Two-way channel irrigation pumping stations are widely used along rivers for irrigation and drainage. Due to fluctuating internal and external water levels, these stations often operate under ultra-low or near-zero head conditions, leading to poor hydraulic performance. This study employs computational fluid dynamics (CFD) to investigate such systems’ pressure fluctuation and chaotic dynamic characteristics. A validated 3D model was developed, and the wavelet transform was used to perform time–frequency analysis of pressure signals. Phase space reconstruction and the Grassberger–Procaccia (G–P) algorithm were applied to evaluate chaotic behavior using the maximum Lyapunov exponent and correlation dimension. Results show that low frequencies dominate pressure fluctuations at the impeller inlet and guide vane outlet, while high-frequency components increase significantly at the intake bell mouth and outlet channel. The maximum Lyapunov exponent in the impeller and guide vane regions reaches 0.0078, indicating strong chaotic behavior, while negative values in the intake and outlet regions suggest weak or no chaos. This integrated method provides quantitative insights into the unsteady flow mechanisms, supporting improved stability and efficiency in ultra-low-head pumping systems.