1 Introduction The application of AC drives in high-performance applications began with the introduction of the concept of vector control, including direct field orientation and indirect field orientation control. Although this concept appeared as early as the 1960s and was formally proposed by Dr. Blaschke of Siemens in 1972[1], its real application was only twenty years after the development of microelectronics technology. Vector control can achieve excellent dynamic and static characteristics in principle, but the sensitivity of motor parameters has become a problem that must be solved in practical applications. The driver adapts to changes in motor parameters and maintains the dynamic and static performance of vector control through self-tuning before startup and online tuning during operation. These complex adaptive control algorithms must be completed by a powerful signal processor. In recent years, with the development of semiconductor technology and the popularization of digital control, the application of vector control has expanded from the high-performance field to general-purpose drive and special-purpose drive applications, and even home appliances. AC drives have been widely used in industrial robots, automated publishing equipment, processing tools, transmission equipment, elevators, compressors, rolling mills, fans and pumps, electric vehicles, lifting equipment and other fields. With the rapid development of semiconductor technology, power devices are constantly being optimized, switching speeds are increasing while losses are decreasing, and the power density of power modules is continuously increasing. Digital signal processors are becoming increasingly powerful, with ever-improving processing speeds. AC drives are fully capable of handling complex tasks and implementing complex observation and control algorithms, thus achieving unprecedented performance levels in modern AC drives. Taking AC servo drives, representing the highest level of AC drive control, as an example, their demand is rapidly growing with the emergence of new production technologies and new processing materials. According to relevant statistics, the annual growth rate of the number of high-performance AC servo drives exceeds 12%. The most commonly used motors in servo drives are asynchronous and synchronous motors, with rated power ranging from 50W to 200kW. The typical bandwidths of the position loop, speed loop, and torque loop are 60Hz, 200Hz, and 1000Hz, respectively. Most problems in AC motor drives have been solved in today's drives. Related mature technologies provide widely accepted solutions and have been successfully applied in many fields. Therefore, from a basic structural perspective, the existing design schemes of AC drives will not undergo significant changes in the next few years. Currently, a key focus in AC drive development is how to organically integrate the drive with the motor to develop a lower-cost, higher-reliability, and higher-performance "drive module." Based on this idea, to further reduce costs and improve reliability, developers have conducted in-depth research on how to eliminate shaft-side sensors and motor phase current sensors. In particular, the implementation of high-performance sensorless vector control (SVC) has attracted widespread attention from researchers worldwide and has become a hot topic in drive control research in recent years. With the advent of powerful digital signal processors, the high robustness, adaptive parameter estimation, and nonlinear state observation required for this control method have become possible, leading to the continuous emergence of new sensorless control schemes. Well-known companies such as Siemens, Yaskawa, Toshiba GE, Rockwell, Mistubishi, and Fuji have all launched their own SVC control products (SVC referred to in this article refers to asynchronous motors), and their control characteristics are constantly improving. SVC is currently widely used in printing, dyeing, textile machinery, steel production lines, hoisting, and electric vehicles, playing an increasingly important role in high-performance AC drives. 2. Advantages of Sensorless Vector Control In summary, sensorless vector control achieves near-closed-loop control performance while eliminating the need for a speed sensor, resulting in lower maintenance costs. Compared to traditional V/Hz control, sensorless vector control offers improved low-speed operating characteristics, enhanced speed regulation under varying loads, and high starting torque, which is particularly advantageous for starting high-friction and inertial loads. Due to these drive characteristics, this control technology has gradually become the choice for general constant torque drive applications. In fact, virtually all AC drive manufacturers offer this control mode. Susan Bowler, Drive Market Manager at Schneider Electric, believes the appeal of this control mode lies in achieving significantly enhanced performance, including low-speed characteristics, torque response, and positioning capabilities, with minimal additional cost. Because its performance is close to that of servo drives, the company is expanding into applications requiring more precise load positioning control. The company's third-generation Altivar sensorless drive products feature self-tuning capabilities, ensuring that the drive continuously optimizes motor operating characteristics even as motor operating parameters change over time. The control algorithm calculates the optimized motor voltage at the set speed to achieve maximum torque output. The motor model already considers the effects of thermal effects. Kirkpatrick, Siemens' AC drive product manager, believes that most AC drive products currently default to SVC control. Closed-loop flux vector control (FVC) is only used in applications requiring more stringent speed control and zero-speed torque control. Due to the higher cost of FVC and the complexity of issues related to encoders, cables, and installation wiring, its sales volume is not large. 3. The Current Status of Sensorless Vector Control Sensorless control, this advanced drive method for induction motors, fills the gap between high-performance closed-loop control and simple open-loop control, with a price commensurate with the drive performance it provides. Although it omits the speed sensor used in closed-loop control, SVC still requires voltage and current sensors to control the motor. On a high-speed processor platform, it uses a complex motor model and intensive mathematical calculations to process the sensor input signals to obtain the flux and torque components required for motor control. Then, an adaptive magnetic field vector method is used to achieve decoupled control and obtain good dynamic response. It should be said that there is currently no standard solution for this control method. In the past ten years, researchers have published many papers [1]-[16], proposing many different ideas. In fact, many companies have also adopted their own different sensorless control schemes in their general-purpose frequency converters. Their drive performance is not the same. This is related to the fact that the core of the scheme is based on V/Hz or field orientation. Most sensorless AC drives belong to the sensorless vector type, while direct torque control (DTC) belongs to another type. Kerkman of Rockwell believes that high-performance sensorless control originates from closed-loop field orientation flux vector control, which is based on rotor flux vector; while relatively low-performance schemes are based on stator flux vector and some simple control algorithms. The accurate estimation of slip frequency in SV control technology is the difficulty. The amount required to calculate this frequency is a basic control quantity for SVC, so it involves many aspects. Eckardt of Siemens Standard Drives R&D believes that while the magnetic field of a high-speed motor can be directly calculated from the motor's back EMF, calculating stator flux is more difficult at low speeds (especially near zero frequency); and at zero frequency, theoretically, stator flux is unobservable. Figure 1 shows the control block diagram of Rockwell's FORCE series products. This product uses a simplified voltage model on which the parameters are insensitive to changes in motor operating temperature rise. Figure 1 SVC control scheme (Rockwell) At Mitsubishi, advanced flux vector control represents the latest sensorless control technology. This technology further optimizes the technology previously developed by the company in 1993, aiming to improve output torque and operational stability during low-speed sensorless operation. Kantarek, the company's AC drive market manager, believes that the excellent characteristics of SVC control can be applied to most constant torque operation applications, especially those requiring high starting torque and smooth low-speed operation, and that SVC drives have now evolved to the point where they can replace DC drives. According to Kantarek's introduction, Mitsubishi's SVC control first adopts the internal characteristic model of the motor, and then samples the motor model every few milliseconds through self-tuning. The driver decomposes the output current into excitation and torque current. The motor stator flux is kept at a stable value through the corresponding voltage compensation, and the slip frequency is further calculated. Direct torque control (DTC) is another sensorless control solution that has attracted widespread attention today [30]. ABB launched its direct torque control product ACS600 in 1995, and it has now been upgraded to ACS800. Its control block diagram is shown in Figure 2. DTC uses a separate loop to control the speed and torque of the motor. Gokhale, ABB's AC drive R&D manager, explained, "DTC has been a sensorless control structure since its development. It is essentially a torque control scheme, not a vector control." It can be clearly seen from Figure 2 that DTC removes the current regulator or voltage command generation link in typical vector control. Instead, there are two hysteresis control links, which estimate and control the flux and torque every 25μs. In this control structure, the integral drift of low-speed flux identification and the influence of stator resistance changes directly limit the minimum operating range of the driver. Since the system lacks intermediate torque current and flux current control links, DTC lacks direct current control. Generally speaking, DTC directly controls torque and indirectly controls current. Figure 2 shows the DTC control block diagram (ABB). Due to these characteristics, some researchers refer to DTC as essentially "advanced scalar control." Due to space limitations, this paper will not elaborate further; the following will only discuss SVC. The key to SVC control lies in accurate speed estimation and decoupling control, but these two are mutually coupled. The accuracy of speed estimation depends not only on the measured stator voltage and current but also on the motor parameters. In digital motor control systems, the accuracy of speed estimation is related to the sampling frequency and the resolution of the feedback signal. The accuracy of speed estimation not only affects the accuracy of speed control but also the design of the speed loop compensator. These issues are interconnected, and even slight errors can affect the stability of the system. To make SVC technology practical, several basic problems must be solved: flux identification, speed estimation, and parameter adaptability. Over the past decade, researchers have developed a variety of flux identification and speed estimation methods. The most widely used flux identification models include: open-loop voltage model [9], closed-loop composite model [3], and adaptive flux observation model [2]. The open-loop voltage model has integral drift at low speeds and is sensitive to parameters. The divergence problem caused by integral drift is solved by introducing a low-pass circuit or multiple cascaded low-pass circuits, but it will introduce amplitude and phase distortion. Therefore, high-performance sensorless control must introduce appropriate compensation methods. The closed-loop composite model completes the smooth switching between high and low speed models by calculating the estimation error between the voltage model and the current model. In actual design, it is usually necessary to select an appropriate gain. The adaptive flux observation model eliminates the influence of parameter changes on flux observation through an adaptive circuit and can be applied to direct rotor field-oriented control. Some speed estimation methods estimate the speed based on the motor terminal voltage and current, while others use an observer to estimate the speed. The basic idea of ​​speed estimation is to use stator voltage, current and frequency to calculate rotor speed. These methods can be basically divided into: (1) Speed ​​estimation method based on slip frequency [17]-[19]; ​​(2) Speed ​​estimation method based on field orientation [20]-[28]; (3) Speed ​​estimation method based on adaptive control [29]; (4) Speed ​​estimation method based on observer [31]-[36]. Among them, the speed estimation method based on field orientation has become the mainstream of industry design due to its speed and high accuracy. Both flux identification and speed estimation are highly dependent on parameters. It is precisely because of this that SVC is more sensitive to changes in motor parameters than closed-loop flux vector control (FVC) which uses speed or position sensors. It lags behind FVC control in dynamic indicators such as speed regulation and torque response. At present, the industry design for SVC parameter tuning includes two types: initial tuning and online tuning. In initial tuning, some manufacturers only need to input the motor nameplate parameters, while others require separate identification of stationary and rotating parameters (offline identification). For example, the GE Fuji AF-300 G11 dynamic torque vector control driver offers both offline and online tuning methods. This product has a subroutine that tracks the motor's operating status and observes parameter changes caused by temperature or load variations. By continuously refreshing the motor parameters during operation and utilizing its unique mathematical model to adjust voltage and current, it optimizes the motor's low-speed performance. Typical online tuning methods include EKF, MRAC, and directly solving the motor's DQ model equations. It is well known that the rotor time constant plays a crucial role in field orientation. In sensorless control, independently identifying the rotor speed and rotor time constant is essential. One approach is to inject higher harmonics, but this requires attention to the resulting speed and torque fluctuations, as the harmonic amplitude needs to be relatively large for effective identification. Alternatively, some researchers have proposed using motor rotor slot harmonics to independently identify the speed. Research on parameter adaptation is still ongoing, and improving the adaptability and robustness of SVC systems is undoubtedly an important research topic. Overall, because it eliminates the need for speed sensors, the motor control model for SVC requires high precision. In terms of computation, SVC control is more complex than FVC, making sensorless control significantly more difficult than closed-loop control. Since motor parameters vary greatly during operation, the self-tuning capability of the SVC driver is crucial for obtaining accurate motor parameters, directly determining the performance of vector control. In fact, adapting to changes in motor operating conditions and maintaining model accuracy is key to avoiding high torque fluctuations; the model's adaptive capability is also the most important factor when the motor is operating near zero speed, as the error of the motor reference model increases significantly at this point. Due to the use of an enhanced motor model that adapts to changes in motor operating conditions, GE Toshiba reported that its products, under certain slip and load conditions, have reduced torque ripple from 7% to less than 2%; torque regulation accuracy is within the range of 1-2%, while speed stability accuracy is within 0.1% of the rated speed. Despite employing adaptive, precise motor models, the current highest-level SVC control still lags behind FVC in both dynamic and static characteristics, particularly noticeable in the low-speed operating range. The limits of SVC's low-speed capability are also related to factors such as load inertia and variations; torque control is relatively easier at 1Hz, and possibly possible around 0.5Hz, depending on the specific application, but torque control at speeds far below this will be quite difficult for SVC. To achieve full torque and very precise torque control near zero speed (typically 5% below base speed), or to achieve a speed stability of 0.01% of the rated speed, encoder feedback is essential. When selecting an SVC driver, its dynamic response must be examined, and the response speed of SVC can differ from FVC by up to 15 times; these differences must be carefully considered for high-performance applications. The attached table compares the performance of SVC with other control methods. Commercialized SVCs still require addressing many details. Achieving high-performance SVC control and stable operation in complex industrial environments necessitates meticulous research into these issues, requiring considerable effort from R&D personnel at various companies. The following are some typical key issues: (1) Low-speed operating region (2) Field weakening operating region (3) Regenerative mode operation (4) Dead zone compensation (5) Digital integral method (6) Selection of PI controller type (7) Steady-state accuracy of speed identification (8) Speed ​​variation of dynamic load (9) Consideration of sampling delay effect (10) Stability of system with respect to parameter changes (11) Magnetic saturation (12) Skin effect 4 Development direction of sensorless vector control In summary, further improvement of the dynamic and static characteristics of sensorless vector control in the future requires a more complete inverter/motor model, which comprehensively considers factors such as motor magnetic circuit saturation, winding skin effect, inverter nonlinearity, and motor parameter changes under different operating conditions. Based on a more accurate adaptive motor model, low-speed torque ripple will be further reduced, speed stability will be further improved, response to load disturbances will be faster, and stability to motor parameter changes will be further strengthened. In particular, the gap between SVC control systems, which feature wide-range speed regulation (including zero speed) and high-precision speed regulation and torque control (not just torque limitation), and FVC control systems will gradually narrow, and SVC is expected to replace some servo applications. Future advancements will also be reflected in high-speed processors and peripherals. The DSP+ASIC/FPGA controller architecture makes the system's signal parallel processing capability more powerful, enabling the core program to run at very high speeds, ensuring faster response of the SVC system to speed commands and load changes, which is crucial for high-performance digital control systems. Furthermore, multi-machine operation under sensorless control and applications in high-power, low-speed operation will also become future development directions. 5 Conclusion Sensorless vector control (SVC) eliminates speed sensors and related encoder wiring, reducing system maintenance costs and improving system reliability, laying the foundation for integrated inverter/motor design. Advanced SVC control, on a high-speed digital signal processing platform, significantly improves the dynamic and static performance of the driver by establishing accurate motor models and introducing advanced control strategies, and is poised to replace some closed-loop vector control and servo control applications. SVC has become the de facto standard for general-purpose frequency converters, and its application areas will be further expanded.