Ensemble Kalman filter (EnKF) is one of the most widely used methods in automatic history matching, but the algorithm will produce problems such as pseudo-correlation and filter divergence during application. A new localized ensemble Kalman filter method of single-well sensitivity region based on connectivity analysis for automatic history matching was established, which solved the problem of mismatch between the traditional distance truncation method and the actual reservoir when dealing with pseudo-correlation. In this method, the reservoir grid was regarded as a connected directed map, and the shortest path between any two grid points was calculated by connectivity analysis and the Floyd algorithm to determine the sensitive area of a single well and the correlation coefficient matrix from the well point to each grid point, and then it combined with the ensemble Kalman filter method, in the end the pseudo-correlation problem was effectively weakened. The improved algorithm is programmed and verified by examples, and the results showes that the EnKF method based on localization of connectivity analysis is superior to the standard EnKF method in terms of production dynamic matching and model parameter field inversion.