Hybrid LE Systems for Simulation of an Activated Sludge Process

Industrial wastewater treatment plants (WWTP), especially in pulp and paper mills, have operational problems resulting fluctuations in effluent discharges. The composition and flow rates of industrial wastewater fluctuate depending on the production schedules of the upstream processes. A lot of process measurements are available, but measurement sets do not include sufficient information on the special features of the influent nor on the microbial composition of the sludge. The populations of microorganisms have an essential effect on the operation of the treatment process. Basic dynamic simulation is done with linguistic equation (LE) models, where interactions are handled with linear equations, and nonlinearities are taken into account by scaling functions. Intelligent indices provide useful information for the detection of process conditions and the system, which consists of three interactive models: load, biomass population and treatment. Process insight is maintained while data-driven tuning relates the measurements to the operating areas. A wide range of operating conditions is handled with a multimodel system and a fuzzy decision module, which defines the weights of the subsystems. Hybrid models with a cascade approach are needed in biological wastewater treatment to cover different operating conditions.