Advanced Intelligent Hybrid Approach for State of Charge Estimation of Li-Ion Batteries in Electric Vehicles under Various Drive Cycles
Abstract
Lithium-ion batteries are currently used in various fields for environmental and resource concerns. An accurate estimation of the state of charge (SOC) facilitates the battery's functioning effectively. Battery management systems (BMS), more frequently used in electric vehicles (EV) and smart grid technologies, need to estimate their SOC to run safely and efficiently. However, the issue of SOC estimation at ambient temperatures with different drive cycles has not received much attention from existing approaches. In this research, an advanced intelligent hybrid approach consisting of stacked long short-term memory (SBLSTM) networks and convolutional neural networks (CNN) was presented. The features are automatically extracted by using a two-dimensional convolutional neural network model which has been developed. In addition, an estimator called an SBLSTM captures the backward and forward temporal dependency of battery sequential states. The combined CNN-SBLSTM network is studied. SOC estimation is conducted under different drive cycles to check the network acceptability. The root mean square error and maximum SOC error are both estimated to be less than 0.1% and 0.2%, respectively. Furthermore, the computational time is achieved at 23.083 and 13.531 seconds. The results of the experiments demonstrate that the approach is more accurate, stable, and computationally efficient.