Combining Newton-Raphson and Stochastic Gradient Descent for Power Flow Analysis
Napoleon Costilla-Enriquez, Yang Weng, Baosen Zhang
Abstract
The power flow problem is an indispensable tool to solve many of the operation and planning problems in the electric grid and has been studied for the last half-century. Currently, popular algorithms require second-order methods, which may lead to poor performance when the initialization points are poor or when the system is stressed. These conditions are becoming more common as both the generation and load profiles changes in the grid. In this paper, we present a hybrid first-order and second-order method that effectively escapes local minima that may trap existing algorithms. We demonstrate the performance of our algorithm on standard IEEE benchmarks.
Topics & Concepts
InitializationMaxima and minimaNewton's methodMathematical optimizationGridComputer scienceElectric power systemPower flowGradient descentFlow (mathematics)AlgorithmPower (physics)MathematicsNonlinear systemArtificial neural networkArtificial intelligenceMathematical analysisGeometryPhysicsQuantum mechanicsProgramming languageOptimal Power Flow DistributionPower System Optimization and StabilityElectric Power System Optimization