Using Xinxue River Constructed Wetland as the study object, the wetland prediction models based on BP neural network were established through the seasonal division of the wetland, and the maximum influent water load was determined on the constraint condition that effluent water quality achieved class Ⅲ of surface water quality. Then nonlinear functions of water quality and quantity scheduling were constructed by Origin software. The optimal influent load was determined adopting prediction results of the models as constraint conditions of the functions. Thus the water quality and quantity scheduling scheme of the wetland was established. The results show that optimal influent load for Feb. ~ May: the influent water quantity is no more than 8560m3/d, CODCr is 25.47mg/l~26.37mg/l, ammonia nitrogen 0.11mg/l~1.0mg/l, TN 10.28mg/l~10.51mg/l, TP 0.16mg/l; for Jun. ~ Sept.: the water quantity is no more than 31750m3/d, CODCr is 26mg/l~32.36mg/l, or 37.15mg/l~45.37mg/l, ammonia nitrogen 0.48 mg/l~1.78mg/l, TN 5.15mg/l~6.18mg/l, TP 0.07mg/l~0.09mg/l; for Oct. ~ Dec.: the water quantity is no more than 11070m3/d, CODCr is 24.55mg/l~26.91mg/l, ammonia nitrogen no more than 0.75, TN no more than 8.61 mg/l, TP 0.10mg/l~0.12mg/l, or 0.16mg/l~0.17mg/l.