The effects of precision and small impedance branches on power flow robustness

Renewed interest at EPRI on improved power flow techniques is due to the recent improvements in the robustness of fast power flow algorithms. In this paper the authors investigate two interacting parameters which affect convergence in power flow algorithms, limited precision and small impedance branches. Analytical methods and experimental results are given for predicting the minimum convergence tolerance achievable by an algorithm as a function of the infimum of the set of branch impedance values, and the number of bits of mantissa precision. Methods are suggested for increasing the virtual precision at minimal computational cost and for scanning power flow input data to determine whether available precision is sufficient