Reservoir operation using the neural network and fuzzy systems for dam control and operation support

Abstract

This paper is concerned mainly with multipurpose dam with drainage area relatively smaller compared with dam capacity. A comparison is made between reservoir operation using the fuzzy and neural network systems and actual one by operator, using examples of floods during flood and non-flood seasons. Then, practical utility and usefulness of reservoir operation by this control system are considered and evaluated. As a result, the main conclusions of this paper are obtained. (1) As a result of applying the fuzzy system and neural network–fuzzy system to dam operation support system, the fuzzy system is an effective operation system, when water use is the main objective, and the neural network–fuzzy system is effective primarily for flood control. (2) Analyses have been made using flood examples of flood season and non-flood season, but there is a structural difference in components for determining discharge. Consequently, the study reveals that there is a structural difference in decision of outflow discharge depending on flood season and non-flood season. That is, for non-flood season, good result has been obtained by using, as input for storage, forecasted inflow in place of change in inflow. From this, it is seen that it is necessary to change structure identification for determining operation quantities depending on the difference in objectives: water use (non-flood season) or flood control (flood season).