Water Injection Optimization Using Historical Production and Seismic Data

Abstract

Abstract This work proposes a method to optimize waterflooding by using production and time-lapse seismic data. Using time-lapse seismic difference as an indication of inter-well influence, inter-well connectivities or influence functions are determined using a combinatorial algorithm (stochastic hillclimbing) by constraining the computation with time-lapse seismic. The water-cut is estimated using a fitting function. By perturbing the injection rate in the injectors, the oil production is optimized. The field in this study is mature with 15 years of production history and with 70% water-cut at the present time. Two legacy repeated seismic surveys, of in 1991 and 2001 vintages respectively, are reprocessed to enhance the repeatability from the original data. The reprocessing is done to eliminate the survey, source and noise differences. The connectivity or influence function is estimated using a window of balanced flooding. It is found that seismic constraining for the connectivity calculation increases the consistency of the inter-well relationship with the spatial data. In order to have an operational waterflooding scheme, a multiple realization approach is used. That is, the optimized injection is randomly perturbed under specified upper and lower bounds. The production rate is computed using the inter-well influence function. It is found that the optimal injection configuration can increase oil production. With a pilot implementation, it is shown that the new injection scheme can delay the oil production decline. Introduction Optimization of water-flooding is a critical process for reservoir management. Generally, the optimization can be conducted through an adjustment of the injection pattern or injection scheduling by reservoir simulation or by intuitive analysis. However, to perform this process, the injector-producer relationship or inter-well relation needs to be understood. To understand this relationship, the production history data are normally used either through a correlation or material balance-type analysis. For example Soeriawinata et al. used a correlation approach to build the injector-producer relationship.1 Jansen used a wavelet transform to analyze the inter-well relationship using producer-injector data.2 Albertoni et. al. proposed an approach of linear regression using injector-producer production history data.3,4 For quantitative analysis, reservoir simulation can be used with even seismic data in the process of history matching.5 However, reservoir simulation is quite time-consuming. On the other hand, using the production history sometimes has difficulties to fully account for the spatial heterogeneity. It would be critical to incorporate the production history data and other data such as seismic to analyze the producer-injector relationship. In this work, we propose using the time-lapse seismic attributes to constrain the estimation of the injector-producer relationship.