Differential evolution-based nonlinear system modeling using a bilinear series model

This paper presents a new modeling method for nonlinear dynamic systems based on using bilinear series model. Basically, bilinear model is an extension of infinite impulse response (IIR) filter and belongs to the recursive nonlinear system model, i.e., its past output signals will heavily affect the present output. This kind of model can efficiently approximate a large class of nonlinear systems with fewer parameters than other non-recursive models. To adjust the model kernels, we here adopt an evolutionary computation called the differential evolution (DE) algorithm. This algorithm is based on real-valued manipulations and has a good convergence property for finding the global solution or the near global solution of optimized problem. Design steps of DE-based nonlinear system modeling are clearly given in this study. Finally, two kinds of digital systems are illustrated to demonstrate the efficiency of the proposed method.

.Figure optionsDownload as PowerPoint slideHighlights
► This paper presents a new modeling method for nonlinear dynamic systems based on using bilinear series model.
► To adjust the model kernels, we here adopt an evolutionary computation called the differential evolution (DE) algorithm.
► This algorithm is based on real-valued manipulations and has a good convergence property for finding the global or near global solution of optimized problem.
► Two kinds of digital systems are illustrated to demonstrate the efficiency of the proposed method.