[Paper Review] Forecasting-Based State Estimation for Three-Phase Distribution Systems with Limited Sensing.

Roel Dobbe, Werner van Westering+4 more|arXiv (Cornell University)|Jun 15, 2018

Power System Optimization and Stability9 references5 citations

TL;DR

This paper proposes a Bayesian state estimation method for three-phase distribution networks that fuses limited real-time PMU measurements with voltage predictions from linear forecast models of unbalanced 3-phase power flow. By formulating estimation as a linear least-squares problem, it enables high-resolution state updates with a priori and a posteriori uncertainty quantification, validated on an IEEE benchmark and a real Alliander network testbed.

ABSTRACT

State Estimation is an essential technique to provide observability in power systems. Traditionally developed for high-voltage transmission networks, state estimation requires equipping networks with many real-time sensors, which remains a challenge at the scale of distribution networks. This paper proposes a method to complement a limited set of real-time measurements with voltage predictions from forecast models. The method differs from the classical weighted least-squares approach, and instead relies on Bayesian estimation formulated as a linear least squares estimation problem. We integrate recently developed linear models for unbalanced 3-phase power flow to construct voltage predictions as a linear mapping of load predictions. The estimation step is a linear computation allowing high resolution state estimate updates, for instance by exploiting a small set of phasor measurement units. Uncertainties can be determined a priori and smoothed a posteriori, making the method useful for both planning, operation and post hoc analysis. The method is applied to an IEEE benchmark and on a real network testbed at the Dutch utility Alliander.

Motivation & Objective

• Address the challenge of observability in distribution networks due to limited real-time sensing infrastructure.
• Overcome the high cost and impracticality of deploying dense sensor networks in distribution systems.
• Enable accurate state estimation using only a sparse set of PMUs by integrating forecasted voltage data.
• Provide both real-time operational state estimates and post-hoc analysis capabilities with quantified uncertainty.
• Develop a computationally efficient, linear estimation framework compatible with existing distribution system models.

Proposed method

• Formulate state estimation as a Bayesian linear least-squares problem, enabling efficient computation and uncertainty propagation.
• Use recently developed linear models for unbalanced three-phase power flow to map load forecasts to voltage predictions as a linear transformation.
• Integrate real-time PMU measurements with forecasted voltages through a linear estimation framework to produce high-resolution state estimates.
• Propagate uncertainties a priori from load and forecast models and smooth them a posteriori using measurement data.
• Apply the method to both an IEEE benchmark system and a real-world testbed at the Dutch utility Alliander for validation.
• Ensure computational efficiency by avoiding iterative solvers, relying instead on direct linear algebraic solutions.

Experimental results

Research questions

• RQ1 Can a limited set of PMU measurements be effectively combined with forecasted voltages to achieve accurate state estimation in three-phase distribution networks?
• RQ2 How does the proposed Bayesian linear estimation framework compare to classical weighted least-squares in terms of accuracy and uncertainty quantification?
• RQ3 To what extent can linearized three-phase power flow models accurately predict voltage behavior for use in state estimation?
• RQ4 Can the method support both real-time operation and post-hoc analysis with consistent uncertainty representation?
• RQ5 How does the method perform on real-world distribution networks with unbalanced loading and complex topology?

Key findings

• The proposed method achieves high-resolution state estimation using only a sparse set of PMU measurements, significantly reducing the need for dense sensor deployment.
• The integration of forecasted voltages with real-time measurements enables accurate state estimates even with limited sensing infrastructure.
• Uncertainty is quantified a priori from load forecasts and refined a posteriori using measurements, supporting both operational and post-analysis use cases.
• The method was successfully validated on both an IEEE benchmark system and a real distribution network testbed at Alliander, demonstrating practical applicability.
• The linear estimation framework ensures computational efficiency, enabling fast updates suitable for real-time operation.
• The approach maintains accuracy across unbalanced three-phase conditions, demonstrating robustness to network complexity.