Author(s): Shourian Mojtaba; Mousavi S. Jamshid
Linked Author(s):
Keywords: River basin management; Simulation; Optimization; MODSIM; PSO; ANN. 1
Abstract: Artificial neural network is used as an effective tool for function approximation in a simulation-optimization model. In the first step, a simulation-based optimization model is developed for optimized design and operation of the Upstream Sirvan River Basin system in Iran. This model integrates MODSIM, the generalized river basin network flow model with the capability of simulating various characteristics and features of water resources in a river basin, and the Particle Swarm Optimization (PSO) algorithm. In the developed PSO-MODSIM model, the size of dams at potential storage sites and water transfer systems, as design variables, and the relative priorities for meeting reservoir target storage levels, as operational variables, are varied and evolved using PSO algorithm while MODSIM is called to simulate the system’s performance and to evaluate the fitness of each set of those design and operational variables with respect to the model’s objective function considered. But the direct incorporation of MODSIM, as a complex simulation model, into the PSO optimization framework is computationally prohibitive. To overcome this problem, in the second step, a artificial neural network (ANN) is trained to approximate the simulation model developed for the considered system. In the resulting PSO-MODSIM~ANN model, the off-line trained backpropagation neural network is used as a submodel in PSO programming algorithm to find optimal solutions. Results of the models are further compared. Short computational time needed in the PSO-MODSIM~ANN model simplifies the analysis of various PSO performances by changing the model's critical parameters, resulting in a better solution in comparison with the results of the PSO-MODSIM model.
Year: 2007