Parallel Optimization with Applications to Power Systems

A new pseudo-conjugate direction algorithm for the solution of the unconstrained optimization problem on a parallel computer is presented. At each iteration, n one-dimensional searches along n directions are performed simultaneously on n processors. These directions are so chosen that they progressively become more nearly conjugate as a minimum is approached. The algorithm does not require the evaluation or computation of any derivatives and it is suitable for execution on an existing multiprocessor computer (Siemens SMS 201). To assess the effectiveness of the proposed algorithm in solving optimization problems that occur in power systems analysis, (i.e. load flow, state estimation and optimal load flow type), its implemenation on the SMS-201 (128-processor system) is analyzed.