Hybrid Models in Dynamic Simulation of a Biological Water Treatment Process

In biological wastewater treatment in pulp and paper industry. A lot of process measurements are available, but measurement sets do not include sufficient information on special features of the influent nor on microbial composition of the sludge. Populations of microorganisms are highly important. Basic dynamic simulation is based on LE models. The models consist of two parts: interactions are handled with linear equations, and nonlinearities are taken into account by membership definitions. Process insight is maintained, while data-driven tuning relates the measurements to the operating areas. Genetic algorithms are well suited for linguistic equation (LE) models based on nonlinear scaling and linear interactions. Activated sludge models provide a basis for phenomenological modelling and can be linked to process expertise. Hybrid models with a cascade approach are needed in biological wastewater treatment to cover different operating conditions.