Today's energy systems are undergoing remarkable changes in the way electrical energy is produced, transmitted and consumed. The operation of traditional electric power systems which is based on production of electric energy in large centralized power plants, usually located away from load centers is facing many challenges due to this transition. Main factors for this transition are use of renewable energy resources, integration of distributed generation into the grid, environmental considerations, increased customer participation and technological development. Further, demand for more and more services to be provided by these energy systems has resulted in complex energy systems. As time passes there will be more and more renewable and distributed energy resources at the lower level of the system. This distributed generation mainly comprising renewable is transforming the power grid from vertical structure to a horizontal one. In order to guarantee continuity of electricity supply to the customers, this transformation is required to be smooth. New methods and paradigms need to be investigated to fulfill the requirements of future grid. This increased use of distributed energy sources at distribution level is contributing in complexity of its operation and control and decentralization, autonomy and active distribution management have become the most important features for the smooth modernization of the present distribution grid. Multi-Agent System(MAS) having features like distributed intelligence, autonomy etc provide an alternative approach for smooth transition of the present grid to the smart grid. Motivated with the above mentioned considerations, the research objective of this dissertation was to develop and study different control architectures for distributed energy resources in the power distribution network using the agent based paradigm. Centralized, decentralized and distributed voltage control schemes developed and suitability of these structures was investigated. As the study involves three domains i.e. the power system, communication and the MAS system, co-simulation framework was developed including tools from these three domains. Proposed control algorithms based on MAS were simulated using this co-simulation framework. A voltage sensitivity based centralized voltage control algorithm was developed and a contract-net-protocol based plan for agent coordination was formulated. Comparisons between agent based decentralized and centralized schemes were made using the Key Performance Indices (KPI). An agent based distributed iterative algorithm for consensus between agents using neighboring communication was presented. Further, a zone based algorithm was developed and system performance was investigated by gradually increasing number of zones in a given network. This gives insight into system performance if we move from centralization towards decentralization. This study investigates the operation of future power distribution network including ICT infrastructure which can increase the flexibility and controllability. It presents different agent based control techniques which constitutes a step forward in determining the best control and management structure including physical and logical structure. It enables us to look into the trade-off between decentralized, distributed and centralized control. Application of such techniques will contribute in smooth modernization of the power grid and explore the design space for alternatives that are insensitive to changes in the system and can maintain their stability and performance in the presence of partial system faults.