You are on page 1of 3

RSS SIMULATOR TRAJECTORY CONTROL BY REINFORCEMENT LEARNING

INTRODUCTION:
o Creating a digital twin of drilling by using RSS Simulator established in UIS.
o Optimizing path control by using Reinforcement Learning with varying drilling parameters.

OBJECTIVES:
o Seek best drillable path closest to Planned Well Path.
o Optimizing WOB and RPM values through hundreds of episodes.
o Achieve highest ROP while controlling path.

AIM:
o Prior to drilling finding best parameters for proposed surveying. Help to make efficient, fast
and cost-effective drilling by the aid of AI support.
Affect of WOB and RPM to drilling path Paths
ROP_a = 0.158 m/s ROP_b = 0.391 m/s ROP_c = 0.59 m/s

Figure 1: Comparison of planned path and simulated path by RSS Simulator with different parameters. (a) Simulated
path with WOB = 80 kN, RPM= 100. (b) Simulated path with WOB = 200 kN, RPM= 100. (c) Simulated path with
WOB = 200 kN, RPM= 150.
Survey Data(V,N,E) Reward Inputs:
- Toolface Real
Planned Well Path Calculator - Toolface drill
- ROP
PWP matrix (V, N, E, MD, Inc, - Radius of Ellipse
Azi, DLS) - DLS

Reinforcement RSS function Reward


Matlab

Python

Python
Learning Function
Toolbox Calculates
• Drilling Observations:
matrix(i-1) drilling_matrix, - TVD
Returns - North
Loop i=1 till • PWP(i) - East
• WOB observation as Step wise
len(PWP)-1 corresponding reward
• RPM
to PWP(i)

• Drilling
matrix(i)

You might also like