Quasi-Newton Power Flow Using Partial Jacobian Updates

We present a quasi-Newton power flow methodology that incorporates several strategies to obtain substantial computing savings. Newton steps are combined with constant Jacobian (or "simple") steps and partial Jacobian updates to get an efficient quasi-Newton method. The methodology proposed includes the possibility of selecting the next best step by measuring the residuals. Partial Jacobian updates (PJUs) are included in the quasi-Newton power flow using LU factorization updates and/or the matrix modification lemma. The method has been tested with systems ranging in size from 14 to 6,372 buses. For large power systems we have obtained savings (in flops) in the order of 50% compared to Newton's method.