Research and Analysis of Oil-Water Interface Algorithms in Crude Oil Storage Tanks

This study focuses on data processing of segmented capacitance sensors and multiphase interface algorithms for oil, gas and water phase boundary detection in crude oil storage tanks. In practical operating environments, sensor signals are susceptible to interference from turbulent flow disturbances inside the tank and heterogeneous distribution of multiphase media, resulting in significant noise contamination and pseudo-signal aliasing in detection data. To enhance the accuracy of interface characterization, this research systematically compares multiple data preprocessing algorithms based on sensor noise mechanism analysis, including peak-valley filtering, median shielding, exponentially weighted moving average (EWMA), Kalman smoothing, and wavelet-Savitzky-Golay hybrid denoising. A denoising framework comprising five differentiated processing schemes is established, with emphasis on systematic comparative analysis of critical metrics such as signal smoothness, feature preservation, and accuracy. The optimal preprocessing strategy is selected through robustness evaluation, laying a high signal-to-noise-ratio mathematical foundation for dynamic multiphase interface identification. Finally, a hybrid algorithm integrating K-means clustering with BP neural network is implemented for oil-gas-water multiphase interface recognition.