State and Topology Estimation for Unobservable Distribution Systems using Deep Neural Networks

Area of Science:

• Electrical Engineering
• Power Systems
• Artificial Intelligence

Background:

• Reconfigurable distribution networks present challenges for time-synchronized state estimation due to limited real-time observability.
• Accurate state estimation is crucial for reliable grid operation and control.

Purpose of the Study:

• To develop a deep learning (DL)-based approach for topology identification (TI) and unbalanced three-phase distribution system state estimation (DSSE).
• To address challenges posed by incomplete real-time observability from synchrophasor measurement devices (SMDs).

Main Methods:

• Formulation of a DL-based methodology using two deep neural networks (DNNs) for time-synchronized TI and DSSE.
• Development of a data-driven approach for optimal SMD placement.
• Testing robustness against non-Gaussian noise in SMD measurements.

Main Results:

• The proposed DNN-based approach achieves accurate TI and DSSE in incompletely observed systems.
• The data-driven SMD placement strategy facilitates reliable TI and DSSE.
• The DL-based DSSE demonstrates superior accuracy compared to conventional methods, requiring fewer SMDs.

Conclusions:

• Deep learning offers a powerful solution for time-synchronized state estimation in reconfigurable distribution networks.
• The proposed methodology enhances accuracy and efficiency, even under noisy conditions and with limited measurements.