Real‐time control of a system of large hydropower reservoirs

This paper describes a real-time optimal control approach for operating a system of large hydropower reservoirs. The operating objective is to track specified power output targets, subject to a variety of physical constraints. The constraints describe the hydrologic behavior of the tributary watershed and the dynamics of the reservoir system. The decision variables are monthly average releases from each of the system reservoirs. These releases are derived in real time, as functions of available measurements of reservoir storage and tributary inflow. Inflow and measurement uncertainty are explicitly included in the state and measurement equations. The stochastic operations problem is first formulated in general terms and then simplified to allow the use of classical linear-quadratic stochastic control concepts. The solution to the simplified control problem is implemented by combining a linear deterministic control law with a linear estimation algorithm. The resulting stochastic controller is applied to a two-reservoir system in the Caroni River basin of Venezuela. Preliminary tests indicate that the controller performs well in this application, even when some of its underlying assumptions are violated.