Steam generator (boiler) is main and very critical equipment in power plant and converts water into superheated steam at rated pressure and temperature as per load demand. The coal is pulverized in mills and fired through burners. Boiler operating performance depends upon numerous non controllable parameters like coal quality, burner imbalance, burner erosion etc. and controllable parameters burner tilt, primary air fuel ratio, secondary air fuel ratio, excess oxygen, furnace fouling, mill combination and loading, furnace to wind box DP etc.

Presently off line boiler performance optimization is done through excess oxygen, burner tilt, mills combination, mill loading furnace to wind box DP, mill PA flow, mill outlet temp, PA header pressure. So these control variables are changed with limited knowledge of their impact on boiler performance. It cannot be assessed accurately whether said changes led to run boiler at best optimized level. So in present condition it is not possible to establish most optimum boiler operating control variable set points, which would lead to optimization for the improvement of the thermal efficiency and environmental performance. So there is huge potential for reducing power generation cost which remains unutilized due to non availability of online steam generator operation optimization schemes in power plants.

In this thesis, an innovative hybrid AI methodology for on line excess air optimization is presented. Genetic algorithm is used for optimization. Since genetic algorithm requires a steam generator simulator to find optimal set points, BP based ANN has been used for boiler operation modeling. Target for ANN has been calculated using on line input data taken from 210 MW plant .The trained ANN has been used for operation modeling of steam generator and model performance has been validated by design performance curves. This trained ANN has been combined with genetic algorithm to find most optimal excess air set points which will maximize boiler thermal efficiency

This thesis also presents an application of NOx emission optimization and multi back propagation neural network has been proposed for GA+ANN based excess air optimization for the improvement of thermal efficiency and emission performance of boiler.