Abstract
In water systems, service reservoirs enable transmission flows to be smoothed by buffering temporal variation of water usage with changing their levels. Skilled operators balance between the variation ranges of levels of service reservoirs and the smoothing of transmission flows. In this study, we propose an operation scheduling method which replicates the balance by learning from the operation records. We analyzed the level records of service reservoirs in a real system and described the effect of characteristics of the reservoirs on the histograms and the quantiles of the records. Based on the analysis, we formulated an operation scheduling model of intake and distribution pumps which minimizes deviations of reservoir levels from their usual ranges, computed by quantiles of level records of service reservoirs. In a case study, the proposed method computed operation schedules which replicated the balance between the variation ranges of levels of service reservoirs and the smoothing of flows exhibited in the operation records. In addition, it is suggested that the proposed method could reduce the violation of a service reservoir compared with current operators.
