Abstract
Fast power system state estimation (SE) solution is of paramount importance for achieving real-time decision making in power grid operations. Semidefinite programming (SDP) reformulation has been shown effective to obtain the global optimum for the nonlinear SE problem, while suffering from high computational complexity. Thus, we leverage the recent advances in nonconvex SDP approach that allows for the simple first-order gradient-descent (GD) updates. Using the power system model, we can verify that the SE objective function enjoys nice properties (strongly convex, smoothness) which in turn guarantee a linear convergence rate of the proposed GD-based SE method. To further accelerate the convergence speed, we consider the accelerated gradient descent (AGD) extension, as well as their robust versions under outlier data and a hybrid GD-based SE approach with additional synchrophasor measurements. Numerical tests on the IEEE 118-bus, 300-bus and the synthetic ACTIVSg2000-bus systems have demonstrated that FGD-SE and AGD-SE, can approach the near-optimal performance of the SDP-SE solution at significantly improved computational efficiency, especially so for AGD-SE.
Introduction
State Estimation (SE) serves as the "brain" of power grid operations, converting noisy raw measurements—such as voltage magnitudes and power flows—into the actual complex voltage phasors of the grid. Traditionally, the Gauss-Newton (GN) method has been the industry standard. However, GN is prone to divergence, particularly under heavy grid loading or in the presence of "bad data" (outliers).
While Semidefinite Programming (SDP) reformulations offer a convex alternative capable of finding global optima, they scale poorly ($O(N^{4.5})$), making them impractical for large-scale systems. This research introduces a high-speed approach that reformulates the SE problem as a nonconvex SDP. By leveraging Burer-Monteiro factorization and accelerated gradient descent, the proposed method achieves the near-global optimality of convex SDP solvers at a fraction of the computational cost, enabling real-time monitoring of large-scale grids.
Proposed Methodology
The core of the approach lies in reducing the dimensionality and complexity of the SDP problem:
- Burer-Monteiro Factorization: Instead of solving for a high-dimensional $N \times N$ matrix $V$, the problem is reformulated by representing $V$ as $V = uu^H$, where $u$ is a vector. This eliminates the computationally expensive positive semi-definite (PSD) constraint, transforming the problem into an unconstrained (though nonconvex) optimization over $u$.
- Factored Gradient Descent (FGD-SE): Simple first-order gradient descent is applied to this factored objective. The authors demonstrate that for power systems, the objective function is "well-behaved" within realistic voltage ranges (typically $\pm 5\%$). This property guarantees a linear convergence rate.
- Nesterov Acceleration (AGD-SE): To address the ill-conditioned nature of power flow equations, a momentum-based acceleration scheme is incorporated. This significantly reduces the iterations required to reach a solution compared to standard FGD.
Robustness and Scalability
To handle outliers resulting from sensor failures or cyber-attacks, the solver utilizes a "hard thresholding" operator. In each iteration, measurements with the highest mismatch errors are discarded, preventing bad data from biasing the estimate. Furthermore, the framework naturally incorporates high-accuracy synchrophasor data (PMUs) as a linear regularization term.
The algorithms were evaluated using standard benchmarks: the IEEE 118-bus, IEEE 300-bus, and the synthetic ACTIVSg2000-bus system (representing a footprint similar to the Texas grid).
Key Results
The numerical tests highlight significant improvements in both speed and reliability:
- Computational Efficiency: On the 300-bus system, the AGD-SE solver reached a solution in just 0.277 seconds, a 350x speedup over standard SDP solvers (98.5 seconds). For the 2000-bus system, where standard SDP solvers often fail, AGD-SE finished in 35.7 seconds.
- Convergence Reliability: In the 2000-bus case, the traditional Gauss-Newton method converged only 30% of the time, whereas the proposed FGD and AGD methods achieved 100% convergence.
- Outlier Rejection: The robust version (RAGD) significantly outperformed classical estimators like Least Absolute Value (LAV), achieving a 72% success rate in identifying outliers in the 2000-bus system.
Discussion and Future Work
The research demonstrates that nonconvexity in power systems is manageable because physical constraints restrict the search space to a region where the problem behaves predictably. By moving from matrix-space to vector-space, the computational burden per iteration is reduced to simple vector-matrix multiplications.
While AGD-SE performs exceptionally well, a rigorous theoretical proof of its convergence rate for this specific nonconvex objective remains an open challenge. Future work will explore incorporating hard constraints, such as strict voltage or line flow limits, directly into the gradient descent framework to further enhance the model's practical utility.
Summary updated