1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Title:	State of charge estimation for lithium-ion battery using artificial intelligent algorithm based heuristic optimization techniques
Authors:	Molla Shahadat Hossain Lipu (P85848)
Supervisor:	Aini Hussain, Prof. Dr.
Keywords:	Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia Artificial intelligent Algorithm Lithium-ion battery
Issue Date:	8-Feb-2019
Description:	State of charge (SOC) is one of the crucial assessment indexes in the battery energy storage management system. As such, SOC estimation has gained attention for lithium-ion battery due to its lucrative characteristics of fast charging, high voltage, high energy density, and long-life cycle. SOC estimation is essential to control battery charging, discharging and extend the battery lifespan. However, accurate SOC estimation is a serious concern due to the lithium-ion battery nonlinear characteristics and complex electrochemical reactions. The existing SOC estimation techniques suffer from high computational efforts and cannot deliver accurate results due to various uncertainties such as temperature, noise and aging. Therefore, the objective of this research is to develop an enhanced SOC estimation method using an artificial intelligent algorithm. Recurrent nonlinear autoregressive with exogenous inputs (RNARX) neural network algorithm is a well-known subclass of artificial intelligent algorithm that has received popularity in designing complex and nonlinear system due to its improved learning capability, convergence speed, generalization performance and high accuracy. In addition, RNARX algorithm does not require battery model, parameters and can be designed without the information about battery internal characteristics. However, the performance of the RNARX algorithm depends on the training algorithm, activation function, and the selection of the appropriate hyper parameters, which are determined by an ineffective trial and error method. To overcome this problem, the heuristic optimization techniques are applied to find the optimal hyper parameters and improve the computation capability. In this study, RNARX algorithm based lightning search algorithm (LSA) is developed to enhance SOC estimation accuracy. The developed method is designed using input information, objective function, and optimization constraints. To evaluate the reliability and efficiency of LSA, it is compared with backtracking search algorithm (BSA), gravitational search algorithm (GSA) and particle swarm optimization (PSO). Furthermore, the performance of RNARX based LSA (RNARX-LSA) algorithm for SOC estimation is compared with state of art artificial intelligent algorithms including back-propagation neural network (BPNN), radial basis function neural network (RBFNN), extreme learning machine (ELM), and deep recurrent neural network (DRNN) and random forests (RFs) algorithm. The proposed method is validated by developing a battery test bench model with BTS 4000, associated software, lithium-ion battery, and the host computer. The different battery experimental tests are conducted to check the SOC estimation results. The results demonstrate that the RNARX-LSA method produces a significant improvement in reducing the objective function. The accuracy and robustness of the RNARX-LSA model are further verified under different EV drive cycles, temperatures, and noise effects. The obtained results clearly indicate that the proposed model outperforms other optimized artificial intelligent based SOC estimation method in terms of accuracy and computational time. Moreover, the accuracy of the proposed SOC estimation results is checked under aging effects. Two different chemistry of lithium-ion cell is employed to evaluate the discharge capacity, cycle life and SOC under four milestone aging cycles; 50 cycles, 100 cycles, 150 cycles and 200 cycles. The results show that the developed method has demonstrated to become an accurate and robust method that can examine SOC accurately under different aging cycles. Therefore, RNARX-LSA based SOC estimation model has great potential to be implemented in real time electric vehicle battery storage systems.,Ph.D.
Pages:	236
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.