Method for estimating residual electric quantity of lithium iron phosphate power battery

Abstract

The invention discloses a method for estimating the remaining capacity of a lithium iron phosphate power battery, and belongs to the technical field of power battery state monitoring. The method comprises the following steps: establishing a second-order RC equivalent circuit model of the lithium iron phosphate power battery, and constructing a continuous state space equation based on the Kirchhoff's law; discretizing the continuous state space equation through a circuit three-element method, and obtaining a discrete state space equation in combination with an ampere-hour integral method; identifying model parameters on line by using a forgetting factor recursive least square method; and inputting the identified model parameters into a state estimator constructed based on a multi-information square root unscented Kalman filtering algorithm in real time, and realizing SOC real-time estimation through Sigma point sampling, multi-information matrix updating and square root filtering. Through verification under DST and FUDS cycle working conditions, the maximum errors are respectively as low as 1.5% and 1.8%, the root mean square errors are 0.16% and 0.32%, and it is proved that high-precision and high-stability estimation of the SOC of the battery is achieved under the complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of power battery state monitoring technology, specifically to a method for estimating the remaining capacity of a lithium iron phosphate power battery. Background Technology

[0002] With the rapid development of the new energy industry, lithium iron phosphate power batteries have been widely used in passenger cars, commercial vehicles, construction machinery, robots, and other fields due to their advantages such as low noise, high energy efficiency, low maintenance costs, and environmental friendliness. The state of charge (SOC) of a battery is a core parameter of the battery management system, and its estimation accuracy directly affects the battery's efficiency, safety, and lifespan.

[0003] Currently, common SOC estimation methods include the ampere-hour integration method, the open-circuit voltage method, and the Kalman filter algorithm. The ampere-hour integration method is simple and easy to implement, but it suffers from initial SOC error and accumulated error. The open-circuit voltage method requires the battery to be stationary for a long time and cannot be used for online estimation. The Kalman filter algorithm and its improved forms, such as the extended Kalman filter, are suitable for nonlinear systems, but for systems like lithium iron phosphate batteries with a flat open-circuit voltage platform and complex operating conditions, the estimation accuracy and stability still face challenges. This is mainly reflected in: 1) Battery model parameters (such as internal resistance, polarization resistance / capacitance) dynamically change with temperature, SOC, and aging degree; models with fixed parameters cannot accurately describe the dynamic characteristics of the battery; 2) Traditional state estimation algorithms are prone to large estimation errors when dealing with complex nonlinear systems. Summary of the Invention

[0004] The purpose of this invention is to overcome the problems of existing technologies in accurately describing the dynamic characteristics of lithium iron phosphate power batteries or easily resulting in large estimation errors when estimating their power capacity. This invention provides a method for estimating the remaining power capacity of lithium iron phosphate power batteries that can identify battery model parameters online and achieve high-precision and high-stability SOC estimation.

[0005] According to one aspect of the present invention, a method for estimating the remaining capacity of a lithium iron phosphate power battery is provided, comprising the following steps: S1, establishing a second-order RC equivalent circuit model of the lithium iron phosphate power battery, and obtaining its continuous state-space equation based on Kirchhoff's laws; S2, discretizing the continuous state-space equation, and constructing a discrete state-space equation with polarization voltage and state of charge (SOC) as state variables by combining the ampere-hour integral method; S3, using the forgetting factor recursive least squares (FFRLS) method to identify the model parameters in the discrete state-space equation online, and obtaining real-time updated model parameters; S4, inputting the model parameters in real time to a state estimator constructed based on the multiple innovation square root unscented Kalman filter (MISRUKF) algorithm to estimate the SOC of the lithium iron phosphate power battery in real time.

[0006] Based on the above, more specifically, the second-order RC equivalent circuit model in step S1 includes a series-connected battery ohmic internal resistance, a first RC parallel network for simulating electrochemical polarization, and a second RC parallel network for simulating concentration polarization; the continuous state-space equations include equations describing the first polarization voltage of the first RC parallel network, equations describing the second polarization voltage of the second RC parallel network, and equations describing the battery terminal voltage.

[0007] More specifically, step S2 includes: discretizing the equations for the first polarization voltage and the second polarization voltage using the three-element circuit method to obtain discrete equations for the first polarization voltage and the second polarization voltage; obtaining the discrete equation for the state of charge (SOC) by combining the ampere-hour integration method; and obtaining the discrete equation for the battery terminal voltage based on the equation for the battery terminal voltage. The four discrete equations constitute the discrete state-space equation.

[0008] More specifically, step S3, which utilizes FFRLS to perform online identification of model parameters in the discrete state-space equations, includes: converting the continuous state-space equations into transfer functions expressed by model parameters using a Laplace transform, wherein the transfer function takes battery current as input and the difference between battery terminal voltage and battery open-circuit voltage as output; discretizing the transfer function using a bilinear transform to obtain a discrete transfer function; constructing an FFRLS regression equation based on the discrete transfer function, wherein the regression equation takes the difference between battery open-circuit voltage and battery terminal voltage as the system measurement value, and uses the difference from the previous two time points, the current time point, and the battery current from the previous two time points as regression quantities to establish a linear relationship between the two; setting an FFRLS forgetting factor, iteratively updating the model parameters using the FFRLS recursive formula, and back-deriving the model parameters based on the preset mapping relationship between the discrete transfer function and the model parameters.

[0009] Based on the above, more specifically, the forgetting factor of FFRLS in step S3 ranges from 0.95 to 1, and the model parameters include the battery ohmic internal resistance, electrochemical polarization resistance, concentration polarization resistance, electrochemical polarization capacitance, and concentration polarization capacitance.

[0010] Based on the above, more specifically, the state estimation process of the state estimator in step S4 is as follows: a) Initialization: Set an initial state vector and its initial covariance matrix, and perform Cholesky decomposition on the initial covariance matrix to obtain the square root of the covariance matrix, wherein the initial state vector includes the initial first polarization voltage, the initial second polarization voltage, and the initial SOC; b) Unscented transformation: Based on the initial state vector, the square root of the covariance matrix, and a preset scaling ratio, construct 2n+1 Sigma points through unscented transformation, where n is the dimension of the state variables; c) State prediction: Substitute all Sigma points into the discrete state space equation to perform state prediction, and obtain the predicted Sigma point corresponding to each Sigma point; perform weighted summation on all predicted Sigma points according to a preset mean weight to obtain the state. d) Predicted mean: Based on the difference between each predicted Sigma point and the state prediction mean, the state prediction error is obtained; d) Square root filtering: Construct a state prediction error matrix based on all state prediction errors, and obtain the square root matrix of the state prediction covariance through QR decomposition; e) Observation prediction: Substitute the predicted Sigma points into the discrete equation of the battery terminal voltage to obtain the observation prediction value, calculate the observation prediction mean according to the mean weight, calculate the observation prediction covariance by combining the covariance weight and the observation noise covariance, and simultaneously calculate the cross-covariance between the state prediction error and the observation prediction error, where the observation prediction error is the difference between each observation prediction value and the observation prediction mean; f) Multi-information update: Construct a multi-information matrix composed of the current time residual and the residuals of the previous M-1 time steps, where the residual is the difference between the current actual observation value and the observation prediction mean, M denoted as the sequence length of the multi-information matrix; Calculate the multi-information gain matrix composed of the Kalman gains at the current time and the previous M-1 time steps, where the Kalman gains are based on the cross-covariance and the observation-prediction covariance; Update the state vector and the square root of the covariance matrix based on the multi-information matrix and the multi-information gain matrix; g) Output: Extract the SOC at the current time from the updated state vector as the real-time estimate of the battery charge.

[0011] Based on the above, more specifically, in step f), the covariance square root matrix is ​​updated using the Cholesky update algorithm based on the predicted covariance square root matrix, Kalman gain, and observed predicted covariance square root matrix.

[0012] According to another aspect of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method for estimating the remaining power of a lithium iron phosphate power battery according to the present invention when executing the computer program.

[0013] According to another aspect of the present invention, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the method for estimating the remaining power of a lithium iron phosphate power battery according to the present invention.

[0014] According to another aspect of the present invention, a computer program product is provided, comprising computer instructions that, when executed by a processor, implement the method for estimating the remaining power capacity of a lithium iron phosphate power battery according to the present invention.

[0015] The beneficial effects of this invention are as follows: 1. This invention uses the Forgotten Factor Recursive Least Squares (FFRLS) method to identify battery model parameters online, which can track changes in battery dynamic characteristics in real time, overcome the problem of accuracy degradation of fixed parameter models under varying operating conditions, and provide an accurate model basis for state estimation.

[0016] 2. This invention employs the multi-innovation square root unscented Kalman filter (MISRUKF) algorithm for state estimation. Unscented transformation can more accurately handle the nonlinearity of the system; square root filtering ensures the symmetric positive definiteness of the covariance matrix, improving the numerical stability and convergence of the algorithm; the multi-innovation method fully utilizes historical observation data, effectively improving estimation accuracy and anti-interference capability.

[0017] 3. This invention organically combines FFRLS and MISRUKF to form a closed-loop optimization system for online parameter identification and accurate state estimation. FFRLS provides MISRUKF with a real-time and accurate model, which MISRUKF uses for high-precision state estimation. The two work together to ensure that SOC estimation maintains high accuracy and strong robustness even under complex dynamic conditions. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method for estimating the remaining power of a lithium iron phosphate power battery according to the present invention.

[0019] Figure 2 This is a schematic diagram of the second-order RC equivalent circuit model of the lithium iron phosphate power battery according to the present invention.

[0020] Figure 3 This is a flowchart illustrating the overall workflow of the method for estimating the remaining power capacity of a lithium iron phosphate power battery according to the present invention.

[0021] Figure 4 This is a comparison curve of the SOC values ​​calculated by the three combined algorithms under DST conditions and the actual SOC values.

[0022] Figure 5 This is a comparison chart of the SOC error curves calculated by the three combined algorithms under DST conditions.

[0023] Figure 6 This is a comparison curve of the SOC values ​​calculated by the three combined algorithms under FUDS conditions and the actual SOC values.

[0024] Figure 7 This is a comparison chart of the SOC error curves calculated by the three combined algorithms under FUDS conditions.

[0025] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0026] like Figure 1 As shown, this invention provides a method for estimating the remaining capacity of a lithium iron phosphate power battery, comprising the following steps: S1, establishing a second-order RC equivalent circuit model of the lithium iron phosphate power battery, and obtaining its continuous state-space equation based on Kirchhoff's laws; S2, discretizing the continuous state-space equation, and constructing a discrete state-space equation with polarization voltage and state of charge (SOC) as state variables by combining the ampere-hour integral method; S3, using the forgetting factor recursive least squares (FFRLS) method to identify the model parameters in the discrete state-space equation online, and obtaining real-time updated model parameters; S4, inputting the model parameters in real time into a state estimator constructed based on the multiple innovation square root unscented Kalman filter (MISRUKF) algorithm to estimate the SOC of the lithium iron phosphate power battery in real time.

[0027] It should be noted that, due to the nonlinear relationships between the various parameters within a lithium iron phosphate (LFP) power battery, establishing a relevant mathematical model is a crucial step. However, the operating conditions of LFP batteries are complex and variable, affecting numerous parameters, necessitating the selection of a relatively simple model. Therefore, the equivalent circuit model should possess characteristics of structural stability, computational simplicity, and good accuracy. This invention chooses to use a second-order RC equivalent circuit model because, compared to a third-order RC equivalent circuit, the second-order RC equivalent circuit model requires less computation and has sufficient accuracy. For example... Figure 2 As shown, the second-order RC equivalent circuit model used in this invention includes the series-connected ohmic internal resistance of the battery. A first RC parallel network for simulating electrochemical polarization and a second RC parallel network for simulating concentration polarization.

[0028] In step S1, the state variables of the continuous state-space equations include the first polarization voltage of the first RC parallel network. The second polarization voltage of the second RC parallel network And the battery's state of charge (SOC). The continuous state-space equations are as follows: (1) (2) (3) in, for Rate of change over time; for Rate of change over time; and These are the polarization internal resistances of the first RC parallel network and the second RC parallel network, respectively. and These are the polarization capacitors of the first RC parallel network and the second RC parallel network, respectively. This is the battery open-circuit voltage; The current flowing through the battery; This is the battery terminal voltage.

[0029] The above expression can be transformed into the following expression using state equation transformation: (4) Since information from lithium iron phosphate power batteries is collected at intervals in practical engineering applications, it is necessary to establish a discrete time-domain expression for the model. In step S2, the collection period is set to... By transforming formulas (1) and (2) using the three-element method of circuit, the first polarization voltage is obtained. Second polarization voltage Discrete equations: (5) Combining the ampere-hour integration method: (6) in, Let be the battery level at time t; This is the initial battery level. The efficiency is typically 1 during discharge and 0.98~1 during charging. The current at any given time is positive during charging and negative during discharging, and the unit is ampere (A). This represents the maximum total usable capacity of the battery, measured in ampere-hours (Ah).

[0030] The discrete equations of SOC can be obtained: (7) in, for The current at any given time is positive during charging and negative during discharging, and the unit is ampere (A). The time k is the moment in seconds (s).

[0031] Combining formulas (4), (5), and (7), the discrete equations of the second-order RC equivalent circuit model can be obtained: (8) It is worth mentioning that the discrete equation of the battery terminal voltage is part of the discrete equation of the second-order RC equivalent circuit model. This is the state observation equation.

[0032] The Recursive Least Square (RLS) algorithm is an identification algorithm based on adaptive filtering theory. Its prerequisite is minimizing the sum of squared errors. As an online parameter identification algorithm, RLS is widely used in dynamic, time-varying systems. However, RLS suffers from the problem of repeated use of previously used data due to the increasing recursion, leading to a decrease in the weight of new data, affecting data correction, and ultimately causing the error to increase. This invention employs the Forgetting Factor Recursive Least Square (FFRLS) method, which adds a forgetting factor to the original algorithm to ensure iterative data updates.

[0033] The recursive formula for FFRLS is as follows: (9) in, This is the estimate of the parameters at time k; These are system measurements; This is the system output value; This represents the prediction error; Let k be the covariance matrix at time k; This is the gain matrix; It is the identity matrix; The forgetting factor ranges from 0.95 to 1. When the value is 1, the algorithm becomes the RLS algorithm. When it is less than 1, it has a good forgetting effect on historical data, but the smaller the factor, the more likely it is that the result will not converge.

[0034] Convert formula (3) to a formula usable by FFRLS: (10) To simplify the formula, set , .

[0035] To simplify again, here are the settings: (11) In step S3, the continuous state-space equation (10) from step S1 is transformed into a transfer function using the Laplace transform: (12) The transfer function is discretized by (12) using a bilinear transformation, wherein, for the bilinear transformation, Let z be a discrete-domain operator. Assume: (13) Reset: (14) The final regression equation for FFRLS is obtained as follows: (15) in, , , , , (16) Where T is the sampling duration, substituting formula (15) into the FFRLS algorithm, we get: (17) in, These are the identification parameters for FFRLS.

[0036] set up Calculate the parameter expression of the model: (18) The unscented Kalman filter (USF) is a filtering algorithm for state estimation of nonlinear systems. It combines the Kalman filter with the unscented transform, samples representative points, and uses these points to describe the state of the nonlinear system. Compared to the Kalman filter, it has the advantage of not requiring linearization of the nonlinear function and can estimate state variables more accurately.

[0037] The principle of the multi-innovation identification method is to expand the residual vector of the filtering algorithm into an innovation matrix using data from multiple time points. This process makes good use of historical observation data, thus enabling it to be effectively applied to nonlinear systems. Specifically, the residual vector represents the difference between the actual and estimated values ​​of the system observations, and is used for updating state estimates, estimating observation noise, detecting outliers, and evaluating filter performance.

[0038] according to Figure 2 The battery current is used as the model input u, and the battery terminal voltage is used as the model output y, with the first polarization voltage as the model input u. Second polarization voltage Using SOC as the state variable, and combining it with formula (8), we obtain the state prediction equation: (19) 1) Initialize the state vector and covariance matrix: (20) (twenty one) 2) Unscented transformation: Constructing the Sigma point: (twenty two) The number of Sigma points and the state variables Dimensional related, for According to formula (19), we know =3, so the number of Sigma points is 7, and the weights of the Sigma points satisfy: (twenty three) in, For scaling ratio, and , , and These are empirical parameters, where It is a non-negative number; The value of must be guaranteed It is a positive definite matrix; The weights are the mean of the predicted values; Weights for the covariance of predicted values; The first root of the matrix List.

[0039] 3) Substitute the Sigma point into the state prediction equation (19) to predict the state of the Sigma point: (twenty four) 4) The predicted state value and covariance are obtained through weighted calculation: (25) in, This is the predicted state value; This represents the covariance value; Let be the process noise covariance matrix.

[0040] 5) After unscented transformation, a set of Sigma points is obtained: (26) 6) Substitute the new Sigma point set into the state observation equation (8) to obtain the observed prediction values: (27) 7) Calculate the system observation predictions, the covariance of the sampling point observation predictions, and the cross-covariance between the sampling point observation predictions and the state predictions: (28) 8) Calculate the Kalman filter gain: (29) in, This represents the Kalman filter gain.

[0041] 9) Update system state and covariance: (30) The following describes the combination of multiple innovation identification methods with unscented Kalman filtering.

[0042] A single residual is defined according to formula (30): (31) By extending the individual residuals, we obtain the multi-inspiration matrix: (32) Where M is the sequence length of the multi-information matrix.

[0043] The gain of formula (29) Expanded into a matrix : (33) Combining the above formulas, formula (30) can be transformed into a state update equation: (34) Combining the square root filtering algorithm can represent the covariance matrix in square root form, avoiding numerical instability, ensuring stability and convergence, while also reducing computational complexity and improving overall running speed.

[0044] The following describes the state estimation process in step S4, which combines the multi-innovation identification method, the unscented Kalman filter method, and the square root filter algorithm.

[0045] 1) Define the initial state vector and its initial covariance matrix: (35) in, The function to perform Cholesky decomposition.

[0046] 2) Unscented Transformation: Constructing Sigma sample points: (36) 3) Predict the Sigma sample points to obtain the predicted Sigma points; and sum all the predicted Sigma points according to the preset mean weight to obtain the state prediction mean: (37) (38) 4) Square root filtering: Based on the difference between each predicted Sigma point and the mean of the state prediction, the state prediction error is obtained. A state prediction error matrix is ​​constructed based on all state prediction errors, and the square root matrix of the state prediction covariance is obtained through QR decomposition. (39) in, This is a matrix factorization method used for extracting matrix eigenvalues; This is the update function derived from the square root method.

[0047] Update a set of Sigma points using an unscented transformation: (40) 5) Substitute the predicted Sigma point into the discrete equation of the battery terminal voltage to obtain the observed predicted value, and calculate the observed predicted mean value according to the mean weight: (41) The observation prediction covariance is calculated by combining the covariance weights and the observation noise covariance, and the cross-covariance between the state prediction error and the observation prediction error is also calculated: (42) 6) Multi-information update: Construct a multi-information matrix consisting of the residual at the current time step and the residuals at the previous M-1 time steps, where the residual is the difference between the current actual observation and the observed predicted mean, and M is the sequence length of the multi-information matrix. (43) Calculate the multi-inspiration gain matrix consisting of the Kalman gains at the current time and the previous M-1 time steps, where the Kalman gains are based on the cross-covariance and the observation-prediction covariance: (44) Update the state vector and the square root of covariance matrix based on the multi-inspiration matrix and the multi-inspiration gain matrix: (45) 7) Output: The updated state vector Extract the current SOC (State of Charge) as the real-time estimate of the battery capacity.

[0048] The above steps describe the calculation process of the multi-innovation square root unscented Kalman filter algorithm. To calculate the battery's SOC value more accurately and promptly, it is necessary to combine it with the forgetting factor recursive least squares method, ultimately forming the FFRLS-MISRUKF joint algorithm. The overall workflow of this algorithm is as follows: Figure 3 As shown.

[0049] To further verify the advantages of the method of this invention, comparative experiments were conducted with other mainstream Kalman filter joint algorithms, specifically the Forgot Factor Recursive Least Squares-Kalman Filter Joint Algorithm (FFRLS-KF) and the Forgot Factor Recursive Least Squares-Extended Kalman Filter Joint Algorithm (FFRLS-EKF). The State of Charge (SOC) values ​​of the power lithium battery were calculated and verified under DST and FUDS cycle conditions.

[0050] 1. Validation of Dynamic Stress Test (DST) Conditions Using the current and battery voltage under this operating condition as input signals, the battery SOC value is calculated in real time using three combined algorithms: FFRLS-KF, FFRLS-EKF, and FFRLS-MISRUKF, resulting in the following numerical curve. Figure 4 As shown, the corresponding error between the algorithm's estimated value and the actual value is as follows: Figure 5 As shown in the figure, the data was exported, and the final SOC estimation error results are shown in Table 1.

[0051] Table 1 shows the SOC error results calculated by the three combined algorithms under DST conditions according to the present invention; Table 1

[0052] Depend on Figure 4 and Figure 5 The simulation results show that, compared to FFRLS-KF and FFRLS-EKF, the values ​​calculated by FFRLS-MISRUKF are closer to the actual SOC values. Table 1 shows that the error curve of FFRLS-MISRUKF is relatively stable. The maximum error of FFRLS-KF is 3.2%, and the root mean square error is 0.75%; the maximum error of FFRLS-EKF is 2.5%, and the root mean square error is 0.31%; while the maximum error of FFRLS-MISRUKF is 1.5%, and the root mean square error is 0.16%.

[0053] 2. Federal Urban Driving Schedule (FUDS) Cyclic Test Validation Using the current and battery voltage under this operating condition as input signals, the battery SOC value is calculated in real time using three combined algorithms: FFRLS-KF, FFRLS-EKF, and FFRLS-MISRUKF, resulting in the following numerical curve. Figure 6 As shown, the corresponding error between the algorithm's estimated value and the actual value is as follows: Figure 7 As shown in the figure, the data was exported, and the final SOC estimation error results are shown in Table 2.

[0054] Depend on Figure 6 and Figure 7 The simulation results show that, compared to FFRLS-KF and FFRLS-EKF, the values ​​calculated by FFRLS-MISRUKF are closer to the actual SOC values. Table 2 shows that the error curve of FFRLS-MISRUKF fluctuates more smoothly. The maximum error of FFRLS-KF is 2.8%, and the root mean square error is 0.81%; the maximum error of FFRLS-EKF is 2.7%, and the root mean square error is 0.55%; while the maximum error of FFRLS-MISRUKF is 1.8%, and the root mean square error is 0.32%.

[0055] Table 2 shows the SOC error results calculated by the three joint algorithms under FUDS conditions according to the present invention; Table 2

[0056] As shown in Tables 1 and 2, the forgetting factor recursive least squares method-based multi-innovation square root unscented Kalman filter joint algorithm proposed in this invention can estimate the SOC value more accurately and stably under complex working conditions compared with other mainstream algorithms.

[0057] Figure 8 A schematic diagram of the physical structure of the electronic device of the present invention is shown, such as... Figure 8 As shown, the electronic device may include: a processor 110, a communication interface 120, a memory 130, and a communication bus 110, wherein the processor 110, the communication interface 120, and the memory 130 communicate with each other through the communication bus 110. The processor 110 can call the logic instructions in the memory 130 to execute the remaining power estimation method of the lithium iron phosphate power battery of the present invention. The method includes: S1, establishing a second-order RC equivalent circuit model of the lithium iron phosphate power battery and obtaining its continuous state-space equation based on Kirchhoff's laws; S2, discretizing the continuous state-space equation and constructing a discrete state-space equation with polarization voltage and state of charge (SOC) as state variables by combining the ampere-hour integral method; S3, using the forgetting factor recursive least squares (FFRLS) method to identify the model parameters in the discrete state-space equation online and obtain real-time updated model parameters; S4, inputting the model parameters in real time to a state estimator constructed based on the multiple innovation square root unscented Kalman filter (MISRUKF) algorithm to estimate the SOC of the lithium iron phosphate power battery in real time.

[0058] Furthermore, when the logical instructions in memory 130 can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0059] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the remaining power estimation method for lithium iron phosphate power batteries of the present invention. The method includes: S1, establishing a second-order RC equivalent circuit model of the lithium iron phosphate power battery and obtaining its continuous state-space equation based on Kirchhoff's laws; S2, discretizing the continuous state-space equation and constructing a discrete state-space equation with polarization voltage and state of charge (SOC) as state variables by combining the ampere-hour integral method; S3, using the forgetting factor recursive least squares (FFRLS) method to identify the model parameters in the discrete state-space equation online and obtain real-time updated model parameters; S4, inputting the model parameters in real time to a state estimator constructed based on the multiple innovation square root unscented Kalman filter (MISRUKF) algorithm to estimate the SOC of the lithium iron phosphate power battery in real time.

[0060] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program performs the remaining power estimation method for lithium iron phosphate power batteries of the present invention. The method includes: S1, establishing a second-order RC equivalent circuit model of the lithium iron phosphate power battery and obtaining its continuous state-space equation based on Kirchhoff's laws; S2, discretizing the continuous state-space equation and constructing a discrete state-space equation with polarization voltage and state of charge (SOC) as state variables by combining the ampere-hour integral method; S3, using the forgetting factor recursive least squares (FFRLS) method to identify the model parameters in the discrete state-space equation online and obtain real-time updated model parameters; S4, inputting the model parameters in real time to a state estimator constructed based on the multiple innovation square root unscented Kalman filter (MISRUKF) algorithm to estimate the SOC of the lithium iron phosphate power battery in real time.

[0061] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for estimating the remaining capacity of a lithium iron phosphate power battery, characterized in that, Includes the following steps: S1. Establish a second-order RC equivalent circuit model of lithium iron phosphate power battery, and obtain its continuous state-space equation based on Kirchhoff's laws. S2. Discretize the continuous state-space equations and, in conjunction with the ampere-hour integration method, construct discrete state-space equations with polarization voltage and state of charge (SOC) as state variables. S3. The model parameters in the discrete state space equation are identified online using the Forgotten Factor Recursive Least Squares (FFRLS) method to obtain the model parameters updated in real time. S4. Input the model parameters in real time into the state estimator constructed based on the MISRUKF algorithm (multiple innovation square root unscented Kalman filter) to estimate the SOC of the lithium iron phosphate power battery in real time.

2. The method for estimating the remaining capacity of a lithium iron phosphate power battery according to claim 1, characterized in that, The second-order RC equivalent circuit model in step S1 includes a series battery ohmic internal resistance, a first RC parallel network for simulating electrochemical polarization, and a second RC parallel network for simulating concentration polarization; the continuous state-space equations include an equation describing the first polarization voltage of the first RC parallel network, an equation describing the second polarization voltage of the second RC parallel network, and an equation describing the battery terminal voltage.

3. The method for estimating the remaining capacity of a lithium iron phosphate power battery according to claim 2, characterized in that, Step S2 specifically includes: discretizing the equations of the first polarization voltage and the second polarization voltage using the three-element circuit method to obtain discrete equations of the first polarization voltage and the second polarization voltage; combining the ampere-hour integration method to obtain the discrete equation of SOC; obtaining the discrete equation of the battery terminal voltage based on the equation of the battery terminal voltage, and the four discrete equations constitute the discrete state-space equation.

4. The method for estimating the remaining capacity of a lithium iron phosphate power battery according to claim 1, characterized in that, Step S3, which uses FFRLS to perform online identification of the model parameters in the discrete state-space equations, includes: The continuous state-space equations are transformed into transfer functions expressed by model parameters through Laplace transform. The transfer function takes the battery current as input and the difference between the battery terminal voltage and the battery open-circuit voltage as output. The transfer function is discretized using a bilinear transform to obtain a discrete transfer function. Based on the discrete transfer function, a regression equation for FFRLS is constructed. The regression equation uses the difference between the battery open-circuit voltage and the battery terminal voltage as the system measurement value, and the difference between the previous two time points, the battery current at the current time point and the previous two time points as the regression values ​​to establish a linear relationship between the two. The forgetting factor of FFRLS is set, and the model parameters are obtained by iteratively updating the model through the recursive formula of FFRLS and back-deriving the model parameters based on the preset mapping relationship between the discrete transfer function and the model parameters.

5. The method for estimating the remaining capacity of a lithium iron phosphate power battery according to claim 4, characterized in that, The forgetting factor of FFRLS in step S3 ranges from 0.95 to 1, and the model parameters include battery ohmic internal resistance, electrochemical polarization resistance, concentration polarization resistance, electrochemical polarization capacitance, and concentration polarization capacitance.

6. The method for estimating the remaining capacity of a lithium iron phosphate power battery according to claim 1, characterized in that, The state estimation process of the state estimator in step S4 is as follows: a) Initialization: Set the initial state vector and its initial covariance matrix, and perform Cholesky decomposition on the initial covariance matrix to obtain the square root of the covariance matrix. The initial state vector includes the initial first polarization voltage, the initial second polarization voltage and the initial SOC. b) Unscented transformation: Based on the initial state vector, the square root of covariance matrix and the preset scaling ratio, 2n+1 Sigma points are constructed through unscented transformation, where n is the dimension of the state variable; c) State prediction: Substitute all Sigma points into the discrete state space equation to perform state prediction, and obtain the predicted Sigma point corresponding to each Sigma point; perform weighted summation of all predicted Sigma points according to the preset mean weight to obtain the state prediction mean; obtain the state prediction error based on the difference between each predicted Sigma point and the state prediction mean. d) Square root filtering: Construct a state prediction error matrix based on all state prediction errors, and obtain the square root matrix of state prediction covariance through QR decomposition; e) Observation and prediction: Substitute the predicted Sigma point into the discrete equation of the battery terminal voltage to obtain the observation and prediction value. Calculate the observation and prediction mean according to the mean weight. Calculate the observation and prediction covariance by combining the covariance weight and the observation noise covariance. At the same time, calculate the cross-covariance between the state prediction error and the observation and prediction error, where the observation and prediction error is the difference between each observation and prediction value and the observation and prediction mean. f) Multi-information update: Construct a multi-information matrix consisting of the residual at the current time step and the residual at the previous M-1 time steps, where the residual is the difference between the current actual observation and the observed prediction mean, and M is the sequence length of the multi-information matrix; Calculate the multi-information gain matrix consisting of the Kalman gains at the current time and the previous M-1 time points, where the Kalman gains are based on the cross-covariance and the observation-prediction covariance; Update the state vector and the square root of covariance matrix based on the multi-information matrix and the multi-information gain matrix; g) Output: Extract the current SOC from the updated state vector as a real-time estimate of the battery charge.

7. The method for estimating the remaining capacity of a lithium iron phosphate power battery according to claim 6, characterized in that, In step f), the covariance square root matrix is ​​updated using the Cholesky update algorithm based on the predicted covariance square root matrix, Kalman gain, and observed predicted covariance square root matrix.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable by the processor, characterized in that, When the processor executes the computer program, it implements the method for estimating the remaining power of a lithium iron phosphate power battery as described in any one of claims 1 to 7.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for estimating the remaining power of a lithium iron phosphate power battery as described in any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the method for estimating the remaining power of a lithium iron phosphate power battery as described in any one of claims 1 to 7.

Images