Temporally-Adaptive Robust Data-Driven Sparse Voltage Sensitivity Estimation for Large-Scale Realistic Distribution Systems With PVs

This letter proposes a new robust data-driven sparse voltage sensitivity estimation approach for large-scale distribution systems with PVs. It has a high statistical efficiency to mitigate the impacts of PV stochasticity and unknown measurement noise under various system operating conditions. A new adaptively-weighted l(1) sparsity-promoting regularization is developed, exploiting the temporal characteristic of time-varying sensitivities for better accuracy. The l(2) regularization is used to mitigate collinearity impacts. The Huber loss function and a concomitant scale estimate are adopted to mitigate the impacts of unknown and non-Gaussian noise. These techniques are implemented in a fast recursive parallel computing framework. The proposed estimator is tested by quasi-static time series simulations of a large three-phase unbalanced system with PVs and various discrete time-delayed control devices. Results validate the superior robustness and efficiency of the proposed estimator over existing alternatives.

Automated summary

This paper proposes a new data-driven approach for estimating sparse voltage sensitivity in large-scale distribution systems with PVs. The approach effectively mitigates the impact of PV stochasticity and unknown measurement noise under different system operating conditions. The proposed method includes adaptively-weighted l(1) sparsity-promoting regularization, l(2) regularization, Huber loss function, and concomitant scale estimate, implemented in a fast recursive parallel computing framework. Simulation results demonstrate the robustness and efficiency of the proposed estimator compared to existing alternatives.