Intelligent real-time operation of a pumping station for an urban drainage system

In this study, we apply artificial intelligence techniques to the development of two real-time pumping station operation models, namely, a historical and an optimized adaptive network-based fuzzy inference system (ANFIS-His and ANFIS-Opt, respectively). The functions of these two models are the determination of the real-time operation criteria of various pumping machines for controlling flood in an urban drainage system during periods when the drainage gate is closed. The ANFIS-His is constructed from an adaptive network-based fuzzy inference system (ANFIS) using historical operation records. The ANFIS-Opt is constructed from an ANFIS using the best operation series, which are optimized by a tabu search of historical flood events. We use the Chung-Kong drainage basin, New Taipei City, Taiwan, as the study area. The operational comparison variables are the highest water level (WL) and the absolute difference between the final WL and target WL of a pumping front-pool. The results show that the ANFIS-Opt is better than the ANFIS-His and historical operation models, based on the operation simulations of two flood events using the two operation models.