Performance Prediction of A Power Generation Gas Turbine Using An Optimized Artificial Neural Network Model

Abstract

This paper introduces an innovative application of
an Artificial Neural Network (ANN) based model for the
performance prediction of a power generation gas turbine. this
approach optimizes the ANN model by utilizing a comprehensive
database to compare various ANN topologies. Based on
optimization results, a two-layer Multi-Layer Perceptron (MLP)
was constructed and used as the best-optimized topology for such
applications. The training dataset comprises historical
operational data from a Rolls-Royce (RB21-24G) gas turbine
unit. Notably, this model shows substantial accuracy for different
ambient conditions and variable power ratings. Furthermore, a
sensitivity analysis using various methods was introduced to
study the impact of each input on the model outputs. To validate
the model's reliability and novelty, we introduce a degradation
study, comparing one-year-later on-site operational data with
predicted values generated by the ANN model. Remarkably, the
results demonstrate strong consistency between measured data
and model predictions