Back propagation neural network (BPNN) prediction model and control strategies of methanogen phase reactor treating traditional Chinese medicine wastewater (TCMW)

Based on the prototype experiment of treating herb wastewater by Up flow Anaerobic Sludge Bed and Anaerobic Filter reactor (UASBAF), an artificial neural network (ANN) model which adopts a back propagation algorithm with momentum and adaptive learning rate was established. And the effect of each parameter to the performance of the reactor was compared, using the method of partitioning connection weights (PCW). The result is pH values > influent of chemical oxygen demand (COD) > hydraulic retention time (HRT) > alkalinity. In addition, many strategies were proposed to optimize the working condition of the system. Adding some alkali to increase pH value when raising influent COD, was an effective way to avoid negative effect to system; low influent COD had a negative impact on the performance of the reactor; pH was suggested to be controlled more than 7.5 when the influent COD was increased over 6000 mg L−1. The best influent COD concentration was 6000–8000 mg L−1 when the conditions were that pH was 7.5, alkalinity was 2000 mg L−1 and HRT was 35–50 h; HRT was suggested to be controlled more than 50 h to maintain good performance of the reactor with high influent COD (8000–10,000 mg L−1). These strategies provided an effective way of controlling UASBAF simply.