Application of the adaptive neuro-fuzzy inference system (ANFIS) for simulating water fluid level control systems on horizontal separator

Abstract

Control system process is an important process that occurs in the branch of industrial world, one of which is in the realm of the oil and gas industry in production of the upstream process. One of the main instrument in the upstream oil and gas process is a separator which has the function of separating the fluid content of crude oil which flows through the pipe into several phases. In a three-phase separator, the separator will separate the heavy content of crude oil into three phases, namely the gas, water and oil phases before being distributed to the gathering station. In fact, almost all control processes separator instrument at PT. Pertamina EP still using the conventional PID control model which must be continuously monitored by human resources 24 hours per day. Sometimes also with a manual control system like this causes many factors in the calculation of daily logging data errors. Therefore, this research designed an intelligent system- based control method, which is a neuro-fuzzy control. This neuro-fuzzy control method is designed using Adaptive Neuro- Fuzzy Inference System (ANFIS) algorithm model with input in the form of setpoint, error, and error difference from the process of fluid separator variable, namely fluid level (h). The research was conducted using the Simulink / MATLAB application by entering the transfer function of the separator mathematical model and then making a comparison by looking at the response graph and parameters between the PID and ANFIS controller models. The results of this research conclude that the performance of the ANFIS model controller on average has a much better overshoot than the PID model because it is always close to zero in each set point condition and the ANFIS model has a better error value when the set point is 5 with a difference in error 0.712 instead of the error value of PID controller model.