M.S. AAI Capstone Chronicles 2024

ELECTRICITY DISTRIBUTION TOPOLOGY CLASSIFICATION

16

increased model complexity, precisely without needing to exhaustively augment the existing dataset or apply the dropout and normalization to the degree to which the current approaches are taken. Each model design and approach provides a viable method. Still, additional data either confirm the efficacy of the architecture or highlight the flaws in the approach if the model does not improve with the larger dataset. To productionize the approach, the continual intake of meter data could help to retrain the model with more extensive and current usage data. This constant influx of data into the training dataset would enhance the demand for deep-learning approaches.

107