Intelligent data science enabled reactive power optimization of a distribution system

Abstract

Recently, distributed generators (DG) and electric vehicles (EV) are commonly employed in recent days in spite of the power flow of the distributed network (DN) being influenced by the intermittency and arbitrariness of DGs and electric vehicles (EVs). The massive increase in the distinct varieties of controllable devices has complicated controlling needs and is combined in the DN. Reactive power optimization (RPO) of the DNs helps to minimize the power loss, enhance the voltage quality, and inexpensive functioning of the DNs. ROR can be considered as a complicated high dimension non-linearity problem. The recent advances of data science approach containing machine learning (ML) and deep learning (DL) aid to make effective decisions. With this motivation, this paper presents a parameter tuned deep learning model for reactive power optimization (PTDL-RPO) in distributed systems. The proposed model involves a bi-directional long short term memory (Bi-LSTM) technique for learning the non-linear complex relationships among the system characteristics and reactive power control solution. The past DG data are used for training the Bi-LSTM model for learning the relativity among the system characteristics and reactive power control solution. Besides, for improving the performance of Bi-LSTM method, search and rescue optimization algorithm (SROA) was employed as a hyperparameter tuning technique. The design of optimal BiLSTM model for RPO of the DNs show the novelty of the study. For examining the enhanced outcomes of the PTDL-RPO technique, a wide range of simulations take place and the experimental outcome highlighted the enhanced performance of the PTDL-RPO technique.