Identification of a Constrained Nonlinear Hydrological System Described by Volterra Functional Series
This paper addresses the problems of identification of a constrained nonlinear system (CNLS) described by Volterra functional series. A workable and practical method in terms of a penalty function principle and orthogonal expansion was developed, which involves an extension of the classical least squares solution to include a penalty function with arbitrarily defined weights which are adjusted to ensure that the constraints are satisfied. A total of nine basins across a range of climates and catchment areas in China were selected for examination of both daily and hourly rainfall-runoff forecasting. It was found that the method described in this paper can provide a more reasonable and robust response function where hydrologic constraints are required. The nonlinear model yields a better streamflow forecasting than the linear model, particularly in peak flows.