CNN-CNN Methodology for Fault Assessment and Categorization in an Active Distribution System Integrated with Distributed Generations

Le résumé fourni par la source

In the period of active implementation of the active distribution and the creation of Distributed Generation (DG), the number of locations that are associated with faults has acquired a significantly difficult-to-detect aspect due to the floating and uncontrollable nature of the power that is injected. The paper gives a hybrid Convolution Neural Network (CNNCNN) implementation of a multi-fault study of the 6-bus IEEE having decentralized RES. The specified model is going to be set up in a way that it will carry out three functionalities primarily as follows: by the kind of fault, by the line of the fault and the exact position of the fault. Depending on the measurement of voltage and current values in critical locations within the grid, the initial CNN learns necessary spatial and temporal affairs, than the second CNN introduces criticality in the identification of faults. This multilayering increases the sensitivity of the subtle variations in the signal dynamics occasioned by the different possible types of fault and grid combinations. According to the various cases, i.e., balanced and unbalanced faults, different ground resistances, and dynamic conditions of continuous RES, significant tests of simulation on PSCAD were performed. The framework showed strong versatility and consistent dynamics in the fluctuation of fault types and system (state). The CNN-CNN is competent in relation to fault diagnosis, boundary identification of lines, as well as, the position of the line, which plays a large role towards real-time monitoring of smart grids. It is efficient and stable, and will be sure to be used in future AI-controlled protection systems of a contemporary distribution network.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.