Technological advances have made wireless sensor nodes cheap and reliable enough to be brought into various application domains. The limited energy capacity of sensor nodes is the key factor that restricts their lifespan. In this paper, a Predictive Control strategy for Dynamic Power Management of a set of wireless sensor nodes is proposed. The control formulation is based on Model Predictive Control with constraints and binary optimization variables, leading to a Mixed Integer Quadratic Programming problem. The control algorithm proposed guarantees services and performances levels with a minimum number of active nodes and/or a minimum load on such components. The strategy is evaluated on a real test-bench with wireless sensor nodes equipped with batteries and harvesting systems. Experimental results show the effectiveness of the control method proposed.