Dealing with Zero Flows in the Simulation of Water Distribution Networks with Low-Resistance Pipes Using the Global Gradient Algorithm

This paper shows that the zero flow treatment algorithm and the convergence criteria of EPANET 2.2, the latest version of the EPANET 2 open-source software package, may be responsible for convergence to distinctly inaccurate results in the case of networks with low-resistance pipes. As an alternative, the paper suggests a zero flow treatment method proposed earlier, which is based on the smallness of the flow rather than the flow derivative of the head loss as in EPANET 2.2. To avoid spurious convergence, the EPANET 2.2 convergence criteria may be complemented with a criterion based on the flow residuals of the energy balance equations. The monotony of the head loss vs. flow relationship allows these residuals to be checked for a specified tolerance without solving the energy balance equations for the flow.

Automated summary

This paper demonstrates that the zero flow treatment algorithm and convergence criteria in EPANET 2.2 can lead to significantly inaccurate results in networks with low-resistance pipes. Instead, the paper proposes an alternative zero flow treatment method that is based on the smallness of flow rather than flow derivative of head loss as in EPANET 2.2. To avoid spurious convergence, the paper suggests supplementing the convergence criteria of EPANET 2.2 with a criterion based on the flow residuals of the energy balance equations, which can be checked without solving the energy balance equations for flow.