Abstract
A drilling system consists of a rotating drill string, which is placed into the well. The drill fluid is pumped through the drill string and exits through the choke valve. During drilling, the pore pressure (minimum limit) and the fracture pressure (maximum limit) define mud density range and pressure operational window. Several disturbances affect bottom hole pressure; for example, as the length of the well increases, the bottom hole pressure varies for growing hydrostatic pressure levels. In addition, the pipe connection procedure causes severe fluctuations in well fluids flow, changing well pressure. Permeability and porous reservoir pressure governs native reservoir fluid well influx, affecting flow patterns inside the well and well pressure. The objective being tracked is operating under desired pressure levels, which assures process safety, also reducing costs. In this scenario, modelling techniques are important tools for narrow operational windows, commonly observed at deepwater and pre-salt layer environments. The major objective of this paper is real time building and comparing model performance for predicting annulus bottom hole pressure, using real time flow, choke index and ROP (rate of penetration) data, available from a drilling site. Neural Network (NN) based models were used successfully for on-line identification purposes, using an adaptive methodology. The proposed methodology can be employed at drilling sites, through the use of PWD (Pressure While Drilling) and mud-logging tools, providing real time data and helping operators to make important decisions concerning safety of the drilling process.
