Probabilistic Well Time Estimation Revisited: Five Years On

Abstract

Abstract In 2010, Nexen UK published a paper on well time estimation (Adams et al. 2010), which for the first time allowed accurate calculation of probabilistic duration for non-critical wells. The present paper extends the method and data coverage to high step-out and HPHT wells. The method has now been in use at Nexen for five years. Actual and predicted durations are given for each year's drilling, showing that the accuracy is historically within 2-3% for stable train-wreck rates. The historical well database upon which the statistics are based now stands at 190 wells, 72 more than the previous paper. To the authors' knowledge, it remains the only large-sample timings database published in the open literature. To allow others to use the method, the updated activity timings, mechanical NPT and waiting on weather data for semi-sub drilled wells are given in full. It is shown that the commonly used distributed train-wrecks model has too high a sampling uncertainty for accurate time estimation. The lumped train-wrecks approach presented here does not suffer from this limitation, and is therefore the only method that delivers the required predictive accuracy for practical dataset sizes.