Groundwater depletion in the North China Plain threatens both aquifer sustainability and food production. Inter-basin water diversion, especially the South-to-North Water Diversion Project, may provide an alternative irrigation source, but its ability to support groundwater recovery while maintaining winter wheat yield remains poorly quantified. Here, we develop <em>irrigmoo</em>, a multi-objective optimization framework that scripts a three-dimensional numerical groundwater model into the non-dominated sorting genetic algorithm II (NSGA-II) and couples it with a statistical winter wheat yield model to optimize zoned irrigation. Applied to Cangzhou, a groundwater-depleted agricultural region in the North China Plain, <em>irrigmoo</em> represents hydrolithological heterogeneity through three irrigation-management zones and evaluates the joint effects of replacement-water capacity and spatial irrigation allocation. Results show that irrigation reduction alone cannot reverse groundwater depletion without substantial yield penalties. By contrast, replacement water shifts the Pareto frontier toward simultaneous groundwater recovery and yield stabilization. Under the maximum replacement scenario, the balanced solution produces 0.40 m a⁻¹ of deep groundwater-level recovery and a 4.29 % yield increase, while the base scenario fails to meet both targets simultaneously. The largest relative benefit occurs in groundwater-scarce Zone III, indicating that uniform irrigation policies miss important spatial differences in groundwater response and crop benefits. Future groundwater-level projections further show that reliable replacement water is necessary for sustaining recovery beyond the optimization period. These findings demonstrate that coordinated water replacement and zoned irrigation can transform the groundwater-yield trade-off into a feasible pathway for integrated water-food management in depleted aquifer systems.