Publications

As one of the mature fields in Indonesia, Offshore Mahakam’s executives are always exploring for new ideas to be developed which could optimize maintaining their production targets. The benefits of this mature field are to have a multitude of existing wells which are no longer producing and have already been plugged and abandoned. The existing well slots are being re-entered and utilized to access new subsurface targets; this has become common practice in Mahakam which has brought significant cost and material savings. Offshore Mahakam has drilled 32 re-entry wells in the last two years and yet it is still unable to formulate the milling roadmap that could produce a consistent milling duration. For a dual casing exit, the fastest is four hours while the longest is 16 hours. Differences between a field engineer’s methods and the response in applying parameters during milling play a significant role.

This paper describes the creation of a comprehensively supervised machine learning model from the drilling parameters of the five fastest wells (dataset input) under seven hours of milling time, after which the model created is applied to the blind dataset consisting of the Alpha-13.G1 well (short milling time), the Alpha-102.G1 well (average milling time) and the Alpha-7.G1 well (long milling time). Parameters such as WOB, RPM, Torque, Flowrate and ROP are optimized by the model and playback matched with the actual response of the field engineer is analyzed to determine the deviation points of each parameter, where the field engineer on site has difficulty in recognizing. With optimized ROP, milling time is re-calculated and shows significant improvement. Artificial intelligence (AI) application through simple machine learning modelling has uncovered the unseen lost time and the ideal parameters that should be applied to obtain a positive result. A milling roadmap of these wells will be a reference for further milling casing exit operations in Offshore Mahak