Multi Time-scale Imputation aided State Estimation in Distribution System

TL;DR

This paper introduces a multi-task Gaussian process framework to effectively impute and aggregate unevenly sampled, multi-scale sensor data in distribution systems, enhancing situational awareness and state estimation accuracy.

Contribution

It proposes a novel multi-task Gaussian process method for multi-scale data imputation, improving distribution system state estimation over traditional linear interpolation.

Findings

Outperforms linear interpolation in imputation accuracy.

Provides confidence bounds on data estimates.

Enhances state estimation in IEEE 37 bus system.

Abstract

With the transition to a smart grid, we are witnessing a significant growth in sensor deployments and smart metering infrastructure in the distribution system. However, information from these sensors and meters are typically unevenly sampled at different time-scales and are incomplete. It is critical to effectively aggregate these information sources for situational awareness. In order to reconcile the heterogeneous multi-scale time-series data, we present a multi-task Gaussian process framework. This framework exploits the spatio-temporal correlation across the time-series data to impute data at any desired time-scale while providing confidence bounds on the imputations. The value of the imputed data for distribution system operation is illustrated via a matrix completion based state estimation strategy. Results on the IEEE 37 bus distribution system reveals the superior performance of the proposed approach relative to linear interpolation approaches.