Machine Learning-Based Formation Evaluation in Horizontal Wells of Complex Carbonate Reservoirs Using Drilling Cuttings Data

Summary Depleted reservoirs lacking aquifer support or artificial pressure maintenance pose significant drilling challenges, such as differential sticking and the need for smaller bit sizes in horizontal sections. Conventional geosteering and formation evaluation LWD tools are not available for slim holes (e.g., 3-7/8” bit size), requiring alternative methods to obtain critical reservoir data for evaluation, completion, and stimulation planning. This study explores a machine learning approach to achieve comprehensive formation evaluation in horizontal wells using cuttings XRD and LWD resistivity data. While log prediction workflows based on nearby wells work for vertical or deviated wells, they are less effective in horizontal wells. To address this, a machine learning model was trained in the landing section using cuttings XRD and the predicted logs from nearby wells, then applied to predict porosity, permeability, and lithology in the horizontal section where only LWD resistivity and XRD data were available. Predictions were later validated against cased-hole interpretation, which showed strong agreement.