An Intelligent System for Process Plant and Well Production Control with Problem Diagnosis

Abstract

Abstract In an oilfield production, the operation of a process plant is very sensitive to the disturbance such as variations in the flow rate and the pressure from the wells. Moreover, variations in the parameters of the oil production could represent abnormalities in the well production. Identification of problems in operation of process plant and, especially oil well production operation demand mainly symbolic reasoning. In gas lift well, gas surface card provides important information but not all data required for decision making then, symbolic reasoning is needed. Many kinds of problems can occur in a production process plant and well operation. In the present paper, neural nets are used to develop an intelligent system whose purpose is to support intelligent control of well production and process operation. Laboratory and field test results are presented. The main purpose of the intelligent system is to help operators in making the correct decision in contingency situations, in an effort to avoid problems and production losses. Introduction In a well production system which includes the process plant, the oil/gas well itself with production devices, disturbances that produce oil and gas load disconnection or other emergency situations have to be localized as quick as possible. To localize a disturbance quickly and exactly is extremely important to start the process plant reconfiguration to restore normal oil and gas production. Moreover, variations of oil and gas production parameters normally indicate abnormality situations in the well production system. However, the identification of the faulted points in that system is not always an easy task which frequently delays the start of correction procedures. It usually occurs when the protection system does not behave as expected. The petroleum process plants in commissioning phase or even ones already in operation with complex constructive and operational natures, they have high incidence of failures in their protection system. On those plants, disturbance localization can take a long time due to a great amount of information to be analyzed, even visual inspection can be required. The difficulty and the time necessary in order to identify faulted points significantly increases in non-conventional process plants and wells. The present paper describes the main characteristics of algorithms here proposed to identify faults and to make diagnosis in a process plant and well production (Patricio, 1997). Symbolic neural nets are used for fault diagnosis and control the well production. Finally, results collected from the field at different operational conditions are discussed.