A penalty-based optimization for reservoirs system management

A model is presented for planning the operations of interconnected surface water resource facilities equipped with hydroelectric power stations. These facilities are operated for the purpose of water supply, hydroelectric power generation and low flow augmentation for navigation purpose. The model is formulated as a deterministic large-scale nonlinear program, with decision variables being the monthly reservoir releases, and canal and pipeline flows through turbines. A solution procedure is proposed based on penalty method in which the nonlinear constraints are assigned penalty weights and moved to the objective function. The resulting problem with linear constraints is solved using reduced gradient techniques. The calculations of the reduced gradient is carried out efficiently by utilizing the special structure of the matrix involved in the linear equality constraints. The results of this procedure are compared against those obtained using a subgradient optimization algorithm. The method has been validated using a five-reservoir system of the Senegal River Basin in West Africa, where it serves as the model-base of the planning decision support system.