The insufficient inertia and reserve in renewable energy source (RES)-dominated power systems significantly challenge frequency security. This paper introduces a distributionally robust frequency constrained unit commitment (DR-FCUC) scheme at the system operation level to tackle this issue, optimizing day-ahead unit commitment, generation dispatch, and demand-side reserve procurement. The proposed DR-FCUC accounts for frequency dynamic constraints under the most extensive power disturbance scenarios and employs a DR chance constrained (DRCC) approach using a Wasserstein-metric ambiguity set to manage RES uncertainty. Furthermore, we utilize alternate support vector machine decision trees (ASVMTREE) to convert the high-dimensional frequency nadir constraint into a set of linear constraints and introduce a two-stage sampling method to enhance the ASVMTREE training dataset. Consequently, the proposed DR-FCUC is formulated as a mixed-integer linear programming (MILP) model. Case studies on modified IEEE 39-bus and IEEE 118-bus test systems demonstrate the necessity of incorporating frequency constraints into dispatch schemes, the critical role of demand-side frequency support, the effectiveness of the proposed DR-FCUC scheme, and the accuracy of the constraint convexification method. © 2024 Elsevier B.V.