There are several architectural configurations possible for an Energy Management System. This case study describes a situation in which a centralized configuration was changed to a decentralized configuration following initial operation. An initial system conforming to the operating precepts and desires of Niagara Mohawk Power Corporation was based upon a centralized configuration and became operational in October of 1985. The initial system had many communication linkages: to RTU's; to a power pool, to several regional control centers; to several power plant control rooms; to corporate computer facilities, and to distribution dispatch offices. Subsequently the system architecture was modified to a more decentralized configuration to better address concerns of remote regional control centers regarding catastrophic communication outages. The modified design represents an evolution from the initial centralized system to a decentralized one. The decentralized architecture needed to support all of the terminals of the originally centralized system. It also needed to contain a strong skeleton for centralized and decentralized operation. This paper describes the two systems, the motivation in the transition, and the attributes of the specific architecture which lent themselves to making the transition a feasible one. © 1990 IEEE