A novel approach for dynamic flow simulation of gas pipelines using teaching–learning-based optimization algorithm

Abstract

Simulation of a natural gas network operation is a prerequisite for optimization and control tasks. Treating gas in a transient manner is necessary for accurate simulation of gas networks. However, solving the governing nonlinear partial differential equations of pipe flows is a challenging task. In this paper, a novel approach is proposed based on using an intelligent algorithm called teaching–learning-based optimization. This approach simplifies transient simulation of gas networks with a specified type of boundary conditions. Teaching–learning-based optimization estimates different values for network inlet flow rates. Then by knowing the inlet boundary conditions of the network, the discretized flow equations become linear and the flow equations of each pipe can be solved independently. Thus, the network outlet flow variables can be easily obtained. The differences of obtained and actual network outlet flow rates are considered as a cost function or error. Finally, this intelligent algorithm determines the optimum inlet flow rates at each time level, which minimize the error. The proposed approach is implemented on the in-service gas network. To validate the simulation results, a conventional gradient-based method called trust region dogleg is also used for simulation of the gas network. The comparison of numerical results confirms the accuracy and efficiency of this approach, while it is more computationally efficient. Moreover, the substitution of teaching–learning-based optimization with another powerful intelligent optimization algorithm would not improve the performance of the proposed approach.