Comparing Different Approaches for Solving Large Scale Power-Flow Problems With the Newton-Raphson Method

This paper focuses on using the Newton-Raphson method to solve power-flow problems. Since most computationally demanding part of is linear equations at each iteration, this study investigates different approaches both central processing unit (CPU) and graphical (GPU). Six have been developed evaluated in paper: two these run entirely CPU while other GPU, remaining are hybrid that GPU. All six direct solvers use either $LU$ or $QR$ factorization equations. Two hardware platforms used conduct experiments. The performance results show version with gives better compared GPU standard library called cuSOLVER even for larger Moreover, it has proven best achieved a where Jacobian matrix assembled preprocessing sparse high solver KLU performed first equation factorized solved CPU. Maximum speed up obtained largest case 25000 buses. shows speedup factor 9.6 NVIDIA P100 13.1 V100 comparison baseline an Intel Xeon Gold 6132