Battery State of Charge Estimation using Neural Network Aided Kalman Filter
Abstract (300 words maximum)
Precise estimation of Battery state of charge (SOC) poses a major challenge in managing and operating electric vehicles (EVs). This is mainly due to its complex behaviour that has a nonlinear dependency on parameters like temperature and SOC. Despite numerous SOC estimation approaches, their accuracy can be enhanced which is crucial for ensuring an optimized energy consumption. This research aims to increase the SOC estimation accuracy by investigating a neural network aided Kalman filter (KalmanNet), for the first time, to estimate battery SOC in EVs. The results demonstrate lower estimation error compared to well-known conventional filters used for SOC estimation. Specifically, the proposed approach is compared to the extended Kalman filter (EKF) and sigma-point Kalman filter (SPKF).
Academic department under which the project should be listed
SPCEET - Electrical and Computer Engineering
Primary Investigator (PI) Name
Yousef Mahmoud
Battery State of Charge Estimation using Neural Network Aided Kalman Filter
Precise estimation of Battery state of charge (SOC) poses a major challenge in managing and operating electric vehicles (EVs). This is mainly due to its complex behaviour that has a nonlinear dependency on parameters like temperature and SOC. Despite numerous SOC estimation approaches, their accuracy can be enhanced which is crucial for ensuring an optimized energy consumption. This research aims to increase the SOC estimation accuracy by investigating a neural network aided Kalman filter (KalmanNet), for the first time, to estimate battery SOC in EVs. The results demonstrate lower estimation error compared to well-known conventional filters used for SOC estimation. Specifically, the proposed approach is compared to the extended Kalman filter (EKF) and sigma-point Kalman filter (SPKF).
