Three-Phase State Estimation

Models the 3x3 primitive impedance matrix for each line segment, capturing self and mutual impedances between phases A, B, and C. This is critical because distribution lines are rarely transposed, causing phase-to-phase inductive coupling that significantly impacts voltage profiles. The formulation includes off-diagonal terms in the bus admittance matrix, enabling accurate calculation of zero-sequence and negative-sequence currents during unbalanced faults or heavy single-phase loading.

Explicitly represents single-phase and two-phase tap lines that branch off the main three-phase trunk. Unlike transmission state estimation which assumes balanced three-phase topology, distribution DSSE must handle missing phases natively:

• Single-phase laterals serve residential neighborhoods
• V-phase configurations supply rural loads
• Open-delta transformer banks create phase asymmetry The measurement Jacobian is constructed with variable dimension per bus, accommodating 1, 2, or 3 phase nodes at each electrical point.

Replaces the balanced constant-power PQ model with per-phase load allocation. Distribution loads are inherently unbalanced due to:

• Uneven distribution of single-phase customers across phases
• Distributed generation injecting asymmetrically on specific phases
• Time-varying phase imbalances from EV charging State estimation incorporates pseudo-measurements derived from AMI data aggregated per phase, weighted by customer count and historical load profiles to initialize the per-phase injection estimates.

Represents transformer banks with their winding connection matrices rather than simplified tap ratios. Key configurations include:

• Delta-Grounded Wye: Common in distribution substations, introduces 30° phase shift
• Open-Wye Open-Delta: Serves single-phase laterals from three-phase primaries
• Grounded Wye-Grounded Wye: Provides neutral path for zero-sequence currents The primitive admittance matrix is built from short-circuit test data, then transformed by the connection matrix to derive the terminal bus admittance contribution.

Extends the state vector to include neutral conductor voltages and ground return currents. In multi-grounded neutral systems, the earth acts as a parallel conductor, creating a four-wire equivalent circuit. The Carson equations model the frequency-dependent earth return impedance, critical for:

• Accurate zero-sequence impedance calculation
• Detecting high-impedance faults through neutral current residuals
• Modeling stray voltage conditions This adds rows to the measurement Jacobian for neutral current magnitude constraints.

Defines the state vector x as the complex voltage at each phase node rather than positive-sequence magnitude and angle. For an N-bus network with mixed phase configurations, the state vector dimension is 3N minus missing phases. The measurement function h(x) computes:

• Three-phase real and reactive power injections per bus
• Per-phase current magnitude on each line segment
• Voltage magnitude for each present phase This formulation enables direct comparison with smart meter voltage readings and PMU phasor data without sequence transformation.

The foundational statistical engine for three-phase state estimation. WLS minimizes the sum of weighted squared residuals between measured and estimated values. Weights are inversely proportional to measurement error variance, giving high-precision sensors like PMUs more influence than pseudo-measurements. In the three-phase context, the WLS objective function expands to account for phase-voltage magnitudes, three-phase power injections, and current flow measurements simultaneously. The iterative Gauss-Newton method solves the resulting non-convex optimization problem by linearizing the three-phase power flow equations at each iteration.

A critical pre-processing step that determines whether a unique three-phase state estimation solution exists. The algorithm analyzes the measurement Jacobian matrix to identify observable islands and unobservable branches. In unbalanced distribution networks, observability is more complex than transmission systems because single-phase laterals and missing phase measurements create partial observability conditions. Numerical observability analysis evaluates the rank of the three-phase gain matrix, while topological observability traces measurement paths through the network graph to ensure every complex bus voltage can be uniquely determined.

Synthetic data points that supplement real-time sensor data to achieve numerical observability in under-instrumentated distribution grids. Common sources include:

• Historical load profiles scaled by feeder-level measurements
• Forecasted distributed generation from rooftop solar PV systems
• Zero-injection constraints at unmonitored junction nodes In three-phase estimation, pseudo-measurements must be generated per-phase to capture unbalanced loading conditions. Their high variance (low weight) reflects their inherent uncertainty compared to physical sensor data.

The mathematical core of the WLS algorithm, defined as G = Hᵀ·R⁻¹·H, where H is the measurement Jacobian and R⁻¹ is the inverse covariance matrix. The gain matrix represents the information content of all measurements mapped to the state variables. Its condition number dictates numerical stability—ill-conditioned gain matrices cause convergence failure. In three-phase networks, mutual coupling between phases and high R/X ratios create matrices that are more prone to ill-conditioning than transmission equivalents, often requiring orthogonal decomposition methods for robust inversion.

Statistical techniques that identify gross measurement errors before they corrupt the three-phase state estimate. The primary method is the Normalized Residual Test, which flags measurements whose residuals exceed a statistical threshold (typically 3σ). The Chi-Square test evaluates the overall fit by comparing the sum of squared residuals against a critical value. In unbalanced networks, bad data detection must account for phase-specific anomalies—a faulted current transformer on Phase A should not cause rejection of valid Phase B and C measurements from the same physical device.

A module that translates the physical node-breaker model of a substation into a computational bus-branch model by processing real-time switch and breaker statuses. For three-phase state estimation, the topology processor must correctly map single-phase, two-phase, and three-phase connectivity. Incorrect breaker status—such as a closed switch reported as open—creates topology errors that manifest as large normalized residuals. Advanced processors cross-validate switch status against analog measurements to detect and correct such errors before estimation begins.