Advanced State Estimations for Gravitational Oil/Water Separator Tanks using a Kalman Filter and Multi-Model Hypothesis Testing

This paper presents a new application of the Kalman filter with Hypothesis testing for a fast and robust model-based estimator for measuring level interfaces of atmospheric gravitational oil-water separator tanks. A newly developed semi-empirical linearized model is applied in the estimator algorithm. A multi-model hypothesis-testing algorithm for covering more scenarios was deployed. The proposed method provides a cost-effective and straightforward solution for estimating all state variables in an oil-water separator. Our evaluation results demonstrate that the proposed algorithm achieves high accuracy with an observation error of less than 2% and a false alarm rate of 3.3% under 50-70% working conditions. Furthermore, the estimator can effectively handle process noise with a 10% feed offset. The proposed platform requires only a few installed sensors yet can accurately estimate unknown parameters. The proposed approach offers a robust and practical soft sensor solution for gravitational oil/water separators