Dynamic load modeling of an Egyptian primary distribution system using neural networks

Abstract

Power systems planning, operation and control are tightly related to load characteristics and voltage magnitudes. Voltage stability is directly dependent on load behavior with voltage and frequency variations. Therefore, exact detection of the occurrence of voltage instability phenomena is mainly dependent on the correctness of the loads power/voltage and reactive power/voltage relations. The main objective of this paper is to validate the load mathematical representation by performing real field measurements on an actual primary distribution system. The load power/voltage characteristics have been derived from the load responses (voltage/time and power/time) characteristics recorded in the power substation chosen using a digital fault recording (DFR) device. Data recorded during disturbances confirmed the adequacy of the model chosen for the residential/industrial feeder monitored. Also, three load models for the chosen primary distribution system have been proposed and tested using artificial neural networks (ANN). They differ in the number of neurons and the number of inputs. The active and reactive power values obtained from the proposed ANN models are compared with the actual active and reactive power values obtained from the DFR. The results verify that ANN models emulate load dynamics accurately.