How to Enhance the Robustness of Vector Control and Direct Torque Control
In AC motor speed control systems, Vector Control (VC) and Direct Torque Control (DTC) are two mainstream high-performance control strategies. However, they differ in sensitivity to parameter perturbations, load disturbances, and external interferences. Targeted optimizations are required to improve their robustness—defined as a system’s ability to maintain stability and control accuracy under disturbances. Below is a detailed breakdown of “core challenge analysis” and “strategy-specific optimization methods,” combined with technical pathways and theoretical details, to systematically explain robustness enhancement solutions.
I. First, Clarify: Core Robustness Challenges of VC and DTC
The root cause of insufficient robustness lies in the “deviation between the control model/target and the actual system state.” The core pain points of the two strategies differ and must be addressed specifically:
| Control Strategy | Core Control Logic | Robustness Weaknesses | Typical Disturbance Scenarios |
|---|---|---|---|
| Vector Control (VC) | Based on “field orientation,” it decomposes the stator current into an excitation component (id) and a torque component (iq), which are independently controlled via PI regulators. It relies on accurate motor parameters (e.g., stator resistance Rs, rotor resistance Rr, leakage inductance σL, mutual inductance Lm) to build a mathematical model. | 1. Strong parameter dependence: Rotor resistance Rr changes with temperature (e.g., Rr can increase by 2–3 times at a temperature rise of 200°C), and mutual inductance Lm changes with magnetic saturation—both directly causing field orientation deviations. 2. Dynamic lag of the current loop: Traditional PI regulators have a slow response to sudden load changes, easily leading to overshoot or steady-state errors. 3. Grid voltage disturbances: Voltage sags or imbalances cause coupling between id and iq, disrupting decoupled current control. |
Long-term motor operation (parameter drift), frequent sudden load changes (e.g., machine tool cutting, elevator start/stop), low-voltage grid fluctuations. |
| Direct Torque Control (DTC) | Does not rely on precise field orientation. It directly selects inverter switching states via “stator flux linkage observation” and “torque estimation,” aiming for hysteresis tracking of torque and flux linkage. | 1. Flux linkage observation errors: At low speeds, the voltage drop across the stator resistance Rs accounts for a large proportion (Rs voltage drop ≈ stator voltage). Errors in Rs cause deviations between the observed and actual flux linkage, leading to torque ripple. 2. Significant torque ripple: The “variable switching frequency” of hysteresis controllers and “discreteness of voltage vector selection” intensify ripple at low speeds or under load disturbances. 3. Poor stability at zero/low speeds: Back EMF is small at low speeds, resulting in a low signal-to-noise ratio (SNR) of the observer and an increased risk of control instability. |
Low-speed motor operation (e.g., conveyor startup), load impacts (e.g., crusher loading), stator resistance drift due to temperature rise. |
II. Robustness Enhancement Methods for Vector Control (VC)
The core of VC robustness optimization is to “reduce parameter dependence” and “improve dynamic anti-disturbance capabilities.” Specific pathways are as follows:
1. Accurate Parameter Identification: Eliminate Deviations Between the Model and Reality
Parameter errors are the root cause of poor VC robustness. Online/offline parameter identification is required to correct model parameters in real time. Common methods include:
- Offline identification (calibration before startup):
- Static test: Apply a DC voltage to the motor and measure the stator resistance Rs (Rs = U/I).
- Locked-rotor test: Lock the motor rotor, apply a low-frequency AC voltage, and measure the leakage inductance σL (calculated based on short-circuit impedance).
- No-load test: Operate the motor under no-load conditions, and calculate the mutual inductance Lm via back EMF (E = 4.44fNΦ, where Φ is related to Lm).
- Online identification (real-time correction during operation):
- Model Reference Adaptive System (MRAS): Use the error between the “actual motor output (e.g., speed, current)” and the “reference model output” as input for the adaptive law to adjust the rotor resistance Rr and mutual inductance Lm in real time (suitable for Rr drift caused by temperature rise).
- Extended Kalman Filter (EKF): Incorporate motor parameters (Rr, Lm) into the filtering model as state variables, and iteratively estimate parameters using observed current and speed values (it has strong anti-noise capability and is suitable for complex working conditions).
2. Improved Current Loop Control: Enhance Dynamic Anti-Disturbance
Traditional PI regulators have delayed responses to sudden load changes and voltage disturbances, so they need to be replaced with high-dynamic controllers:
- Proportional Resonant (PR) Controller: It features infinite gain at specific frequencies (e.g., grid fundamental frequency), enabling zero-static-error tracking of AC currents while suppressing harmonic interference caused by grid voltage imbalances (resolving id/iq coupling under voltage disturbances).
- Sliding Mode Control (SMC): It forces the system to move along a preset “sliding surface” via “discontinuous control signals,” providing “invariance” to parameter perturbations and load disturbances (i.e., disturbances do not affect sliding mode motion). However, a “boundary layer method” is needed to suppress high-frequency chattering.
- Predictive Current Control (PCC): It predicts the current for the next 1–2 cycles based on the discrete motor model and selects the voltage vector that minimizes the “current error.” Its dynamic response is 3–5 times faster than that of PI (suitable for scenarios with frequent load changes).
3. Optimized Field Orientation: Reduce Orientation Deviations
Inaccurate field orientation disrupts id/iq decoupling, requiring orientation compensation to improve accuracy:
- Improvements to Field-Oriented Control (FOC): Traditional FOC relies on rotor flux linkage observation, which has large errors at low speeds. Switching to “stator field orientation” or “air-gap field orientation” reduces dependence on the rotor resistance Rr.
- Cross-Coupling Compensation: Add “coupling term compensation” (e.g., back EMF corresponding to iq, coupling term between id and speed) to the output of the current loop to actively offset coupling effects between id and iq (Formula: Δud = -ωLqiq, Δuq = ωLdid + ωψr, where ψr is the rotor flux linkage).
III. Robustness Enhancement Methods for Direct Torque Control (DTC)
The core of DTC robustness optimization is to “reduce flux linkage observation errors” and “suppress torque ripple.” Specific pathways are as follows:
1. High-Precision Flux Linkage Observation: Resolve Low-Speed Observation Inaccuracy
Stator flux linkage observation errors are the key cause of DTC instability at low speeds, so the observer design needs to be improved:
- Limitations of Traditional Observers: Observers based on “stator voltage integration” (ψs = ∫(us – Rs is)dt) suffer from a significant Rs is voltage drop at low speeds (us ≈ Rs is), where Rs errors cause integral drift. Additionally, initial integral errors accumulate over time.
- Improved Observer Solutions:
- Voltage-Current Model Switching Observer:
- High-speed range (speed > 5% of the rated speed): Use the voltage model (∫(us – Rs is)dt) to observe flux linkage (the back EMF is large, so the impact of Rs errors is minimal).
- Low-speed/zero-speed range: Switch to the current model (ψs = Ls is + Lm ir, where ir is the rotor current estimated via speed and torque) to avoid integral drift of the voltage model.
- Sliding Mode Observer (SMO): It uses the “discontinuous characteristics” of sliding mode control to suppress parameter perturbations and generates flux linkage observations via stator current errors. Its robustness to Rs changes and load disturbances is over 50% higher than that of traditional integral observers.
- Adaptive Flux Linkage Observer: It treats Rs as an adaptive parameter and corrects Rs in real time via flux linkage observation errors (e.g., designing an adaptive law based on Popov’s hyperstability theory) to address Rs drift at low speeds.
- Voltage-Current Model Switching Observer:
2. Torque Ripple Suppression: Reduce the Impact of Switching Discreteness
DTC torque ripple stems from the “discreteness of voltage vector selection” and “variable switching frequency of hysteresis controllers.” Optimizations are needed from both the “control strategy” and “topological structure”:
- Control Strategy Improvements:
- Discrete Space Vector Modulation (DSVM): It expands the traditional 6 non-zero voltage vectors into “virtual vectors” (e.g., combinations of two adjacent vectors), increasing the degree of freedom in voltage vector selection. This smooths changes in flux linkage and torque, reducing ripple by 30%–50%.
- Model Predictive Direct Torque Control (MPDTC): It predicts torque and flux linkage for the next several cycles based on the motor model and selects the optimal voltage vector via a “cost function” (e.g., torque error + flux linkage error + switching loss). This achieves multi-objective optimization of “torque ripple” and “switching frequency.”
- Variable Hysteresis Bandwidth Control: It dynamically adjusts the hysteresis bandwidth based on the magnitude of the torque error (wider bandwidth for large errors to ensure fast response; narrower bandwidth for small errors to suppress ripple). This avoids “slow response to large errors” or “frequent switching for small errors” caused by fixed bandwidth.
- Topological Structure Optimization:
- Replace traditional two-level inverters with three-level inverters: Three-level inverters output 12 voltage vectors (compared to 6 for two-level inverters), providing higher vector density and improved regulation accuracy of flux linkage and torque, which further reduces ripple.
- Add filtering components: Install LC filters at the inverter output to suppress the impact of switching frequency harmonics on current, indirectly reducing torque ripple (suitable for scenarios with strict ripple requirements, such as precision machine tools).
3. Enhance Stability at Zero/Low Speeds
Back EMF is small at low speeds, resulting in a low observer SNR. Targeted enhancements are required:
- High-Frequency Signal Injection: Inject high-frequency voltage signals (e.g., 1–5 kHz) into the stator windings. Estimate the rotor position and flux linkage by detecting changes in the amplitude/phase of the high-frequency current, enabling stable control at zero speed (suitable for permanent magnet synchronous motors and induction motors).
- Torque Bias Control: Add a small “bias value” to the torque command at low speeds to avoid instability caused by frequent torque switching near zero.
- Sliding Mode Speed Observer: Use sliding mode control to estimate speed, replacing traditional photoelectric encoders (sensorless control). This avoids speed fluctuations caused by insufficient encoder resolution at low speeds, improving the robustness of sensorless DTC.
IV. Common Optimizations: Robustness Enhancement Technologies for Both VC and DTC
In addition to the strategy-specific methods above, two types of common technologies can improve the robustness of both control strategies simultaneously:
1. Anti-Disturbance Control Algorithms
- Disturbance Observer (DOB): It treats “load disturbances” and “parameter perturbations” as “equivalent disturbances.” Estimate disturbance values via “control input” and “system output,” and add a “disturbance compensation term” to the control signal to actively offset disturbances (Formula: u = u0 – ˆd, where u0 is the original control signal and ˆd is the observed disturbance). This significantly improves anti-disturbance capabilities against sudden load changes.
- Adaptive Control: It adjusts control parameters (e.g., PI proportional gain kp, integral gain ki) online to adapt to changes in system characteristics. Common approaches include “fuzzy adaptive control” (adjust kp/ki based on fuzzy rules and error magnitude) and “neural network adaptive control” (approximate system nonlinearity via neural networks and dynamically adjust control parameters).
2. Hardware and Signal Processing Optimization
- High-Precision Sampling: Use 16-bit or higher-precision ADCs (Analog-to-Digital Converters) to collect current and voltage, reducing sampling errors. Simultaneously, adopt “synchronous sampling” (sampling timing synchronized with PWM carriers) to avoid control errors caused by sampling delays.
- Temperature Compensation: Install temperature sensors (e.g., PT100) near the motor stator windings to detect temperature in real time and correct the stator resistance Rs and rotor resistance Rr (temperature coefficient of resistance: α ≈ 0.004/°C). This directly eliminates parameter errors caused by temperature drift.
- EMC Design: Optimize the PCB layout of the controller (e.g., separate high-voltage and low-voltage regions, reduce loop area) and add EMI filters to suppress the impact of external electromagnetic interference on control signals, preventing switching malfunctions caused by interference.
V. Summary: Core Differences and Selection of Robustness Optimization for VC and DTC
| Dimension | Core of VC Robustness Optimization | Core of DTC Robustness Optimization |
|---|---|---|
| Core Objective | Reduce parameter dependence and improve current decoupling accuracy | Reduce flux linkage observation errors and suppress torque ripple |
| Key Technologies | Online parameter identification (MRAS/EKF), PR/PCC controllers | High-precision flux linkage observers (SMO/MPDTC), DSVM |
| Suitable Scenarios | High-speed/constant-speed loads sensitive to parameter drift (e.g., fans) | Low-speed loads requiring fast torque response and low ripple (e.g., elevators) |
Through the above methods, the parameter perturbation tolerance of VC can be improved by over 40%, and the torque ripple of DTC can be reduced by over 50%. Both control strategies can achieve robustness that meets industrial-grade high-performance speed control requirements (e.g., new energy vehicles, CNC machine tools, rail transit).