We consider heterogeneous wireless networks in the physical interference model and introduce a new formulation of the optimization problem underlying their management. This formulation targets the minimization of power consumption by integrating base-station activation and many-to-many associations into the same mixed-integer nonlinear programming (MINLP) problem. We also introduce HetNetGA, a genetic algorithm tailored to solving the MINLP problem without further smoothing or approximation. We give computational results in scenarios involving both macrocells and picocells. Some results align well with what is expected, but others are unexpected, pointing essentially to the need to better understand the role of macrocells in such networks.