Electrical power network analysis and computation play an important role in the planning and operation of the power grid, and they are modeled mathematically as differential equations and network algebraic equations. The direct method based on Gaussian elimination theory can obtain analytical results. Two factors affect computing efficiency: the number of nonzero element fillings and the length of elimination tree. This article constructs mapping correspondence between eliminated tree nodes and quotient graph nodes through graph and quotient graph theories. The Approximate Minimum Degree (AMD) of quotient graph nodes and the length of the elimination tree nodes are composed to build an Approximate Minimum Degree and Minimum Length (AMDML) model. The quotient graph node with the minimum degree, which is also the minimum length of elimination tree node, is selected as the next ordering vector. Compared with AMD ordering method and other common methods, the proposed method further reduces the length of elimination tree without increasing the number of nonzero fillings; the length was decreased by about 10% compared with the AMD method. A testbed for experiment was built. The efficiency of the proposed method was evaluated based on different sizes of coefficient matrices of power flow cases.

Nowadays, power quality problems are affecting people's daily life and production activities. With an aim to improve disturbance detection accuracy, a novel analysis approach, based on multiple impact factors, is proposed in this paper. First, a multiple impact factors analysis is implemented in which two perspectives, i.e., the wavelet analysis and disturbance features are simultaneously considered. Five key factors, including wavelet function, wavelet decomposition level, redundant algorithm, event type and disturbance intensity, and start and end moment of disturbance, have been considered. Next, an impact factor based accuracy analysis algorithm is proposed, through which each factor's potential impact on disturbance location accuracy is investigated. Three transforms, i.e., the classic wavelet, lifting wavelet and redundant lifting wavelet are employed, and their superiority on disturbance location accuracy is investigated. Finally, simulations are conducted for verification. Through the proposed method, the wavelet based parameters can be validly selected in order to accurately detect power quality disturbance.