Identificação e controle em linha de processo de perfuração de poços de petróleo utilizando redes neuronais

Abstract

A well that presents a good oil production rate is directly related to the drilling stage being the control of the pressure is the most important step. Thus the major objective of this work is the implementation of neural network-based controllers to regulate the annulus bottomhole pressure (BHP) during drilling, in the event of disturbances such as gas kick, circulation and pipe connection procedure. Such disturbances, in addition to causing pressure fluctuations that can damage the well, can lead to environmental, financial and human resource damage in the most extreme cases. In this study, the choke valve was used as the manipulated variable. For the purpose of identification and control in real time, a feedforward neuronal network was used with a layer of hidden neurons, presenting as input signals: annular pressure, choke pressure, water pump and mud frequency, choke opening , annular flow, time and set point, and a neuron in the output layer. Neural controllers are attractive because they have the ability to deal with nonlinear and inherently transient systems, as is the case with of oil well drilling process. Neuronal controllers were compared with classical PI controller (Ziegler Nichols (1942) and Cohen-Coon (1953)). It is noteworthy that simulation studies and experiments in drilling unit were carried out. The developed closed loop scheme showed to be eficient to regulate annulus bottomhole pressure