An interactive reservoir management system for Lake Kariba

This paper presents a user-interactive decision support system (DSS) for the management of the Lake Kariba reservoir. Built in the fourth-generation computer language IFPS, the system takes into account relevant reservoir characteristics and parameters, such as the amount of hydropower generated, reservoir storage throughout the year, and the amount of water released for down-stream usage. The system blends water release rules determined previously using optimization and simulationbased scenario analyses with expert input from an experienced reservoir manager, yielding a n intuitive and realistic DSS with which the reservoir manager may easily identify. The DSS also includes a BoxJenkins time series model that forecasts future inflows. Each month, the system provides the manager with a proposed release schedule, which the manager then uses to explore and evaluate the consequences in terms of the decision criteria, over an extended period of time. The types of information provided to and sought from the manager correspond closely with actual reservoir management practice. An important characteristic of the system is that the manager can quickly explore various different potential release decisions a priori, for a variety of potential inflow scenarios, including predicted inflows for average hydrological years, as well as inflows reflecting extreme events such as drought and flood periods. The manager can compare the results of the release decisions made in the scenario analysis, both with the release strategy proposed by the system and with historical release decisions, thus aiding the manager in establishing effective reservoir management policies in practice. Thus, rather than a mechanical value, our DSS offers the manager a flexible problem analysis with suggested courses of action. We illustrate the system using example sessions with an experienced reservoir manager. While the system is designed specifically to support the management of Lake Kariba, its extension to a more general class of reservoir management problems is straightforward.