Lithium ion battery state-of-charge online optimization estimation method based on coulombic efficiency real-time updating
Through cyclic aging experiments and machine learning algorithm training models, the capacity and Coulomb efficiency of lithium-ion batteries are updated in real time, and the state of charge estimation error problem caused by ignoring the impact of Coulomb efficiency in the existing technology is solved, and a higher precision state of charge is achieved Online optimization estimation.
Patent Information
- Application Number
- CN202411090257.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-12-03
AI Technical Summary
The prior art ignores the influence of Coulomb efficiency with battery aging in lithium-ion battery state of charge estimation, resulting in an increase in the calculation error of state of charge calculation, making it difficult to improve the estimation accuracy at different aging stages of the battery.
Current and voltage data were collected through cyclic aging experiments, simplified electrochemical model was identified using genetic algorithms, correlation coefficients between model parameters and capacity and Coulomb efficiency were calculated, capacity and Coulomb efficiency estimation models were trained, capacity and Coulomb efficiency were updated in real time, and Coulomb counting was used to calculate the correlation coefficients of model parameters and capacity and Coulomb efficiency. Calculate the state of charge.
It significantly improves the estimation accuracy of the state of charge of lithium-ion batteries, and can accurately update the capacity and Coulomb efficiency at different aging stages of the battery, reducing the estimation error of the state of charge.
Smart Images

Figure CN119064786A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of lithium-ion batteries, and in particular relates to an online optimization estimation method for the state of charge of a lithium-ion battery based on real-time updating of coulombic efficiency. Background Art
[0002] With the rapid development and large-scale promotion and application of electric vehicles, it is extremely important to accurately estimate the state of lithium-ion power batteries throughout their life cycle. Just like the remaining fuel level of a fuel vehicle, the state of charge indicates the remaining power of the battery. The difference is that the fuel level can be measured by sensors, but the state of charge can only be estimated by the current and voltage of the battery. In engineering, the ampere-hour integration method is usually used, that is, the power consumption is calculated by accumulating the current value per unit time, and then subtracting it from the battery capacity to obtain the remaining power.
[0003] The prior art discloses the following patents:
[0004] CN107748336A discloses a method and system for online estimation of the state of charge of a lithium-ion battery, wherein the method includes: obtaining the highest terminal voltage Vmax of a battery cell in a lithium-ion battery pack and / or a system; obtaining the lowest terminal voltage Vmin of a battery cell in a lithium-ion battery pack and / or a system; estimating the state of charge SOC1 according to the highest terminal voltage Vmax, current and other parameters of the single cell, and estimating the state of charge SOC2 according to the lowest terminal voltage Vmin, current and other parameters of the single cell; fusing the state of charge SOC1 and the state of charge SOC2 to obtain the current state of charge of the lithium-ion battery pack and / or the system. The method can obtain the state of charge SOC1 corresponding to the highest terminal voltage Vmax and the state of charge SOC2 corresponding to the lowest terminal voltage Vmin respectively and simultaneously, and then obtain the current state of charge according to the fusion algorithm, thereby effectively improving the working efficiency of the estimation, and is simple and easy to implement.
[0005] CN112083334 A discloses a data-driven lithium-ion battery state of charge estimation method, which relates to the technical field of battery management systems. The present invention uses the collected battery voltage, current and temperature as well as the calculated health state value and charge state value to train a charge state estimation model based on an extreme learning machine. Then, the charge state value estimated by the extreme learning machine is used as the observation value of the Kalman filter, and the observation value is filtered by the Kalman filter to filter out the interference of the measurement noise, so as to improve the estimation accuracy of the charge state.
[0006] However, both the ampere-hour integration method and the above-mentioned patented technology ignore the impact of capacity and coulombic efficiency as the battery ages, resulting in increasing errors in the calculation of the state of charge. Traditional state-of-charge estimation can take into account the decline of battery capacity, but usually ignores the impact of coulombic efficiency and regards it as a constant. In fact, coulombic efficiency gradually decreases as the battery ages, which is an important reason for the increase in the error in the estimation of the state of charge of lithium-ion batteries throughout their life cycle. Therefore, how to improve the estimation accuracy of the state of charge of lithium-ion batteries has become a difficult problem that technicians in the field of lithium-ion batteries need to solve urgently. Summary of the invention
[0007] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is to provide an online optimization estimation method for the state of charge of a lithium-ion battery based on real-time updating of coulomb efficiency, which can significantly improve the estimation accuracy of the battery state of charge at different battery aging stages.
[0008] In order to solve the above technical problems, the technical solution adopted by the present invention is: an online optimization estimation method of the state of charge of a lithium-ion battery based on real-time update of coulomb efficiency, comprising the following steps:
[0009] Step 1: Select a batch of lithium-ion batteries of the same type, perform a cycle aging experiment on the lithium-ion batteries until the battery capacity decays to 80%, and collect the current and voltage data of the battery during cycle aging. Extract the charging current, voltage data, capacity and coulomb efficiency of each cycle from the above data;
[0010] Step 2: With the above current data as input and voltage data as output, a genetic algorithm is used to identify a simplified electrochemical model, identify the model parameters of all batteries in each cycle, and then calculate the correlation coefficients between the identified model parameters and the battery capacity and coulombic efficiency. The model parameters with high correlation coefficients are selected as health features ① and ②, respectively, to train the capacity and coulombic efficiency estimation models;
[0011] Step 3: Take the capacity, coulomb efficiency and health characteristics of the battery at different cycles as the training set, take the health characteristic ① of the data set as input and the corresponding battery capacity as output, and use the machine learning algorithm to train the capacity estimation model; take the health characteristic ② of the data set as input and the corresponding coulomb efficiency as output, and use the machine learning algorithm to train the coulomb efficiency estimation model;
[0012] Step 4: Collect the battery current and voltage data during charging, identify the model parameters in the simplified electrochemical model described in step 1, substitute the parameters into the capacity estimation model and coulomb efficiency estimation model to calculate the battery capacity and coulomb efficiency, update the capacity and coulomb efficiency in real time during the battery aging process, and use the coulomb counting method to calculate the battery state of charge.
[0013] In the above-mentioned online optimization estimation method of the state of charge of a lithium-ion battery based on real-time update of coulombic efficiency, in step 2, the specific process of establishing a simplified electrochemical model is as follows:
[0014]
[0015] Among them, U p and U n are the equilibrium potentials of the positive and negative electrodes, I L is the battery constant charging current, R ohm Represents internal resistance, Q p and Q n Represents the positive electrode capacity and negative electrode capacity, SOC p,0 and SOC n,0 Represent the initial lithium insertion rates of the positive and negative electrodes, respectively.
[0016] In the above-mentioned online optimization estimation method of the state of charge of a lithium-ion battery based on real-time update of coulomb efficiency, in the step 3, a capacity estimation model and a coulomb efficiency estimation model are trained, and the support vector regression algorithm is used to train the above two models. The specific process is as follows:
[0017]
[0018] Where x and f(x) represent the input and the corresponding output respectively. is a nonlinear mapping function, w and b are weight coefficients and bias factors respectively, which can be calculated by the following convex optimization problem;
[0019]
[0020] Where C is a parameter that determines the complexity of the model, ξi and Represents the slack variable, which is ξ when the training error is less than -ε i , when it is greater than ε In order to simplify the calculation, the dual optimization problem of the above formula is usually solved by the following formula:
[0021]
[0022] Where α i and is the Lagrange multiplier; the final calculated weight coefficient and deviation factor are as follows:
[0023]
[0024] In the formula, k(x i ,x j ) is the kernel function.
[0025] In the above-mentioned online optimization estimation method of lithium-ion battery state of charge based on real-time update of coulomb efficiency, in step 4, the coulomb counting method is used to calculate the battery state of charge, and the formula is:
[0026]
[0027] Where SoC(t) and i(t) represent the state of charge and current at time t, respectively, SoC0 represents the initial state of charge, η and C cur Indicates the coulombic efficiency and current capacity to be updated.
[0028] The advantages of the on-line optimization estimation method for the state of charge of a lithium-ion battery based on real-time update of coulomb efficiency of the present invention are:
[0029] 1. By identifying model parameters from charging data, healthy features for estimating capacity and coulombic efficiency can be extracted from the parameters at the same time, and then the battery capacity and coulombic efficiency estimation models can be trained separately, which improves the efficiency of model training;
[0030] 2. The trained capacity estimation model and coulombic efficiency estimation model are embedded in the battery management system. The battery capacity and coulombic efficiency can be estimated by simply collecting charging data to identify the model parameters. At different stages of battery aging, the estimation accuracy of the battery state of charge can be significantly improved by updating the capacity and coulombic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 A schematic diagram of the capacity parameters of each battery at each cycle extracted from the cycle aging data of 7 batteries;
[0032] Figure 2 This is a schematic diagram of the relationship between capacity and coulomb efficiency;
[0033] Figure 3 Schematic diagram of the estimated results of battery SoC at the 700th aging cycle. DETAILED DESCRIPTION
[0034] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0035] In the present invention, unless otherwise stated, directional words such as "upper" and "lower" generally refer to the upper and lower parts of the device in actual use or working state, specifically the drawing direction in the attached drawings; while "inside" and "outside" refer to the outline of the device. In addition, in the description of this application, the term "including" means "including but not limited to". The terms first, second, third, etc. are used only as labels and do not impose numerical requirements or establish an order. The term "plurality" means "two or more".
[0036] The present invention discloses an online optimization estimation method for the state of charge of a lithium-ion battery based on real-time updating of coulomb efficiency, comprising:
[0037] Firstly, according to the current and voltage data of the charging process under different aging cycles, a genetic algorithm is used to identify the parameters of a simplified electrochemical model. Then, the correlation between the model parameters and the capacity and coulombic efficiency is calculated respectively, and the model parameters with strong correlation are used as health features. Furthermore, with the health features as input, a machine learning algorithm is used to train the capacity and coulombic efficiency estimation models respectively. Finally, under different aging cycles, the battery state of charge is calculated by real-time updating of the capacity and coulombic efficiency using the ampere-hour integration method.
[0038] The specific process is as follows:
[0039] Step 1: Select a batch of lithium-ion batteries of the same type, perform a cycle aging experiment on the lithium-ion batteries until the battery capacity decays to 80%, and collect the current and voltage data of the battery during cycle aging. Extract the charging current, voltage data, capacity and coulomb efficiency of each cycle from the above data;
[0040] Step 2: With the above current data as input and voltage data as output, a genetic algorithm is used to identify a simplified electrochemical model, identify the model parameters of all batteries in each cycle, and then calculate the correlation coefficients between the identified model parameters and the battery capacity and coulombic efficiency. The model parameters with high correlation coefficients are selected as health features ① and ②, respectively, to train the capacity and coulombic efficiency estimation models;
[0041] Step 3: Take the capacity, coulomb efficiency and health characteristics of the battery at different cycles as the training set, take the health characteristic ① of the data set as input and the corresponding battery capacity as output, and use the machine learning algorithm to train the capacity estimation model; take the health characteristic ② of the data set as input and the corresponding coulomb efficiency as output, and use the machine learning algorithm to train the coulomb efficiency estimation model;
[0042] Step 4: Collect the battery current and voltage data during charging, identify the model parameters in the simplified electrochemical model described in step 1, substitute the parameters into the capacity estimation model and coulomb efficiency estimation model to calculate the battery capacity and coulomb efficiency, update the capacity and coulomb efficiency in real time during the battery aging process, and use the coulomb counting method to calculate the battery state of charge.
[0043] In step 2, the specific process of establishing a simplified electrochemical model is as follows:
[0044]
[0045] Among them, U p and U nare the equilibrium potentials of the positive and negative electrodes, I L is the battery constant charging current, R ohm Represents internal resistance, Q p and Q n Represents the positive electrode capacity and negative electrode capacity, SOC p,0 and SOC n,0 Represent the initial lithium insertion rates of the positive and negative electrodes, respectively.
[0046] In step 3, the capacity estimation model and the coulomb efficiency estimation model are trained. The support vector regression algorithm is used to train the above two models. The specific process is as follows:
[0047]
[0048] Where x and f(x) represent the input and the corresponding output respectively. is a nonlinear mapping function, w and b are weight coefficients and bias factors respectively, which can be calculated by the following convex optimization problem;
[0049]
[0050] Where C is a parameter that determines the complexity of the model, ξ i and Represents the slack variable, which is ξ when the training error is less than -ε i , when it is greater than ε In order to simplify the calculation, the dual optimization problem of the above formula is usually solved by the following formula:
[0051]
[0052] Where α i and is the Lagrange multiplier; the final calculated weight coefficient and deviation factor are as follows:
[0053]
[0054] In the formula, k(x i ,x j ) is the kernel function.
[0055] In step 4, the coulomb counting method is used to calculate the battery state of charge, and the formula is:
[0056]
[0057] Where SoC(t) and i(t) represent the state of charge and current at time t, respectively, SoC0 represents the initial state of charge, η and C cur Indicates the coulombic efficiency and current capacity to be updated.
[0058] The present invention extracts health features related to battery capacity decay and coulombic efficiency decline from the charging curve, and then uses a machine learning algorithm to train capacity estimation and coulombic efficiency estimation models respectively; further, by updating capacity and coulombic efficiency in real time, an optimized estimation of the battery state of charge over the entire life cycle is achieved, thereby realizing online precise optimized estimation of the lithium-ion battery state of charge.
[0059] Embodiment 1:
[0060] The method of the present invention is described in detail below by taking 8 LiNCM batteries with a rated capacity of 32.5Ah as an example:
[0061] The simplified electrochemical model is as follows:
[0062]
[0063] Among them, U p and U n are the equilibrium potentials of the positive electrode and the negative electrode, respectively, obtained through the battery static experiment test, I L is the battery constant charging current, R ohm Represents internal resistance, Q p and Q n Represents the positive electrode capacity and negative electrode capacity, SOC p,0 and SOC n,0 Represent the initial lithium insertion rates of the positive and negative electrodes, respectively.
[0064] The charging current, voltage data, capacity and coulombic efficiency of each battery in each cycle are extracted from the cycle aging data of 7 batteries. The parameters of the above model are identified by genetic algorithm based on the charging current and voltage data, and their correlation coefficients with capacity and coulombic efficiency are calculated, such as Figure 1 and Figure 2 The model parameters with correlation coefficients higher than 0.8 are taken as health characteristics. Specifically, SOC n,0 , Q n , R ohm As a health feature① used to train the capacity estimation model, Q n , R ohm As the health feature ②, it is used to train the Coulomb efficiency estimation model, and the Gaussian process regression algorithm is used to train the above two models.
[0065] Among them, the training capacity estimation model and the Coulomb efficiency estimation model use the support vector regression algorithm to train the above two models. The specific process is as follows:
[0066]
[0067] Where x and f(x) represent the input and the corresponding output respectively. is a nonlinear mapping function, w and b are the weight coefficient and bias factor respectively, which can be calculated by the following convex optimization problem.
[0068]
[0069] Where C is a parameter that determines the complexity of the model, ξ i and Represents the slack variable, which is ξi when the training error is less than -ε and is In order to simplify the calculation, the dual optimization problem of the above formula (3) is usually solved by the following formula:
[0070]
[0071] Where α i and is the Lagrange multiplier. The final calculated weight coefficient and deviation factor are as follows:
[0072]
[0073] In the formula, k(x i ,x j ) is the kernel function.
[0074] Under different aging cycles, the state of charge of the 8th battery at the 700th cycle is calculated by real-time updating of capacity and coulomb efficiency using the coulomb counting method, as shown in the following formula:
[0075]
[0076] Where SoC(t) and i(t) represent the state of charge and current at time t, respectively, SoC0 represents the initial state of charge, η and C cur Represents the coulombic efficiency and capacity to be updated. The estimated result of battery SoC at the 700th aging cycle is as follows: Figure 3 Obviously, when both the capacity and the coulombic efficiency are updated, the SoC estimation error is very low and close to the true value.
[0077] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Any changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention should fall within the protection scope of the present invention.
Claims
1. A method for online optimization estimation of the state of charge of a lithium-ion battery based on real-time update of coulombic efficiency, characterized in that: The steps include: Step 1: Select a batch of lithium-ion batteries of the same type, perform a cycle aging experiment on the lithium-ion batteries until the battery capacity decays to 80%, and collect the current and voltage data of the battery during cycle aging. Extract the charging current, voltage data, capacity and coulomb efficiency of each cycle from the above data; Step 2: With the above current data as input and voltage data as output, a genetic algorithm is used to identify a simplified electrochemical model, identify the model parameters of all batteries in each cycle, and then calculate the correlation coefficients between the identified model parameters and the battery capacity and coulombic efficiency. The model parameters with high correlation coefficients are selected as health features ① and ②, respectively, to train the capacity and coulombic efficiency estimation models; Step 3: Take the capacity, coulomb efficiency and health characteristics of the battery at different cycles as the training set, take the health characteristic ① of the data set as input and the corresponding battery capacity as output, and use the machine learning algorithm to train the capacity estimation model; take the health characteristic ② of the data set as input and the corresponding coulomb efficiency as output, and use the machine learning algorithm to train the coulomb efficiency estimation model; Step 4: Collect the battery current and voltage data during charging, identify the model parameters in the simplified electrochemical model described in step 1, substitute the parameters into the capacity estimation model and coulomb efficiency estimation model to calculate the battery capacity and coulomb efficiency, update the capacity and coulomb efficiency in real time during the battery aging process, and use the coulomb counting method to calculate the battery state of charge.
2. The method for online optimization estimation of the state of charge of a lithium-ion battery based on real-time update of coulombic efficiency according to claim 1 is characterized in that: In the step 2, the specific process of establishing a simplified electrochemical model is as follows: Among them, U p and U n are the equilibrium potentials of the positive and negative electrodes, respectively, and I L is the battery constant charging current, R ohm Represents internal resistance, Q p and Q n Represents the positive electrode capacity and negative electrode capacity, SOC p,0 and SOC n,0 Represent the initial lithium insertion rates of the positive and negative electrodes, respectively.
3. The method for online optimization estimation of the state of charge of a lithium-ion battery based on real-time update of coulombic efficiency according to claim 1 is characterized in that: In the step 3, the capacity estimation model and the coulomb efficiency estimation model are trained, and the support vector regression algorithm is used to train the above two models. The specific process is as follows: Where x and f(x) represent the input and the corresponding output respectively. is a nonlinear mapping function, w and b are weight coefficients and bias factors respectively, which can be calculated by the following convex optimization problem; Where C is a parameter that determines the complexity of the model, ξ i and Represents the slack variable, which is ξi when the training error is less than -ε and is In order to simplify the calculation, the dual optimization problem of the above formula is usually solved by the following formula: Where α i and is the Lagrange multiplier; the final calculated weight coefficient and deviation factor are as follows: In the formula, k(x i ,x j ) is the kernel function.
4. The method for online optimization estimation of the state of charge of a lithium-ion battery based on real-time update of coulombic efficiency according to claim 1 is characterized in that: In step 4, the battery state of charge is calculated using the coulomb counting method, and the formula is: Where SoC(t) and i(t) represent the state of charge and current at time t, respectively, SoC0 represents the initial state of charge, η and C cur Indicates the coulombic efficiency and current capacity to be updated.
