State of Charge Estimation Analysis of Karman Filter, Unscented Karman Filter, and Recurrent Neural Network for Electric Double Layer Capacitors as Solar Energy Storage

Electric Double-Layer Capacitors (EDLCs) present a viable solution for efficient renewable energy storage systems, offering high power density and long lifecycle. Despite their advantages, there is a significant research gap in the methodologies of estimating the State of Charge (SoC) for EDLCs, as most studies have focused on lithium-ion batteries. To achieve this, a simulated solar panel generation model is utilized, serving as the basis for applying the Kalman Filter (KF), the Unscented Kalman Filter (UKF), and the Recurrent Neural Network (RNN) methods. The results indicate that the KF achieves a Mean Absolute Error (MAE) of 0.0295 and a Mean Squared Error (MSE) of 0.0123, demonstrating high accuracy with minimal deviation. The UKF, while designed to handle non-linearities better, shows an MAE of 0.0680 and an MSE of 0.1265, which are higher compared to the KF. The RNN model, although capable of capturing temporal dependencies, has an MAE of 0.2727 and an MSE of 0.1974, indicating the need for further development to reduce estimation errors. The KF outperforms the UKF and RNN in SoC estimation for EDLCs, providing the best accuracy and reliability for real-time applications in solar energy storage systems. The results from this study will enable further development of efficient energy storage systems and optimizing the performance and lifespan of EDLCs.