Deep data fusion model for risk perception and coordinated control of smart grid

Abstract

This paper presents a deep data fusion model for risk perception and coordinated control in a regional power system control center. A knowledge learning data fusion approach has been used to find an efficient state representation based on prior knowledge from cross-domain energy management systems. In particular, a kernel principal components analysis technique is presented for nonlinear dimensionality reduction of knowledge learning. The control strategy we study is based on cross-domain global optimization approach, which regards the contingencies and control actions of mutual backup systems as constraints. The objective function is defined as the product of cross-domain assessment and control factors. The method for obtaining optimal solution is given by interior point code. To show the applicability, different machine learning method has been studied. The experimental results show that the proposed knowledge learning approach consistently outperforms the traditional machine learning method. In addition, the proposed coordinated control approach is verified effective on large-scale smart grid decision support system for East China project.