Research on the state of charge estimation method of lithium‐ion batteries based on novel limited memory multi‐innovation least squares method and SDE‐2‐RC equivalent model

摘要

Summary Because of the common data redundancy phenomenon in the current least‐squares parameter identification algorithm and the complex offline parameter identification process, this research innovatively proposes a Limited Memory Multi‐Innovation Least Squares (LM‐MILS) ternary lithium‐ion battery (LIB) parameter identification algorithm that uses a limited set of data to estimate model parameters and attenuates the effects of old data. To improve the parameter fidelity of the equivalent circuit model (ECM) of the LIB, considering that the open‐circuit voltage of the lithium‐ion battery will gradually decrease with the self‐discharge when it is not in use, based on a large number of experiments, a model considering the self‐discharge of the LIB is constructed. The experimental results show that the self‐discharge effect‐2‐RC (SDE‐2‐RC) model can achieve higher accuracy in simulating the working state of the battery and the peak error of the simulated voltage is only 0.04342?V, and the accuracy can reach more than 98.966. Using LM‐MILS and adaptive Kalman filtering algorithm (AEKF) for the state of charge (SOC) estimation, the results show that the algorithm has a fast convergence speed and strong tracking performance. The maximum SOC estimation errors in HPPC, DST, and BBDST three operating conditions are 0.00929, 0.01273, and 0.01002, respectively. The fluctuation range is small, and the maximum estimation error is less than 2, which verifies that the improved parameter identification algorithm has good performance in improving the SOC estimation accuracy of LIB.