Chance-Constrained Optimization for Non-Linear Network Flow Problems

Many engineered systems, such as energy and transportation infrastructures, are networks governed by non-linear physical laws. A primary challenge for operators of these networks is to achieve optimal utilization while maintaining safety and feasibility, especially in the face of uncertainty regarding the system model. To address this problem, we formulate a Chance Constrained Optimal Physical Network Flow (CC-OPNF) problem that attempts to optimize the system while satisfying safety limits with a high probability. However, the non-linear equality constraints representing the network physics introduce modelling and optimization challenges which make the chance constraints numerically intractable in their original form. The main contribution of the paper is to present a method to obtain tractable polynomial approximations to the chance constraints using Semidefinite Programming (SDP). The method uses a combination of existing semi-algebraic techniques for projection and volume computation in combination with novel set manipulations to provide conservative inner approximations to the chance constraints. In addition, we develop a new two-step procedure to improve computational speed. While the method is applicable to general physical network flow problems with polynomial constraints, we use the AC optimal power flow problem for electric grids as an example to demonstrate the method numerically. Keywords—Stochastic/Uncertain Systems, Nonlinear Systems, Algebraic/Geometric Methods, Optimization, Transportation Networks