Water system risk refers to the cumulative effects of water-related threats driven by natural and anthropogenic pressures. In light of this challenge, this study develops an evaluation-optimization framework for water resources management in the Beijing-Tianjin-Hebei urban agglomeration. It represents the first attempt to incorporate water system risk into a multi-level programming model while simultaneously assessing water resources utilization efficiency. Water system risk is assessed through three key dimensions: water disasters, environmental pollution, and resource utilization, as quantified by water disaster risk, water environmental risk, and water resources ecological footprint. The multi-level programming model is constructed with the goals of maximizing economic benefits, minimizing social inequality coefficient, and minimizing water pollutant emissions in a hierarchical order, which is solved by interactive fuzzy satisfaction algorithm. Results reveal that a majority of cities especially in Hebei facing increased water system risk, of which water disaster risk and water environmental risk would annually rise by 1.52% and 3.23%, respectively. Water resources utilization efficiency annually increased by 6.08%. Beijing and Tianjin had lower efficiency yet higher growth compared with Hebei. The optimal amounts of allocated water resources and economic benefits would respectively increase by 16.04% and 35.80%, while pollutant emissions and inequality coefficient would decline. Sensitivity analysis shows that water conservation curbs pollutant emissions (especially in domestic sector), yet restricts economic benefits and social equality. The marginal benefit of the multi-level decision is higher than that of single-level decisions. Findings offer insights for implementing targeted interventions in high-risk and water-inefficient cities while balancing economic-environmental-social priorities.