Nowaday, in domains with a uniform hierarchical structure, like Power System networks domain, the advantages of the introduction of knowledge based systems to automate information processing are well known. In the specific case of alarm processing, the majority of studies have been focused on alarms in central energy management systems, in which only one system is required. The still unsolved problem appears when trying to automate a task at all local nodes (in our case, at substation level), i.e., when a distributed solution must be find. To go one step further, the solution method must be capable of the automatic generation of similar expert systems, taking into account not only conceptual similarities, but also physical differences, different topologies and different tasks. With the purpose to identify the REUSABLE POINTS of an intelligent substation alarm processor, in order to be modified and reconfigured to correspond to any particular real substation, a prototypical hybrid rule-based object-oriented expert system for the treatment of alarm messages in power substation plants has been developed. This work would lead to state the necessary features of a Distributed Network of intelligent substation alarm processors, installed at each substation of a Power System network. The paper describes the complete, really operational on-line alarm processor, which has embedded a prototype of expert system. The attention is focused on the mentioned features of the distributed network target (actual AI limitations are discussed). Finally, software reusability, final installation constraints, automatic validation and maintenance are mentioned.