The biological nutrient removal in anaerobic-anoxic-oxic (A/A/O) process in a full-scale wastewater treatment plant is simulated and later optimized by means of an integrated system model, which combines an extended model derived from activated sludge model No. 3 with support vector machine (SVM) and accelerating genetic algorithm (AGA). The data generated by the mechanistic model are used for establishing the relationship between the operating conditions and effluent water quality by SVM. The AGA approach is used to optimize the operating conditions by taking into account the effluent quality. Through the optimization of the integrated model, the volume of the anoxic tank could be reduced by about 11% and the internal recycle ratio could be decreased from 300% to 250-260%, while the effluent quality would still meet the discharging standards. The results demonstrate that the integration of the mechanistic model, SVM model and AGA approach is an effective strategy for the optimization of complex biological processes. (C) 2011 Elsevier B.V. All rights reserved.