Gas Turbine Combustion Optimization Using Neural Network Model and Wavelet Analysis

Abstract

Abstract Rapid development in data science provides new methods for combustion tuning. This paper describes an artificial neural network (ANN) model that can accurately predict the key parameters in gas turbine combustion tuning and optimization, including NOx emission, combustor vibrational acceleration (ACC), and combustor dynamic pressure (DP). Wavelet denoising method was used in data preprocessing to improve the signal-to-noise ratio (SNR), which greatly improved the prediction accuracy of the neural network model. A combustion tuning simulation was then conducted to optimize NOx emissions using the acquired accurate mappings. By adjusting controllable parameters, optimization can be realized within necessary constraints. The effects of user-defined initialization parameters in the simulation were investigated for fast combustion tuning. An operating window was given considering the tradeoff between optimization results and computing time.