Optimization on FNN based on genetic algorithm and its application on CCR soft sensor

The Continuous Catalyst Reforming Unit plays an important role in the refinery. In the regeneration tower, oxygen content which is very important to catalyst regeneration process is hard to measure. So T-S fuzzy model is used on the estimation of oxygen content. Based on NARMAX model identification of T-S fuzzy-neural-network (FNN), the genetic algorithm is applied to optimize the membership functions and network parameters, rule sets which can only be acquired by experiences. So that it can make better performance with the optimized parameters. The improved GA combines the advantages of GA´s strong search capacity and conventional optimization technologies´s fast convergence and accuracy merits. Therefore, the algorithm achieves a trade-off between accuracy, reliability and computing time in global optimization. Finally, validity and accuracy of the present model is verified by the on-field performance of oxygen content soft sensor modeling that was put into operation recently.