Vehicle clustering has been utilized for reducing the complexity of vehicle-to-everything (V2X) communications that would ultimately improve road traffic efficiency. Using a car-following model in a V2X communication system, we propose a vehicle clustering method for dynamically classifying vehicles and adjusting cluster size in real time. The especially important issue for the selected cluster-head vehicles (CHVs) can achieve an optimal trade-off between the CHVs' relative speed and power allocation. Furthermore, in order to balance the power allocation among the CHVs to further increase the downlink throughput, a power control approach for non-orthogonal multiple access (NOMA) is proposed. In this approach, the power allocation coefficients are obtained by maximizing the achievable rate while meeting the predefined target rate of each NOMA user. Finally, numerical simulations are provided to confirm the theoretical results and demonstrate the superior performance of the proposed approach. Note that through the numerical simulations, we can find the critical point of maximizing the minimum achievable rate among the CHVs at low transmission power of base station (BS).