APPLICATION OF NEUROFUZZY SPEED AND LOAD CONTROL FOR GAS TURBINE POWER UNITS

摘要

A neurofuzzy PI controller applied to a Gas Turbine Power Unit (GT) for speed and load control is presented. The capacity for empirical knowledge acquisition from artificial intelligence systems was utilized in the development of the strategy. The PI is a neurofuzzy system obtained from process data. The Gas Turbine GE5001 type is the selected nonlinear process, for speed control during startup operation, where the GT has to follow a specific speed path that imposes tight regulation requirements for the control system, including fast response and precision. This controller is calculated and tuned from normal operation data, automatically modifying the input-output mappings of the neurofuzzy system. Simulation tests were carried out with a mathematical model of a GE-5001 Gas Turbine Power Unit. Once validated the control strategy, startup and load tests were made in the Laguna-Chavez unit 2 Gas Turbine Power Unit from Comision Federal de Electricidad (CFE), the Mexican electric utility company. The neurofuzzy control system was inserted in a modular control system for Gas Turbine Power Units developed in the Control and Instrumentation Department of the Instituto de Investigaciones Electricas (IIE) with close client's collaboration CFE. The main functions were implemented according to the state-of-the-art IEC 1131-3 programming standard. Comparisons between conventional and neurofuzzy Proportional-Integral (PI) controller were made for startup and load phases, using the Integral of Absolute Error (IAE) performance index and fuel consumption values. The analysis of these two factors shows better results for the neurofuzzy PI controller; in general, it exhibited improvements to reference changes and operation disturbances at any single point of operation in the startup phase and generation phase as well.