An Efficient Starting Point to Adaptive Holomorphic Embedding Power Flow Methods

Abstract

This paper proposes a new approach for accelerating the convergence runtime of a modified Holomorphic Embedding Load-Flow Method (HELM). In this adaptive HELM it is not used a flat start, nor an aleatory Initial Guess, but a previous starting solution provided by an iterative method based on Newton Krylov subspace. It is proposed, to improve this Adaptive HELM-based strategy, applying the use of the BiCGStab (Bi-Conjugate Gradient Stabilized) method, preconditioning, incomplete LU factorization, and reordering strategy. The goal is to optimize this previous starting solution to assist Adaptive HELM methods for getting the final solution in the fastest way. The results obtained are also compared with the traditional NR-method with flat start, an assisted NR-method, and the original HELM.