Optimized implementation of power dispatch in the OPA model and its implications for dispatch sensitivity for the WECC power network

Abstract

The social and economic costs of large blackouts in power transmission networks make it critical to properly understand their dynamics. The OPA model was developed with this objective in mind and has previously been applied to power grids of small and medium size, some of them properly modeling realistic cases such as the simplified WECC network, covering the Western region of the US. The bulk of the OPA model's computational cost comes from the repeated solution of a linear programming problem using the Simplex method which is difficult to parallelize. In this paper we introduce important improvements to the modeling part of the linear problem, accelerating the previous implementation by a factor of up to 200, depending on the network. These improvements make it possible, from a practical point of view, to simulate the largest, most detailed, WECC network consisting of 19,402 nodes, reducing the wall-clock time of the simulation from two years to only 10 days. The first simulations show an interesting result: the detailed 19,402 nodes network displays a reduced sensitivity of the dynamics to the dispatch, when compared to the previously used simplified WECC models containing only 1553 and 2504 nodes.