Robust Simulation of CO2-Air Conditioning System Using SwarmIntelligent Based Ensemble Neural Network

In this paper, we present a new approach which combines adaptive particle swarm optimization(APSO) with ensemble techniques to simulation a dynamic and nonlinear model of CO2-airconditioning system. APSO algorithm is based on the original PSO algorithm which is a new,simple and robust optimization tool, has been used to train standard datasets for regressionpurposes. An ensemble neural network (ENN) exhibits higher generalization performancecompared to a single neural network. As error derivatives don't implemented in PSO method, allof objective functions can be used in this algorithm. A robust objective function is used in thiswork to overcome noisy condition and then a comparison with other common methods has beenperformed to show the excellency of APSO. First ENN is trained using optimization evolutionaryalgorithm, APSO, then in validation stage predicted output is compared with source data. Tocreate data set a physically based model is used and the well-trained ENN predict the effects ofthe two input parameters.

کلیدواژه ها:

CO(2) air condition ، ensemble neural network ، particle swarm optimization ، simulation