Fuzzy Model-Based Predictive Control of a Heat Exchanger

Abstract

Abstract In this contribution, fuzzy model-based predictive control is applied to temperature control of a cross-flow water/air heat exchanger. After a review of Takagi-Sugeno type fuzzy systems and their identification from measurement data the on-line adaptation of these models is discussed. If the nonlinearity of the process is assumed to keep its structure, the linear parameters in the rule consequents can be locally updated by a recursive least-squares algorithm. Since this algorithm is computationally inexpensive it can be utilized in nonlinear model predictive control (NMPC). If the process is influenced by unmeasurable or unmodeled disturbances, the fuzzy model is adapted to these quantities. It can be distinguished between disturbances and model deficiencies which require the adaptation of all linear parameters in the rule consequents and model offsets in the static mapping which can be canceled by local adaptation of one offset parameter per rule. The second case is particularly simple to implement because there is no need for persistent excitation. Moreover, this mode of operation copes with the challenges in control which are specific to the heat exchanger. The effectiveness of the adaptive nonlinear model predictive controller is proven by application to an industrial-scale pilot plant.