In past years, the industry focused on improved horizontal well placement, identifying the most favorable target window with the best petrophysical properties as well as the appropriate geomechanical properties to ensure the rock is conducive to hydraulic fracturing. Beyond this first and unique optimized landing point, having a set of pertinent low-cost data before fracking to estimate the geological variations along the laterals and evaluating systematically the well production performance via a PLT proxy opens the possibility of initiating a machine learning process to guide completion strategy choices. The outcome of the proposed learning model is an automated engineered completion design adapted to various facies along the lateral, resulting in prediction of stage-production performance.
A new type of log acquisition that mixes well testing and logging techniques has been developed in over the past years. When run before fracturing in an open hole, it highlights formation heterogeneities and helps selection of the best zones to be fracked. If run after fracturing in a cased hole, this new type of log acquisition can be considered as a PLT proxy for identifying the producing clusters and evaluating the fracking operation. Avoiding a black box type learning process, the results of the well test log are combined with all available data to sort various frackability facies. Productivity distribution, measured at a cluster scale determined by the PLT proxy, will then feed a data base of facies distribution versus fracs productivity paving the way for a stage optimization process.
Field results show that for the same design and proppant loading, well production could be reduced by a factor of 2 as the result of the non-adapted completion design for the lateral heterogeneity. As any conventional wells for which formation heterogeneity is considered before perforating, reservoir knowledge will bring added value to unconventional developments. Low-cost data acquisition opens the possibility of initiating a machine-learning process to guide completion strategy choices by better locating each cluster, adapting the cluster number by stage, and possibly applying different pumping schedules by geological type. This learning process could lead to an overall significant productivity improvement and contribute to the reduction of environmental impact by reducing fracking intensity.
Optimizing fracturing operations requires a systematic evaluation at the appropriate scale to be able to fulfil the potential of putting poorly producing stages back on track. Most of what has been done in the literature covers the improvement of fracking design at the well scale. The novelty of the proposed approach relies on new low-cost acquisitions at the cluster scale. Productivity distribution measured at the cluster scale feeds a database of facies distribution versus fracture productivity paving the way for a stage-optimization process. The outcome of the proposed learning model is a new, automated, engineered completion design that locates and group clusters by stages that are adapted to facies type along the formation, and includes a prediction of stage-production performance.
