| CPC G01R 31/367 (2019.01) [G01R 31/392 (2019.01)] | 2 Claims |
|
1. A method for estimating a state of health of lithium-ion batteries considering user charging behavior, comprising the following steps:
S1, collecting data in a cycle charging process of an actual vehicle, wherein the data collected in S1 is obtained by carrying out a cyclic charging test on different types of lithium-ion batteries until a discharge capacity of each type of batteries is lower than 80% of a nominal capacity, and further comprises real-time recording data of battery charging voltage, charging capacity, start charging voltage and stop charging voltage of charging;
S2, analyzing the collected charging data;
an analysis process in S2 is as follows:
S2.1, reading a frequency of occurrence of the start charging voltage and the stop charging voltage during the cyclic charging process of lithium-ion batteries;
S2.2, calculating the frequency of occurrence of different start charging voltage and stop charging voltage;
S2.3, establishing a clustered bar chart based on the frequency of occurrence of the start charging voltage and the stop charging voltage in S2.1, and integrating it with the frequency of occurrence of the start charging voltage and the stop charging voltage in S2.2 to create a heat map;
S3, extracting data from a battery charging process during a cycling process as health features;
a specific process of setting the health features in S3 is as follows:
S3.1, establishing a IC curve during the battery charging process;
the process of establishing the IC curve in S3.1 is as follows:
S3.1.1, determining a relationship between capacity and voltage, with an expression as follows:
 wherein, Qa represents a charging capacity, I represents charging current, V represents battery voltage, ƒ(Qa) represents a mapping function from Qa to V, ƒ−1 represents an inverse function of ƒ, G represents a derivative of the inverse function ƒ−1, ∫Idt represents an integral of the current over time t, dQa represents a differential of the charging capacity Qa, dV represents a differential of the battery voltage V, dt represents a differential of time t;
S3.1.2, introducing fitting error, and replacing dV in S3.1.1 with a fixed voltage interval ΔV;
S3.2, denoising the data during S3.1;
in a denoising process in S3.2, selecting a Kalman filter algorithm for denoising, and establishing a state equation and a measurement equation; the expressions are deduced as follows:
 wherein, xk represents incremental capacity data at time k, ωk represents noise of a control system, xk-1 represents incremental capacity data at time k−1 yk represents measurement of noise pollution for xk, and vk represents measured noise;
 wherein, Q and R represent covariance matrices of process noise and measurement noise, respectively, xk− represents a prior estimation state value at time k, xk-1− represents a prior estimation state value at time k−1, xk represents a posterior estimation state value at time k, Pk− represents a posterior estimation covariance at time k, Pk represents a posterior estimation covariance at time k, Pk-1 represents a posterior estimation covariance at time k−1, and Kk represents a Kalman gain matrix under optimal estimation conditions, and the IC curve is obtained after the calculation is completed;
S3.3, determining the health features;
the data selected for the health features in S3.3 are corresponding peak values of two peaks and one valley of the denoised IC curve in S3.2, as well as a median voltage value selected within a voltage range of 3.718V to 4.0V;
S4, establishing a state of health estimation model based on the health features obtained in S3; wherein a process of S4 comprises:
S4.1, establishing a TCN-BiGRU model;
S4.2, taking the first 70% of the cyclic data features obtained in S3 as inputs for a TCN-BiGRU network, and taking the corresponding state of health values as outputs for the TCN-BiGRU network, calculating the battery health values using the TCN-BiGRU network to complete the model construction;
S5, estimating a state of health of the battery using the state of health estimation model established in S4.
|