2. Modeling of Power-System Components

Introduction and Context

Power-system analysis is meaningful only when physical equipment is represented by mathematical models that preserve the electrical behavior relevant to the study. A generator, transformer, transmission line, cable, load, or interconnected network contains electromagnetic, thermal, mechanical, and control phenomena operating over widely different time scales. No single model is suitable for every investigation. A transmission line may be represented by a simple series reactance in a preliminary power-flow study, by a distributed-parameter model in travelling-wave analysis, and by a frequency-dependent model in electromagnetic-transient simulation.

The central engineering task is therefore model selection: deciding which physical effects must be retained, which may be neglected, and how the resulting parameters can be obtained and validated. This chapter develops steady-state and dynamic representations of principal power-system components. It also introduces per-unit normalization, parameter estimation, network equivalents, and the formation of bus admittance and impedance matrices. These ideas provide the mathematical foundation for load-flow, short-circuit, transient-stability, protection, and operational studies.

Learning Objectives

After studying this chapter, the student should be able to:

1. Explain the purpose, assumptions, and limitations of component models used in power-system analysis.

2. Derive and interpret steady-state models of synchronous machines, transformers, lines, cables, and loads.

3. Distinguish transient, subtransient, electromechanical, and electromagnetic models.

4. Convert nameplate and test data into per-unit parameters and estimate uncertain parameters from measurements.

5. Formulate bus admittance and bus impedance representations of an interconnected network.

6. Select an appropriate model for a specified engineering study and assess the consequences of simplification.

7. Recognize practical, regulatory, numerical, and data-quality issues associated with model development.

Modeling as an Engineering Approximation

A model is a structured approximation of a physical system. Its quality is not determined by complexity alone; it is determined by whether it reproduces the quantities of interest over the relevant operating range and time scale. A useful model specifies:

• the state variables, algebraic variables, inputs, and outputs;
• the parameter values and their sources;
• the assumptions under which the equations are valid; and
• the limits beyond which the model should not be used.

Power-system models are commonly classified as steady-state, dynamic phasor, electromechanical, or electromagnetic-transient models. Steady-state phasor models assume sinusoidal operation at one frequency and neglect transients. Electromechanical models retain rotor-angle and speed dynamics but represent stator electrical quantities algebraically. Electromagnetic-transient models retain rapidly changing fluxes, currents, and voltages, often using differential equations in the time domain.

The per-unit system is used extensively because it simplifies equations, reduces numerical scaling problems, and makes parameters of machines and transformers comparable. For a selected three-phase base apparent power \(S_B\) and line-to-line base voltage \(V_B\),

\[

I_B=\frac{S_B}{\sqrt{3}V_B},\qquad Z_B=\frac{V_B^2}{S_B},\qquad Y_B=\frac{1}{Z_B}.

\]

A physical quantity divided by its base value is its per-unit value. When voltage bases on two sides of a transformer are selected in the turns ratio, the per-unit impedance is numerically unchanged across the ideal transformer. This is one reason why transformers can often be eliminated from the network topological formulation after consistent base selection.

Synchronous Machines: From Phasor Representation to Dynamic Equations

A synchronous generator converts mechanical input into electrical output through a rotating magnetic field. In steady-state balanced analysis, the machine is represented by an internal generated voltage behind an impedance. For a cylindrical-rotor machine, a commonly used generator equation is

\[

\mathbf{E}=\mathbf{V}+\mathbf{I}(R_a+jX_s),

\]

where \(\mathbf{E}\) is the internal generated emf, \(\mathbf{V}\) is terminal voltage, \(\mathbf{I}\) is stator current, \(R_a\) is armature resistance, and \(X_s\) is synchronous reactance. Since resistance is usually small, it may be neglected in preliminary studies. The corresponding internal impedance is then \(jX_s\).

For salient-pole machines, direct- and quadrature-axis reactances differ. The two-reaction model uses \(X_d\) and \(X_q\), with rotor angle and current resolved into d- and q-axis components. This model is important for accurate power-angle calculations, reactive-power capability assessment, and stability studies.

The steady-state power-angle relation for a simplified cylindrical-rotor machine connected to an infinite bus through reactance \(X\) is

\[

P_e=\frac{EV}{X}\sin\delta,

\]

where \(\delta\) is the angle of the internal emf relative to the bus voltage. This expression illustrates the physical connection between torque production, rotor angle, and synchronism. It is not valid without qualification when resistance, saliency, saturation, or severe disturbances are significant.

Dynamic machine models

During a disturbance, rotor speed and flux linkages change. The swing equation describes the electromechanical motion:

\[

\frac{2H}{\omega_s}\frac{d^2\delta}{dt^2}=P_m-P_e-D\frac{d\delta}{dt},

\]

where \(H\) is the inertia constant in seconds, \(\omega_s\) is synchronous angular speed, \(P_m\) is mechanical input power, \(P_e\) is electrical output power, and \(D\) represents damping. In state-variable form, rotor angle and speed deviation are states. The equation is the basis of transient-stability analysis.

Different dynamic models are appropriate to different studies. A classical model uses a constant internal emf behind transient reactance and is suitable for introductory transient-stability calculations. A one-axis or two-axis model includes transient emf states. More detailed models include subtransient effects, damper windings, saturation, field-voltage dynamics, excitation systems, turbine governors, and power-system stabilizers.

The reactance hierarchy generally satisfies

\[

X_d''<X_d'<X_d,

\]

where the superscripts denote subtransient, transient, and synchronous values. Immediately after a fault, subtransient reactance determines the large initial current; after several cycles, transient reactance becomes more relevant; in sustained steady state, synchronous reactance is used. Confusing these reactances is a frequent source of incorrect fault-current and stability results.

Machine parameter data are obtained from open-circuit and short-circuit tests, sudden-short-circuit tests, standstill frequency-response tests, and manufacturer identification procedures. Field tests must be conducted under controlled conditions because temperature, magnetic saturation, frequency, and operating point affect the inferred parameters.

Transformers and the Meaning of Equivalent Circuits

An ideal transformer changes voltage and current levels according to its turns ratio while transferring apparent power. A practical transformer has winding resistance, leakage reactance, magnetizing reactance, and core-loss resistance. Its approximate equivalent circuit contains a series impedance

\[

Z_{eq}=R_{eq}+jX_{eq}

\]

and a shunt branch representing excitation current and core loss. The series branch is dominant in load-flow and short-circuit studies, whereas the shunt branch becomes important in no-load current, voltage regulation, energization, and harmonic studies.

Transformer parameters are commonly determined from open-circuit and short-circuit tests. The open-circuit test estimates the shunt branch because the applied voltage is near rated value while current is small. The short-circuit test estimates series impedance because a small voltage produces rated current with the secondary shorted. If the measured short-circuit voltage is \(V_{sc}\) at rated current, the per-unit magnitude of impedance is approximately \(V_{sc}/V_{rated}\).

For a two-winding transformer with off-nominal complex tap ratio \(t\), the transformer modifies both series admittance and bus injections. A phase-shifting transformer may be represented by \(t=\tau e^{j\phi}\). Such a representation is essential in power-flow studies because tap position and phase shift affect active-power transfer and voltage control.

The ideal transformer is not merely a circuit detail. It is also a network-modeling device that changes the relationship between bus voltages and branch currents. Errors in tap conventions, phase-shift signs, or base voltages can produce apparently plausible but physically wrong power-flow results.

Transmission Lines and Cables

A transmission line is characterized by distributed series impedance and shunt admittance. For a line of length \(l\), resistance and inductance arise from conductors, while capacitance results from electric fields between conductors and between conductors and ground. Leakage conductance is generally small for overhead lines but may be more significant in cables.

For short lines, shunt capacitance may be neglected and the model is simply a series impedance. For medium-length lines, the nominal-\(\pi\) model places half the total shunt admittance at each end:

\[

Z=zl,\qquad Y=yl,

\]

where \(z\) and \(y\) are per-unit-length series impedance and shunt admittance. For long lines, the distributed model uses the propagation constant and characteristic impedance:

\[

\gamma=\sqrt{zy},\qquad Z_c=\sqrt{\frac{z}{y}}.

\]

The ABCD parameters of a uniform line are

\[

A=D=\cosh(\gamma l),\quad B=Z_c\sinh(\gamma l),\quad C=\frac{1}{Z_c}\sinh(\gamma l).

\]

The nominal-\(\pi\) model is usually adequate for power-frequency load flow over moderate distances, whereas the distributed model is preferable for long lines, voltage-profile calculations, switching surges, and high-frequency phenomena.

Cable models require particular care. Underground cables have much larger capacitance than overhead lines because conductors are closely spaced and surrounded by insulation with relatively high permittivity. At long lengths, charging current can become substantial, causing receiving-end voltage rise and reactive-power effects. Frequency-dependent resistance, sheath currents, bonding arrangements, dielectric losses, and mutual coupling may be required in transient studies.

Line parameters are calculated from conductor geometry, material properties, earth-return approximations, and sequence networks. They may also be estimated from synchronized voltage and current measurements. Parameter validation should compare predicted voltage drops, charging currents, and losses with field observations under several loading conditions.

Loads: Algebraic, Dynamic, and Composite Representations

A load model describes how active and reactive power respond to voltage and frequency. The three basic static forms are constant power, constant current, and constant impedance. For a constant-power load,

\[

P=P_0,\qquad Q=Q_0.

\]

For a constant-current load, power varies approximately in proportion to voltage magnitude, while for a constant-impedance load it varies with the square of voltage magnitude. A ZIP model combines the three:

\[

P=P_0[a_p(V/V_0)^2+b_p(V/V_0)+c_p],

\]

with an analogous expression for reactive power, where the coefficients sum to unity within each component group.

Constant-power models are convenient in load flow but can exaggerate voltage sensitivity during disturbances. In reality, residential, commercial, and industrial demand is composite. Motors introduce electromechanical states, consume reactive power, and may stall or disconnect during voltage depressions. Electronic power supplies may behave as approximately constant-power devices over a control range but may cease operation below a threshold.

Dynamic load models may include induction motors, thermostatically controlled loads, discharge lighting, converter-interfaced devices, and frequency-sensitive demand. Correct load representation can materially change voltage-stability margins, fault recovery, and frequency response. Load-model identification uses feeder measurements, smart-meter data, disturbance records, and controlled tests. Privacy, aggregation, and nonstationary consumer behavior complicate this process.

Network Equivalents and Matrix Formulation

A large interconnected system is often reduced to a smaller equivalent for planning, real-time analysis, or studies focused on a particular area. A Thevenin equivalent represents an external network by a voltage source behind an impedance; a Norton equivalent uses a current source in parallel with an admittance. For small-signal or fault analysis, the equivalent must preserve the behavior seen at the retained terminals over the operating conditions of interest.

The bus admittance matrix, \(Y_{bus}\), is formed from branch admittances and shunt elements. For a branch with series admittance \(y_{ij}\), the diagonal entries receive the connected admittances and the off-diagonal entries generally receive \(-y_{ij}\):

\[

\mathbf{I}_{bus}=\mathbf{Y}_{bus}\mathbf{V}_{bus}.

\]

A shunt admittance is added to the corresponding diagonal element. Transformer tap ratios modify the relevant diagonal and off-diagonal terms. The matrix is usually sparse, symmetric for reciprocal networks without phase-shifting devices, and complex-valued. Sparsity is essential for solving very large systems efficiently.

The bus impedance matrix is formally

\[

\mathbf{Z}_{bus}=\mathbf{Y}_{bus}^{-1},

\]

when the inverse exists. Its diagonal elements are driving-point impedances and its off-diagonal elements are transfer impedances. Explicit inversion is generally avoided for large systems; factorization and selected-inverse techniques are preferred.

Network reduction can be performed by partitioning the nodal equations and applying a Schur complement. If retained buses are denoted by subscript r and eliminated buses by e,

\[

Y_{red}=Y_{rr}-Y_{re}Y_{ee}^{-1}Y_{er}.

\]

This reduction preserves terminal behavior under the assumptions embedded in the original network. It may, however, hide internal constraints, control interactions, and topology changes. An equivalent created for one operating condition should not automatically be used for another.

Parameter Estimation, Validation, and Uncertainty

Model parameters are obtained from design calculations, laboratory tests, manufacturer data, commissioning tests, and operational measurements. Estimation can be formulated as an optimization problem. Given measurements \(z\), a model output \(h(x,\theta)\), measurement covariance \(R\), and parameters \(\theta\), weighted least squares seeks to minimize

\[

J(\theta)= [z-h(x,\theta)]^T R^{-1}[z-h(x,\theta)].

\]

The covariance matrix weights measurements according to reliability. Robust estimators may reduce the influence of bad data, while Bayesian methods can combine prior manufacturer information with field measurements. Identifiability is important: if several parameters produce nearly identical outputs, numerical fitting does not make them independently trustworthy.

Validation should use data not used for calibration. The engineer should examine residuals, parameter plausibility, sensitivity to operating point, and performance during disturbances. Uncertainty should be reported rather than hidden. For example, a line resistance estimated from a cold commissioning test may understate resistance at operating temperature, while a load model fitted during mild weather may fail during extreme conditions.

Selecting Models for Engineering Studies

Model selection is a decision involving accuracy, computational cost, data availability, and risk. A practical sequence is:

1. Define the study question, outputs, frequency range, and time horizon.

2. Select the least complex model capable of answering that question.

3. Identify parameters and their sources.

4. Establish consistent voltage, power, frequency, and phase conventions.

5. Implement the model and test limiting cases.

6. Compare predictions with analytical expectations, test results, or field measurements.

7. Document assumptions, uncertainty, version, and valid operating range.

For a balanced load-flow study, a synchronous generator may be represented by a scheduled active power and voltage magnitude, a transformer by series impedance and tap ratio, a line by a nominal-\(\pi\) model, and a load by constant power or ZIP behavior. For transient stability, generator rotor dynamics, exciters, governors, and frequency-dependent loads must be added. For electromagnetic transients, detailed winding, saturation, frequency-dependent line, cable, breaker, and surge-arrester models may be required.

Illustrative Case Study: Voltage-Recovery Analysis of a Cable-Rich Distribution Area

Background

A fictional 132/33-kV distribution area supplies industrial plants and urban feeders through several long underground cable circuits. The planning team observes that voltages rise during light load and recover slowly after faults.

Situation and problem

An initial load-flow model uses only series reactance for the 33-kV cables and represents all demand as constant power. The model predicts acceptable voltages. Field measurements, however, show significant charging current at light load and deeper voltage depression during a motor-starting event.

Analysis

The series-only cable model omits shunt capacitance, so it underestimates reactive-power injection and receiving-end voltage rise. The constant-power load model also fails to represent voltage-sensitive lighting, induction motors, and converter loads. During motor starting, the induction motors draw large reactive current; the simplified model cannot reproduce this transient demand.

Approach and solution

The engineering team replaces the cable representation with a nominal-\(\pi\) model using capacitance calculated from geometry and checked against field charging-current measurements. The load is divided into ZIP demand, induction-motor groups, and converter-interfaced industrial loads. Transformer tap limits and tap-control delays are included. A dynamic simulation is then performed for motor starting and a temporary feeder fault.

Outcome

The revised model predicts light-load voltage rise and identifies a post-fault recovery problem that was absent from the original representation. The utility evaluates switched reactors, revised tap-control settings, motor-starting limits, and feeder reinforcement. The case demonstrates that greater model complexity is justified when the neglected phenomenon directly affects the engineering decision.

Learning points

A model should be improved in response to a specific discrepancy, not simply made more detailed. Measurement-based validation is indispensable. Cable capacitance, load composition, control delays, and transformer tap limits can interact; treating them independently may conceal the actual mechanism.

Common Errors, Professional Responsibilities, and Emerging Practice

Common errors include mixing line-to-line and phase-voltage bases, changing the per-unit base without transforming impedances, using synchronous reactance for the first instant of a fault, omitting transformer tap effects, ignoring cable charging, and applying a constant-power load model to a dynamic voltage-recovery problem. Another serious error is using a model outside its validated frequency or operating range.

Model governance is a professional responsibility. Utilities and consultants should maintain controlled data sources, revision histories, parameter ownership, test evidence, and documented assumptions. Incorrect models can influence protection settings, investment decisions, reliability assessments, and public safety. Cybersecurity is also relevant when model databases and operational measurements are exchanged across digital platforms; access control and data integrity must be preserved.

Modern wide-area measurement systems, phasor measurement units, smart meters, and digital substations support continuous model validation. Machine-learning methods can detect discrepancies and estimate aggregate load behavior, but they should complement—not replace—physical understanding. Explainability, training-data quality, extrapolation risk, and protection of operational data must be considered. Emerging digital twins combine real-time measurements with validated network and equipment models, enabling predictive maintenance and scenario analysis.

Best practice is to maintain a hierarchy of models: a fast operational model, a planning model with validated dynamic controls, and detailed specialist models for fault and transient studies. Interfaces between these models should be tested so that simplification does not introduce inconsistent initial conditions or contradictory parameter conventions.

Chapter Summary

Power-system component modeling converts physical equipment into mathematical forms suitable for particular studies. Per-unit normalization simplifies calculations and preserves transformer-side consistency. Synchronous machines require different reactances and dynamic states depending on whether the study concerns steady state, fault inception, transient stability, or electromagnetic transients. Transformer models include series leakage impedance, shunt excitation effects, and tap or phase-shifting ratios. Lines and cables range from series models to nominal-\(\pi\), distributed, and frequency-dependent representations. Loads may be modeled as constant impedance, current, power, ZIP, or dynamic composite demand.

Interconnected systems are represented through \(Y_{bus}\), \(Z_{bus}\), Thevenin/Norton equivalents, and Schur-complement reductions. Parameters should be obtained from tests and measurements, estimated with appropriate statistical methods, and validated against independent data. Model choice must reflect the engineering question, time scale, available data, and consequences of error. Sound model governance, uncertainty reporting, and controlled use of digital analytics are essential professional practices.

Key Terms

Per-unit system: A normalized system in which electrical quantities are divided by selected base quantities.

Synchronous reactance: The steady-state reactance of a synchronous machine associated with its synchronous flux response.

Transient reactance: A machine reactance governing behavior after the subtransient interval and before steady state.

Subtransient reactance: The effective machine reactance immediately after a severe disturbance or fault.

Swing equation: The differential equation relating rotor acceleration to the difference between mechanical and electrical power.

Power angle: The angular displacement between a machine internal emf and a reference voltage.

Transformer leakage reactance: Reactance caused by flux that links one winding but not the other.

Off-nominal tap ratio: A transformer voltage ratio differing from its nominal turns ratio.

Nominal-π model: A line model with a series impedance and two half-shunt admittances.

Distributed-parameter model: A line representation retaining continuous spatial distribution of impedance and admittance.

Characteristic impedance: The voltage-to-current ratio associated with a travelling wave on a uniform line.

ZIP load model: A load model combining constant-impedance, constant-current, and constant-power components.

Thevenin equivalent: A voltage source and series impedance representing a network at selected terminals.

Norton equivalent: A current source and parallel admittance representing a network at selected terminals.

Bus admittance matrix: The nodal matrix relating bus current injections to bus voltages.

Bus impedance matrix: The inverse nodal matrix whose elements represent driving-point and transfer impedances.

Schur complement: A matrix operation used to eliminate internal network buses while preserving terminal behavior.

Parameter identifiability: The degree to which available measurements uniquely determine model parameters.

Model validation: Comparison of model predictions with independent measurements or accepted reference behavior.

Review Questions

Short Answer Questions

1. Why is model selection dependent on the purpose of a power-system study?

2. State the principal bases used in a three-phase per-unit system.

3. Distinguish synchronous, transient, and subtransient reactance.

4. What physical effects are represented by the shunt branch of a transformer equivalent circuit?

5. Why can cable capacitance be more important than overhead-line capacitance?

Long Answer Questions

1. Derive and explain the simplified power-angle equation for a synchronous generator connected to an infinite bus, and discuss its assumptions.

2. Compare short-line, nominal-π, and distributed-parameter transmission-line models and identify suitable applications for each.

3. Explain the ZIP load model and discuss why dynamic load models are required for voltage-recovery studies.

4. Describe the formation and properties of the bus admittance matrix, including the effect of transformer tap ratios.

5. Discuss a complete parameter-estimation and validation procedure for a power-system component.

Analytical / Application Questions

1. A generator has a subtransient reactance of 0.18 per unit and a synchronous reactance of 1.8 per unit. Explain which value should be used immediately after a fault and why using the other value would distort the calculated current.

2. A long underground cable produces unacceptable voltage rise at light load. Identify the omitted or poorly represented phenomena in a simple series-impedance model and propose a modeling and mitigation approach.

3. An external network is to be reduced while preserving behavior at two boundary buses. Explain how a Schur-complement reduction can be used and state two limitations of the resulting equivalent.

22. Multiple Choice Questions with Answers

Question 1

Which model is generally most appropriate for a moderate-length transmission line in a power-flow study when shunt capacitance cannot be neglected?

A. A purely series impedance model

B. A nominal-π model

C. A constant-power load model

D. A synchronous-machine swing model

Correct Answer: B. A nominal-π model

Question 2

Immediately after a severe three-phase fault, which synchronous-machine reactance primarily determines the initial fault current?

A. Synchronous reactance

B. Transient reactance

C. Subtransient reactance

D. Zero-sequence reactance only

Correct Answer: C. Subtransient reactance

Question 3

What is the principal advantage of selecting voltage bases in the transformer turns ratio?

A. It eliminates all transformer losses

B. It makes per-unit impedance transfer consistently across the transformer

C. It makes all bus voltages equal

D. It removes the need for a tap ratio

Correct Answer: B. It makes per-unit impedance transfer consistently across the transformer

Question 4

In a ZIP load model, the constant-impedance component causes power magnitude to vary approximately as

A. The inverse of voltage magnitude

B. The first power of voltage magnitude

C. The square of voltage magnitude

D. A fixed value independent of voltage

Correct Answer: C. The square of voltage magnitude

Question 5

For a reciprocal network without phase-shifting devices, the bus admittance matrix is generally

A. Real and diagonal

B. Complex and symmetric

C. Antisymmetric and real

D. Singular for every operating condition

Correct Answer: B. Complex and symmetric

Question 6

A line model uses γ = √(zy) and Zc = √(z/y). This is a

A. Constant-power load model

B. Distributed-parameter line model

C. Classical generator model

D. Norton load equivalent

Correct Answer: B. Distributed-parameter line model

Question 7

Which test primarily estimates a transformer’s shunt excitation branch?

A. Open-circuit test

B. Short-circuit test

C. Sudden-short-circuit generator test

D. Rotor inertia test

Correct Answer: A. Open-circuit test

Question 8

A model calibrated using one operating condition performs poorly during a disturbance. The most appropriate first response is to

A. Increase every parameter arbitrarily

B. Validate it against independent disturbance data and reassess its valid range

C. Delete all shunt elements

D. Replace every load with constant impedance

Correct Answer: B. Validate it against independent disturbance data and reassess its valid range

Question 9

In network reduction, the expression Yred = Yrr − Yre Yee⁻¹ Yer is used to

A. Calculate synchronous speed

B. Eliminate internal buses while preserving terminal behavior

C. Estimate transformer core loss

D. Convert phase quantities to line quantities

Correct Answer: B. Eliminate internal buses while preserving terminal behavior

Question 10

Why may a constant-power load model be unsuitable for a voltage-recovery study?

A. It cannot represent any active power

B. It ignores voltage sensitivity and dynamic behavior of actual loads

C. It always predicts zero reactive power

D. It is valid only for transmission lines

Correct Answer: B. It ignores voltage sensitivity and dynamic behavior of actual loads