Lithium ion battery state-of-charge estimation method, system, equipment and medium

Abstract

The invention relates to a lithium ion battery charge state estimation method and system. The method comprises the following steps: acquiring initial current, initial voltage and temperature data of a battery under different working conditions; inputting the initial voltage and the initial current data into a mechanism model to obtain voltage simulation data and current simulation data; inputting the voltage simulation data, the current simulation data and the temperature data into a neural network for training, constructing an SOC estimation model, and obtaining an SOC output result; and according to the SOC output result and the SOC true value, respectively calculating a root mean square error RMSE and a mean absolute error MAE for verification and testing so as to evaluate the estimation performance of the SOC estimation model. According to the method, the calculation speed of the SOC estimation process of the battery is high, the relative precision is high, a solid foundation is provided for a battery management system to guarantee safe and reliable operation of a battery power supply system, and the applicability of the SOC estimation model under different operation conditions and temperatures is improved.

Description

Technical Field

[0001] The present application relates to the field of battery management and control, and in particular to a method, system, device and medium for estimating the state of charge of a lithium-ion battery. Background Art

[0002] With the continuous development of new energy technologies, more and more application scenarios are using new energy technologies, and the widespread use of lithium-ion batteries is one example. Lithium-ion batteries are widely used in the electric vehicle industry due to their long cycle life, high energy density, and stable performance. The battery management system (BMS) can provide protection for the safe and healthy use of batteries in electric vehicles. Among them, the battery state of charge (SOC) is a very important status indicator, which provides a basis for the detection and control of other battery management systems. However, there is currently a lack of direct measurement methods for SOC, so how to accurately estimate SOC is a very challenging proposition.

[0003] The research on SOC estimation methods mainly includes four categories: open circuit voltage method, ampere-hour integration method, mechanism model-based method and data-driven method. The open circuit voltage method is to establish a lookup table relationship by measuring the mapping between open circuit voltage and battery SOC. This method requires precise experimental measurement to obtain a very accurate mapping relationship, and the voltage needs to be kept stable during measurement to eliminate the influence of overpotential, which makes each measurement take a lot of time. These problems make this method not easy to implement in practical applications. The ampere-hour integration method uses the integral of current release to calculate the battery capacity change to obtain the battery SOC. This method has the characteristics of simple calculation and high efficiency, but as an open-loop method, its performance is affected by sampling noise, that is, there are high initial estimation requirements and increased cumulative errors. The method based on the mechanism model combines the battery model with the nonlinear state space algorithm to overcome the shortcomings of the above methods. It can be specifically divided into equivalent circuit model and electrochemical model. The equivalent circuit model simulates the voltage change process by equivalently treating the battery working process as a circuit composed of devices such as voltage source, resistor, capacitor, etc. This method has a relatively simple structure and high calculation efficiency, but the overall accuracy is not high and is limited by factors such as ambient temperature. The typical electrochemical model mainly describes the working behavior of the battery by establishing the microscopic process inside the battery. The model established by this method contains many partial differential equations and a large number of corresponding parameters. Therefore, although it is relatively accurate, the calculation speed is slow.

[0004] As computer technology has been developing continuously in recent years, data-driven SOC estimation methods have attracted more and more attention from the public. This method can directly extract the corresponding information from the input of current and voltage and map it to SOC, thus avoiding the complex modeling process of battery behavior and achieving reasonable accuracy and practicality. As an important part of machine learning, deep learning (DL) has also been used for SOC estimation, especially recurrent neural networks (RNNs), which are good at processing data sequentially. However, although the SOC estimation method based on deep learning is powerful and easy to implement, its accuracy still needs to be improved. Since lithium-ion batteries are a very complex nonlinear system, the existing deep learning SOC estimation method, as a purely data-driven "black box" model, ignores the electrochemical principles of the battery. This defect makes it difficult for the SOC estimation method based on deep learning to further improve its accuracy.

[0005] Therefore, there is an urgent need to provide a lithium-ion battery state of charge estimation method to solve the above-mentioned technical problems. Summary of the invention

[0006] The present application provides a lithium-ion battery state of charge estimation method, system, device and medium, which combine the mechanism information of the equivalent circuit model with the neural network to estimate the battery SOC, taking into account the applicable scenarios of the battery under different operating conditions and temperatures, and has a fast calculation speed and high relative accuracy, providing a solid foundation for the battery management system to ensure the safe and reliable operation of the battery power supply system.

[0007] In a first aspect, the present application provides a method for estimating the state of charge of a lithium-ion battery, comprising the following steps: obtaining initial current, initial voltage and temperature data of the battery under different operating conditions; inputting the initial voltage and the initial current data into a mechanism model to obtain voltage simulation data and current simulation data; inputting the voltage simulation data, the current simulation data and the temperature data into a neural network for training, constructing a SOC estimation model, and obtaining an SOC output result; and calculating the root mean square error RMSE and the mean absolute error MAE based on the SOC output result and the true SOC value for verification and testing to evaluate the estimation performance of the SOC estimation model.

[0008] Optionally, in the step, the initial voltage and the initial current data are input into the mechanism model to obtain voltage simulation data and current simulation data, and the mechanism model includes: the mechanism model selects an equivalent circuit model; the equivalent circuit model specifically adopts a second-order Thevenin model, and the second-order Thevenin model consists of two series-connected RC loops, a resistor and an ideal voltage source, and is described by formula (1):

[0009]

[0010] Where U 1 , U 2 R 1 C 1 and R 2 C 2 The polarization voltage corresponding to the branch, U 0 For R 1 The corresponding ohmic voltage.

[0011] Optionally, the initial voltage and the initial current data are input into a mechanism model to obtain voltage simulation data and current simulation data, including: inputting the initial voltage and the initial current data into the equivalent circuit to obtain a voltage signal and a current signal; deconstructing the voltage signal output by the equivalent circuit into an open circuit voltage Uocv to obtain voltage simulation data; calculating the current signal output by the equivalent circuit according to the ampere-hour integration method to obtain the cumulative current ΔI to obtain current simulation data.

[0012] Optionally, the voltage signal output by the equivalent circuit is deconstructed into an open circuit voltage Uocv to obtain voltage simulation data, including: discretizing the equivalent circuit to obtain the Laplace equation of the equivalent circuit model, which is described by formula (2):

[0013]

[0014] Where s represents the response of the system in the complex frequency domain; the Laplace equation of the equivalent circuit model is bilinearly transformed to obtain the differential equation of the system input and output of the equivalent circuit model, which is described by formula (3):

[0015] y(k)=U ocv (k)-U(k)

[0016] a1y(k-1)+a2(k2)+a3I(k)+a4I(k-1)+asI(k-2)#(3)

[0017] Where I(k) is the current input of the system, and y(k) is the system output;

[0018] According to the differential equation, the input and output matrix at a specific time and the parameter matrix to be estimated are obtained, which are described by formulas (4) and (5):

[0019]

[0020] θ(k)=[a 1 a 2 a 3 a 4 a 5 ]#(5),

[0021] In the formula, refers to the input and output matrix at time K, and θ(k) refers to the parameter matrix to be estimated;

[0022] The recursive equation of the equivalent circuit model is obtained according to the input-output matrix and the parameter matrix, and the parameter identification result is obtained from the equivalent circuit model by sampling least squares method to obtain the open circuit voltage Uocv.

[0023] Optionally, the current signal output by the equivalent circuit is calculated according to the ampere-hour integration method to obtain the cumulative current ΔI, and the current simulation data is obtained, including: integrating the current signal and time according to the ampere-hour integration method to obtain the cumulative current ΔI, so as to obtain the change value of the SOC, which is described by formula (6):

[0024]

[0025] In the formula, η represents the discharge efficiency of the battery, and Qn represents the rated capacity of the battery.

[0026] Optionally, the voltage simulation data, the current simulation data and the temperature data are input into a neural network for training, a SOC estimation model is constructed, and a SOC output result is obtained, including: inputting the voltage simulation data, the current simulation data and the temperature data into a neural network for training, and a SOC estimation model is constructed; modifying the loss function of the SOC estimation model, comparing the SOC estimation value with the SOC true value based on the original mean square error MSE and then adding a penalty term to obtain the SOC output result to meet the SOC accuracy estimation.

[0027] Optionally, the loss function of the SOC estimation model is modified, and a penalty term is added to compare the SOC estimated value with the SOC true value based on the original mean square error MSE to obtain an SOC output result to meet the SOC accuracy estimation, including: comparing the SOC true value SOCi with the SOC estimated value obtained according to the SOC estimation model Compare; if the SOC estimation value If the SOC is greater than the true value SOCi, the penalty term W1 is added to the calculated loss function; otherwise, the penalty term W2 is added to the calculated loss function; wherein W1 and W2 are both user-defined parameters, and the penalty term W2 is less than the penalty term W2.

[0028] Optionally, the root mean square error RMSE and the mean absolute error MAE are calculated based on the SOC output result and the true SOC value for verification and testing to evaluate the estimation performance of the SOC estimation model, including: evaluating the standard deviation of the estimation error of the SOC estimation model under different operating conditions and temperatures according to the root mean square error RMSE; evaluating the robustness of the SOC estimation model under different operating conditions and temperatures according to the mean absolute error MAE.

[0029] In a second aspect, the present application provides a lithium-ion battery state of charge estimation system, which includes: an acquisition module, used to obtain the initial current, initial voltage and temperature data of the battery under different operating conditions; a processing module, used to input the initial voltage and the initial current data into a mechanism model to obtain voltage simulation data and current simulation data; a model construction module, used to input the voltage simulation data, the current simulation data and the temperature data into a neural network for training, construct a SOC estimation model, and obtain an SOC output result; an evaluation module, used to calculate the root mean square error RMSE and the mean absolute error MAE according to the SOC output result and the true value of the SOC for verification and testing, so as to evaluate the estimation performance of the SOC estimation model.

[0030] In a third aspect, the present application also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-described method when executing the computer program.

[0031] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned method are implemented.

[0032] This application has at least the following advantages:

[0033] The above steps are mainly to obtain the initial current, initial voltage and temperature data of the battery under different operating conditions, input the initial voltage and the initial current data into the mechanism model, and further extract the mechanism information of the battery and input it into the neural network, so as to improve the estimation accuracy and the calculation speed. The battery SOC is estimated by combining the mechanism information of the equivalent circuit model with the neural network, taking into account the applicable scenarios of the battery under different operating conditions and temperatures, and the calculation speed is fast and the relative accuracy is high, which provides a solid foundation for the battery management system to ensure the safe and reliable operation of the battery power supply system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A diagram showing an application environment of a method for estimating a state of charge of a lithium-ion battery in one embodiment;

[0035] Figure 2 A schematic diagram showing the steps of a method for estimating the state of charge of a lithium-ion battery in one embodiment;

[0036] Figure 3 A flowchart showing a method for estimating a state of charge of a lithium-ion battery in one embodiment;

[0037] Figure 4 A schematic diagram showing a second-order Thevenin model in one embodiment;

[0038] Figure 5 A schematic diagram showing a second-order Thevenin model in one embodiment;

[0039] Figure 6 A diagram showing SOC estimation results at 25° C. in one embodiment;

[0040] Figure 7 is a graph showing an SOC estimation error at 25° C. in one embodiment;

[0041] Figure 8 A diagram showing SOC estimation results under different working conditions in one embodiment;

[0042] Fig. 9 A structural block diagram showing a lithium-ion battery state of charge estimation system in one embodiment;

[0043] Fig.10 FIG. 4 is a schematic structural diagram of a computer device in one embodiment. DETAILED DESCRIPTION

[0044] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0045] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form also means

[0046] The figures include plural forms, and further, it should be understood that when the terms “include” and / or “comprise” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0047] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. However, it will be appreciated by those skilled in the art that in the embodiments of the present application, many technical details are proposed in order to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present application can also be implemented. The markings of the following embodiments are for convenience of description and should not constitute any limitation to the specific implementation of the present application. The various embodiments can be combined with each other and quoted from each other without contradiction.

[0048] Figure 2 A schematic diagram of a flow chart of a method for estimating the state of charge of a lithium-ion battery provided in an embodiment of the present application. The method can be performed as follows: Figure 1 The user space file server is executed in the system shown. Figure 2 , Figure 3 As shown, the method may include the following steps:

[0049] S201, obtaining initial current, initial voltage and temperature data of the battery under different working conditions;

[0050] S202, inputting the initial voltage and initial current data into the mechanism model to obtain voltage simulation data and current simulation data;

[0051] S203, inputting the voltage simulation data, the current simulation data and the temperature data into a neural network for training, constructing an SOC estimation model, and obtaining an SOC output result;

[0052] S204, calculating the root mean square error RMSE and the mean absolute error MAE according to the SOC output result and the true SOC value for verification and testing to evaluate the estimation performance of the SOC estimation model.

[0053] In the embodiment of the present application, the battery's mechanism information is obtained by inputting the initial voltage and initial current data of the battery into the mechanism model, that is, the voltage simulation data and the current simulation data are used as the input data of the neural network, so as to estimate the SOC of the battery by combining the mechanism information with the neural network, so as to achieve better estimation performance and scope of application.

[0054] The following is a detailed description of each step:

[0055] Please refer to Figure 2 , Figure 3 As shown, step S201, obtaining initial current, initial voltage and temperature data of the battery under different working conditions.

[0056] In this embodiment, it should be noted that the initial current, initial voltage and temperature data of the battery under different operating conditions are obtained, wherein the initial current, initial voltage and temperature data of the battery are specifically a public data set collected by Dr. Phillip Kollmeyer (phillip.kollmeyer@gmail.com) of the University of Wisconsin-Madison.

[0057] In one example, for batteries under different working conditions, the battery model is Panasonic's 18650PF battery with a rated capacity of 2.9Ah. The test temperature range is -20 to 25°C, and the test conditions include intense conditions with high vehicle speeds (US06), standard test conditions for fuel economy under high-speed conditions in the United States (HWFET), urban road cycle conditions (UDDS), driving simulation conditions (LA92) formulated based on the urban traffic characteristics of Los Angeles, USA, and a variety of combined conditions.

[0058] Please continue to refer to Figure 2 , Figure 3 As shown, in step S202, the initial voltage and initial current data are input into the mechanism model to obtain voltage simulation data and current simulation data.

[0059] In this embodiment, it should be noted that the open circuit voltage has a more direct mapping relationship with the SOC than the terminal voltage, so the terminal voltage replacement input method proposed in this embodiment is implemented through an equivalent circuit model. The mechanism model selects the equivalent circuit model. The reason for selecting the equivalent circuit model instead of the electrochemical model is that there is a better balance between the complexity and accuracy of the equivalent circuit model. Specifically, Figure 4 As shown in the figure, the equivalent circuit model adopts the second-order Thevenin model, which consists of two series-connected RC loops, a resistor, and an ideal voltage source. In the battery system, the two series-connected RC loops of the equivalent circuit model can simulate the electrochemical polarization and concentration polarization processes of the battery respectively. The series resistor R 0 It can represent the ion transfer resistance inside the battery system, and the ideal voltage source represents the amount of power that the battery can provide under ideal conditions. The model is described by formula (1):

[0060]

[0061] Where U 1 , U 2 R 1 C 1 and R 2 C 2 The polarization voltage corresponding to the branch, U 0 For R 1 The corresponding ohmic voltage.

[0062] After the equivalent circuit model is determined, the initial voltage data and the initial current data are input into the equivalent circuit to obtain voltage simulation data and current simulation data, specifically including: inputting the initial voltage data and the initial current data into the equivalent circuit to obtain voltage signals and current signals; deconstructing the voltage signal output by the equivalent circuit into the open circuit voltage Uocv to obtain voltage simulation data; calculating the current signal output by the equivalent circuit according to the ampere-hour integration method to obtain the cumulative current ΔI to obtain current simulation data. The initial voltage and initial current data of the battery are replaced with the open circuit voltage Uocv through the equivalent circuit model, and the cumulative current ΔI is the input data of the subsequent SOC estimation model, thereby improving the estimation accuracy.

[0063] In one example, the voltage signal output by the equivalent circuit is deconstructed into the open circuit voltage Uocv to obtain voltage simulation data. Specifically, the Laplace equation of the equivalent circuit model is obtained by discretizing the equivalent circuit, which is described by formula (2):

[0064]

[0065] Where s represents the response of the system in the complex frequency domain;

[0066] The Laplace equation of the equivalent circuit model is bilinearly transformed to obtain the differential equation of the input and output of the equivalent circuit model, which is described by formula (3):

[0067] y(k)=U ocv (k)-U(k)

[0068] a 1 y(k-1)+a 2 y(k-2)+a 3 I(k)+a 4 I(k-1)+a 5 I(k-2)#(3),

[0069] Where I(k) is the current input of the system, and y(k) is the system output;

[0070] According to the differential equation, the input and output matrix at a specific time and the parameter matrix to be estimated are obtained, which are described by formulas (4) and (5):

[0071]

[0072] θ(k)=[a 1 a 2 a 3 a 4 a 5 ]#(5),

[0073] In the formula, refers to the input and output matrix at time K, and θ(k) refers to the parameter matrix to be estimated;

[0074] The recursive equation of the equivalent circuit model is obtained according to the input-output matrix and the parameter matrix, and the parameter identification result is obtained from the equivalent circuit model by sampling least squares to obtain the open circuit voltage Uocv. Figure 5 As shown, specifically, the obtained input-output matrix and the parameter matrix to be estimated are mainly used as recursive equations. During the operation of the system, each time a new set of data is measured, the predicted value of the previous moment is corrected by the recursive formula to obtain the estimated value of the parameter value at the current moment, that is, the new parameter estimate = the old parameter estimate + the corrected value, and finally the open circuit voltage Uocv is obtained.

[0075] Regarding the calculation of accumulated current, since the ampere-hour integration method is a simple and effective method for calculating SOC, the ampere-hour integration method is used to accumulate the current and time to obtain the change value of SOC, which is described by formula (6):

[0076]

[0077] Among them, η represents the discharge efficiency of the battery, and Qn represents the rated capacity of the battery. The ampere-hour integration method is a simple and effective method to calculate SOC. Here, the current input is replaced with the accumulated current input to obtain the integral information of the current, which further helps the neural network learn the input information characteristics to improve the estimation accuracy of the SOC estimation model.

[0078] Reference Figure 2 , Figure 3 As shown, step S203, inputting the voltage simulation data, the current simulation data and the temperature data into the neural network for training, constructing the SOC estimation model, and obtaining the SOC output result, includes: inputting the voltage simulation data, the current simulation data and the temperature data into the neural network for training, and constructing the SOC estimation model; modifying the loss function of the SOC estimation model, comparing the SOC estimation value with the SOC true value on the basis of the original mean square error MSE and then adding a penalty term, to obtain the SOC output result to meet the SOC accuracy estimation.

[0079] In this embodiment, it should be noted that the ampere-hour integration method itself has some defects in practical applications, that is, there is the influence of cumulative errors, and the error becomes larger towards the end of the estimation. Therefore, when the ampere-hour integration method is used to obtain current integration information, the loss function during the training of the SOC estimation model will be modified accordingly, and the error offset term at the end will be penalized, thereby improving the accuracy of the estimation.

[0080] Specifically, the original mean square error MSE expression is as follows:

[0081]

[0082] In the formula, n represents the total number of samples, SOC i Indicates the reference SOC value, The penalty term is added on the basis of the original mean square error MSE, and the true value of SOC SOC i Compared with the SOC estimation value obtained according to the SOC estimation model For comparison, if the SOC estimate Greater than the actual SOC value SOC i , then the loss function is calculated by adding the penalty term W 1 Otherwise, the penalty term W is added to the loss function 2 , described by the formula:

[0083]

[0084] Where W 1 and W 2 are user-defined parameters, and the penalty term W 1 <W 2 .

[0085] Reference Figure 2 , Figure 3 As shown, step S204, based on the SOC output result and the true value of the SOC, the root mean square error RMSE and the mean absolute error MAE are respectively calculated for verification and testing to evaluate the estimation performance of the SOC estimation model, including: evaluating the standard deviation of the estimation error of the SOC estimation model under different operating conditions and temperatures according to the root mean square error RMSE; evaluating the robustness of the SOC estimation model under different operating conditions and temperatures according to the mean absolute error MAE.

[0086] In this embodiment, it should be noted that RMSE can evaluate the standard deviation of the estimation error, and MAE can evaluate the robustness of the model, which are respectively expressed as follows:

[0087]

[0088] In order to verify the SOC estimation performance of this embodiment under complex driving conditions, we first verified and discussed the results of the method proposed in this embodiment on the data set collected at 25°C for the battery, and then discussed the influence of different ambient temperatures, and obtained good results under different test conditions.

[0089] Application 1: Results at 25°C under normal conditions

[0090] According to the above operation steps, the SOC estimation results of the proposed method at 25°C were first verified. The battery model used is Panasonic's 18650PF battery with a rated capacity of 2.9Ah. For comparison, the LSTM estimation model using only three inputs of current, voltage and temperature is used as the benchmark, called LSTM; the estimation model using open circuit voltage parameters is marked as LSTM-1, the estimation model using cumulative current and loss function is marked as LSTM-2, and the estimation model used by all improved methods is recorded as LSTM-1&2. The test uses a total of four working conditions including US06, HWFET, UDDS, and LA92, of which the training set is the first three cycles of cycle1 to cycle4 (the four batches of data are a mixture of the first four working conditions), the validation set is the remaining cycle4, and the test set is Neural Network drive cycles, NN.

[0091] Table 1 shows the estimation results of the method proposed in this embodiment. It can be seen that the RMSE of the traditional three-input LSTM based on current, voltage and temperature is as high as 2.639%. Figure 6 , Figure 7 It can be observed that the estimated value fluctuates greatly and the result is inaccurate. The RMSE of LSTM-1 with expanded open circuit voltage input is 1.857%, and the corresponding RMSE of LSTM-2 is 1.521%. The RMSE of LSTM-1&2 with the two improved methods is the smallest, which is only 0.726%. Therefore, it can be considered that the method proposed in this embodiment is effective in the current test.

[0092] Table 1 - RMSE and MAE of different estimation methods

[0093] Item LSTM LSTM-1 LSTM-2 LSTM-1&2 RMSE / % 2.639 2.121 1.490 0.726 MAE / % 2.035 1.706 1.018 0.565

[0094] Application 2: Testing process under complex working conditions

[0095] For the test process under complex working conditions, cycle 1 to 4 are used as test sets under four different working conditions, and the rest are used as training data to analyze the effect of the proposed method under different working conditions. The specific estimation results are as follows: Figure 8As shown, the four groups of data a, b, c, d under LSTM are compared with the four groups of data A, B, C, D under LSTM-1&2. It can be seen that whether it is cycle1, cycle2, cycle2 or cycle4, the estimation model LSTM-1&2 that combines open circuit voltage parameters, cumulative current and loss function has obtained higher accuracy than the traditional three-input LSTM estimation model based on current, voltage and temperature. Table 2 shows the RMSE and MAE estimation results of the proposed method under different working conditions. It can also be seen that the RMSE and MAE calculated by the corresponding cycle1, cycle2, cycle2 or cycle4 data under LSTM-1&2 are smaller than the RMSE and MAE calculated by the cycle1, cycle2, cycle2 or cycle4 data under LSTM, so it can be considered that the proposed method is effective in the current test.

[0096] Table 2-RMSE and MAE of the proposed method under different working conditions

[0097]

[0098] It can be seen from the ablation test that the method of this embodiment has higher estimation accuracy under different test conditions, which means that the method of this embodiment can meet different test conditions and has certain versatility.

[0099] Implementation principle of this embodiment: In this implementation, SOC estimation is the basic content of the battery management system to ensure the safe and reliable operation of the battery power system, and the research on the estimation method of the neural network has been verified. In the Panasonic 18650PF lithium battery test data set, the accuracy of the LSTM model with only three inputs is low, and the deep learning algorithm combined with the equivalent circuit model proposed in the present invention is used for SOC estimation by further extracting the battery mechanism information and inputting it into the neural network. In the test results of the 25°C battery data, lower RMSE and MAE are achieved, which are 72.49% and 72.24% lower than those of the ordinary three-input LSTM model, respectively. Therefore, the research results of the present invention show that applying battery mechanism information to the battery state estimation problem of the data-driven method is an effective method.

[0100] The specific process of each step above is mainly to obtain the initial current, initial voltage and temperature data of the battery under different working conditions, input the initial voltage and initial current data into the mechanism model, further extract the mechanism information of the battery and input it into the neural network, and improve the loss function of the SOC estimation model, so as to improve the estimation accuracy and fast calculation speed. The battery SOC is estimated by combining the mechanism information of the equivalent circuit model with the neural network, taking into account the applicable scenarios of the battery under different operating conditions and temperatures, with fast calculation speed and high relative accuracy, which provides a solid foundation for the battery management system to ensure the safe and reliable operation of the battery power supply system.

[0101] Reference Fig. 9 As shown, the present application also provides a lithium-ion battery state of charge estimation system, which may include: an acquisition module 301, a processing module 302, a model building module 303, and an evaluation module 304. The main functions of each component module are as follows:

[0102] An acquisition module 301 is used to acquire initial current, initial voltage and temperature data of the battery under different working conditions;

[0103] The processing module 302 is used to input the initial voltage and initial current data into the mechanism model to obtain voltage simulation data and current simulation data;

[0104] A model building module 303 is used to input the voltage simulation data, the current simulation data and the temperature data into the neural network for training, build an SOC estimation model, and obtain an SOC output result;

[0105] The evaluation module 304 is used to calculate the root mean square error RMSE and the mean absolute error MAE according to the SOC output result and the SOC true value for verification and testing to evaluate the estimation performance of the SOC estimation model.

[0106] like Fig.10 , is a block diagram of a computer device according to an embodiment of the present application. The computer device is intended to represent various forms of digital computers or mobile devices. The digital computer may include a desktop computer, a portable computer, a workbench, a personal digital assistant, a server, a mainframe computer, and other suitable computers. The mobile device may include a tablet computer, a smart phone, a wearable device, etc.

[0107] like Fig.10 As shown, the device 600 includes a computing unit 601, a ROM 602, a RAM 603, a bus 604, and an input / output (I / O) interface 605, and the computing unit 601, ROM 602 and RAM 603 are connected to each other through the bus 604. The input / output (I / O) interface 605 is also connected to the bus 604.

[0108] The computing unit 601 can perform various processes in the method embodiment of the present application according to the computer instructions stored in the read-only memory (ROM) 602 or the computer instructions loaded from the storage unit 608 to the random access memory (RAM) 603. The computing unit 601 can be various general and / or special processing components with processing and computing capabilities. The computing unit 601 may include, but is not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. In some embodiments, the method provided in the embodiment of the present application can be implemented as a computer software program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 608.

[0109] RAM 603 may also store various programs and data required for the operation of device 600. Part or all of a computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609.

[0110] The input unit 606, output unit 607, storage unit 608 and communication unit 609 in the device 600 can be connected to the I / O interface 605. The input unit 606 can be, for example, a keyboard, a mouse, a touch screen, a microphone, etc. The output unit 607 can be, for example, a display, a speaker, an indicator light, etc. The device 600 can exchange information, data, etc. with other devices through the communication unit 609.

[0111] It should be noted that the device may also include other components necessary for normal operation, or may only include components necessary for implementing the solution of the present application, rather than all the components shown in the figure.

[0112] Various implementations of the systems and techniques described herein can be realized in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof.

[0113] The computer instructions for implementing the method of the present application can be written in any combination of one or more programming languages. These computer instructions can be provided to the computing unit 601, so that when the computer instructions are executed by the computing unit 601 such as a processor, the steps involved in the method embodiment of the present application are executed.

[0114] The computer-readable storage medium provided in the present application may be a tangible medium that may contain or store computer instructions for executing the various steps involved in the method embodiments of the present application. The computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, and other forms of storage media.

[0115] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for estimating the state of charge of a lithium-ion battery, characterized in that: The following steps are involved: Obtain the initial current, initial voltage and temperature data of the battery under different working conditions; Inputting the initial voltage and the initial current data into a mechanism model to obtain voltage simulation data and current simulation data; Inputting the voltage simulation data, the current simulation data and the temperature data into a neural network for training, constructing an SOC estimation model, and obtaining an SOC output result; The root mean square error (RMSE) and the mean absolute error (MAE) are calculated based on the SOC output result and the true SOC value for verification and testing to evaluate the estimation performance of the SOC estimation model.

2. The method for estimating the state of charge of a lithium-ion battery according to claim 1, characterized in that: In the step, the initial voltage and the initial current data are input into the mechanism model to obtain voltage simulation data and current simulation data. The mechanism model includes: The mechanism model selects an equivalent circuit model; The equivalent circuit model specifically adopts a second-order Thevenin model, which consists of two series-connected RC loops, a resistor, and an ideal voltage source, and is described by formula (1): Where U1 and U2 are the polarization voltages corresponding to the R1C1 and R2C2 branches respectively, and U0 is the ohmic voltage corresponding to R1.

3. The method for estimating the state of charge of a lithium-ion battery according to claim 2, characterized in that: The step of inputting the initial voltage and the initial current data into a mechanism model to obtain voltage simulation data and current simulation data specifically includes: Inputting the initial voltage and the initial current data into the equivalent circuit to obtain a voltage signal and a current signal; Deconstruct the voltage signal output by the equivalent circuit into an open circuit voltage Uoc v , get voltage simulation data; The current signal output by the equivalent circuit is calculated according to the ampere-hour integration method to obtain the accumulated current ΔI, thereby obtaining current simulation data.

4. The method for estimating the state of charge of a lithium-ion battery according to claim 3, characterized in that: Deconstructing the voltage signal output by the equivalent circuit into an open circuit voltage Uocv to obtain voltage simulation data, including: After discretizing the equivalent circuit, the Laplace equation of the equivalent circuit model is obtained, which is described by formula (2): Where s represents the response of the system in the complex frequency domain; The Laplace equation of the equivalent circuit model is bilinearly transformed to obtain the differential equation of the input and output of the equivalent circuit model, which is described by formula (3): <h2 style=";text-align:left;direction:ltr">y(k)=U<h2 style=";text-align:left;direction:ltr"> ocv <h2 style=";text-align:left;direction:ltr"> (k)-U(k) = a1y(k-1)+a2y(k-2)+a3I(k)+a4I(k-1)+a5I(k-2)#(3), Where I(k) is the current input of the system, and y(k) is the system output; According to the differential equation, the input and output matrix at a specific time and the parameter matrix to be estimated are obtained, which are described by formulas (4) and (5): <h2 style=";text-align:left;direction:ltr">θ(k) = [a1 a2 a3 a4 a<h2 style=";text-align:left;direction:ltr"> s <h2 style=";text-align:left;direction:ltr"> ]#(5), In the formula, refers to the input and output matrix at time K, and θ(k) refers to the parameter matrix to be estimated; A recursive equation of an equivalent circuit model is obtained according to the input-output matrix and the parameter matrix, and a parameter identification result is obtained from the equivalent circuit model by sampling least squares to obtain the open circuit voltage Uocv.

5. The method for estimating the state of charge of a lithium-ion battery according to claim 3, characterized in that: The step of calculating the current signal output by the equivalent circuit according to the ampere-hour integration method to obtain the accumulated current ΔI and obtain current simulation data includes: The current signal and time are integrated according to the ampere-hour integration method to obtain the accumulated current ΔI, so as to obtain the change value of the SOC, which is described by formula (6): In the formula, η represents the discharge efficiency of the battery, and Qn represents the rated capacity of the battery.

6. The method for estimating the state of charge of a lithium-ion battery according to any one of claims 1 to 3, characterized in that: Inputting the voltage simulation data, the current simulation data and the temperature data into a neural network for training, constructing an SOC estimation model, and obtaining an SOC output result, including: Inputting the voltage simulation data, the current simulation data and the temperature data into a neural network for training to construct an SOC estimation model; The loss function of the SOC estimation model is modified, and a penalty term is added to compare the SOC estimation value with the SOC true value on the basis of the original mean square error MSE, so as to obtain the SOC output result to meet the SOC accuracy estimation.

7. The method for estimating the state of charge of a lithium-ion battery according to claim 6, characterized in that: The modification of the loss function of the SOC estimation model, comparing the SOC estimation value with the SOC true value based on the original mean square error MSE and then adding a penalty term to obtain the SOC output result to meet the SOC accuracy estimation, includes: The real value of SOC SOC i and the SOC estimation value obtained according to the SOC estimation model Make comparisons; If the SOC estimation value Greater than the actual SOC value SOC i , then the penalty term W1 is added to the loss function; Otherwise, the loss function is calculated to add a penalty term W2; wherein, W1 and W2 are both user-defined parameters, and the penalty item W1 is smaller than the penalty item W2.

8. The method for estimating the state of charge of a lithium-ion battery according to claim 1, characterized in that: The step of respectively calculating the root mean square error RMSE and the mean absolute error MAE based on the SOC output result and the true SOC value for verification and testing to evaluate the estimation performance of the SOC estimation model includes: Evaluate the standard deviation of the estimation error of the SOC estimation model under different operating conditions and temperatures according to the root mean square error RMSE; The robustness of the SOC estimation model under different operating conditions and temperatures is evaluated based on the mean absolute error (MAE).

9. A lithium-ion battery state of charge estimation system, characterized in that: include: An acquisition module is used to obtain the initial current, initial voltage and temperature data of the battery under different working conditions; A processing module, used for inputting the initial voltage and the initial current data into a mechanism model to obtain voltage simulation data and current simulation data; A model building module is used to input the voltage simulation data, the current simulation data and the temperature data into a neural network for training, build an SOC estimation model, and obtain an SOC output result; The evaluation module is used to calculate the root mean square error RMSE and the mean absolute error MAE according to the SOC output result and the SOC true value for verification and testing to evaluate the estimation performance of the SOC estimation model.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images