A battery state of charge estimation method and system considering health status

Abstract

The present invention discloses a battery state of charge estimation method and system considering the health state, belonging to the technical field of battery state of charge estimation, the method comprises the steps of: obtaining battery charging and discharging data, constructing a data set; establishing a corresponding battery model; estimating the open circuit voltage of the battery based on the data set and the battery model; using the estimated open circuit voltage and the battery health state as inputs of a neural network model to preliminarily estimate the SOC; and obtaining the final estimated value of the SOC through a filtering algorithm in combination with current information. The present invention combines a battery model method and a data-driven method to achieve accurate estimation of the battery state of charge under different health states.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery state of charge estimation, and in particular relates to a battery state of charge estimation method and system taking health status into consideration. Background Art

[0002] The battery state of charge (SOC) is an important parameter used to characterize the short-term power changes of the battery. It is generally defined as the ratio of the current remaining power to the current maximum available capacity. SOC estimation is one of the core functions of the battery management system. Accurate SOC estimation can improve the battery energy utilization rate, prevent the battery from overcharging and overdischarging, ensure the safety and reliability of the battery during use, and extend the battery life. However, due to the nonlinearity and time-varying characteristics of the battery itself and the complex working environment, it is difficult to obtain high-precision SOC estimation results.

[0003] Commonly used battery SOC estimation methods mainly include: ampere-hour integration method, open circuit voltage method, equivalent model method and data-driven method. Among them, the calculation of the ampere-hour integration method is simple and efficient, but there is an obvious error accumulation effect; the open circuit voltage method requires the battery to be static for more than half an hour, which is difficult to apply in practice; the equivalent model method usually uses a battery model plus a filter algorithm observer to achieve SOC estimation, which can achieve a good balance between estimation accuracy and computational complexity, but the applicability and generalization ability are limited; the data-driven method usually uses current and voltage information as input to establish a machine learning or deep learning model to directly estimate SOC, with high estimation accuracy, but this method has the risk of open-loop divergence, large amount of computation and difficult to run online, and the established model is a black box model, lacking interpretability, and the interpretability of the data-driven method to estimate SOC has not been fully studied.

[0004] Most of the above estimation methods do not consider how to continue to ensure SOC accuracy when battery capacity decay occurs, that is, as the battery health state changes, the SOC estimation error gradually increases. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a battery state of charge estimation method and system taking into account the health status, which can effectively solve the technical problems of increased error in SOC estimation as the health status changes, the risk of open-loop divergence of the data-driven method, and poor interpretability.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention discloses a method for estimating a battery state of charge (SOC) taking into account a state of health (SOH), comprising the following steps:

[0008] Step 1: Use public data or charging and discharging equipment to obtain battery charging and discharging data and its corresponding health status to build a data set;

[0009] Step 2: Establish the corresponding battery model;

[0010] Step 3: Estimate the open circuit voltage of the battery based on the data set in step 1 and the battery model in step 2;

[0011] Step 4: Build a neural network model with open circuit voltage (OCV) and health status as input and battery state of charge as output;

[0012] Step 5: Combine the current information in the data set of step 1 and the initial value of the battery state of charge estimated by the neural network model of step 3, and fuse them through a filtering algorithm to obtain the final battery state of charge estimate.

[0013] Preferably, in step 1, the battery charging and discharging data and the corresponding health status are obtained by using public data or charging and discharging equipment.

[0014] Preferably, the charging and discharging equipment is a battery detection system or a charging pile.

[0015] Preferably, the battery is a lead-acid battery, a nickel-metal hydride battery, a nickel-chromium battery or a lithium-ion battery.

[0016] Preferably, in step 2, constructing a battery model comprises the following steps:

[0017] Step 21: Establish the state space equation based on the first-order RC equivalent circuit model of the battery:

[0018]

[0019] Where, E0 is the open circuit voltage, R1 is the ohmic internal resistance, R2 is the polarization internal resistance, C2 is the polarization capacitance, U2 is the polarization voltage, I is the current, and U0 is the battery terminal voltage;

[0020] Step 22, calculate the transfer function of the state space equation, add a zero-order holder link, and calculate the transfer function G(s):

[0021]

[0022] Where, T is the sampling interval, H(s) is the transfer function of the zero-order holder, s is a variable symbol; e is an irrational number;

[0023] Step 23: Perform an inverse Laplace transform on the transfer function G(s) and perform discretization processing:

[0024]

[0025]

[0026] Obtain the difference equation at time k:

[0027]

[0028] Step 24: Combine formula (1) to calculate the open circuit voltage at time k:

[0029]

[0030] in,

[0031] Step 25: Let V k =E 0(k) -U 0(k) , establish the controlled autoregressive moving average model of the battery, namely the CARMA model:

[0032]

[0033] Among them, θ1=a, θ2=[R1+R2(1 / a-1)], θ3=aR0, q -1 is a one-step translation operator, ξ k is the system noise.

[0034] Preferably, in step 3, the recursive least square method with a forgetting factor is used to estimate the battery CARMA model parameters online, including the following steps:

[0035] Step 31: Since the open circuit voltage changes very little in a very short time, it is assumed that the open circuit voltage remains unchanged in a sampling interval, that is, E 0(k) =E 0(k-1) , according to the battery CARMA model, we get:

[0036] U 0(k) =[U 0(k-1) I k -I k-1 1][θ 1k θ 2k θ 3k θ 4k ] T (26)

[0037] Among them, θ 4k =(1-a)E 0(k) , and the open circuit voltage E is obtained 0(k) Estimated value:

[0038]

[0039] Step 32: Use the recursive least square method with a forgetting factor to perform online estimation of the open circuit voltage and initialize the estimated value:

[0040]

[0041] Among them, x is the parameter to be estimated, x0 is the initial value of the parameter to be estimated, P0 is the initial value of the covariance matrix of the error, and E is the unit matrix;

[0042] Step 33: perform iterative solution and estimate the open circuit voltage according to formula (9). The iterative process is as follows:

[0043] e k =y k -H k x k-1 (29)

[0044]

[0045] x k =x k-1 +K k e k (31)

[0046]

[0047] Among them, k refers to the k moment, y k is the battery terminal voltage, e k for y k The estimated error, H k =[U 0(k-1) I k -I k-1 1], x k =[θ 1k θ 2k θ 3k θ 4k ] T is the parameter matrix to be estimated, K k is the gain factor, P k is the covariance moment, and λ is the forgetting factor.

[0048] Preferably, in step 4, the specific method of constructing the neural network model includes:

[0049] Construct a BP neural network model including input layer, hidden layer and output layer, and its input feature sequence is x = [OCVSOH] T , that is, the open circuit voltage and health status of the battery. The output sequence is y=SOC, that is, the state of charge of the battery. The number of neurons in the hidden layer is set to 64, and the activation function is the ReLU function. The output of the entire network for any sample is:

[0050]

[0051] Among them, β i is the connection weight between the i-th neuron in the hidden layer and the output layer, g(x) is the activation function, a i is the i-th row of the weight matrix connecting the input layer and the hidden layer, x j is the input matrix of the jth sample, b i is the i-th row of the bias matrix connecting the input layer and the hidden layer, b i i is the output layer bias value, and P is the number of samples.

[0052] Preferably, in step 5, fusion is performed by a filtering algorithm, specifically including:

[0053] Step 51, calculate the relationship between the current and the battery state of charge:

[0054]

[0055] Among them, Q is the current maximum available capacity of the battery, T is the sampling interval, and I is the current;

[0056] Step 52: Combining the SOC result estimated by the BP neural network with formula (16), the recursive formula of the state space equation is obtained:

[0057]

[0058] Among them, SOC k is the estimated SOC value at time k, SOC BP,k is the estimated SOC value at time k obtained by the BP neural network model, I k is the current at time k;

[0059] Step 53: Simplify the state equation and observation equation of the system, and use the Kalman filter algorithm to perform time update and measurement update to obtain the final battery state of charge estimate:

[0060]

[0061] Among them, x k is the system state variable, i.e. SOC, y k is the output, i.e. the SOC estimated by the BP neural network model, u k is the input quantity, A is the state matrix, B is the input matrix, C is the output matrix, ω k is the process noise, v k is the observation noise.

[0062] The present invention also discloses a battery state of charge estimation system considering health status, comprising:

[0063] Dataset building module, used to obtain battery charging and discharging data and corresponding health status;

[0064] Battery model building module, used to build battery model;

[0065] An open circuit voltage estimation module, used to estimate the open circuit voltage of the battery according to the constructed data set and battery model;

[0066] A neural network model building module is used to build a neural network model with open circuit voltage and charge and discharge data health status as input and battery charge status as output;

[0067] The battery state of charge estimation module is used to combine the current information in the data set with the initial value of the battery state of charge estimation output by the neural network model, and fuse them through a filtering algorithm to obtain the final battery state of charge estimation value.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] The estimation method proposed in the present invention estimates the open circuit voltage of the battery through voltage, current data and battery model, and uses it as the input of the neural network, giving the neural network domain knowledge. Compared with the traditional voltage and current input, this method has certain interpretability. The present invention uses a filtering algorithm to fuse the estimation results, which can avoid the open-loop divergence risk brought by the data-driven method, making the method more robust. The present invention takes into account the health status of the battery and uses the health status as the input of the neural network as well, so as to achieve accurate estimation of the battery SOC under different health states, which is of great significance for improving the performance of the battery management system. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 The present invention is a flow chart of the method proposed by the present invention.

[0071] Figure 2 It is a first-order RC equivalent circuit model in an embodiment of the present invention.

[0072] Figure 3 This is a structural diagram of the BP neural network constructed in an embodiment of the present invention.

[0073] Figure 4 This is the state of charge estimation result when the battery SOH is 97.5% in the embodiment of the present invention.

[0074] Figure 5 This is the state of charge estimation result when the battery SOH is 73.2% in the embodiment of the present invention. DETAILED DESCRIPTION

[0075] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0076] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0077] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0078] See also Figure 1 The present invention provides a method for estimating the state of charge of a battery taking into account the health state, comprising the following steps:

[0079] The first step is to use public data or charging and discharging equipment (battery detection system, charging pile, etc.) to obtain battery charging and discharging data and its corresponding health state (SOH) to build a data set;

[0080] The second step is to establish the corresponding battery model. The specific establishment process includes the following steps:

[0081] Step 21: Establish the state space equation based on the first-order RC equivalent circuit model of the battery:

[0082]

[0083] Where, E0 is the open circuit voltage, R1 is the ohmic internal resistance, R2 is the polarization internal resistance, C2 is the polarization capacitance, U2 is the polarization voltage, I is the current, and U0 is the battery terminal voltage;

[0084] Step 22, calculate the transfer function of the state space equation, add a zero-order holder link, and calculate the transfer function G(s):

[0085]

[0086] Where, T is the sampling interval, H(s) is the transfer function of the zero-order holder, s is a variable symbol; e is an irrational number;

[0087] Step 23: Perform an inverse Laplace transform on the transfer function G(s) and perform discretization processing:

[0088]

[0089]

[0090] Obtain the difference equation at time k:

[0091]

[0092] Step 24: Combine formula (1) to calculate the open circuit voltage at time k:

[0093]

[0094] in,

[0095] Step 25: Let V k =E 0(k) -U 0(k) , establish the controlled autoregressive moving average model of the battery, namely the CARMA model:

[0096]

[0097] Among them, θ1=a, θ2=[R1+R2(1 / a-1)], θ3=aR0, q -1 is a one-step translation operator, ξ k is the system noise.

[0098] The third step is to estimate the open circuit voltage (OCV) of the battery based on the data set and the battery model. Specifically, the recursive least squares method with forgetting factor is used to estimate the battery CARMA model parameters online, including the following steps:

[0099] Step 31: Since the open circuit voltage changes very little in a very short time, it is assumed that the open circuit voltage remains unchanged in a sampling interval, that is, E 0(k) =E 0(k-1) , according to the battery CARMA model, we get:

[0100] U 0(k) =[U 0(k-1) I k -I k-1 1][θ 1k θ 2k θ 3kθ 4k ] T (44)

[0101] Among them, θ 4k =(1-a)E 0(k) , and the open circuit voltage E is obtained 0(k) Estimated value:

[0102]

[0103] Step 32: Use the recursive least square method with a forgetting factor to perform online estimation of the open circuit voltage and initialize the estimated value:

[0104]

[0105] Among them, x is the parameter to be estimated, x0 is the initial value of the parameter to be estimated, P0 is the initial value of the covariance matrix of the error, and E is the unit matrix;

[0106] Step 33: perform iterative solution and estimate the open circuit voltage according to formula (9). The iterative process is as follows:

[0107] e k =y k -H k x k-1 (47)

[0108]

[0109] x k =x k-1 +K k e k (49)

[0110]

[0111] Among them, k refers to the k moment, y k is the battery terminal voltage, e k for y k The estimated error, H k =[U 0(k-1) I k -I k-1 1], x k =[θ 1k θ 2k θ 3k θ 4k ] T is the parameter matrix to be estimated, K k is the gain factor, P k is the covariance moment, and λ is the forgetting factor.

[0112] The fourth step is to build a neural network model with open circuit voltage and health status as input and SOC as output. The specific method of building a neural network model includes:

[0113] Construct a BP neural network model including input layer, hidden layer and output layer, and its input feature sequence is x = [OCVSOH] T , that is, the open circuit voltage and health status of the battery. The output sequence is y=SOC, that is, the state of charge of the battery. The number of neurons in the hidden layer is set to 64, and the activation function is the ReLU function. The output of the entire network for any sample is:

[0114]

[0115] Among them, β i is the connection weight between the i-th neuron in the hidden layer and the output layer, g(x) is the activation function, a i is the i-th row of the weight matrix connecting the input layer and the hidden layer, x j is the input matrix of the jth sample, b i is the i-th row of the bias matrix connecting the input layer and the hidden layer, b i i is the output layer bias value, and P is the number of samples.

[0116] The fifth step is to combine the current information in the data set with the initial SOC estimation value output by the neural network model and fuse them through a filtering algorithm to obtain the final accurate SOC estimation value.

[0117] Fusion is performed through filtering algorithms, including:

[0118] Step 51, calculate the relationship between the current and the battery state of charge:

[0119]

[0120] Among them, Q is the current maximum available capacity of the battery, T is the sampling interval, and I is the current;

[0121] Step 52: Combining the SOC result estimated by the BP neural network with formula (16), the recursive formula of the state space equation is obtained:

[0122]

[0123] Among them, SOC k is the estimated SOC value at time k, SOC BP,k is the estimated SOC value at time k obtained by the BP neural network model, I k is the current at time k;

[0124] Step 53: Simplify the state equation and observation equation of the system, and use the Kalman filter algorithm to perform time update and measurement update to obtain the final battery state of charge estimate:

[0125]

[0126] Among them, x k is the system state variable, i.e. SOC, y k is the output, i.e. the SOC estimated by the BP neural network model, u k is the input quantity, A is the state matrix, B is the input matrix, C is the output matrix, ω k is the process noise, v k is the observation noise.

[0127] Specific application examples:

[0128] The test object in this embodiment is a polyfluoro ternary soft-pack lithium battery (DFDPSP1265132-10Ah) with a nominal capacity of 10Ah. It is subjected to pulse characteristic tests and urban road conditions (UDDS) and new European driving cycle (NEDC) discharge tests in different health states. The voltage, current, time and their corresponding SOH during the test are obtained to construct a data set. Among them, the UDDS condition discharge data under different aging degrees is used for modeling, and the NEDC condition discharge data under different aging degrees is used for verification.

[0129] The differential equation mathematical model is established based on the first-order RC equivalent circuit model of the battery, see Figure 2 , where E0 is the open circuit voltage, R1 is the ohmic internal resistance, R2 is the polarization internal resistance, C2 is the polarization capacitance, U2 is the polarization voltage, I is the current, and U0 is the battery terminal voltage. By discretizing and differentiating the mathematical model, a controlled autoregressive moving average model (CARMA model) of the battery is established, as shown in the following formula:

[0130]

[0131] Where: θ1 is θ2 is [R1+R2(1 / a-1)]; θ3 is aR0; q -1 is a one-step translation operator; ξ k is the system noise.

[0132] In order to avoid the "data saturation" problem, the forgetting factor is introduced into the recursive least squares method, taking into account the tracking performance of the program and the error requirements after stabilization. The battery CARMA model parameters are estimated online by the recursive least squares method with forgetting factor (FFRLS method) to obtain the open circuit voltage estimate. The forgetting factor in the recursive least squares method is set to 0.9, and the initial value of the covariance matrix P0 is the unit matrix.

[0133] The BP neural network model is used as the main body to realize the SOC estimation based on OCV and SOH, that is, the battery SOH and the OCV obtained in the third step are used as the input features of the BP neural network, and the battery SOC is used as the output. The BP neural network model is divided into three layers: input layer, hidden layer and output layer. Its structure can be found in Figure 3 The number of neurons in the hidden layer of the BP neural network model is set to 64, and the activation function is the ReLU function.

[0134] The fourth step can preliminarily realize SOC estimation, but the estimation accuracy is relatively low and fluctuates greatly. Considering the open-loop divergence risk of the data-driven method, the estimation accuracy is further improved by combining the current information with the Kalman filter algorithm in the fifth step. The maximum absolute error (MaxAE) and root mean square error (RMSE) are used as quantitative indicators of estimation error.

[0135] In this embodiment, accurate SOC estimation under different SOHs is achieved and compared with a dual adaptive attenuated Kalman filter (DAFEKF) based on a first-order RC model: the MaxAE of the method proposed in the present invention does not exceed 1.5% in the range of battery SOH from 97.7% to 73.2%, and the RMSE is around 0.5%. In contrast, the SOC estimation error of the DAFEKF method gradually increases with the decrease of SOH, indicating the superiority of the method proposed in the present invention.

[0136] Figure 4 and Figure 5 The SOC estimation results and comparisons under two different SOHs are shown in detail. When the SOH is 97.5%, the error of the SOC estimated by the method proposed in the present invention is close to that obtained by the DAFEKF method, and the error does not exceed 2%, and both can estimate the SOC value more accurately. However, when the battery SOH reaches 73.2%, the estimation error of the DAFEKF method has exceeded 6%, while the SOC estimation method proposed in the present invention can still ensure that the maximum absolute error does not exceed 1.5%, and can overcome the initial value dependence problem in common filtering methods, and there is no obvious convergence process in the early stage of calculation. This proves that the estimation accuracy of the proposed SOC estimation method can be guaranteed under different battery health conditions.

[0137] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for estimating a battery state of charge taking into account a health state, characterized in that: The following steps are involved: Step 1: Obtain battery charging and discharging data and corresponding health status to build a data set; Step 2: Build a battery model, including the following steps: Step 21: Establish the state space equation based on the first-order RC equivalent circuit model of the battery: (1) in, is the open circuit voltage, is the ohmic internal resistance, is the polarization internal resistance, is the polarization capacitance, is the polarization voltage, is the current, is the battery terminal voltage; Step 22, calculate the transfer function of the state space equation, add a zero-order holder link, and calculate the transfer function G(s): (2) in, T is the sampling interval, H(s) is the transfer function of the zero-order holder; s is the variable symbol; Step 23: Perform an inverse Laplace transform on the transfer function G(s) and perform discretization processing: (3) (4) Obtain the difference equation at time k: (5) Step 24: Combine formula (1) to calculate the open circuit voltage at time k: (6) in, ; Step 25: , establish the controlled autoregressive moving average model of the battery, namely the CARMA model: (7) in, = , = , = , is a one-step translation operator, is the system noise; Step 3: Estimate the open circuit voltage of the battery based on the data set in step 1 and the battery model in step 2; Step 4: Construct a neural network model with open circuit voltage and charge and discharge data health status as input and battery charge status as output; Step 5: Combine the current information in the data set of step 1 and the initial value of the battery state of charge estimated by the neural network model output in step 3, and fuse them through a filtering algorithm to obtain the final battery state of charge estimate.

2. The method for estimating the battery state of charge considering the health state according to claim 1, characterized in that: Step 1 is to obtain battery charging and discharging data and corresponding health status using public data or charging and discharging equipment.

3. The method for estimating the battery state of charge considering the health state according to claim 2, characterized in that: The charging and discharging equipment is a battery detection system or a charging pile.

4. The method for estimating the battery state of charge considering the health state according to claim 1, characterized in that: The battery is a lead-acid battery, a nickel-metal hydride battery, a nickel-chromium battery or a lithium-ion battery.

5. The method for estimating the battery state of charge considering the health state according to claim 1, characterized in that: In step 3, the recursive least square method with forgetting factor is used to estimate the battery CARMA model parameters online, including the following steps: Step 31: Since the open circuit voltage changes very little in a very short time, it is assumed that the open circuit voltage remains unchanged in a sampling interval, that is, , according to the battery CARMA model, we get: (8) in, , and the open circuit voltage is obtained Estimated value: (9) Step 32: Use the recursive least square method with a forgetting factor to perform online estimation of the open circuit voltage and initialize the estimated value: (10) in, is the parameter to be estimated, is the initial value of the parameter to be estimated, is the initial value of the error covariance matrix, is the identity matrix; Step 33: perform iterative solution and estimate the open circuit voltage according to formula (9). The iterative process is as follows: (11) (12) (13) (14) Among them, k refers to the k-th moment, is the battery terminal voltage, for The estimated error, , is the parameter matrix to be estimated, is the gain coefficient, is the covariance moment, For the forgetting factor.

6. The method for estimating the battery state of charge considering the health state according to claim 1, characterized in that: In step 4, the specific method of constructing the neural network model includes: Construct a BP neural network model consisting of input layer, hidden layer and output layer, and its input feature sequence is , that is, the open circuit voltage and health status of the battery, the output sequence is , which is the state of charge of the battery. The number of neurons in the hidden layer is set to 64, and the activation function is the ReLU function. The output of the entire network for any sample is: (15) in, is the connection weight between the i-th neuron in the hidden layer and the output layer, is the activation function, is the i-th row of the weight matrix connecting the input layer and the hidden layer, is the input matrix of the jth sample, is the i-th row of the bias matrix connecting the input layer and the hidden layer, is the output layer bias value, and P is the number of samples.

7. The method for estimating the battery state of charge considering the health state according to claim 1, characterized in that: In step 5, fusion is performed through a filtering algorithm, which specifically includes: Step 51, calculate the relationship between the current and the battery state of charge: (16) Among them, Q is the current maximum available capacity of the battery, T is the sampling interval, and I is the current; Step 52: Combining the SOC result estimated by the BP neural network with formula (16), the recursive formula of the state space equation is obtained: (17) in, is the estimated SOC value at time k, is the estimated SOC value at time k obtained by the BP neural network model, is the current at time k; Step 53: Simplify the state equation and observation equation of the system, and use the Kalman filter algorithm to perform time update and measurement update to obtain the final battery state of charge estimate: (18) in, is the system state variable, namely SOC, is the output, i.e. the SOC estimated by the BP neural network model, is the input, A is the state matrix, B is the input matrix, C is the output matrix, is the process noise, is the observation noise.

8. An estimation system based on the battery state of charge estimation method considering the health state according to any one of claims 1 to 7, characterized in that: include: Dataset building module, used to obtain battery charging and discharging data and corresponding health status; Battery model building module, used to build battery model; An open circuit voltage estimation module, used to estimate the open circuit voltage of the battery according to the constructed data set and battery model; A neural network model building module is used to build a neural network model with open circuit voltage and charge and discharge data health status as input and battery charge status as output; The battery state of charge estimation module is used to combine the current information in the data set with the initial value of the battery state of charge estimation output by the neural network model, and fuse them through a filtering algorithm to obtain the final battery state of charge estimation value.

Images