Identification of hydraulic resistance parameters in hydraulic network model

The work is devoted to the identification of hydraulic resistance coefficients in hydraulic network model. The prob- lem raised in the paper is relevant for enterprises dealing with either pipeline transport of fluid hydrocarbons, or water supply and central heating. The proposed algorithm can be implemented for corresponding technological modes calcu- lation and prediction. Complexity of real pipeline networks, uncertainty and instability of their parameters make parameters identification procedure indispensable. The network model under research is a system of nonlinear equa- tions, drawn up in accordance with Kirchhoff’s laws for pipeline networks. Equations describe laws of flows conserva- tion for nodes as well as pump heads and head drops along independent circuits of the network. To solve such system one traditionally implements Newton’s method or its modifications, for instance the successive approximation method. Mentioned methods have numerous disadvantages like instability and sensibility to initial approximation of the roots. The system of equations solving is accompanied with their parameters tuning usually with the Nonlinear Least Square Method. Again, the method can be instable due to overdefined and large-scale problem statement. In the paper we pro- pose to substitute solution of the system with nonparametric estimation of the solution. We implement regression type estimate with respect to residuals of the equations calculated for measured data sample. Simultaneously parameters of the equations are identified. Modification of Kiefer-Wolfowitz procedure is used as an identification algorithm. Iden- tification is performed with simultaneous evaluation of the solution of the system of equations. The use of appropriate algorithms is considered at three-loop pipeline network with a single active pressure. Numerical experiments with proposed algorithm demonstrate its applicability for practical identification problems solution.