A large-scale distributed supervisory system that services a large area should ideally be capable of performing automatically optimal control based on various types of data provided by the controlled equipment. However, conventional systems have generally been configured so that pre-ordained transmission and reception processes are the input to the FEP (front end processor) unit according to various criteria of the data provided by the controlled equipment (e.g., the type and significance of the data, its priority, and its designated destination), while the host computer sends back a response to the controlled equipment based on the data transferred from the FEP according to preset optimal control algorithm. But to achieve fundamental improvements of system speed and reduce the load on the host computer, one must adopt the configuration in which the FEP can make automatic judgments regarding the criteria of data provided by a diverse variety of controlled equipment, instead of having to devise system-wide algorithms based on predetermined criteria in the data provided by the controlled equipment. To configure an FEP for this sort of system, we propose a method where a plurality of programs that identify the conditions of a specific type of data are prepared separately and integrated by means of a multi-agent architecture.