# Research on an intelligent drilling parameter optimization method using sliding window segmentation based on the hydraulic-mechanical specific energy model

## TL;DR

This paper introduces an intelligent drilling parameter optimization method that improves drilling efficiency using a novel time-series data approach.

## Contribution

The study introduces a hybrid MSE model with sliding window segmentation and a novel data fusion framework for intelligent drilling optimization.

## Key findings

- The proposed method achieved prediction accuracies of 70% and 90% using ROP and BP neural network.
- Weight on bit, rotary speed, and flow rate were identified as dominant factors influencing mechanical specific energy.
- The simulated annealing algorithm improved drilling efficiency by an average of 43.34%.

## Abstract

Optimizing drilling parameters is essential for improving drilling efficiency and reducing operational costs in oil and gas engineering. This study presents an intelligent optimization approach for drilling parameters based on a hydraulic–mechanical specific energy (MSE) model. A time-series data fusion framework integrating Savitzky–Golay filtering, random forest, and hybrid anomaly detection was established to incorporate hydraulic parameters into the MSE model. The model parameters were further refined by coupling the rate of penetration (ROP) equation with a backpropagation (BP) neural network, achieving prediction accuracies of 70% and 90%, respectively. Field validation using 7,231 datasets from four wells revealed that weight on bit, rotary speed, and flow rate are the dominant factors influencing mechanical specific energy. Moreover, the simulated annealing algorithm was employed to globally optimize key parameters, resulting in an average improvement of 43. 34% in drilling efficiency. Compared with conventional MSE-based approaches, the proposed method innovatively integrates sliding window segmentation with the hybrid MSE (HMSE) technique, significantly enhancing time-series data processing. The developed multi-objective optimization model demonstrates superior prediction accuracy and adaptability under field conditions, providing a practical and effective tool for intelligent drilling parameter optimization.