A data-based KPI prediction approach for wastewater treatment processes

In this paper, the Benchmark Simulation Model No. 1, which is designed for the purpose of simulating actual wastewater treatment processes, is introduced and implemented in SIMULINK environment. Then the partial least squares (PLS) model and its kernel version is studied, and wavelet transform is used to carry out the so called multi-scale kernel partial least squares (KPLS). By means of multi-scale KPLS, the prediction of key performance indicator (KPI)-the COD concentration in effluent-is implemented. Simulation results show that this prediction model has strong generalization ability under the condition that the data collected during the wastewater treatment processes are distributed unevenly and coupled tightly.