A kind of MHE prediction estimation sensorless control system and method for maglev train

Abstract

The present disclosure relates to a kind of MHE prediction estimation sensorless control system and method for maglev train, belong to the control technical field of high-speed maglev train long stator linear synchronous motor.It is established to include the discrete state space model of extended counter electromotive force and lumped disturbance, constructs the estimated value function of rolling time domain, adopts Sage-Husa adaptive filtering to correct the initial state estimation error weight of rolling window;Again, the rolling time domain estimation problem is converted into quadratic programming problem with equality constraint, and the optimal state estimation value is obtained by solving KKT equation, finally, according to the extended counter electromotive force estimation value at the end of window, the electrical angle and electrical angular velocity are obtained by normalizing phase-locked loop, to realize traction current closed-loop control and speed closed-loop control.The present application reduces the cumulative error in the recursive estimation process, and improves the anti-disturbance performance and dynamic response performance of long stator linear synchronous motor position sensorless control.

Description

Technical Field

[0001] This disclosure relates to the field of long stator linear synchronous motor control technology for high-speed maglev trains, and in particular to a sensorless control method and system for MHE predictive estimation of maglev trains. Background Technology

[0002] Normal-conducting high-speed maglev trains use long-stator synchronous linear motors (LSMs) to generate traction thrust. The field-oriented control of these LSMs requires real-time and accurate acquisition of the traveling wave magnetic field position and velocity information. Position acquisition errors directly affect current decoupling, thrust output, starting performance, and train operation safety. During low-speed operation, the normal-conducting train uses a speed measurement and positioning system to acquire position signals. However, when the train speed exceeds 100 km / h, a sensorless control method based on back EMF detection is used to estimate the position signal. Currently, the mainstream sensorless control schemes for long-stator synchronous linear motors in high-speed maglev trains include four main approaches: back EMF observation, high-frequency signal injection, Kalman filtering, and sliding mode / model reference adaptive observer methods. In low-speed and zero-speed conditions, positioning methods mainly rely on high-frequency signal injection or tooth groove detection, position markers, etc.; extended Kalman filter (EKF), unscented Kalman filter (UKF), sliding mode observer (SMO), model reference adaptive (MRAS) and other state observation methods are widely used to improve estimation robustness and dynamic performance; in medium- and high-speed conditions, observation methods based on back electromotive force have been implemented in engineering applications.

[0003] Current mainstream sensorless control schemes for linear synchronous motors in conventional high-speed maglev trains generally suffer from common defects such as narrow applicable speed range, weak anti-disturbance capability, poor low-speed accuracy, excessive sensitivity to models and parameters, and insufficient dynamic robustness, making it difficult to meet the requirements for reliable operation under all operating conditions. Back-EMF-based observation methods suffer from weak signals and extremely low signal-to-noise ratios at low and zero speeds, making effective estimation impossible. While high-frequency signal injection methods can be used at low speeds, they are affected by long stator end effects and high-power, low-switching-frequency conditions, resulting in severe signal distortion, large delays, and phase errors, thus limiting their practical engineering applicability. Traditional state observers such as EKF, SMO, and MARS cannot explicitly observe and compensate for lumped disturbances inside and outside the system in real time, leading to significantly increased estimation errors under complex conditions such as changes in motor parameters, sudden changes in resistance, and track irregularities. Specifically, EKF relies on local linearization, resulting in significant accuracy degradation under strong nonlinearity; SMO suffers from inherent chattering; and MARS is highly sensitive to parameter mismatch, exhibiting slow overall dynamic response and large transient fluctuations. Furthermore, existing methods do not incorporate the strong constraints and disturbances of maglev systems into their optimized design, making it difficult to simultaneously guarantee steady-state accuracy and dynamic stability across the entire speed range and under conditions of strong disturbances. Summary of the Invention

[0004] To address the problems of narrow applicable speed range, weak anti-disturbance capability, poor low-speed accuracy, excessive sensitivity to models and parameters, and insufficient dynamic robustness of the sensorless control method for long stator linear synchronous motors in maglev trains, which makes it difficult to simultaneously guarantee steady-state accuracy and dynamic stability across the entire speed range and under strong disturbance conditions, this disclosure aims to solve at least one of these problems. It proposes a sensorless control system and method for predictive estimation of MHE for maglev trains, and the specific technical solution is as follows.

[0005] In a first aspect, a sensorless MHE predictive estimation control system for maglev trains includes: a coordinate transformation module configured to transform the static coordinates of the long stator linear synchronous motor of the maglev train. Coordinate system current , Combined with electrical angle , to current , Switch to dynamic Coordinate system obtained , The MHE prediction estimation module is configured to estimate the current. and ,Voltage and Electric angular velocity As input, using a pre-defined value function that includes extended back EMF, lumped disturbance, and measurement error, the end of the window is calculated by minimizing the value function within a preset rolling estimation time window. Axial extension back EMF and The extended back EMF estimate is used to calculate the traveling wave magnetic field electric angle using the MathCal electric angle calculation module via a normalized phase-locked loop. and electric angular velocity The speed calculation module is configured to calculate the train's running speed based on the relationship between the electric angular velocity and the motor pole pitch, obtaining a speed estimate; the ASR speed regulator is configured to generate the traction current based on the speed setpoint and the speed estimate; the ACR current regulator is configured to generate the traction current based on the traction current and the current output from the coordinate transformation module. , Adjustment, output voltage control command and The inverter drive module is configured to respond to voltage control commands. and Combined with electrical angle The voltage required to generate a long stator motor and This drives the linear synchronous motor of the maglev train's stator; the aforementioned ASR speed regulator, ACR current regulator, coordinate transformation module, MHE prediction and estimation module, MathCal electrical angle calculation module, speed calculation module, and inverter drive module constitute a dual closed-loop feedback structure.

[0006] Secondly, a sensorless control method for the stator motor of a maglev train conductor based on rolling time-domain estimation is provided, comprising the steps of: [the method is described in the original text, but the provided text is incomplete and requires further context to translate accurately.] Coordinate system current , Combined with electrical angle , to current , Switch to dynamic Coordinate system obtained , Based on current and ,Voltage and Electric angular velocity Using a preset value function that includes extended back EMF, lumped disturbance, and measurement error, the end of the window is obtained within a preset rolling estimation time window. Axial extension back EMF and Extended back EMF estimate; based on the extended back EMF estimate, the traveling wave magnetic field electric angle is calculated using a normalized phase-locked loop. and electric angular velocity The electric angular velocity Used to generate speed estimates, these speed estimates are combined with a speed setpoint to generate the traction current, which is then used in the current... , Adjusting the output voltage control command and According to voltage control commands and Combined with the estimated electrical angle , obtain voltage and This drives the linear synchronous motor of the maglev train's stator, thereby achieving sensorless current and speed closed-loop control.

[0007] The beneficial technical effects of this disclosure include: 1) This invention utilizes rolling time-domain multi-step optimization estimation instead of single-step recursive estimation, which can suppress the cumulative error in angle estimation caused by long-term operation and improve the accuracy of electrical angle observation during zero-speed start-up and low-speed operation. 2) This invention employs Sage-Husa adaptive filtering to correct the initial state estimation error in the arrival cost online, making the reliability of the initial state of the rolling window adaptively change with the operating conditions, avoiding estimation divergence or response hysteresis caused by fixed initial value weights under disturbed operating conditions. 3) This invention transforms rolling time-domain estimation into fixed-dimensional equality constraint quadratic optimization and solves it quickly using KKT equations, making it suitable for real-time online implementation in traction controllers. 4) This invention obtains electrical angle and electrical angular velocity based on the extended back EMF state quantity through a normalized phase-locked loop, eliminating the need to inject additional high-frequency signals during the zero- and low-speed stages of long stator motors, thus reducing additional losses and electromagnetic disturbances. 5) This invention can be used for operating conditions such as constant thrust, constant speed, sudden drag changes, and speed steps, improving the anti-disturbance performance and dynamic response performance of sensorless control of long stator linear synchronous motors. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating the calculation of the electric angle of a maglev train based on MHE in one implementation method.

[0010] Figure 2 This is a diagram of sensorless closed-loop control of a maglev train based on MHE in one embodiment.

[0011] Figure 3 This is a simulation result of the estimated speed tracking of a maglev train at 600 km / h based on MHE in one implementation.

[0012] Figure 4 This is a simulation result of the speed estimation error of a maglev train at 600 km / h based on MHE in one implementation method.

[0013] Figure 5 This is a simulation result of the estimated electrical angle difference for a maglev train traveling at 600 km / h based on MHE in one implementation.

[0014] Figure 6 This is a simulation result of the electrical angle estimation error of a maglev train at 600km / h based on MHE in one implementation.

[0015] Figure 7This is a simulation result of extended back EMF estimation for a maglev train traveling at 600 km / h based on MHE in one implementation method.

[0016] Figure 8 This is a simulation result of the three-phase current amplitude and phase of a maglev train at 600km / h based on MHE in one implementation. Detailed Implementation

[0017] This invention provides a sensorless control method and system for predictive estimation of the MHE (Mechanical High-Electromotive Force) in maglev trains, applicable to traction control scenarios of long stator linear synchronous motors in high-speed maglev trains. The method and system construct a discrete state-space model in a stationary coordinate system using three-phase current and voltage commands from the motor. The extended back electromotive force is used as the estimated state related to the electrical angle, and rolling time-domain estimation is introduced to optimize the state sequence within the window, thereby obtaining electrical angle, electrical angular velocity, and operating speed information. This information is then fed back into the traction current closed loop and the speed closed loop, achieving sensorless closed-loop control of the maglev train. This scheme can be used for zero-speed start-up, constant thrust operation, constant speed operation, constant speed variable resistance operation, or constant resistance variable speed operation. When the rolling window moves forward, the end state of the window obtained at the previous moment is used as the prior state of the next rolling window, and Sage-Husa adaptive filtering is used to correct the initial state estimation error of the rolling window, achieving online updating of the initial state estimation error weight.

[0018] The following provides a clear and complete description of how the technical solution of this case is implemented. Obviously, the described implementation methods are only a part of the implementation methods of this case, and not all of them. Based on the implementation methods in this case, all other implementation methods obtained by those skilled in the art without inventive effort are within the scope of protection of this application.

[0019] like Figure 1 As shown, Figure 1 This is a flowchart of the electric angle calculation for a maglev train based on MHE.

[0020] Step S1: Collect the three-phase armature current of the long stator linear synchronous motor and the voltage command output by the controller, and transform the three-phase armature current and voltage command to the αβ stationary coordinate system to obtain the measured output and input quantities.

[0021] First, the three-phase current of the long stator linear synchronous motor is collected. i a , i b and i c and three-phase voltage command u a , u b and uc The current in the stationary coordinate system is obtained by Clarke transformation. i α , i β and voltage u α , u β .

[0022] αβ The differential expression for the under-shaft current is as follows:

[0023] in, σ α , σ β They are respectively αβ Lumped disturbance on the axis; L d , L q respectively dq Inductance in the axial coordinate system; For armature winding resistance; The velocity of the mover; Stator pole pitch; The electric angle of the traveling wave magnetic field; For excitation flux; E ext To extend the back electromotive force and , i d , i q They are respectively dq Current value in the axial coordinate system . Figure 2 middle for Extended back electromotive force of the shaft ( ), for Extended back electromotive force of the shaft .

[0024] Will α shaft current i α and β shaft current i β As a measurement output, α Shaft voltage command u α and β Shaft voltage command u β As input.

[0025] Step S2: Establish a discrete state-space model of the long stator linear synchronous motor that includes extended back EMF and lumped disturbance, and determine the rolling window length N.

[0026] The discrete state-space model of a long stator linear synchronous motor is represented as follows: , in,

[0027] The values ​​in each vector correspond to the values ​​at time t.

[0028] and

[0029] in, , , , , These are, respectively, state variables, control variables, output variables, lumped disturbance variables, and sampling error variables. and These are the extended back electromotive force along the α-axis and the extended back electromotive force along the β-axis, respectively. ε α , ε β They are state variables i α , i β Sampling error (measurement error); i αm , i βm They are state variables i α , i β Actual sampled values. A is the state transition matrix, B is the control matrix, and C is the observation matrix. It represents the electric angular velocity.

[0030] Discretizing the above equation using the backward Euler method, it can be expressed as follows:

[0031] in, t Sampling time; A d It is a time-varying matrix. , T s The sampling period is I d It is a 4×4 identity matrix; B d It is a constant matrix. ; Represents the lumped disturbance term. .

[0032] The state variables in the state space include α shaft current, β shaft current, α Axial extended back EMF E α = φ 1= E ext sin θ e , β Axial extended back EMF E β = φ 2= E ext cos θ e and with electrical angle θ e Related state variables. The initial value of the electrical angle is obtained through positioning methods, including high-frequency signal injection or tooth groove detection, and position markers. After the speed crosses the switching threshold, it smoothly switches to the rolling time domain estimation mode.

[0033] Scroll window length N Example N=5. MHE will t -1 is truncated to 0. t - N -1 Historical estimation window and t - N arrive t- 1. A scrolling window that moves forward over time. N for tN and t- The rolling estimation time window length between 1 and 2, and MHE only applies to the sliding window. N The step size information processing achieves online estimation of the state variables at the latest time step by minimizing the value function.

[0034] Step S3: Construct a rolling time-domain estimation value function based on the discrete state-space model. The value function includes at least the arrival cost, system disturbance penalty term, and measurement error penalty term.

[0035] Constructing a finite rolling window based on rolling time-domain estimation N Value function within L tThe value function includes an arrival cost term, a system disturbance term, and a measurement error term. Specifically, a rolling window of length N is used as the estimation interval. At each sampling time, the voltage input and current measurement values ​​from the past N sampling periods are introduced into the optimization problem. An optimization objective function is established as the value function, comprising three parts: the first part is the arrival cost term, which describes the deviation between the initial state of the rolling window and the prior estimated state; the second part is the system disturbance term, which constrains the deviation between the model's predicted state and the actual state evolution; and the third part is the measurement error term, which constrains the deviation between the model's output current and the actual sampled current. Through this objective function, historical estimation information, system model constraints, and real-time measurement information can be considered, thereby improving the accuracy of state estimation.

[0036] In one embodiment, the rolling time-domain estimation value function is:

[0037] in, The confidence factor is greater than 0, representing the weight of the impact of the initial state estimation error on the optimization process; These are the initial state sample values; This is the initial state estimate; The estimation error per unit step represents the estimation measurement for each step within the time window. and punish. To account for measurement error, The total disturbance is denoted by N. N is the length of the rolling window, and t is the current time of the window.

[0038] according to Expanding the value function, we get:

[0039] in, This represents the initial sampled value at position tN at time t. This represents the initial estimate of the position of tN at time t. This represents the lumped disturbance at time i, where Ts is the sampling period; Let C represent the output at time i, and let C represent the observation matrix. This represents the state at time i, which consists of the current and the extended back electromotive force. .

[0040] The constructed value function is pre-defined and can be used directly in the system.

[0041] Step S4: Use Sage-Husa adaptive filtering to correct the weights of the initial state estimation error of the rolling window online.

[0042] Since the confidence factor in the value function in step S3 is used to limit the proportion of historical information outside the window in the current rolling window, in order to avoid estimation divergence caused by excessively large initial value weights or error accumulation caused by excessively small initial value weights, in actual train operation, this confidence factor should change with the initial state estimation error of the window. Therefore, this invention introduces Sage-Husa adaptive filtering, which uses measurement information to estimate the confidence factor online and corrects the observation noise covariance in real time, so that the filter gain and state estimation covariance are adaptive to the current measurement noise level, thereby weakening the continuous impact of initial estimation bias and prior statistical inaccuracies, and improving the accuracy of the uncertainty characterization of the value function.

[0043] Sage-Husa adaptive filtering is employed to correct the arrival cost of the Medium-Range Emission (MHE) online. Specifically, the state error covariance matrix is ​​recursively obtained from the Sage-Husa adaptive filtering, and its inverse matrix or equivalent weight matrix is ​​used as the confidence factor for the arrival cost term. When the initial state estimation uncertainty is large, the arrival cost weight is reduced accordingly, making the MHE more dependent on in-window measurement information; when the initial state estimation confidence is high, the arrival cost weight is increased accordingly, maintaining the continuity of the estimation results. This approach reduces the problem of continuous accumulation of initial errors in traditional recursive estimation methods and improves the estimation reliability under zero-speed start-up, low-speed operation, and high-speed weak back EMF disturbance conditions.

[0044] Step S5: Within the rolling window, the state estimation problem is transformed into an optimization problem with equality constraints, and the optimal state sequence within the window is solved using the KKT equations.

[0045] The rolling time-domain estimation value function within a rolling window is transformed into an equality-constrained quadratic programming problem. The equality constraints consist of discrete state equations, and the optimization variables are the state sequence and disturbance sequence within the rolling window. By constructing a set of KKT equations, the optimal state estimate at the end of the window is obtained by minimizing the arrival cost, system disturbance term, and measurement error term. Compared with the traditional point-by-point recursive filtering method, rolling time-domain estimation can simultaneously utilize historical input-output data within a time window, exhibiting better smoothness and robustness in the presence of measurement noise, model uncertainty, and load disturbances.

[0046] Next, based on the extended back EMF component estimated at the end of the window, the electrical angle and electrical angular velocity are calculated via a normalized phase-locked loop and used as feedback information for sensorless control.

[0047] Specifically, the state variables, lumped disturbances, and measurement errors at each sampling time within the rolling window are combined into augmented variables; the discrete state-space model is organized into matrix equality constraints; the value function is organized into a standard quadratic form; a Lagrangian function is constructed and a KKT linear equation system is formed; the optimal state increment and optimal state sequence within the rolling window are obtained by solving the KKT linear equation system.

[0048] ,

[0049] in, It is a Hessian matrix; ; ; For containing 4 N ×4( N +1) dimensional known matrix; X t For state variables to be estimated 4 ( N +1)×1 matrix; For containing ; and 4 N A known 1×1 dimensional matrix.

[0050] Step S6: Extract the extended back EMF state quantity at the end of the window in the optimal state sequence, obtain the traveling wave magnetic field electric angle and electric angular velocity through a normalized phase-locked loop, and feed the electric angle and electric angular velocity back to the traction controller to achieve position-free control.

[0051] Normalized phase-locked loop based on the end of the window α Axial extension back EMF and β The phase error signal is calculated by extending the back electromotive force of the shaft, and the estimated electric angular velocity is obtained by proportional-integral adjustment. The estimated electric angular velocity is then integrated to obtain the estimated electric angle.

[0052] In one embodiment, the extended back electromotive force in the stationary coordinate system can be expressed as: φ 1 and φ 2. When MHE obtains the estimated value at the end of the window. and Then, the sample is normalized to obtain the unit vector component related to the synchronous relationship between the rotor magnetic field direction or the traction traveling wave magnetic field, and the electrical angle estimate is extracted through a phase-locked loop. and estimated electric angular velocity Therefore, this invention can avoid noise amplification and phase jitter caused by directly differentiating or simply calculating the arctangent of the back EMF signal, and improve the continuity and stability of electrical angle estimation under low-speed and high-speed operating conditions.

[0053] One implementation method adopts Figure 2 The diagram shown illustrates the structure of a sensorless closed-loop control system for a maglev train based on the MHE (Mechanical Estimation and Prediction) module. This system includes an ASR (Automatic Speed ​​Regulator), an ACR (Automatic Current Regulator), an inverter drive module, a coordinate transformation module, an MHE prediction and estimation module, a MathCal electrical angle calculation module, and a speed calculation module. The ASR generates traction current commands based on the speed setpoint and speed estimate; the ACR generates commands based on the traction current and the current output from the coordinate transformation module. i d , i q Adjustment, output voltage control command and The inverter drive module controls the voltage according to the command. and The electrical angle estimated by combining the MathCal electrical angle calculation module The voltage required to generate a long stator motor and This drives the linear synchronous motor of the maglev train's stator; the coordinate transformation module is used to convert the three-phase current... i a , i b , i c Transform to αβ Obtained from the stationary coordinate system i α , i β At the same time, combined with the calculated electrical angle, αβ Current conversion from stationary coordinate system to dynamic coordinate system dq Coordinate system obtained i d , i q The MHE prediction and estimation module is used to estimate the current. and ,Voltage and Electric angular velocity Using a preset value function that includes extended back EMF, lumped disturbance, and measurement error as input, the end of the window is obtained within a preset rolling estimation time window. Axial extension back EMF and Extended back EMF estimate; MathCal electric angle calculation module is used to calculate the electric angle based on the extended back EMF estimate. and electric angular velocity This enables sensorless current and speed closed-loop control. The speed calculation module calculates the train's running speed based on the relationship between the electrical angular velocity and the motor pole pitch, obtaining a speed estimate. The ASR speed regulator, ACR current regulator, coordinate transformation module, MHE prediction and estimation module, MathCal electrical angle calculation module, speed calculation module, and inverter drive module together form a dual closed-loop feedback structure.

[0054] exist Figure 2 In the closed-loop control process shown, the speed controller generates a value based on the deviation between the target speed and the estimated speed. q Axle traction current reference value, current controller according to d shaft and q Shaft current error output voltage command and The voltage command is used to drive the inverter and also serves as input to the MHE prediction estimation model. Thus, the MHE estimator, current closed loop, and speed closed loop form a closed-loop coupling relationship, enabling the system to achieve traction control of the maglev train's stator linear synchronous motor without relying on external position sensors.

[0055] In addition, the estimated electric angular velocity output by the MathCal electric angle calculation module As input to the MHE prediction and estimation module, it is used for calculating velocity estimates. On the one hand, it achieves closed-loop speed control; on the other hand, it inputs the MHE state-space model to update the state transition matrix A, thereby achieving online adaptive model parameter control.

[0056] In one embodiment, simulation is performed using the above method and steps to obtain... Figure 3 The figure shows the simulation results of the maglev train's estimated speed tracking at 600 km / h based on the MHE (Mechanical Estimation and Performance) method. This figure illustrates that under high-speed operation at 600 km / h, the speed estimate based on the MHE can follow the actual speed changes of the maglev train. Because the MHE simultaneously utilizes the motor model, historical current measurements, and voltage input information for optimized estimation, it can maintain good speed tracking capability during high-speed operation, providing reliable feedback for speed closed-loop control.

[0057] In one embodiment, simulation is performed using the above method and steps to obtain... Figure 4The figure shows the simulation results of the speed estimation error of a maglev train at 600 km / h based on the MHE method. This figure illustrates the variation of the speed estimation error under high-speed conditions using the MHE prediction estimation method. The simulation results show that after using rolling window optimization estimation, the speed estimation error can be kept within a small range and converges quickly under speed changes or disturbances. This demonstrates that the present invention can improve the speed estimation accuracy and dynamic response capability of a sensorless control system under high-speed operation.

[0058] In one embodiment, simulation is performed using the above method and steps to obtain... Figure 5 The figure shows the simulation results of the estimated electrical angle difference for a maglev train at 600 km / h based on the MHE (Magnetic Helicopter Electromagnetic Elevation) system. This figure characterizes the phase difference change between the estimated electrical angle and the reference or actual electrical angle. For a long stator linear synchronous motor, the accuracy of the electrical angle estimation directly affects coordinate transformation, traction current decoupling control, and electromagnetic thrust output. Figure 5 This demonstrates that the electrical angle estimated based on MHE can maintain good synchronization with the actual electrical angle, avoiding issues caused by angle deviation. d shaft and q Shaft current coupling error.

[0059] In one embodiment, simulation is performed using the above method and steps to obtain... Figure 6 The figure shows the simulation results of the electrical angle estimation error for a maglev train at 600 km / h based on the MHE method. This figure further illustrates the amplitude and convergence characteristics of the electrical angle estimation error. The simulation results show that, under 600 km / h operating conditions, the MHE prediction estimation method can effectively suppress the electrical angle estimation error and ensure the continuity and smoothness of the electrical angle estimation. Because this invention uses extended back EMF estimation combined with a normalized phase-locked loop to obtain the electrical angle, it can reduce the impact of current noise, voltage error, and model parameter deviation on the electrical angle calculation.

[0060] In one embodiment, simulation is performed using the above method and steps to obtain... Figure 7 The figure shows the simulation results of extended back EMF estimation for a maglev train at 600 km / h based on MHE. This figure illustrates the effectiveness of MHE in estimating the extended back EMF components in a stationary coordinate system. The extended back EMF contains information related to electrical angular velocity, electrical angle, and flux linkage, and is an important intermediate variable for extracting electrical angle and velocity in sensorless control. Optimizing the extended back EMF estimation through rolling time-domain estimation can reduce estimation fluctuations of traditional back EMF observers under low-speed, noise, and parameter mismatch conditions, thus improving the stability of subsequent phase-locked loop calculations.

[0061] In one embodiment, simulation is performed using the above method and steps to obtain... Figure 8The figure shows the simulation results of the three-phase current amplitude and phase of a maglev train at 600 km / h based on the MHE (Mechanical Helicopter Equivalent) method. This figure illustrates that under the described sensorless closed-loop control method, the three-phase current of the long stator linear synchronous motor can maintain good amplitude consistency and phase symmetry. The stability of the three-phase current amplitude and phase reflects the coordination between electrical angle estimation, current closed-loop control, and inverter drive control. Simulation results show that the sensorless control method based on the MHE method can provide reliable angle information for coordinate transformation and current regulation, thereby ensuring balanced output of the three-phase traction current and stable generation of electromagnetic thrust.

[0062] In one embodiment, the MHE predictive estimation sensorless control system for maglev trains includes: a data acquisition unit for acquiring the three-phase current of a long stator linear synchronous motor and obtaining three-phase voltage commands; a coordinate transformation unit for transforming the three-phase current and three-phase voltage commands to a stationary coordinate system and a dynamic coordinate system; a model building unit for establishing a discrete state-space model including extended back EMF and lumped disturbances; a rolling time-domain estimation unit for constructing and solving an optimization problem including arrival cost, system disturbance term, and measurement error term; a covariance update unit for online correction of arrival cost weights using Sage-Husa adaptive filtering covariance; a KKT solving unit for solving a quadratic programming problem with equality constraints and outputting the optimal state estimate at the end of the window; an electrical angle calculation unit for obtaining the electrical angle and electrical angular velocity from the extended back EMF estimate via a normalized phase-locked loop; and a closed-loop control unit for completing traction current closed-loop and speed closed-loop control based on the estimated electrical angle, electrical angular velocity, and speed estimate.

[0063] By employing the above technical solution, this invention enables the estimation of the electrical angle, electrical angular velocity, and speed of the linear synchronous motor of the long stator of a maglev train without installing mechanical position sensors or in situations where position sensors fail, experience delays, or suffer from reduced accuracy. Compared to traditional recursive observers, this invention fully utilizes historical input and output information through rolling window optimization estimation and corrects arrival cost weights using covariance, thereby reducing the impact of initial estimation errors and measurement noise on the estimation results. This improves the accuracy and disturbance rejection performance of the sensorless control system under conditions of high-speed operation at 600 km / h, speed step changes, load disturbances, and parameter mismatch.

[0064] This invention can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.

[0065] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk drive or magnetic tape drive. The storage media described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0066] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0067] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0068] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0069] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0070] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0071] The methods disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0072] The features disclosed in the several product embodiments provided by this invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0073] The features disclosed in the several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0074] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various equivalent substitutions or obvious modifications can be made without departing from the concept of the present invention, and all such modifications, achieving the same performance or application, should be considered within the scope of protection of the present invention.

[0075] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of the invention is defined by the appended claims.

Claims

1. A sensorless control system for predictive estimation of magnetic levitation trains (MHE), characterized in that, include: The coordinate transformation module is configured to transform the static coordinates of the stator linear synchronous motor of the maglev train. Coordinate system current , Combined with electrical angle , to current , Switch to dynamic Coordinate system obtained , ; The MHE prediction estimation module is configured to estimate the current. and ,Voltage and Electric angular velocity As input, using a pre-defined value function that includes extended back EMF, lumped disturbance, and measurement error, the end of the window is calculated by minimizing the value function within a preset rolling estimation time window. Axial extension back EMF and Estimated value of axial extended back electromotive force; The MathCal electric angle calculation module is configured to calculate the traveling wave magnetic field electric angle based on the extended back electromotive force estimate via a normalized phase-locked loop. and electric angular velocity ; The speed calculation module is configured to calculate the train's running speed based on the relationship between the electric angular velocity and the motor pole pitch, and obtain a speed estimate. The ASR speed regulator is configured to generate traction current based on the speed setpoint and the speed estimate. The ACR current regulator is configured to adjust based on the traction current and the current output from the coordinate transformation module. , Adjustment, output voltage control command and ; The inverter drive module is configured to respond to voltage control commands. and Combined with electrical angle The voltage required to generate a long stator motor and This, in turn, drives the linear synchronous motor of the maglev train's stator. The aforementioned ASR speed regulator, ACR current regulator, coordinate transformation module, MHE prediction and estimation module, MathCal electrical angle calculation module, speed calculation module, and inverter drive module constitute a dual closed-loop feedback structure.

2. The system according to claim 1, characterized in that, The value function is as follows: in, is a confidence factor greater than 0, representing the weight of the impact of the initial state estimation error on the optimization process; N is the preset rolling window length, and t is the sampling time. This represents the initial sampled value at position tN at time t. This represents the initial estimate of the position tN at time t; This represents the lumped perturbation at time i. T s The sampling period; Let C represent the output at time i, and let C represent the observation matrix. This represents the state at time i, which consists of the current and the extended back electromotive force.

3. The system according to claim 2, characterized in that, The confidence factor is corrected online using Sage-Husa adaptive filtering, which includes: obtaining the state error covariance matrix recursively from Sage-Husa adaptive filtering, and using its inverse matrix or equivalent weight matrix as the confidence factor for the arrival cost term.

4. The system according to claim 2, characterized in that, The steps for constructing the value function include: Establish a discrete state-space model of a long-stator linear synchronous motor that includes extended back electromotive force and lumped disturbances: in, and in, , , , , These are, respectively, state variables, control variables, output variables, lumped disturbance variables, and sampling error variables. and These are the extended back electromotive force along the α-axis and the extended back electromotive force along the β-axis, respectively. , They are state variables , Sampling error; , They are state variables , Actual sampled values; A is the state transition matrix, B is the control matrix, and C is the observation matrix; , They are respectively Inductance in the axial coordinate system; For armature winding resistance; Electric angular velocity; The discrete state-space model is discretized using the backward Euler method: in, t Sampling time; It is a time-varying matrix. , The sampling period is It is a 4×4 identity matrix; It is a constant matrix. , Represents the lumped disturbance term. . Using a preset rolling estimation time window length N as the estimation interval, at each sampling moment, the voltage input and current measurement values ​​from the past N sampling periods are introduced into the optimization problem, and an optimization objective function is established as the value function.

5. The system according to claim 1, characterized in that, Normalized phase-locked loop based on the end of the window α Axial extension back EMF and β The phase error signal is calculated by the shaft extended back electromotive force, and the estimated electric angular velocity is obtained by proportional-integral adjustment. The estimated electric angular velocity is then integrated to obtain the estimated electric angle.

6. The system according to claim 1, characterized in that, The initial value of the electrical angle is obtained through positioning methods, including high-frequency signal injection or tooth groove detection and position markers. After the speed crosses the switching threshold, the mode is switched to the rolling time domain estimation mode.

7. The system according to claim 1, characterized in that, The rolling time-domain estimation value function within the rolling window is transformed into an equality-constrained quadratic programming problem. The equality constraints are composed of discrete state equations, and the optimization variables are the state sequence and disturbance sequence within the rolling window. The optimal state estimate within the rolling window is solved using the KKT equation.

8. A method for position-sensorless control of the long stator motor of a maglev train based on rolling time-domain estimation, characterized in that the steps are as follows: include: The static direction of the stator linear synchronous motor of the maglev train. Coordinate system current , Combined with electrical angle , to current , Switch to dynamic Coordinate system obtained , ; Based on current and ,Voltage and Electric angular velocity Using a preset value function that includes extended back EMF, lumped disturbance, and measurement error, the end of the window is obtained within a preset rolling estimation time window. Axial extension back EMF and Estimated value of axial extended back electromotive force; The electric angle of the traveling wave magnetic field was calculated using a normalized phase-locked loop based on the extended back electromotive force estimate. and electric angular velocity The electric angular velocity Used to generate speed estimates, these speed estimates are combined with a speed setpoint to generate the traction current, which is then used in the current... , Adjusting the output voltage control command and According to voltage control commands and Combined with the estimated electrical angle , obtain voltage and This drives the linear synchronous motor of the maglev train's stator, thereby achieving sensorless current and speed closed-loop control.

Images