Advanced online battery states coestimation using Kalman filter for electric vehicle Applications / Prashant Shrivastava

• Main Author: Prashant , Shrivastava
• Format: Thesis
• Published: 2021
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://studentsrepo.um.edu.my/14570/1/Prashant.pdf; http://studentsrepo.um.edu.my/14570/2/Prashant_Shrivastava.pdf; http://studentsrepo.um.edu.my/14570/
• Institution: Universiti Malaya

Carbon impression and the growing reliance on fossil fuels are two unique concerns for world emission regulatory agencies. These issues have placed electric vehicles (EVs) powered by lithium-ion batteries (LIBs) on the forefront as alternative vehicles. The LIB has noticeable features, including high energy and power density, compared with other accessible electrochemical energy storage systems. However, LIB is exceedingly nonlinear and dynamic; therefore, it requires an accurate state estimation technique in a battery management system (BMS). Due to the existing correlation between the battery states, the co-estimation method for different battery states estimation is preferred over individual state estimation. Though, the trade-off between accuracy and computational burden of the co-estimation method is difficult to maintain in real-time application. This thesis focuses on the development of the co-estimation methods of lithium-ion battery states of interest, which are capable to improve the efficiency of BMS, especially for EV applications. To achieve high estimation accuracy at a low cost, the co-estimation method for state of charge (SOC) and state of energy (SOE) is investigated in the first phase of the thesis. A new dual forgetting factor-based adaptive extended Kalman filter (DFFAEKF) algorithm to concurrently estimate the electrical equivalent circuit model parameters and SOC at high accuracy is first developed. The DFFAEKF algorithm has the feature to reduce the possibility of battery model parameter divergence from the true value under different dynamic conditions with the same order of big O notation complexity as DEKF. Thereafter, with the credible SOC estimation by using DFFAEKF, a co-estimation method for SOC and SOE using a quantitative relationship between SOC and SOE is developed. The simplicity of the proposed co-estimation method can avoid the heavy computational burden required by the individual state estimation of SOC and SOE. Finally, to effectively utilize the correlation amongst battery states and reduce the computational burden of the BMS, a unified frame of co-estimation method for battery states including SOC, SOE, state of power (SOP). actual capacity and maximum available energy is developed. In addition to co-estimation of SOC and SOE in the first method, the SOP estimation is performed by using identified Rint battery model parameters using the forgetting factor recursive least square (FFRLS) algorithm. Next, the actual capacity and maximum available energy estimation are performed by using a new sliding windowapproximate weighted total least square (SW-AWTLS) algorithm at a low computational burden. The performance of the proposed co-estimation methods are experimentally verified with battery cells of different chemistries and dynamic load profiles which suitable for EV. Besides, the low computational burden of the proposed co-estimation, the results demonstrate the high accuracy of the battery states estimation irrespective of the change in battery chemistry under-considered dynamic operating conditions. With the effective utilization of battery states correlation and high estimation accuracy of the battery states co-estimation methods, the performance of the BMS can be significantly improved. Furthermore, the proposed co-estimation methods in this thesis can contribute to the safe, reliable, and efficient utilization of the LIBs used in EV applications.