Enhanced Tensor-Completion Distribution System State Estimation Considering Measurement Outliers and Noises

The tensor completion state estimation (TCSE) based on nuclear norm minimization with the linearized power flow constraints is effective in distribution system state estimation (DSSE) under low-observable conditions. However, measurement outliers and noises compromise the low-rank property of the tensor and thus degrade the performance of TCSE. In low-observable DSSE, measurement outliers cannot be identified based on the redundancy. This paper proposes an enhanced TCSE (ETCSE) framework, aiming to improve the estimation accuracy under measurement outliers and noises in DSSE. First, a prior-stage sparse optimization-based bad data detection (SO-BDD) is formulated to identify outliers and recover true measurements using historical measurements. Then, a rotational invariant estimator (RIE) is developed to clean measurement noises based on random matrix theory. Last, the recovered and cleaned measurements are used by the TCSE to estimate the voltage. Numerical results on the unbalanced three-phase IEEE 123-node test feeder demonstrate that the proposed ETCSE framework can greatly improve the estimation accuracy under the measurement outliers and noises.