To enable real-time detection of lithium deposition in energy storage systems, this paper proposes a portable electrochemical impedance spectroscopy (EIS) device that can be directly connected to battery packs without structural modifications. A1Hz sinusoidal current is applied during charging. The voltage response is analyzed using the Complex Morlet Wavelet Transform (CMWT) to detect lithium plating. A Gated Recurrent Unit (GRU) neural network predicts the 1 Hz impedance's real part evolution. Experimental results show the model achieves 0.21% mean absolute error (MAE) and 0.99 a coefficient of determination (R²). The system enables non-invasive monitoring and outperforms traditional Long Short-Term Memory (LST M) in accuracy and efficiency for energy storage diagnostics.