CN-122017632-A - Lithium battery SOC estimation method based on hybrid algorithm and error compensation network

CN122017632ACN 122017632 ACN122017632 ACN 122017632ACN-122017632-A

Abstract

The application discloses a lithium battery SOC estimation method based on a hybrid algorithm and an error compensation network, which comprises the steps of obtaining an SOC-OCV curve of a lithium battery, carrying out dynamic stress test DST on the lithium battery to obtain terminal voltage and load current of the lithium battery, establishing a second-order RC equivalent circuit model of the lithium battery, establishing an AE-DE hybrid algorithm optimization framework based on an alpha evolution AE algorithm and combining a differential evolution DE algorithm, carrying out parameter identification on the second-order RC equivalent circuit model to obtain model parameters, constructing an extended Kalman filter EKF to obtain a preliminary estimation value, and constructing a long-term memory LSTM network to obtain a final SOC estimation value. The application provides a hybrid optimization algorithm to improve the identification precision of parameters, solves the problem of estimation error accumulation under complex dynamic working conditions by using the combined framework of the EKF and the LSTM, and realizes high-precision SOC estimation.

Inventors

LI JUNHONG
Zhao Gaoqi
GU JUPING
ZHOU XINGTIAN
ZENG XIAORU
WANG YUTIAN

Assignees

南通大学

Dates

Publication Date: 20260512
Application Date: 20260126

Claims (8)

1. The lithium battery SOC estimation method based on the hybrid algorithm and the error compensation network is characterized by comprising the following steps: S1, acquiring an SOC-OCV curve of a lithium battery based on an intermittent constant current discharge experiment; S2, performing Dynamic Stress Test (DST) on the lithium battery to obtain terminal voltage and load current of the lithium battery; s3, establishing a second-order RC equivalent circuit model of the lithium battery based on battery characteristics; S4, based on an alpha evolution AE algorithm and a differential evolution DE algorithm, constructing an AE-DE hybrid algorithm optimization framework, and carrying out parameter identification on the second-order RC equivalent circuit model to obtain model parameters; S5, constructing an extended Kalman filter EKF, and primarily estimating the SOC of the lithium battery to obtain a primary estimated value; s6, constructing a long-term memory LSTM network, and carrying out learning and error compensation on the residual error of the preliminary estimated value to obtain a final SOC estimated value.
2. The method according to claim 1, characterized in that the method of step S3 comprises: Based on a formula (1), the second-order RC equivalent circuit model is obtained, wherein the formula (1) is as follows: (1); Wherein: is the open circuit voltage of a lithium battery, Is the terminal voltage of the lithium battery, And Is that And The voltage across the two terminals of the capacitor, Is the ohmic internal resistance of the lithium battery, And The parallel link of (a) represents the electrochemical polarization reaction of the lithium battery, And Represents concentration polarization reaction of the lithium ion battery, Is a current; based on a formula (2), obtaining the ratio of the residual capacity to the nominal capacity of the lithium battery, wherein the formula (2) is as follows: (2); Wherein, the And T and respectively The SOC value at the moment of time, Is the rated capacity of the battery; Acquiring a continuous state space expression based on formulas (3) - (4), wherein formulas (3) - (4) are as follows: (3); (4)。
3. the method according to claim 1, characterized in that the method of step S4 comprises: s41, acquiring an initial solution based on an AE algorithm; s42, carrying out alpha evolution on the initial solution to obtain an evolution matrix; S43, carrying out boundary constraint and strategy selection on the evolution matrix based on a distance halving method and a greedy selection algorithm to obtain an evolution vector; s44, based on the evolution vector, adding 3 strategies of variation, intersection and selection of a DE algorithm, constructing an AE-DE hybrid algorithm optimization framework, and obtaining the model parameters.
4. A method according to claim 3, characterized in that the method of step S41 comprises: based on a formula (5), acquiring a continuous state space expression, wherein the formula (5) is as follows: (5); Wherein: For the i-th candidate solution, which ranges i=1, 2,3, D represents dimensions, ub and lb represent upper and lower bounds, respectively, Representing the resulting 1 x D dimensional random matrix with elements satisfying a uniform distribution within (0, 1).
5. The method of claim 4, wherein the method of step S42 comprises: based on the formula (6), sampling and replacing the initial solution for N times to obtain an evolution matrix E to be evolved. Wherein, the formula (6) is: (6); Acquiring an alpha operator based on a formula (7), wherein the formula (7) is as follows: (7); Wherein, the For the ith evolution solution at the t+1th iteration, P is the base vector to determine the starting position of the evolution, Is the attenuation coefficient that the control algorithm explores and develops, For the ith random step size, In order to control the parameters of the differential vector, And All are sample solutions in X.
6. The method of claim 5, wherein the method of step S43 comprises: Obtaining a boundary constraint formula based on a formula (8), wherein the formula (8) is as follows: (8); Wherein, the The j-th element of the ith solution in the evolution matrix E. Based on a formula (9), judging and selecting the evolution solution through a greedy selection algorithm, wherein the formula (9) is as follows: (9)。
7. The method according to claim 1, characterized in that the method of step S5 comprises: Based on a formula (10), acquiring a state equation and an observation equation of a nonlinear system, wherein the formula (10) is as follows: (10); Wherein, the As a state variable, a state variable is used, In order to input the variable(s), In order to observe the variables of the object, And Is Gaussian white noise, and the variances are Q and R respectively.
8. The method according to claim 1, characterized in that the method of step S6 comprises: Constructing an LSTM error compensation network; and simultaneously carrying out residual estimation and learning error compensation on the LSTM error compensation network based on the model parameters and the preliminary estimated value, and carrying out network training by taking current, voltage change rate, current change rate and the preliminary estimated value as inputs to obtain a final SOC estimated value.

Description

Lithium battery SOC estimation method based on hybrid algorithm and error compensation network Technical Field The application belongs to the technical field of lithium battery estimation, and particularly relates to a lithium battery SOC estimation method based on a hybrid algorithm and an error compensation network. Background State of charge (SOC) estimation of lithium ion batteries is a core function of Battery Management Systems (BMS). The accurate SOC information not only can ensure the safe and stable operation of the battery system, but also is beneficial to optimizing the battery performance, and provides a key basis for the energy management and the safety control of the electric automobile. The establishment of an accurate and applicable battery model is the basis of SOC estimation, and common battery models include an electrochemical model, an equivalent circuit model, a fractional order model and the like. The equivalent circuit model uses common elements such as a resistor, a capacitor, a constant voltage source and the like to form a network to simulate the dynamic characteristics of the battery. The model has simple structure, easily acquired parameters and better precision, and is widely used. The parameter identification mode is mainly divided into two modes, namely off-line identification and on-line identification. The traditional online identification method can correct parameters and estimate states in real time, but in some cases, the conditions of large errors or failure can occur. The off-line identification can use a large amount of historical data, and the accuracy of the identified parameters is relatively high, so that the method has remarkable advantages in the accuracy aspect. Various intelligent algorithms are widely applied to the field of offline identification, but a single algorithm is easy to fall into local optimization, so that an optimal solution cannot be obtained. The proper SOC estimation method is of great importance, and the mainstream methods at present can be divided into two major categories, namely a model-based method and a data-based method. The model-based method mainly comprises various filtering algorithms, has the advantages of strict theory, strong interpretability and high calculation efficiency, but the accuracy of the model-based method is seriously dependent on the accuracy of a battery model, and has limited adaptability to complex nonlinear characteristics. The data-based method does not need an accurate physical model, can directly learn complex nonlinear mapping and long-term dependency relationship from data, has good generalization, but has poor interpretability and high requirements on data quality and quantity. Disclosure of Invention The application provides a lithium battery SOC estimation method based on a hybrid algorithm and an error compensation network, which aims to solve the technical problem of large SOC estimation error caused by the fact that the traditional method is easy to fall into local optimum. In order to solve the technical problems, the application adopts a technical scheme that the lithium battery SOC estimation method based on a hybrid algorithm and an error compensation network comprises the following steps: S1, acquiring an SOC-OCV curve of a lithium battery based on an intermittent constant current discharge experiment; S2, performing Dynamic Stress Test (DST) on the lithium battery to obtain terminal voltage and load current of the lithium battery; s3, establishing a second-order RC equivalent circuit model of the lithium battery based on battery characteristics; s4, based on an alpha evolution AE algorithm and a differential evolution DE algorithm, constructing an AE-DE hybrid algorithm optimization framework, and carrying out parameter identification on a second-order RC equivalent circuit model to obtain model parameters; S5, constructing an extended Kalman filter EKF, and primarily estimating the SOC of the lithium battery to obtain a primary estimated value; S6, constructing a long-term memory LSTM network, and carrying out learning and error compensation on residual errors of the preliminary estimated value to obtain a final SOC estimated value. Further, the method of step S3 includes: based on the formula (1), a second-order RC equivalent circuit model is obtained, wherein the formula (1) is as follows: (1); Wherein: is the open circuit voltage of a lithium ion battery, Is the terminal voltage of the lithium battery,AndIs thatAndThe voltage across the two terminals of the capacitor,Is the ohmic internal resistance of the lithium battery,AndThe parallel link of (a) represents the electrochemical polarization reaction of the lithium battery,AndRepresents concentration polarization reaction of the lithium ion battery,Is a current. The definition of SOC is the ratio of the remaining capacity to the nominal capacity, and the calculation formula is as follows: (2); Wherein, the AndT and respectivelyThe SOC valu