Considerations and Design Goals for Unbalanced Optimal Power Flow Benchmarks

Distribution network constrained mathematical optimization is key technology to enable advanced distribution network management. In recent years, a growing amount of articles on the development of such models and algorithms have been published. Benchmarking of different approaches is crucial to establish performance trade-offs between accuracy, reliability and computational intensity. Today, practitioners tend to take ad-hoc approaches, building on power flow data sets but adding customer extensions to establish information such as voltage/current/power bounds and to parameterize pre-defined objective functions. To foster progress in this field, in this work we discuss i) a number of design trade-offs and pitfalls related to benchmarking, ii) develop a data model and mathematical specification for (up-to) four-wire optimal power flow, and iii) develop some initial data sets. The data sets are provided through open-access initiatives under a creative commons license, and a reference implementation of the mathematical model is made available with a permissive license. For ease-of-use, and to maximize uptake in the community, we establish a tiered approach to the benchmark development with a multi-year plan.