Multimodel Predictive Control for Wastewater Treatment Process with Multirate Sampling Intervals

Influenced by the biochemical reactions and sensor characteristics, the wastewater treatment process (WWTP) has been regarded as a multirate sampling system. To deal with the asynchronous control problem caused by multirate sampling, a multimodel predictive control (MMPC) scheme is proposed to improve the operation performance of WWTPs. First, a multimodel predictive control structure is constructed such that the controlled variables with multirate sampling intervals of WWTPs are able to be tracked synchronously at each sampling instant. Second, the fuzzy neural networks (FNNs) are adopted as the carrier for multimodel to compute the synchronous prediction outputs of all controlled variables at each sampling instant. Moreover, an adaptive cost function is excogitated to obtain the optimal control action in which the adaptive parameter is to minimize the impact of forecast errors on the tracking effect. Ultimately, the experiment results on the benchmark simulation model 1 (BSM1) have verified the efficiency of the proposed MMPC.