In recent years, efforts have been increasing to develop hybrid power generation systems, which consist of a mixture of renewable, conventional, and storage systems. The variability and non-dispatchability of some renewable sources as well as the attainment of tight performance requirements make the development of control systems for power plants in hybrid systems a remarkably challenging task. A promising approach to solve this problem is to apply the paradigm of Multi-Agent Systems, from Artificial Intelligence, which is currently carried out in two phases: engineering design and engineering implementation. This paper presents the engineering design, including the definition, description, and allocation of the Intelligent Agents of an improved Intelligent Multi-Agent System for Supervision and Control (iMASSC), which is proposed to create intelligent power plants from either renewable or conventional power generation units. In addition, simulations of a wind turbine generator model equipped with iMASSC are provided to demonstrate the feasibility of this methodology. The iMASSC approach uses a Multi-Agent System with a generic structure and takes into account the complexity of current power plants, which is more realistic than the single Intelligent Agent approach commonly found in the literature. This provides a scalable, modular, customizable, and functional framework to create intelligent power plants regardless of being conventional or not. Eventually, the community of intelligent power plants will achieve the objectives of the hybrid power system through autonomous and coherent collaboration. Hence, the proposed iMASSC model is expected to provide feasible solutions to the operation of modern hybrid power generation systems.