-ion battery and is obtained by calculating through the OCV-SOC curve in (1.2).

[0093] (1.10) Determine the ampere-hour integration formula for the SOC of the lithium-ion battery:

[0094]

[0095] Where C b is the capacity of the lithium-ion battery and is obtained from the battery manual.

[0096] (1.11) Take the lumped mass heat model established in (1.9) and the ampere-hour integration formula in (1.10) as the state equation:

[0097]

[0098] (1.12) Use the state-dependent model established in (1.8) as the observation equation:

[0099]

[0100] (2) Online application stage:

[0101] (2.1) The process of online estimating the temperature of a lithium-ion battery is as Figure 2 . First, input the measured ambient temperature, current, and calculated heat generation rate into the state equation to obtain the predicted SOC and temperature of the model. Then, input the predicted SOC and temperature into the observation equation to obtain the predicted terminal voltage of the model. Since there is a difference between the actually measured terminal voltage of the lithium-ion battery and the predicted terminal voltage of the model, through the error value between the two, combined with the feedback correction ability of the unscented Kalman filter, the corrected temperature is output. In this embodiment, the comparison diagram of the true temperature and the estimated temperature is as Figure 4 shown, and the error diagram of the estimated temperature is as Figure 5 shown. As can be seen from Figure 4 , Figure 5 , the estimated temperature obtained by the lithium-ion battery temperature estimation method provided by the present invention for estimating the temperature of the battery to be measured is basically consistent with the true temperature. The positive and negative errors of the estimated temperature obtained by the lithium-ion battery sensorless temperature estimation method do not exceed 1 °C, indicating that the estimation method provided by the present invention can accurately estimate the temperature of a single lithium-ion battery without using a sensor, only using the ambient temperature, current, and terminal voltage collected in the existing battery management system.

[0102] Another aspect of the present invention also provides a lithium-ion battery sensorless temperature estimation device, which includes:

[0103] A lithium-ion battery sensorless temperature prediction module, which obtains the data set of the battery to be measured, inputs it into the ampere-hour integration formula and the parameterized lumped mass heat model for processing to obtain the predicted state of charge and the predicted temperature value; inputs the predicted state of charge and the predicted temperature value into the trained terminal voltage model to obtain the predicted terminal voltage;

[0104] A lithium-ion battery sensorless temperature correction and output module, which combines the actually measured terminal voltage of the battery to be measured and the predicted terminal voltage, and uses the unscented Kalman filter for feedback correction to output the temperature estimation value of the battery to be measured.

[0105] The above has introduced in detail a lithium-ion battery sensorless temperature estimation method and device provided by the present invention. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A sensorless temperature estimation method for a lithium-ion battery, characterized in that, Including: Obtain a data set of the battery to be measured, input it into the ampere-hour integration formula and the parameterized lumped mass thermal model for processing to obtain the predicted state of charge and the predicted temperature value; Input the predicted state of charge and the predicted temperature value into the trained terminal voltage model to obtain the predicted terminal voltage; Combine the measured terminal voltage of the battery to be measured and the predicted terminal voltage, and use unscented Kalman filtering for feedback correction to output the temperature estimated value of the battery to be measured; The terminal voltage model is constructed based on the state-dependent model; The terminal voltage model is: Where, U t is the terminal voltage, I is the current, p and q are the regression orders of the terminal voltage U t and the current I respectively, ζ0, ζ Ut,i , ζ I,i are functional coefficients, w(k) is the state variable, and e is the Gaussian white noise.

2. The sensorless temperature estimation method for a lithium-ion battery according to claim 1, wherein The state quantity w(k)=[SOC(k) I(k) T(k)]; In the formula, SOC is the state of charge, I is the current, and T is the temperature of the lithium-ion battery.

3. The sensorless temperature estimation method for a lithium-ion battery according to claim 1, characterized in that, The functional coefficients ζ0, ζ Ut,i , ζ I,i are determined by a radial basis function neural network: where m is the number of neurons in the radial basis function neural network, and are the linear weights of the radial basis function neural network, is the center of the radial basis function neural network, is the scaling factor of the radial basis function neural network.

4. The sensorless temperature estimation method for a lithium-ion battery according to any one of claims 1 to 3, characterized in that, The lumped mass thermal model is: Where, T is the temperature of the lithium-ion battery, h is the equivalent convective heat transfer coefficient, A is the battery surface area, m q is the battery mass, c is the battery specific heat capacity, Δt is the sampling period, is the heat generation rate, T a is the ambient temperature.

5. The method for estimating the temperature of a lithium-ion battery without a sensor according to claim 4, wherein, The heat generation rate is calculated by a heat generation model ; wherein, OCV is the open circuit voltage, and the open circuit voltage OCV is a function of the state of charge SOC.

6. The sensorless temperature estimation method for a lithium-ion battery according to claim 5, wherein The open-circuit voltage OCV is obtained by calculating through the open-circuit voltage-state of charge curve; the open-circuit voltage-state of charge curve is fitted by the following steps:

001. Perform small current charge and discharge tests on the lithium-ion battery, collect small current charge and discharge data of the lithium-ion battery, and the small current charge and discharge data includes the terminal voltage and the state of charge; 002. Take 1% of the state of charge as the interval as the state of charge points, extract the terminal voltage corresponding to the state of charge points as the open-circuit voltage points, and fit the state of charge points and the open-circuit voltage points into the open-circuit voltage-state of charge curve through a polynomial function.

7. The sensorless temperature estimation method for a lithium-ion battery according to claim 4, characterized in that, The equivalent convective heat transfer coefficient h and the specific heat capacity c are optimized by the particle swarm algorithm.

8. The sensorless temperature estimation method for a lithium-ion battery according to any one of claims 1 to 3, characterized in that The ampere-hour integration formula is: where C b is the capacity of the lithium-ion battery.

9. A sensorless temperature estimation device for a lithium-ion battery, characterized in that, The device includes: a lithium-ion battery sensorless temperature prediction module, which obtains a data set of the battery to be measured, inputs it into the ampere-hour integration formula and the parameterized lumped mass thermal model for processing to obtain the predicted state of charge and the predicted temperature value; inputs the predicted state of charge and the predicted temperature value into the trained terminal voltage model to obtain the predicted terminal voltage; a lithium-ion battery sensorless temperature correction output module, which combines the measured terminal voltage of the battery to be measured and the predicted terminal voltage, and uses unscented Kalman filtering for feedback correction to output the temperature estimated value of the battery to be measured; The terminal voltage model is: Where U t is the terminal voltage, I is the current, p and q are the regression orders of the terminal voltage U t and the current I respectively, ζ0, ζ Ut,i , ζ I,i are functional coefficients, w(k) is the state variable, and e is Gaussian white noise.

Images