A Comprehensive Review of Various Topologies and Control Techniques for DC-DC Converter-Based Lithium-Ion Battery Charge Equalization

Table 5

Comparison of an adaptive method of SOC estimation.

Methods	Principle	Merits	Demerits
NN [98]	NN is a framework for many machine learning algorithms to fulfil various tasks. The highly complex and nonlinear system can be handled using NN. SOC estimation is based on the training data used in NN, wh measured by experiments on charging and discharging batteries.	(i) Simple method; (ii) Minimum error; (iii) Solve nonlinear system	(i) The error will increase if enough trained data is not available; (ii) Need bulky memory unit
SVM [99]	In numerous domains of pattern recognition, SVM is used to classify data. Regression problems can also be solved with SVM. It’s capable of dealing with nonlinear systems.	(i) Simple method; (ii) More robust	(i) Insensitive to small changes in the input signal
FNN [100]	FNN is used to identify unknown systems. FNN can handle nonlinear systems by finding the learning mechanism’s optimal coefficient.	(i) Very effective; (ii) High performance	(i) Complex computation; (ii) Costly; (iii) Need bulky memory unit
KF [101]	KF is a valuable method for estimating SOC for a battery in real-time state estimation, and it’s also used to estimate the state of a dynamic system. The extended kalman filter (EKF) is a KF enhancement. A nonlinearization method expanded the nonlinear system dynamics and model measurement. This will increase the accuracy of SOC estimation.	(i) Self-correcting nature; (ii) More accuracy; (iii) Solve nonlinear dynamic errors	(i) Complex calculation; (ii) Linearization error will occur