Skip to Main Content
This research presents new intelligent approaches for the estimation and comprehensive analysis of the two main power quality parameters (sags and swells) using Neural Networks. Typical power quality (PQ) disturbances include sag, swell, harmonics, transients and temporary, momentary and sustained interruptions in a power distribution network. Among all these disturbances, sags and swells get prime importance, as they can cause sufficient damage to industrial consumer's equipment and can ultimately lead to shut down of their system. In this research Principal Component Analysis technique (PCAT) is used to pre-process the raw PQ data and reduce the number of attributes of real PQ data. Refined data attributes are then processed through Feed Forward Back Propagation (FFBP) & Recurrent Neural Networks (RNN) for the estimation/prediction of sag and swell. Application of RNN on PQ data demonstrates its good estimation abilities (accuracy for sag & swell estimation=96%) as compared to FFBP neural network (accuracy for sag estimation [93.5%] & swell estimation [91.5%]). The results obtained in this paper are compared with the field data of a power company in Melbourne, Australia. This research will facilitate power utilities and industrial consumers on common understandings to set a base line for PQ parameters and also to evolve a comprehensive strategy for better management of PQ problems.