Geometric Matrix Completion for Missing Data Estimation in Power Distribution Systems

With the increasing penetration of renewable energy, the reverse power flow and voltage violation become more and more frequent in power distribution networks. However, due to the lack of measurements in the system, it is difficult to observe the state of the whole system under low-observability scenarios using traditional techniques, which may compromise system operations. To address these issues, we come up with a missing data estimation method based on geometric matrix completion with neural networks to recover data matrix of distribution networks. The proposed method generally has good performance, especially in low-observability scenarios where traditional methods can hardly operate. Numerical results using IEEE 33- and 300-bus illustrate to the effectiveness and efficiency of the proposed approach.