Decentralized decision making is a promising candidate for cost-effective participation of small electric loads to power system operations. In such setup, individual loads take control decisions based on local information. Local control laws are designed to synchronize individual efforts. Large loads groups may thereby contribute to system stability in a predictable way. In this paper, Energy Constrained Loads (e.g., battery charging, night storage heater) participates in primary control in a communication-free setup. Three different controllers and associated actions are considered: (1) delay at start, (2) regular on/off switching, or (3) proportional power modulation. Control laws are respectively based on discrete thresholds, probabilistic algorithms and proportional control. Two linear aggregate models are introduced to render large-scale simulations tractable. They accurately reproduce the group's frequency response in the three proposed control strategies. Long-term system simulations are then run to assess the group's response performance.