Non-point source (NPS) pollution has become a primary threat to the water environment. With the implementation of numerous control projects, striking a balance between economic costs and efficiency has become an urgent issue for decision-makers. To identify the optimization solutions for best management practices (BMPs) at the watershed scale, a genetic algorithm and the Soil and Water Assessment Tool (SWAT) model were combined to evaluate the trade-offs of NPS control strategies. The BMPs considered in this model included soil testing and fertilizer application technology, integrated pest management projects, conservation tillage, lakeside belt ecological construction, forest construction, and vegetation filter strips. The decision variables are the type and size of the BMPs, with the objective of minimizing costs while simultaneously maximizing both total phosphorus (TP) reduction and ecosystem service value. TP reduction was simulated usinga calibrated SWAT model that successfully assessed NPS pollution in the Dianchi lake watershed. Considering NPS generation and transfer, three NPS control scenarios were designed: source reduction, process retention, and integrated control. The upper limits of the costs for three scenarios were 157, 4921, and 5013 million yuan, respectively. It is suggested that the ecological benefits should be taken into account in the BMPs design. This finding is conducive to generating optimized control schemes that can effectively balance the trade-offs between socioeconomic costs and environmental sustainability.