Abstract:

A novel cultural differential evolution algorithm (CDEA) was proposed. In CDEA, differential evolution algorithm (DEA) was applied as population space of cultural algorithm (CA). The designs of belief space and influence functions adopted a variety of knowledge sources which were applied to supervising variation and mutation operation of DEA. The expression of knowledge and the role of knowledge supervising the evolution of population were strengthened. The test results of function showed that CDEA produced more competitive performance of search at lower parameter sensitivity than DEA. The compensatory fuzzy neural network (CFNN) based on CDEA was proposed and applied in soft-sensing model of a rectifying column system in ethylene production. Through analysis of simulation and generalization capability, the results showed that the proposed network was superior to the conventional FNN，CFNN and the compensatory fuzzy neural network based on genetic algorithm(GACFNN) in modeling precision and convergence.