An End-to-End Learning Framework for Real-Time Decomposition and Classification of Aggregate Measurements at Substations

Understanding the composition of measured demand at the substation level is essential for understanding and modeling the highly complicated behaviors of numerous loads, which is very challenging due to their dynamic nature. This article presents a flexible, real-time decomposition framework that separates individual load sources from aggregate substation measurements while providing interpretability through load-type classification. Our approach employs an enhanced dual-path neural network to adaptively decompose load signals, supported by an encoder–decoder-based source-counting mechanism that estimates the number of load sources dynamically. To accommodate varying load counts, we include a multilayer convolutional module for flexible separation mask application. Extensive case studies with aggregate measurements containing different numbers of load sources demonstrate our model’s adaptability, achieving a mean-squared error reduction of up to 15% and a signal-to-distortion ratio improvement of 10% over baseline methods, with computational times suitable for real-time implementation. Comparative studies further show that our solution achieves consistent load-type classification accuracy of over 95%, enabling reliable insights for operational decisions in grid management.