关键词: 最小二乘支持向量机, 在线系统辨识, 时变非线性系统, 多输入多输出系统, Least Squares Support Vector Machines （LSSVM）, online system identification, time-varying nonlinear system, Multi-Input Multi-Output (MIMO) system

Abstract: To tackle the difficulty in identifying timevarying nonlinear MultiInput MultiOutput (MIMO) system online, a fast online system identification approach based on Least Squares Support Vector Machine (LSSVM) was proposed. The existing LSSVM incremental and online learning algorithms were introduced, and some speeding up implementing tactics were designed and adopted in the algorithm; consequently, a fast online LSSVM learning algorithm was obtained. MIMO system was decomposed into multiple MultiInput SingleOutput (MISO) subsystems, and each MISO was modeled online via a LSSVM. The numerical simulation shows the modeling method is faster and the obtained models provide accurate prediction.