An entropy-based online multi-model identification algorithm and generalized predictive control

摘要

Based on the minimum entropy and fuzzy subtractive clustering method, a new specialized algorithm for online multi-model identification is proposed in this paper. Different from the traditional identification model, the structure and parameters of the established model can be recursively updated when new data coming to the system, which makes it a wise choice for online modeling and complex processes control. The entropy-based online fuzzy subtractive clustering method is used to determine the number of the local models and their corresponding memberships. A controlled auto-regressive integrated moving average expression is adopted as the form of linear subsystems, for it not only match the identification process, but also can be used to design the control system easily. The parameters of local models are calculated by weighted recursive least square method, and the nondimensional error index is used to evaluate the performance of the identified model. By applying generalized predictive control strategy to the established model, a fuzzy generalized predictive control system is constructed, and the control law is given in the paper. Finally, a case of the method to "Mackey-Glass difference time delay equation" is studied. The simulation results illustrate the viability and the robustness of the strategy.