Performance improvement of level-of-hydrogen estimation for a metal hydride-based hydrogen storage system

Accurate estimation of the Level-of-Hydrogen (LOH) in Metal Hydride (MH) systems is a critical state indicator for effective energy management. In this study, a real-time optimal estimation framework is proposed to surpass conventional machine learning approaches and traditional mass conservation methods, particularly in terms of accuracy, convergence speed, and robustness. For the first time, an algorithm based on the Extended Kalman Filter (EKF) is developed for LOH estimation. Specifically, an MH system model is established in MATLAB/Simulink®, with its key parameters identified and validated using experimental data. Subsequently, the state-space equations tailored for the MH system are formulated, defining LOH as the state variable and pressure as the observation variable. To address the limitations of standard filtering, the impacts of state noise, measurement noise, and initial value perturbations on estimation fidelity are systematically evaluated. An Adaptive EKF (AEKF) is further introduced to mitigate these effects and enhance tracking performance. The results demonstrate that: 1) State noise significantly influences estimation precision; 2) The EKF-based algorithm achieves high-fidelity results with an accuracy exceeding 97%; 3) The proposed AEKF further refines estimation accuracy while exhibiting superior convergence characteristics. In conclusion, the EKF-based framework provides a robust solution for LOH estimation in MH systems. Given its high reliability, computational efficiency, and practical feasibility, this method holds significant promise for both academic research and large-scale engineering applications.