An adaptive learning method with dynamic error transfer factor for batch processes modeling

Many batch processes can be considered as a class of control affine nonlinear systems. In this paper, a novel adaptive learning approach for batch process modeling is developed. By introducing dynamic error transfer factor associated with mean squared error and using extended recursive least squares approach, the proposed approach can offer an effective fuzzy T-S predication model, resolve the conflicting problem of convergence speed and osciallation existed in recursive least squares method. The proposed modeling scheme is illustrated on a semi-batch reactor, and simulation results show its effectiveness and accuracy.