運用序率動態規劃與系集入流預報於水庫枯旱供水策略擬定

Abstract

臺灣水資源的利用主要仰賴於水庫的配水策略。雖然臺灣的降雨量十分充沛，但因為地形導致時間空間降雨不均，而使得大部份的水無法被利用。特別是在乾季，更是難以同時滿足下游用水需求與儲存水資源。因此，為了使水資源的分配更有效率，入流預報日益發展蓬勃，並且運用於水庫操作上，找尋使缺水清況得以減緩的操作方式。值得一提的是，國家災害防救中心為石門水庫開發了系集入流預報的產品，其可以運用於水庫操作，有助於擬定預報期間內的放水策略。本研究主要目的為建立結合系集入流預報與序率最佳化以進行水庫操作的架構，其中，本研究採用的序率最佳化為隱式與顯式，並且將其運用在石門水庫2009年的實際操作。經過模擬，這兩種模式因其特性所呈現的操作結果產生了些微差異。尤其在面臨乾旱情況時，隱式序率最佳化傾向於分散入流特性的分布，再為預報期間決定出較為保守的決策。而顯式序率最佳化則是採納了預報所有可能的路徑，做出較具信心的放水策略。因此，若預報的準確度得以大幅提升，則顯式序率最佳化可以被水庫操作者採納；反之，若預報的可靠度較低，隱式序率最佳化可以處理較難以定義的不確定性。

In Taiwan, utilization of water resources mainly relies on reservoir. Although the precipitation is abundant, lots of water is unavailable due to steep landform and spatio-temporally unequal distribution of rainfall. Especially in dry season, it is hard to satisfy the demand and conserve water. To allocate limited water resources effectively, inflow prediction has been developed and applied to reservoir operation for mitigating water shortage. National Science and Technology Center for Disaster Reduction (NCDR) produced ensemble streamflow prediction (ESP) for Shihmen Reservoir. It could be helpful information for determining release strategy in the following period with prediction. This study provides a framework for combining ESP with stochastic optimization, which contains implicit and explicit stochastic optimization (ISO and ESO), is applied to practical operation with for Shihmen Reservoir in 2009. Because of the characteristics of ISO and ESO, the release strategy decided by two models are presented as slight difference. While confronting drought, ISO tends to disperse the distribution of inflow in the predicted period and determine a conservative decision; on the contrary, ESO absorbs the traits of prediction and makes the policy with confidence. Therefore, ESO model could be adopted to the process of decision making if the precision of inflow prediction is promoted; on the other hand, ISO model could be used to deal with indefinite uncertainty if the reliability of prediction is insufficient.