Dynamic Modeling and Reference-Frame Theory

Introduction and Context

Electrical machines are nonlinear, coupled electromechanical systems. Their terminal voltages and currents vary with switching, speed, load torque, magnetic saturation, and disturbances. A steady-state phasor model is valuable for rated-frequency analysis, but it cannot adequately predict starting current, torque pulsations, transient stability, controller performance, or the response of a drive to a sudden load change.

Dynamic machine modeling addresses these limitations by expressing electrical, magnetic, and mechanical relationships as differential equations. Reference-frame theory then transforms time-varying phase quantities into coordinates in which important relationships become simpler. The abc, αβ, and dq representations are alternative descriptions of the same physical system when the transformations are applied consistently.

This chapter develops practical models for three-phase machines, with emphasis on induction and synchronous-machine applications. It also introduces per-unit normalization, state-space representation, space vectors, transformations, transient interpretation, and parameter sensitivity.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Explain why steady-state models are insufficient for transient analysis.
2. Select consistent base quantities and formulate machine parameters in per unit.
3. Transform three-phase variables between abc, αβ, and dq coordinates.
4. Develop voltage, flux-linkage, torque, and mechanical state equations.
5. Interpret electromechanical transients using eigenvalues, time constants, and energy flow.
6. Identify the effects of resistance, inductance, inertia, friction, and saturation on model accuracy.
7. Apply a dynamic model to the analysis of a motor-drive disturbance.

1. From Steady-State Phasors to Dynamic Models

A steady-state sinusoidal model assumes that all electrical transients have decayed and that speed is either constant or known. For example, an induction motor equivalent circuit can predict stator current, power factor, and torque at a specified slip. It cannot, by itself, predict the current immediately after energization or the rotor acceleration following a load step.

A dynamic model retains derivatives such as dψ/dt and dω_m/dt. It therefore represents stored magnetic and mechanical energy. The principal state variables are commonly chosen as flux linkages, currents, rotor speed, and rotor position.

For a winding with resistance R and flux linkage ψ, Faraday’s law gives the general voltage relation

where v is voltage in volts, i is current in amperes, R is resistance in ohms, and ψ is flux linkage in webers. The equation applies to a lumped winding model with a chosen voltage polarity and current reference. It shows that terminal voltage is divided between resistive drop and the voltage associated with changing magnetic flux.

For a mechanically rotating system,

where J is the combined inertia in kg·m², ω_m is mechanical angular speed in rad/s, T_e is electromagnetic torque in N·m, T_L is load torque in N·m, and B is viscous-friction coefficient in N·m·s/rad. Coulomb friction, windage, and nonlinear loads may require additional terms. The equation assumes a rigid shaft and represents friction as viscous.

Rotor position is related to speed by

The electromagnetic torque is obtained from magnetic co-energy or from a power balance. For a balanced three-phase machine, mechanical output power is P_mech = T_eω_m, neglecting losses.

2. Per-Unit Quantities

Per-unit (p.u.) notation normalizes a physical quantity by a selected base value:

The actual quantity is recovered using x_actual = x_pu x_base. Per-unit values are dimensionless, although the base quantities retain physical units.

For a three-phase machine, a common choice is:

The voltage and current bases must use compatible line or phase conventions. Mixing line voltage with phase current without the corresponding factor of √3 is a frequent modeling error.

Per-unit impedance conversion between two bases is

This expression assumes the same type of voltage base is used in both terms. If voltage base is unchanged, impedance in per unit scales inversely with the power base.

Per-unit modeling is useful because machine parameters typically fall within moderate numerical ranges, transformer ratios disappear from idealized equations, and machines of different ratings can be compared. It does not eliminate the need to document base values.

Worked Example: Base Quantities

A 15 kVA, 400 V, three-phase motor has a 50 Hz base frequency. Using line-to-line voltage as the voltage base,

I_base = 15000/(√3 × 400) = 21.65 A.

The three-phase impedance base is

Z_base = 400²/15000 = 10.67 Ω.

If the motor has four poles, synchronous mechanical speed at 50 Hz is

and ω_sync = 2π(1500/60) = 157.1 rad/s. A torque base based on synchronous speed is approximately T_base = 15000/157.1 = 95.5 N·m. The chosen speed convention should be stated because rated speed and synchronous speed are not identical.

3. Three-Phase Machine Equations in abc Coordinates

Let the phase variables be represented by vectors

For a symmetrical three-phase winding, the voltage equations can be written compactly as

For a machine with stator and rotor windings,

where each symbol may denote a three-component vector. In a squirrel-cage induction motor, rotor conductor voltages are zero in the rotor reference frame, so v_r = 0. In a wound-rotor machine, rotor terminal voltage may be nonzero.

Flux linkages depend on currents and rotor position:

The inductance matrices vary with rotor angle because mutual alignment between stator and rotor windings changes. This angle dependence is a principal source of time-varying coefficients in abc models.

The electromagnetic torque can be obtained from magnetic co-energy W′_m:

The derivative is taken at constant current, and θ_m is mechanical rotor angle. For linear magnetic systems, co-energy is a quadratic function of currents. When saturation is significant, inductances become current-dependent and the torque calculation must use the nonlinear magnetic energy or co-energy relationship.

An abc model is physically direct and convenient for unbalanced supplies, asymmetrical faults, and harmonics. Its disadvantages are angle-dependent inductances, six or more electrical states, and greater computational complexity.

4. Clarke Transformation: abc to αβ0

The Clarke transformation maps three phase quantities into two stationary orthogonal axes, α and β, plus a zero-sequence component. One amplitude-invariant form is

The inverse transformation is

The αβ0 transformation is linear and is applicable to voltages, currents, flux linkages, and other phase variables. For a balanced three-wire system with no neutral connection,

therefore x₀ = 0, and the zero-sequence state can be omitted. The α and β components then represent a single rotating or pulsating space vector in a stationary reference frame.

The amplitude-invariant transformation preserves the peak amplitude of a balanced sinusoidal set in the space-vector representation. Other conventions, such as power-invariant scaling by √(2/3), are also valid, but voltage, current, power, and torque equations must be adjusted consistently.

5. Park Transformation: αβ to dq

The Park transformation rotates the stationary αβ axes by an angle θ. A commonly used power-invariant form is

The inverse is

The sign of the q-axis and the direction of θ vary among textbooks and software packages. A model is correct only when its transformation, positive rotation direction, phase sequence, and torque convention are mutually consistent.

Differentiating a rotating vector introduces speed-coupling terms. For a vector x,

where ω_k = dθ/dt is reference-frame angular speed and

for the stated transformation convention. Consequently, stator dq voltage equations commonly appear as

The signs change if the q-axis or rotation convention changes. These speed-voltage terms are not optional; omitting them removes the physical effect of a rotating reference frame.

Important reference frames include:

For a balanced sinusoidal machine in a synchronous frame aligned with the rotating field, sinusoidal steady-state quantities become constant. This is why field-oriented control can regulate d- and q-axis currents as approximately decoupled quantities.

6. Space Vectors and Electromagnetic Torque

A complex space vector for a three-phase quantity may be defined, using the amplitude-invariant convention, as

where a = −1/2 + j√3/2 and j is the imaginary unit. For balanced systems, the space vector contains the positive-sequence αβ information. Negative- and zero-sequence components require additional treatment.

In a two-axis model, torque for a three-phase machine is often written

where p is the number of pole pairs, ψ_d and ψ_q are flux linkages in Wb, and i_d and i_q are currents in A. This formula applies to the stated power-invariant dq convention and balanced three-phase representation. A different scaling convention changes the numerical coefficient.

For a surface permanent-magnet synchronous machine with rotor-flux alignment, a frequently used special case is

where ψ_f is permanent-magnet flux linkage, and L_d and L_q are d- and q-axis inductances. The first term is magnet torque; the second is reluctance torque. For a surface machine with approximately L_d = L_q, reluctance torque is negligible.

7. State-Space Formulation

A nonlinear machine model can be expressed as

where x is the state vector, u is the input vector, y is the output vector, and θ_p is the parameter vector. A typical induction-machine state vector is

with stator and rotor flux linkages as electrical states and rotor speed as a mechanical state. Currents are obtained by inverting the flux-current relationship. Inputs may include d- and q-axis stator voltages and load torque.

Near an operating point, let small deviations be denoted by Δ. Linearization gives

where A, B, C, and D are Jacobian matrices evaluated at the operating point. The eigenvalues of A indicate local modes. A mode with eigenvalue λ has approximate time constant τ = −1/Re(λ) when Re(λ) < 0. Complex eigenvalues indicate oscillatory behavior.

State choice affects numerical conditioning and interpretation. Flux states are often advantageous because voltage equations naturally integrate flux. Current states may be convenient for control implementation but can become ill-conditioned when inductance matrices are poorly scaled. Per-unit states usually improve numerical conditioning across electrical and mechanical subsystems.

8. Induction-Machine dq Model

For an arbitrary reference frame rotating at electrical speed ω_k, a commonly used two-axis induction-machine model is

Here ω_r = pω_m is rotor electrical speed. For a squirrel-cage rotor, v_dr = v_qr = 0. The flux equations for a symmetrical machine are

where L_s and L_r are stator and rotor self-inductances and L_m is mutual inductance. These are linear magnetic equations; saturation requires nonlinear inductance functions.

The torque is

and the mechanical equation is as previously stated. This model is widely used in vector control, simulation, parameter identification, and stability studies.

9. Electromechanical Transients

A motor transient contains electrical and mechanical time scales. Electrical transients are associated with resistance and leakage inductance; mechanical transients are usually slower because inertia resists speed change. During a voltage step, current may change rapidly while speed initially remains nearly constant. The resulting torque accelerates the rotor until electromagnetic torque approaches load torque plus friction.

For a simplified first-order electrical subsystem,

where L is inductance in H and R is resistance in Ω. This time constant applies only to a linear RL subsystem with constant parameters. For the mechanical subsystem near constant torque,

before speed-dependent effects become important. With viscous friction and constant net torque, the speed approaches a first-order response with mechanical time constant τ_m = J/B when B > 0.

A load increase causes electromagnetic torque to become temporarily smaller than opposing torque, producing deceleration. In a controlled drive, the controller increases q-axis current to restore torque, subject to current, voltage, thermal, and inverter limits. Excessive current commands can cause saturation, overcurrent trips, or loss of synchronism in some machines.

10. Parameter Sensitivity and Model Validation

Parameter sensitivity measures how an output changes when a parameter changes. A normalized local sensitivity may be written

where p is a parameter and y is an output. A magnitude greater than one indicates that a percentage change in p produces a larger percentage change in y locally. Sensitivity depends on operating point and should not be interpreted as a universal machine property.

Important sensitivities include:

• Stator resistance: strongly affects low-speed voltage drop, copper loss, and flux estimation; it also varies substantially with temperature.
• Rotor resistance: affects induction-motor slip, torque distribution, and rotor time constant.
• Leakage inductance: affects current rise, short-circuit behavior, torque pulsation, and voltage requirements.
• Mutual inductance: affects magnetizing current and torque production; saturation makes it operating-point dependent.
• Inertia: determines acceleration and deceleration rates.
• Friction and windage: affect no-load speed, steady-state loss, and low-speed response.

Validation should compare simulated and measured currents, speed, torque, and power under several operating conditions. A model fitted only to rated steady-state data may predict rated operation well but fail during starting or field weakening. Identification experiments should be designed so that parameters are observable; for example, rotor resistance cannot generally be estimated reliably from a single steady-state measurement.

11. Worked Application: Load-Step Response of a Vector-Controlled Motor

Consider an illustrative 15 kW induction-motor drive operating at constant flux. The controller aligns the d-axis with rotor flux, so the q-axis current primarily controls torque. Before a disturbance, electromagnetic torque is 45 N·m and load torque is also 45 N·m. At t = 0.10 s, load torque steps to 70 N·m. Assume the controller can increase torque rapidly to 70 N·m, while the combined inertia is J = 0.80 kg·m² and viscous friction is neglected over the short interval.

Background and Problem

The load step creates a temporary torque deficit of 25 N·m. If the controller did not respond immediately, the initial acceleration would be

dω_m/dt = (45 − 70)/0.80 = −31.25 rad/s².

The motor would decelerate. The engineering question is how the drive model represents and mitigates this event.

Analysis and Approach

In rotor-flux-oriented coordinates, flux is maintained primarily by i_d, while i_q is increased to produce torque. The torque relationship is approximately

T_e ≈ K_T i_q,

where K_T depends on flux and machine parameters. Thus, a 55.6% increase in required torque, from 45 to 70 N·m, requires approximately a 55.6% increase in q-axis current if flux remains constant and voltage/current limits are not reached.

The state-space simulation contains flux states, rotor speed, and controller states. Immediately after the load step, speed begins to fall, the speed controller produces a positive i_q command, and electromagnetic torque rises. During the brief regulator transient, speed reaches a minimum. Once T_e = T_L and losses are included, the speed returns toward its commanded value.

Outcome and Interpretation

The simplified initial acceleration predicts the direction and severity of the transient, but not its complete waveform. The actual response depends on speed-controller bandwidth, current-loop bandwidth, torque limits, rotor resistance, flux level, sampling delay, and inverter voltage. If the demanded q-axis current exceeds its limit, the motor cannot instantly produce 70 N·m and a larger speed dip results.

Learning Points

1. The mechanical equation determines the immediate speed tendency.
2. The dq model separates flux-producing and torque-producing current approximately, simplifying control design.
3. Parameter errors change the estimated rotor flux and therefore the torque-per-ampere relationship.
4. Current and voltage limits must be included in realistic drive models.
5. A steady-state equivalent circuit cannot predict the speed dip or controller recovery process.

12. Practical Implementation and Best Practices

A reliable simulation or embedded implementation should:

• document phase sequence, polarity, transformation scaling, and positive rotation direction;
• use consistent electrical and mechanical angle definitions, including the pole-pair factor;
• normalize variables when ratings differ greatly;
• initialize flux states consistently with the initial voltage and frequency;
• distinguish line-to-line from phase quantities;
• include inverter dead time, sampling, saturation, and limits when studying a real drive;
• use appropriate numerical solvers and sufficiently small steps for switching models;
• validate energy and power balance;
• test balanced, unbalanced, startup, load-step, and fault cases;
• estimate temperature-dependent resistance when precision is required.

A useful energy check is that electrical input power equals the sum of resistive loss, rate of stored magnetic energy, mechanical power, and other losses, subject to sign conventions. Violations often reveal an incorrect transformation coefficient or missing speed-voltage term.

13. Limitations, Safety, and Contemporary Developments

Classical dq models assume sinusoidally distributed windings, spatial symmetry, and often linear magnetic behavior. They may not capture slotting, space harmonics, rotor-bar skew, magnetic saturation, hysteresis, iron loss, bearing dynamics, or converter common-mode effects. Finite-element models can represent several of these effects but require greater computational cost.

Digital twins increasingly combine reduced-order dq models with measurements and online parameter estimation. Machine-learning methods can identify residual errors or estimate temperature, but they should not replace physically validated models in protection-critical functions. Training data must cover relevant operating conditions, and model predictions should be monitored for extrapolation.

Electrical-machine models also support functional safety and cybersecurity analysis. Incorrect sensor scaling, corrupted angle data, or an adversarial command can produce excessive torque or overvoltage. Drive software should therefore apply plausibility checks, current limits, safe shutdown logic, access control, and traceable verification procedures.

Chapter Summary

Dynamic models extend steady-state machine analysis by representing magnetic and mechanical energy storage through differential equations. Per-unit normalization simplifies comparison and numerical computation, provided that base quantities and phase conventions are explicit. The abc model is physically direct but contains angle-dependent inductances. Clarke transformation produces stationary αβ0 coordinates, while Park transformation produces rotating dq coordinates. In a suitable synchronous or rotor-oriented frame, balanced sinusoidal quantities become constant or slowly varying.

Voltage equations contain resistance, flux derivatives, and reference-frame speed-coupling terms. Flux-current relations determine currents, and co-energy or cross-product expressions determine electromagnetic torque. State-space models support simulation, linearization, eigenvalue analysis, controller design, and parameter identification. Electromechanical transients reflect interactions between fast electrical dynamics and slower inertial dynamics. Resistance, inductance, mutual coupling, inertia, friction, saturation, and temperature all influence accuracy. Professional practice requires consistent conventions, validation against measurements, explicit limitations, and safe implementation of drive controls.

Key Terms

Per unit: A normalized quantity equal to actual value divided by a selected base value.

Reference frame: A coordinate system whose angular speed determines how electrical variables are represented.

abc model: Three-phase phase-coordinate model using the physical winding variables.

Clarke transformation: Transformation from abc quantities to stationary αβ0 coordinates.

Park transformation: Rotation of αβ quantities into dq coordinates.

Space vector: Complex or two-axis representation of a balanced multiphase quantity.

Flux linkage: Magnetic flux linkage associated with a winding, measured in weber-turn equivalents.

Co-energy: Energy-based function useful for calculating torque at constant current.

State variable: A variable whose value, together with inputs, determines future system behavior.

State-space model: First-order differential and algebraic representation of a dynamic system.

Pole pair: Half the number of machine poles, used to convert mechanical to electrical speed.

Slip: Relative speed difference between synchronous field and rotor, commonly normalized by synchronous speed.

Synchronous frame: Reference frame rotating at the selected electrical synchronous speed.

Rotor-flux orientation: dq alignment in which the d-axis follows rotor flux.

Electrical time constant: Characteristic response time associated with resistance and inductance.

Mechanical time constant: Characteristic response time associated with inertia and mechanical damping.

Parameter sensitivity: Change in an output resulting from variation of a model parameter.

Saturation: Nonlinear reduction of incremental inductance as magnetic material approaches its limit.

Field-oriented control: Control method that regulates flux- and torque-producing components separately in rotating coordinates.

Eigenvalue: Quantity describing a local dynamic mode of a linearized state-space model.

Review Questions

Short Answer Questions

1. Why can a steady-state equivalent circuit not predict motor acceleration after a load step?
2. What is the purpose of per-unit normalization?
3. Under what condition is the zero-sequence component zero in a three-wire balanced system?
4. Why do dq voltage equations contain speed-coupling terms?
5. What physical quantities are commonly represented as states in an induction-machine model?

Long Answer Questions

1. Explain the transformation from abc to αβ0 and discuss the physical meaning of each component.
2. Compare stationary, synchronous, and rotor reference frames and identify an application for each.
3. Derive the mechanical speed equation from Newton’s law for a motor with electromagnetic torque, load torque, and viscous friction.
4. Discuss the assumptions and limitations of a linear dq induction-machine model.
5. Explain how resistance, inductance, inertia, and saturation affect dynamic-model accuracy.

Analytical / Application Questions

1. A 30 kVA, 460 V, three-phase motor has a per-unit stator impedance of 0.08 on its own base. Calculate its current and impedance bases, then determine the actual impedance magnitude corresponding to 0.08 p.u.
2. A motor has J = 0.50 kg·m², electromagnetic torque 80 N·m, and load torque 55 N·m. Neglect friction. Determine its initial mechanical acceleration and explain how the result changes if inertia doubles.
3. A vector-controlled motor experiences a load increase while its d-axis flux current remains constant. Explain the expected q-axis current, torque, speed, and current-limit responses, including conditions under which the simplified torque-current relationship fails.

Multiple Choice Questions with Answers

Question 1

Which equation represents the voltage of a lumped winding with resistance and changing flux linkage?

A. v = Ri + dψ/dt

B. v = R/i + ψt

C. v = i/R − dψ/dt

D. v = T_e/ω_m

Correct Answer: A. v = Ri + dψ/dt

Question 2

A 15 kVA, 400 V three-phase system has approximately what current base?

A. 12.5 A

B. 21.7 A

C. 37.5 A

D. 65.0 A

Correct Answer: B. 21.7 A

Question 3

In a balanced three-wire system, which zero-sequence condition normally applies?

A. x₀ = x_a + x_b + x_c

B. x₀ = 1

C. x₀ = 0

D. x₀ = x_α − x_β

Correct Answer: C. x₀ = 0

Question 4

Which reference frame is commonly selected so balanced steady-state sinusoidal variables become constant?

A. Stationary αβ frame

B. Synchronous dq frame

C. Random abc frame

D. Zero-sequence frame

Correct Answer: B. Synchronous dq frame

Question 5

For the stated dq convention, which expression gives three-phase electromagnetic torque?

A. T_e = (3/2)p(ψ_d i_q − ψ_q i_d)

B. T_e = RI²

C. T_e = Jω_m

D. T_e = ψ_dψ_q

Correct Answer: A. T_e = (3/2)p(ψ_d i_q − ψ_q i_d)

Question 6

A machine has inductance 0.20 H and resistance 4 Ω in a simplified RL subsystem. What is its electrical time constant?

A. 0.0125 s

B. 0.05 s

C. 0.80 s

D. 20 s

Correct Answer: B. 0.05 s

Question 7

For a squirrel-cage induction motor represented in rotor dq coordinates, which rotor voltage condition is normally used?

A. v_dr = v_qr = 1

B. v_dr = v_qr = 0

C. v_dr = R_ri_dr

D. v_dr = ω_rψ_qr

Correct Answer: B. v_dr = v_qr = 0

Question 8

If electromagnetic torque is 45 N·m, load torque is 70 N·m, and inertia is 0.80 kg·m², what is the initial acceleration when friction is neglected?

A. −31.25 rad/s²

B. −20.00 rad/s²

C. 31.25 rad/s²

D. 143.75 rad/s²

Correct Answer: A. −31.25 rad/s²

Question 9

Which parameter most directly determines the acceleration rate for a given net torque?

A. Pole-pair number only

B. Combined mechanical inertia

C. Zero-sequence voltage

D. Stator phase sequence only

Correct Answer: B. Combined mechanical inertia

Question 10

What is a major consequence of using inconsistent Clarke/Park scaling conventions?

A. The machine poles physically change

B. Torque and power calculations can have incorrect coefficients

C. Rotor resistance becomes zero

D. Mechanical speed no longer has units

Correct Answer: B. Torque and power calculations can have incorrect coefficients