Energy Prediction of Thermal Power Plants by Using AI Based Approach Author MAMOON UR RASHID NUST201362502MCEME35513F MS-78

Abstract

The rate of decaying of the energy resources with the passage of time is increases and may result in reduction of panic reserves. Efficiency of all resources is crucial both in an environmental and economic sense. Improper use of energy results in waste generation. It has environmental impacts with regional, local and global implications. The key object is to adopt energy management in every field in order to reduce the wastage of energy sources and cost effectiveness without affecting productivity and growth. In a cement plant, nearly 30% heat is lost, primarily from the pre-heater and cooler waste gases. This thermal energy can be tapped by installing a Waste Heat Recovery Power Plant (WHRPP). Prediction of power plant output based on operational parameters is major focus now days. This research present four different types of machine algorithms. These algorithms include feed forwards neural network trained with particle swarm optimization (PSO), simulated annealing, hill Climbing and genetic algorithm. These algorithms takes steam inlet pressure (Mpa), steam inlet temperature (˚C) , steam flow (t/h) , exhaust temperature (˚C) as an input parameters to feed forward neural network to predict hourly output of the power plant. The mean square error (MSE) of Particle swarm optimization is better then the other three for the same number of iterations. The mean square error (MSE) of Particle swarm optimization (PSO) for training is 0.0074034 and 0.0097715 for testing data, Particle swarm optimization (PSO) shows promising results to predict power plant output.