Research on Covariance Localization of Enkf Reservoir-Assisted History Fitting Method Based on Fast Marching Method

The Ensemble Kalman filter (EnKF) is a widely used intelligent algorithm in the field of automatic history fitting. The method has a number of drawbacks, such as inaccurate gradient calculation, filter divergence, and pseudo-correlation of parameters, leading to parameter correction errors and model inversion distortion in the process of historical fitting. A history fitting method based on the fast marching method and covariance-localized Ensemble Kalman filter (FMM-CLEnKF) is established to reduce pseudo-correlation in the calculation process of the traditional distance truncation method. According to the static parameter field information of the reservoir geological model combined with the state equation, the fast marching method (FMM) is used to quickly track the propagation time of the pressure wave in every well, determine the sensitive area of a single well, and construct the localization matrix. Combined with the covariance localization Ensemble Kalman filter method, the gradient correction of the data assimilation method is realized, and the pseudo-correlation of parameters is reduced. Finally, the optimal model is improved by gradually fitting and updating the reservoir parameter model. The calculation results of a field example show that the FMM-CLEnKF method has a higher reservoir parameter inversion accuracy, data fitting speed, and production data fitting accuracy than the ensemble Kalman filter method.