Two-Stage Pumping Control Model for Flood Mitigation in Inundated Urban Drainage Basins

Abstract

Abstract This study proposes a two-stage intelligence-based pumping control (TWOPC) model for real-time pumping operation to solve the complex problem of estimating the desired pump flow and determining the optimal combination of pumps deployed in a flood event. In Stage I of the model, the desired pump flow was forecasted using the multilayer perceptron (MLP) with hydrological information including rainfall and basin runoffs, forebay water levels, and pump flows. In Stage II, the optimal pump combination was forecasted using tree-derived rules obtained from C4.5, classification and regression tree (CART), and chi-squared automatic interaction detection (CHAID) classifiers. The East Chung-Kong pumping station in New Taipei City was used as the study area. The pumping facilities included both submersible and upright axial pumps. The optimal inputoutput patterns, derived from a deterministic pumping operation optimization model, were used to train and validate the proposed TWOPC model. Data for this study were collected from three storms and four typhoons that affected an urban drainage basin. A total of 1,765 records were available. The results indicated that the case with a lag time of 5 min provided the most desirable pump flows in Stage I, and the C4.5 tree-based classifier performed well in Stage II. In addition, Typhoons Sinlaku (2) (2008/9/15) and Jangmi (2008/9/29) were selected for simulating the TWOPC model. The results demonstrated that the TWOPC model provided a more favorable performance than the traditional experienced method did. Overall, the proposed two-stage prediction model successfully addressed the problems of both determining the desired pump flow and optimal pump combination.