This paper presents several techniques and methods for improving inverter performance through control and modulation strategies. These methods can all be implemented in software on a general-purpose hardware platform, which is beneficial for modularization and integration. Some of the technical issues and corresponding solutions are discussed below.
Several techniques in control strategies
1. Compensation technology
Compensation techniques are essential in open-loop control. These include torque compensation, slip compensation, and dead-zone compensation. At low frequencies, the stator resistance voltage drop is significant relative to the inverter's output voltage and must be compensated; otherwise, the output voltage will be insufficient, causing the motor to stop or experience a significant speed drop at low frequencies. Slip compensation is primarily designed to address the issue of the motor's actual output speed being lower than the set speed under heavy load. Both of these compensation methods can be implemented using simple fixed values. Improved methods include calculations based on the three-phase motor current, but this is only based on the current amplitude and is essentially scalar compensation. A more precise method involves vector decomposition of the three-phase motor AC current, incorporating motor losses into the calculation, resulting in better compensation. However, this method is computationally complex and dependent on certain motor parameters, presenting some challenges in implementation.
Dead-zone compensation technology plays a crucial role in open-loop control. It effectively improves the smoothness of the output current waveform and reduces harmonics, while also increasing the effective value of the output voltage and reducing motor current oscillations. Especially in environments requiring quiet operation, artificially increasing the carrier frequency without dead-zone compensation can prevent the motor from running even under no-load conditions at low frequencies. Commonly used dead-zone compensation techniques include direct current zero-crossing compensation, current decomposition based on stator magnetic field orientation, dead-zone voltage pulse width compensation, and sensorless dead-time prediction compensation. While the current zero-crossing judgment method is simple and easy to implement, the large noise component in the current waveform, coupled with load fluctuations and external interference, can cause errors in zero-crossing judgment. A dead-zone plateau at the zero-crossing point significantly impacts low-frequency compensation, particularly at higher carrier frequencies. The stator magnetic field orientation method does not directly determine the current zero-crossing point but instead decomposes the stator current in a rotating coordinate system to obtain the relationship between the current vector angle and the dead-zone voltage vector for compensation. Combining this method with dead-zone voltage pulse width compensation yields even better results. The dead time compensation method based on phase angle prediction is a fixed compensation method that eliminates the need for a current sensor. This method first predicts the current phase angle and then compensates for the dead time accordingly. The predicted angle can be set in software according to the different output capacities of the frequency converter, or it can be modified externally.
The advantage of this method is that it eliminates the need for a current sensor, reducing costs and system size. However, the compensation does not adjust accordingly to changes in the external load, which reduces accuracy and dynamic performance.
2. Current oscillation suppression technology
When an AC motor is powered by PWM, it may experience local instability across a wide frequency range under light or no-load conditions. This instability manifests as significant current amplitude fluctuations and changes in output frequency. Current oscillations can trigger false alarms due to overcurrent, hindering stable and reliable system operation. The causes of oscillations are multifaceted, but a common view is that they occur during energy exchange between the motor and the inverter , and are closely related to the dead-zone effect. Compensating for the dead-zone effect can effectively reduce the oscillation amplitude, but it cannot fundamentally suppress it. One effective method is to adjust the actual output frequency or voltage when oscillation occurs, creating a simple negative feedback system through the current to suppress the oscillation. However, this method has limitations. Since different motors have different oscillation frequency ranges, varying from approximately 5Hz to 30Hz, and current amplitude control is only a scalar, the control effect is poor, reducing system robustness. Decomposing the stator current and directly controlling the magnetic flux excitation current component that affects energy exchange can significantly improve the suppression effect. A more precise and effective method is to use intelligent control, but the algorithm is complex and difficult to implement on a general V/f control platform.
3. Simple Flux Vector Control Method
Conventional V/f control is based on a steady-state motor model, neglecting the stator resistance voltage drop. Therefore, it cannot control the motor's dynamic state. As it is open-loop control, it is insensitive to load fluctuations or changes in motor parameters, resulting in poor dynamic performance. The simple flux vector control method, based on conventional V/f control, controls the motor current. Specifically, it calculates the torque current component and the excitation current component by vector decomposition of the inverter's output current. Then, it adjusts the voltage to match the motor current with the load torque, thereby improving low-speed torque characteristics. This method can provide 200% of the rated torque at 6Hz. Some motor parameters used in the vector calculation are pre-stored in the controller's RAM; these parameters are essentially constants for a given motor model.
4. Vector control technology based on speed sensorless operation
For high-performance AC speed control systems, a speed closed loop is essential. This loop requires real-time motor speed monitoring. Currently, speed feedback is primarily detected using photoelectric pulse encoders, rotary transformers, or tachogenerators. However, speed sensors are relatively expensive, significantly increasing the system's hardware cost; they also have poor environmental adaptability, making them unsuitable for high-temperature or vibrating environments; and their signal transmission distance is limited, preventing reliable operation over long distances. Therefore, researching sensorless AC speed control systems is crucial for improving system reliability, environmental adaptability, and expanding the application range of AC speed control systems. This research has become a hot topic in academic and engineering circles both domestically and internationally in recent years.
The ultimate goal of sensorless control is to simultaneously and accurately estimate motor speed, rotor flux linkage, and motor parameters. There are various methods for estimating motor speed and flux linkage. Methods based on ideal models include: open-loop flux linkage estimation and compensated flux linkage estimation; Model Reference Adaptive (MRAS); and closed-loop observer methods. Methods based on non-ideal characteristics include: speed identification using tooth harmonic signals; high-frequency injection of rotor salient poles for detection; leakage inductance pulsation detection; dq impedance difference orientation; and saturated salient pole detection. Motor parameter detection can be performed offline or online.
There are several aspects of sensorless vector control technology that deserve special attention in its implementation, as they have a significant impact on the system's control performance and accuracy. These aspects are:
(1) Detection and signal processing technology of current and voltage signals
The signal processing techniques mainly involve effectively and accurately filtering the detected current and voltage signals to reproduce the valid signals without causing amplitude attenuation or phase lag. Practical methods include simplified extended Kalman filters and shape filters.
(2) Problem of online adjustment of stator resistance
The stator resistance value changes significantly with increasing temperature during motor operation, with the maximum change reaching 150% of the rated value. Therefore, online detection of the stator resistance during operation and adjustment of corresponding control quantities are crucial to the system performance.
(3) Dead zone compensation techniques
(4) The problem of establishing an accurate dynamic motor model
Motor parameters measured online or offline are only obtained at a certain moment. If the parameters change during operation, the motor model should also be changed accordingly to achieve the best control effect. Currently, the model reference adaptive method is more commonly used in practical research.
(5) Inverter model reconstruction problem
This technology is primarily designed for operation at 0Hz under extreme conditions. In this case, the time relationship between the saturation voltage drop and the collector current of the power device must be taken into account.
III. Several Techniques of PWM Modulation Strategies
Early PWM modulation methods were primarily generated through hardware circuit simulation, mainly using sinusoidal pulse width modulation. Later, a hybrid analog and digital circuit control approach emerged, and current modulation techniques are largely implemented directly through software algorithms. Software implementation offers significant advantages: flexible programming, easy modification, the ability to implement multiple modulation strategies under the same hardware conditions, convenient maintenance, and strong anti-interference capabilities. From initially pursuing a sinusoidal voltage waveform, to a sinusoidal current waveform, and then to a sinusoidal magnetic flux; from optimal efficiency and minimal torque ripple to noise elimination, the development of PWM control technology has undergone a continuous process of innovation and improvement. Based on control principles, PWM control technology can be divided into four categories: constant-width PWM, sinusoidal PWM (SPWM), flux tracking PWM, and current tracking PWM. The recently proposed discontinuous SVPWM and random PWM methods will be highlighted here.
1. SPWM method
SPWM (Simultaneous Dynamic Width Modulation) method, from the perspective of motor power supply, focuses on how to generate a frequency- and voltage-adjustable three-phase symmetrical sinusoidal power supply. Specifically, it uses a sine wave as a reference wave (called the modulation wave) and intersects it with a series of equal-amplitude triangular waves (called the carrier wave). The switching mode of the inverter is determined by their intersection point. To improve the output voltage amplitude of the inverter, quasi-optimized switching modes have been proposed for SPWM.
PWM (Pulse Width Modulation) is a method that uses the third harmonic superposition technique. By injecting a certain proportion of the third harmonic into a sine wave, the amplitude of the modulating wave is significantly reduced. Without overmodulation, the fundamental wave amplitude can exceed the triangular wave amplitude, achieving modulation with a modulation coefficient greater than 1. Under this modulation method, the maximum modulation ratio can be increased to approximately 1.15, and the corresponding DC bus voltage utilization rate can be increased by up to 15%.
2. SVPWM method
Flux tracking PWM, also known as space vector pulse width modulation (SVPWM), differs from SPWM in that it focuses on the motor's perspective, aiming to achieve a circular magnetic field. It uses the ideal flux linkage of an AC motor powered by a three-phase symmetrical sinusoidal voltage as a reference. The actual flux linkage vectors generated by different inverter switching modes are used to track this reference flux linkage circle. The tracking result determines the inverter's switching mode, forming the PWM wave. The inverter's switching modes have eight space voltage vectors, with V0 and V7 being zero-voltage vectors. SVPWM not only reduces motor torque ripple and current waveform distortion but also significantly improves DC bus voltage utilization compared to SPWM technology. Under this modulation mode, DC bus voltage utilization can be increased by up to 15%, and it is easily implemented digitally.
3. Discontinuous SVPWM strategy (DHPWM)
Discontinuous SVPWM is a novel voltage space vector pulse width modulation strategy proposed in recent years. It is referred to as discontinuous SVPWM (DHPWM) in foreign literature and as hybrid modulation (HPWM) or low switching loss mode modulation in some domestic articles. For continuous PWM modulation, the three-phase modulated waves are all located between the peak values of their corresponding carrier waves. Therefore, the switching losses of the inverter are the same for all continuous PWM modulation methods and are independent of the phase angle of the load current. The simplest way to reduce switching losses is to make the switching devices inactive or minimize their operation within a cycle. In traditional SVPWM methods, the positions of the zero vectors (V0 and V7) are symmetrically present during pulse width generation. The conduction times of the zero vectors are equal, and their positions are fixed and cannot be changed. If the effective conduction time of the zero vector remains unchanged, the effective value of the synthesized space voltage vector will not be affected. Simultaneously, changing the positions of the zero vectors V0 and V7 in the pulse width distribution reduces the number of switching operations; this is the discontinuous SVPWM method.
Different zero vector allocations and positions result in different modulation effects. If a suitable zero vector is fixed within the sector sandwiched by three adjacent vectors, each group can have a 120° sector without switching within one cycle. Since the non-switching range for each phase is a continuous 120° area, the switching losses of the upper and lower bridge arms are inconsistent, and the waveform distortion is much greater than that of SVPWM. If different zero vectors are selected in adjacent 60° intervals, there are three zero vector allocation schemes. In practical applications, each phase switching device should be switched in the range with lower load current as much as possible, and the sector with higher load current should be arranged in the non-switching sector. This not only reduces the number of switching operations but also effectively reduces the maximum switching current of the switching devices, thereby minimizing switching losses. This method can reduce the number of switching operations to 1/3 of the original SVPWM, greatly reducing switching losses. Simultaneously, because the position of the inserted zero vector changes the inverter's freewheeling process, it also has a certain effect on suppressing current waveform oscillations and distortion. In engineering, implementing dead-time effect compensation techniques during modulation using this method presents certain difficulties. An effective method is to compensate for the effective conducting main vector within each sector.
4. Random Pulse Width Modulation (RPWM)
Noise has long been a concern in AC drive systems powered by frequency converters. In some low-noise environments, the noise emitted by frequency converters and motors can be unbearable. Frequency converter noise is primarily caused by the pulse width modulation (PWM) method used in the inverter. In conventional PWM methods, the inverter's power switches are turned on and off in a deterministic manner. While this control method effectively suppresses low-order harmonics in the voltage waveform, it generates some high-order harmonics with large amplitudes. These harmonics are mainly concentrated around one and two times the carrier frequency, producing significant noise and vibration. In recent years, random pulse width modulation (RPWM) has emerged, offering a novel approach to solving inverter noise problems. The basic idea of RPWM is to replace the fixed switching pattern in conventional PWM with a random switching strategy, thereby distributing the harmonic spectrum of the inverter output voltage evenly over a wider frequency range, achieving the goal of suppressing noise and mechanical vibration.
There are currently three feasible RPWM schemes:
(1) Randomized switching frequency
In traditional SPWM, by randomly varying the slope of the triangular carrier wave, the number of switching cycles per week can be randomly varied, thus achieving the goal of random switching frequency.
(2) Randomized pulse position
In this scheme, the random quantity is the position of the switching signal pulse within each on/off cycle. The simplest version uses only two random bits, one at the beginning and one at the end.
(3) Random switch
The random wave is compared with a sinusoidal reference signal, and the result of the comparison forms a digital RPWM signal.
Based on existing space vector pulse width modulation technology, random PWM can be achieved by using randomized pulse position.
The optimized SVPWM above analyzed how different zero vector positions reduce system switching losses. If these optimized SVPWMs are randomly modulated, the pulse positions will randomly change due to the different zero vector insertion positions in each carrier cycle, achieving the purpose of random PWM modulation. Currently, a relatively simple and practical method uses only two fixed zero vectors that are randomly switched. A random function generates two random states, 0 and 1. If it is 0, the zero vector V0 is applied at both ends of the switching cycle; if it is 1, the zero vector V7 is applied in the middle of the switching cycle. This method is essentially a random switching of two low-switching-loss modulations.
The better the randomness of the states generated by the random function, the more modulation states are switched, and the better the continuous distribution of system harmonic energy, resulting in a better performance of random PWM. However, this also increases the difficulty of algorithm implementation and the real-time requirements of the controller. Using the RPWM method can effectively reduce inverter system noise and transform certain concentrated, discrete high-frequency components into continuous, evenly distributed frequency components, thus reducing propagating electromagnetic interference.
5. Overmodulation technique
Overmodulation technology is primarily based on Space Vector Pulse Width Modulation (SVPWM). For high-performance AC drive systems, fully utilizing the DC voltage to obtain the maximum output electromagnetic torque is a crucial factor. Especially during the field weakening phase, it is necessary to control the inverter to operate within the overmodulation range to obtain sufficient voltage. With traditional SPWM control, the inverter output voltage can only reach 78.54% of the square wave condition, while SVPWM can increase the output voltage to 90.69% of the square wave condition. To obtain a higher output voltage, the inverter must operate in the overmodulation region until it reaches the square wave condition.
Currently, there are many space voltage vector overmodulation methods proposed in academia, with varying algorithmic complexity and effectiveness. However, output voltage vector modulation methods generally fall into two categories: dual-mode control, which divides the overmodulation interval into two parts for separate modulation; and single-mode control, which treats the overmodulation interval as a whole. Essentially, single-mode is simply an engineering simplification of dual-mode, making it easier to implement, but it produces a lower fundamental voltage and higher harmonic content compared to dual-mode. If the controller's processing speed and storage space are sufficient, dual-mode control can be used to improve the system's output characteristics. The relationship between the angle and modulation ratio can be obtained offline and stored in RAM, or calculated online using a fitted curve.
A typical V/f frequency converter system consists of three parts: control, modulation, and main circuit. The control and pulse width modulation (PWM) parts are entirely implemented using software algorithms. This control method is derived from the steady-state model of AC motors and is independent of motor parameters and their changes, making it simple and easy to implement. However, its speed range is relatively narrow, suitable only for loads such as fans and pumps with low speed control requirements. To improve the system's low-speed load-carrying capacity and dynamic performance, and to meet the needs of actual industrial environments, existing control methods and PWM strategies must be improved and enhanced.
