More demand-side flexible resources (DFRs) are participating in the frequency regulation of renewable power systems, whose heterogeneous characteristics have a significant impact on the system frequency response. Consequently, selecting suitable DFRs poses a formidable challenge for independent system operators (ISO). In this paper, a reserve allocation methodology for heterogeneous DFRs is proposed to manage the risk of power system frequency. Firstly, a performance curve is developed to describe the cost, capacity, and response speed of DFRs. Moreover, a clustering method for multiple distributed DFRs is conducted to calculate the aggregated performance curves and uncertainty coefficients. Then, the frequency security criterion considering DFRs’ performance is constructed, whose linearity makes it can be easily coupled into the system scheduling model and solved. Furthermore, a risk management model for DFRs considering frequency-chance-constraint is proposed to make a trade-off between cost and frequency security. Finally, the model is transformed into mixed integer second-order cone programming (MISOCP) and solved by the commercial solver. The proposed model is validated by the IEEE 30 and IEEE 118 bus systems.