Solution: We developed a soft sensor based on an Artificial Neural Network (ANN) model as a predictor for fuel gas composition. The inputs of the model were automatically selected and the proposed technique was tested on a 29.114 MW gas turbine engine.
Problem: Gas turbine engines are commonly used for power generation in industries. The quality and composition of fuel gas have a tremendous impact on operational performance of the engine. Accurate determination of fuel gas composition allows for optimal adjustments of the air to fuel ratio, enabling the combustion turbine to operate at its most efficient point. The current technique to determine the gas composition of the natural gas is gas chromatography (GC) which can provide the gas composition information every 4 min at best. However, the efficiency of combustion control would be greatly improved if such determination time could be reduced.
Benefit: The soft sensor can reduce the determination time from 4 min to 1 min, while maintaining the capability of correctly predicting the gas composition.