抽水站水位預測及系統操作之研究

Abstract

抽水站為都市防洪最重要的一環，當河川水位高漲市區內水無法利用重力閘門排水時，只能依靠各抽水站的操作將市區內水排放至河川，因此本研究針對雨水下水道系統進行水位預報與抽水機組操作規線為研究方向；選用台北市文山區中港抽水站集水區作為研究目標，研究主要分成兩部分，第一部分應用倒傳遞類神經網路架構中港三多時刻的水位預報模式，將北政國中雨量站與中港三水位站及中港抽水站閘門與抽水機組的操作記錄等為輸入資訊，作為模式建構之用，並針對輸入項不同所得到的結果做分析與討論，同時探討資料雜訊之影響及提供資料雜訊處理的方法，研究結果顯示倒傳遞類神經網路可有效的預報未來的水位，而雜訊的處理可提升水位站多時刻的水位預報精度；第二部分為模擬抽水機組操作規線，藉由歷史資料計算出來的入流量，根據抽水機組操作規線模擬操作，並改變其規線比較模擬結果，結果顯示中港抽水站內抽水機組操作規線有改善的空間。

The pumping station is one of the most important facilities in preventing flooding events takes place for the Taipei city. When the water level of the river arises a certain level, we can only rely on the operation of pumping stations to discharge water from the city area to the rivers. The objective of this study include is to construct a forecast model and to investigate the suitability of the operating rule. The first main constructs a BPNN model to predict one-step-ahead and two-step-ahead sewer drainage system of Chung-Kang during flood events and investigate the influence of amplitude or noise data on the model and provide effective solution. The results show that the BPNN can suitably predict multi-step-ahead water level in the sewer system, while the data processing can improve the accuracy of the model. The second part is to investigate the suitability of the existing operating rule lines of the pumping station and comparing the results obtained by simulating a number of adjusted operating rule lines based on historical data sets with the historical information. The results indicate that the operation of the pumping station of Chung-Kang can be improved.