Unifying Power Flow and Electromagnetic Transient Modeling

• SALT is a fully physics-based steady-state solver for transmission analysis.Source: Highlights · p. 2
• SALT achieves EMT fidelity at the computational efficiency of power flow.Source: Highlights · p. 2
• PF reported 75% fewer line violations than SALT in contingency analysis.Source: Highlights · p. 2

• PF steady-state solutions can miss violations detected by EMT, as the two tools produce different results under the same grid conditions.In the illustrated 39-bus case, PF reports 0.995 PU at bus 9 while EMT reports 0.923 PU after a 13% load increase.Source: Figure 1 · p. 3
• PF's static device abstractions differ from EMT models derived from first principles, while its slack generator absorbs mismatches to sustain 60 Hz.Source: §1 · p. 3
• Reduced inertia and reserve margins make PF–EMT disparity a grid-security concern and motivate steady-state analysis consistent with EMT.PF–EMT model inconsistencies also require extra computation when initializing EMT simulations.Source: §1 · pp. 3–4
• EMT steady-state computation is costly because microsecond timesteps may continue until slow device dynamics settle over minutes to hours.Harmonic Balance also inherits transient-response costs when modeling dynamic devices.Source: §1 · p. 4
• SALT constructs algebraic steady-state models unified with physics-based EMT device models by exploiting power-device structure in single-harmonic, balanced systems.Its frequency-domain solution avoids stepping through time.Source: §1 · p. 4
• SALT solutions are within local truncation error of settled EMT waveforms, with around a 1 × 10^6 times speedup and PF-like computational complexity.In a 39-bus contingency analysis, PF-based methods reported 75% fewer rated-line violations, as well as fewer Q-limit and voltage violations.Source: §1 · p. 4

2. Literature Review

• Accurate steady-state analysis is difficult because, in theory, it requires the EMT response at infinite time; prior work spans PF enhancements, faster EMT, and co-simulation.Source: §2 · p. 4

2.1. PF Enhancements

• PF enhancements add phase detail, distributed slack, dynamic controls, post-contingency response, or voltage-dependent ZIP loads, but retain abstract device models.SALT instead uses physics-based steady-state device models.Source: §2.1 · p. 4

2.2. EMT Speedup

• EMT speedups use co-simulation or decoupled, parallel matrix solves, but still simulate transients; TSA uses larger steps while sacrificing fidelity.Source: §2.2 · p. 4

2.3. Harmonic Balance: Frequency-domain and EMT Co-simulation

• Harmonic Balance bridges time and frequency domains through DFT, but conversion overhead and dynamic-device memory can reintroduce EMT simulation costs.Prior work either omits device memory or applies DFT to settled time-domain simulations.Source: §2.3 · p. 4

• SALT forms a frequency-domain steady-state framework from EMT state-space equations to produce an operating point consistent with grid physics.Source: §3 · p. 4
• The steady state is the settled simulation-to-infinity response, represented in the frequency domain because transients have decayed into periodic waveforms.Source: §3 · p. 4
• SALT reduces the transmission-system spectrum to one system frequency because non-fundamental frequencies decay aggressively, and adds a reference bus for phase and full rank.The frequency-domain formulation includes complex bus voltages, frequency-dependent admittance, and nonlinear device models.Source: §3 · p. 5
• SALT builds component-level frequency-domain device models, replacing derivatives with jω_s0 for linear terms while addressing nonlinear dynamic couplings that generally lack closed forms.The stated approach contrasts with Harmonic Balance's DFT conversion and time-domain simulation overhead.Source: §3 · p. 5

4.1. Definitions

• EMT device models share a dq-structure in which a dq-transform maps three-phase network quantities to dq-quantities that drive an internal circuit.The dq-transform projects sinusoidal quantities into a synchronous reference frame; alternative device formulations map through linear transformations of the dq-quantities.Source: §4.1 · p. 5
• SALT assumes a single-harmonic, balanced transmission steady state, where three-phase quantities have equal magnitude and frequency and are 120° apart.The paper motivates these conditions by harmonic attenuation across long lines and balance enforcement to minimize loss.Source: §4.1 · p. 5

4.2. Methodology

• SALT integrates classical steady-state dq-machine results with harmonic-balance theory to provide a generalized frequency-domain framework for EMT devices.The resulting models are characterized by a memoryless dq-circuit connected to the transmission grid via a phase shift.Source: §4.2 · pp. 6–7
• At steady state, synchronous rotation makes dq-quantities constant, allowing transients to decay and the internal dq-circuit to be solved as a memoryless system.For DC inputs, capacitors are opened and inductors are shorted.Source: §4.2 · p. 6
• SALT connects the memoryless dq-circuit to the frequency-domain grid through a phase shift, bypassing the dq-transform and DFT used by conventional harmonic-balance methods.Its construction DC-constrains dq-inputs, collapses dynamic elements into memoryless components, and interfaces with the network via phase shift.Source: §4.2 · pp. 6–7

5. Results

• SALT matches EMT steady-state fidelity with a 1 × 10^6-times speedup, while PF methods miss security violations and introduce initialization errors.The comparison includes traditional PF and enh-PF, which adds distributed slack and static ZIP loads for abnormal operating conditions.Source: §5 · p. 8

5.1. Disparities in Device Modeling

• SALT generator and load responses track EMT, while PF misses generator droop and load voltage dependence and enh-PF retains device-modeling errors.Enh-PF captures droop through distributed slack but omits losses and nonlinearities; its ZIP load fit still misses air-gap power transfers and loss.Source: §5.1 · p. 8, Figure 4 · p. 8, Figure 5 · p. 8

5.2. Accuracy of System-Level Solutions

• SALT/EMT deviations stayed near numerical precision, while PF errors grew with system size, reaching 0.187 PU, 129 degrees, and 0.786 Hz.SALT/EMT deviations were 1 × 10−5 PU in voltage magnitude, 1 × 10−3 degrees in phase, and 1 × 10−5 Hz in frequency; enh-PF errors were around 0.01 PU, 0.4 degrees, and 0.0014 Hz.Source: §5.2 · p. 8, Figure 6 · p. 9

5.3. Impact on Security Assessment

• SALT violations exactly agreed with EMT; PF reported 6 fewer rated-line violations, while convergence-matched counts reverse its Q-limit implication.PF mostly overstated grid stress and failed to converge to a feasible solution for the largest-generator loss; its under-reported rated-line violations reflected mischaracterized flows near the slack.Source: Table 1 · p. 9, §5.3 · p. 9

5.4. Errors in EMT Initialization

• SALT initialization needed no transient correction, while PF required at least 100,000 timesteps to settle, and enh-PF did not shorten settling.The PF initialization burden scaled with network size; enh-PF's closer initial frequency did not prevent a first swing of the same magnitude as PF.Source: Table 2 · p. 10, §5.4 · p. 10

5.5. Computation and Runtime

• SALT's total runtime reached 2.7× PF and up to 7× PF's compute time on large benchmarks, despite maintaining 3-4 NR iterations.PF grew to 6 NR iterations; SALT's additional cost was attributed to more sophisticated device equations, frequency-dependent branches, and a larger matrix.Source: §5.5 · pp. 10–11, Table 3 · p. 11

• SALT derives frequency-domain equivalents for dynamic nonlinear devices and embeds them into transmission network analysis.The framework directly solves EMT steady-state systems while maintaining the computational speed of PF.Source: §7 · p. 11
• PF-based solvers reported 75% fewer rated line violations, 18% fewer Q-limit violations, and 7% fewer voltage violations than SALT in contingency analysis.SALT captures post-outage violations consistently.Source: §7 · p. 11
• EMT simulation requires 1 × 10^6 times more runtime, while SALT maintains runtimes on the same order-of-magnitude as PF.The passage attributes EMT's cost to hundreds-of-thousands of timesteps.Source: §7 · p. 11

• The authors used Anthropic Opus 5 (2026) to generate the graphical abstract, then reviewed and edited it and took responsibility for its content.Source: §8 · p. 11

• Tina Gao is credited with methodology and original drafting; Aayushya Agarwal with investigation and review; Timothy McNamara with conceptualization; and Lawrence Pileggi with conceptualization, review, and funding acquisition.Source: CRediT authorship contribution statement · p. 11