Improving the efficiencies of simultaneous organic substance and nitrogen removal in a multi-stage loop membrane bioreactor-based PWWTP using an on-line Knowledge-Based Expert System.

The results of the use of an expert system (ES) to control a novel multi-stage loop membrane bioreactor (MLMBR) for the simultaneous removal of organic substances and nutrients are reported. The study was conducted at a bench-scale plant for the purpose of meeting new discharge standards (GB21904-2008) for the treatment of chemical synthesis-based pharmaceutical wastewater (1200-9600 mg/L COD, 500-2500 mg/L BOD5, 50-200 mg/L NH4+-N and 105-400 mg/L TN in the influent water) by developing a distributed control system. The system allows various expert operational approaches to be deployed with the goal of minimizing organic substances and nitrogen levels in the outlet while using the minimum amount of energy. The proposed distributed control system, which is supervised by a Knowledge-Based Expert System (KBES) constructed with G2 (a tool for expert system development) and a back propagation BP artificial neural network, permits the on-line implementation of every operating strategy of the experimental system. A support vector machine (SVM) is applied to achieve pattern recognition. A set of experiments involving variable sludge retention time (SRT), hydraulic retention time (HRT) and dissolved oxygen (DO) was carried out. Using the proposed system, the amounts of COD, TN and NH4+-N in the effluent decreased by 55%, 62% and 38%, respectively, compared to the usual operating conditions. These improvements were achieved with little energy cost because the performance of the treatment plant was optimized using operating rules implemented in real time.