<p>Industrial production faces the challenges of optimizing allocations in resource-constrained scenarios to reduce operating costs. Capacity-constrained clustering is a commonly encountered problem in resource-constrained allocation. While some prior studies have incorporated object demands into capacity constraints, they often neglect the direct impact of these demands on the cost computation in the objective function, which is crucial in scenarios like logistics cost optimization. This paper studies a nonuniform capacity-constrained clustering problem that imposes constraints on the total demand of objects within each cluster. We reformulate the problem into an equivalent network flow model, which provides an easily solvable framework for satisfying the capacity constraints. Then, we propose a rapid-converging iterative algorithm that alternately optimizes the decision variables. The experiments on five commonly used clustering datasets show the excellent performance of our algorithm on capacity constraint satisfaction and intra-cluster aggregation. Furthermore, two real-world applications demonstrate the effectiveness of our method on challenging cases involving objects with high-deviation demands or distributed in non-convex regions. Our method provides an effective solution with theoretical guarantees for the nonuniform capacity-constrained clustering problem in resource-constrained allocation scenarios.</p>