Model simulation has demonstrated its usefulness in evaluation and decision-making for improving preliminary versions of artefacts before production. Particularly, one of the main goals of simulation is to verify model properties based on data collected from its execution. In this paper, we present the simulation capabilities of our REFAS framework for specifying requirements models for dynamic software products lines and self-adaptive systems. The simulation is controlled by a feedback loop and a reasoning engine that operates on the functional and non-functional requirements. The paper contribution is threefold. First, REFAS allows developers to evaluate and improve requirements models through their simulation capabilities. Second, REFAS provides rich feedback in its interactive simulations for the human modeller to make informed decisions to improve her model. Third, REFAS automates the generation of simulation scenarios required to verify the model adequacy and correctness. We evaluate our contribution by comparing the application of REFAS to a case study used in other approaches.