SMART CANDIDATE SELECTION FOR HYDRAULIC FRACTURING USING KNN (K-NEAREST NEIGHBOR) METHOD

• Main Author: Fara Daniella, Khobita
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/40062
• Institution: Institut Teknologi Bandung

With an increasing demand for hydrocarbon, it is essential to improve hydrocarbon production. A lot of efforts has been done, including hydraulic fracturing. Hydraulic fracturing is a process of creating a conductive path from a formation to the wellbore that stimulates the flow of oil or gas, increasing the volume that can be recovered. However, a large number of hydraulic fracturing operations resulted in underperforming wells even with improved techniques. In order to efficiently alleviate the risk of a hydraulic fracturing treatment, choosing the best candidate well selection is vital as the success rate of hydraulic fracturing greatly relies upon candidate well selection. Studies have demonstrated that poor candidate selection is more likely to yield even worse outcomes than randomly selected. As asserted in the literature, although a common practice, candidate well selection is not a straightforward process and until now, there has not been a systemic approach to address this candidate well selection process. This study discusses the execution of the hydraulic fracturing candidate selection process. A customized candidate selection process is developed to identify candidates from a pool of 353 wells in two phases for hydraulic fracturing treatment. The author summarizes the criteria which must be evaluated during the candidate well screening. Wells are then ranked according to incremental hydrocarbon production pre- and post-hydraulic fracturing treatment. In the absence of an exact initial production rate of the wells pre-hydraulic fracturing treatment, due to limited data available, a practical alternative is introduced, K-Nearest Neighbor, to predict initial production rate. K-Nearest Neighbor algorithm predicts a value, in this case, production rate, by calculating the distance between each data point and parameters of interest and searching for K-nearest neighbors based only on interpreted log data. The KNN model presented in this study is proven to give a reliable prediction of initial production rate when there is little or no prior knowledge compared to the deterministic model. Post fracture initial production rate is determined by using wells Fold of Increase approach. Wells FOI represents the incremental dimensionless productivity index after hydralic fracturing treatment. The novelty of this paper resulted in an easier and faster candidate selection when there is limited data and no simulation model available. As treatment is expensive, it was critical to identify the most suitable candidates with the available dataset. This is addressed through a systemic workflow to perform production performance analysis before and after hydraulic fracturing treatment. For further implementation, this paper can be used in the selection procedure of oil and gas wells for hydraulic fracturing treatment of identifying a proper candidate for this job.