ARCHIVES
VOL. 10, ISSUE 1 (2026)
Quantifying the optimal look-back window for rate of penetration (ROP) prediction in drilling operations
Authors
Ichenwo John Lander, Marvellous Amos Dornubari
Abstract
In real time, prediction of Rate of
Penetration (ROP) has gained great importance in optimisation of drilling,
automation, and edge-based intelligent systems. Although ROP prediction using
machine learning methods has shown promising outcomes, the choice of temporal
look-back windows, the period of history used as input to the model, is still
mostly arbitrary across the literature. This paper involves systematic
experimental framework to measure the best temporal horizon at which ROP
prediction can be used based on high-frequency drilling data of a series of
wells. We compared the performance of prediction in terms of time window, i.e.
5 seconds to 10 minutes, between various model structures, i.e. XGBoost, Random
Forest, and LSTM networks. We observed that the prediction accuracy diminishes
after 30-60 seconds, and there is no statistical improvement in forecasting by
considering more historical contexts (p > 0.05).This observation implies
that in the steady-state conditions of drilling, the dynamics of ROP are
near-Markovian in the sense that the current state variables are adequate to
characterise the dynamics of the processes. The computational efficiency of the
best short window enables deployment at very low latency on drilling rigs with
up to 90% lower memory demand compared to 5-minute windows, while maintaining
the same predictive accuracy (RMSE of the order of 2.3 ft/hr).The implications
of these findings on the oil and gas industry are important to real-time
drilling automation system and edge computing application.
Download
Pages:47-53
How to cite this article:
Ichenwo John Lander, Marvellous Amos Dornubari "Quantifying the optimal look-back window for rate of penetration (ROP) prediction in drilling operations". International Journal of Advanced Engineering and Technology, Vol 10, Issue 1, 2026, Pages 47-53
Download Author Certificate
Please enter the email address corresponding to this article submission to download your certificate.