Data-driven control of water reservoirs using El Niño Southern Oscillation indexes

Abstract

Advanced modeling and control contribute to the design of efficient and sustainable water management strategies, which are challenged by the increasing frequency and intensity of extreme events often associated with large-scale climate signals, such as El Niño Southern Oscillation (ENSO). Despite ENSO-related information provides a great opportunity to make the operations of water systems more flexible and adaptive, incorporating it into an operating policy still represents a major challenge for optimal control algorithms. In this work, we contribute a framework combining Input Variable Selection techniques for ENSO detection with a data-driven control strategy to use this information for improving the system operations. Our framework is demonstrated on the control of the multipurpose Hoa Binh reservoir (Vietnam), showing that ENSO teleconnection represents a valuable information for addressing the performance tradeoffs between energy production, water supply, and flood protection.