TY - JOUR
T1 - TEXTURED MODEL FOR COMPUTATIONALLY EFFICIENT REACTIVE POWER CONTROL AND MANAGEMENT.
AU - Zaborszky, J.
AU - Huang, G.
AU - Lu, K. W.
PY - 1985
Y1 - 1985
N2 - The texture of the power system which governs; the interplay of reactive power and voltage is emulated by a textured model which assembles local groups of buses into a multi-leaf structure. Groups on the same leaf of the model are not coupled with each other; groups on different leaves overlap partially and are thus coupled. The paths of computational information are organized in an efficient manner and inefficient computation and information paths are eliminated; however, the computation converges to the exact solution, not an approximate one. The resulting model is ideally suited for parallel processing, especially since there is no sequential component in the computation, no computation overhead, and (if the size of the groups and their numbers per leaf are uniform) no waiting time. A 100-fold saving in computation time has been observed in experiments on steepest descent algorithms with systems of around 100 buses. Computation times also compare favorably with existing speedup techniques such as block pivoting. Computation times for common algorithms (like matrix manipulations, Newton-Raphson, linear and nonlinear programming) increase with the system size at a fast nonlinear rate. The computation times remain essentially constant for the textured model in parallel processing. Thus, very large computation time savings are implied on larger systems.
AB - The texture of the power system which governs; the interplay of reactive power and voltage is emulated by a textured model which assembles local groups of buses into a multi-leaf structure. Groups on the same leaf of the model are not coupled with each other; groups on different leaves overlap partially and are thus coupled. The paths of computational information are organized in an efficient manner and inefficient computation and information paths are eliminated; however, the computation converges to the exact solution, not an approximate one. The resulting model is ideally suited for parallel processing, especially since there is no sequential component in the computation, no computation overhead, and (if the size of the groups and their numbers per leaf are uniform) no waiting time. A 100-fold saving in computation time has been observed in experiments on steepest descent algorithms with systems of around 100 buses. Computation times also compare favorably with existing speedup techniques such as block pivoting. Computation times for common algorithms (like matrix manipulations, Newton-Raphson, linear and nonlinear programming) increase with the system size at a fast nonlinear rate. The computation times remain essentially constant for the textured model in parallel processing. Thus, very large computation time savings are implied on larger systems.
UR - http://www.scopus.com/inward/record.url?scp=0022090708&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0022090708&partnerID=8YFLogxK
U2 - 10.1109/TPAS.1985.319204
DO - 10.1109/TPAS.1985.319204
M3 - Article
AN - SCOPUS:0022090708
SN - 0018-9510
VL - PAS-104
SP - 1718
EP - 1727
JO - IEEE transactions on power apparatus and systems
JF - IEEE transactions on power apparatus and systems
IS - 7
ER -