Spread Rate Forecasting in Well Cost Estimation – A Study of Methods and Applications

Abstract

Abstract In the realm of well cost estimation, the accurate forecasting of spread rates is pivotal, given the substantial financial implications of erroneous assumptions. This paper, "Spread Rate Forecasting in Well Cost Estimation – A Study of Methods and Applications," delves into the uncertainty inherent. Through a thorough examination of predictive methodologies, the research harnesses both econometric and machine learning models, which are commonly utilized in forecasting crude oil prices. The study formulates models based on publicly available data, such as ‘West Texas Intermediate’ (WTI) and the ‘Baker Hughes Rig Count’, to predict the Spread Rate. The empirical results underscore the efficacy of the proposed models, with the predicted spread rates closely mirroring actual figures. Notably, the models’ precision wanes when extending the forecast horizon beyond a year, a limitation accentuated by the unforeseen WTI and Spread Cost fluctuations during the COVID-19 pandemic. A comparative analysis shows the superiority of RNN, LSTM, Bayesian, and OLS models over the ARIMA model, evidenced by lower Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) metrics. The paper advocates for a probabilistic approach to navigate the uncertainties prevalent in long-term forecasting endeavors.