A Li-Ion Battery SOC Estimation Method Based On Decision Tree Optimized Fuzzy Logic Equivalent Circuit Model

A novel schema of Lithium ion Battery or Li-ion Battery SOC estimation is proposed in this dissertation to solve the problem of battery equivalent circuit model parameters variation, which is a major barrier of the precise estimation of battery SOC. The main idea is to identify the equivalent circuit model parameters concurrently at the same time of battery SOC estimation.The external characteristics of Li-ion Battery is briefly investigated, according to which the battery equivalent circuit model is deduced and interpreted. Thereafter, a bat-tery terminal voltage integral based method is introduced to derive the specific values of battery model parameters. An experiment platform is built up to collect the data needed for the battery model parameters derivation. With the data collected at different ambient temperatures, the battery equivalent circuit model parameters are calculated.The parameters identification is implemented with a Fuzzy Logic model (FLM) due to its simplicity and satisfying efficiency. The inputs of the FLM are the ambient tem-perature and the battery SOC, whereas the outputs are the battery equivalent circuit model parameters. Mamdani reasoning law and Center of Gravity Defuzzifier is em-ployed in the fuzzy implication and defuzzification process.The fuzzy rule base is obtained through the Decision Tree algorithm instead of from human assignation for the sake of improving the correspondence of the Fuzzy Logic model to the reality. In the Decision Tree training process, C4.5 algorithm is utilized to construct the decision tree.Extended Kalman Filter (EKF) is recruited to accomplish the job of battery SOC es-timation. As a recursive algorithm, EKF is normally recruited on the basis of a discrete system model, so the battery state space mathematical model is deduced and discretized. Thereafter, it is combined with EKF to implement the battery SOC estimation. Each time before EKF starts a new round of calculation, the battery model parameters are identified and updated ensuring the model to emulate the practical battery behavior pre-cisely. Consequently, the battery SOC can be estimated accurately.