A permanent magnet motor model-free predictive control method and system embedded with a generalized integral observer

Abstract

The application discloses a permanent magnet motor model-free predictive control method and system with embedded generalized integral observer, and belongs to the field of permanent magnet motor control. A super-local model of the permanent magnet motor is constructed, a new variable is introduced to integrate disturbance variables of mechanical angular velocity and q-axis current to form a lumped disturbance, a model is reconfigured, a speed prediction model is obtained by observing the disturbance through a three-order extended state observer with embedded quasi-second-order generalized integrator, a d-axis current prediction model is obtained by observing the disturbance through an observer for a d-axis current model, a control rate equation of d-q axis voltage is established based on the two prediction models, and voltage control of the d-q axis of the motor is calculated through the control rate equation based on collected system parameters. The disturbance of the speed outer loop and the q-axis current loop is integrated into the lumped disturbance, the design cost is reduced, and the dynamic response is improved. The TESO architecture with embedded QGI considers dynamic anti-disturbance and periodic disturbance suppression, and guarantees system stability.

Description

Technical Field

[0001] This invention relates to permanent magnet motor control, specifically to a model-free predictive control method and system for permanent magnet motors with an embedded generalized integrator observer. Background Technology

[0002] PMSMs (Power Management Systems) have been widely used in high-precision drive fields such as industrial servo robot systems due to their high power density, high operating efficiency, and wide speed range. To achieve high-performance control, Model Predictive Control (MPC) has become a research hotspot due to its intuitive multi-objective optimization capabilities, good handling of nonlinear constraints, and ease of digital implementation. Particularly in the field of speed control, Model Predictive Direct Speed ​​Control (MPDSC) aims to eliminate the response hysteresis and speed ripple problems caused by proportional-integral regulators in traditional cascaded control structures by directly generating the optimal voltage vector to track speed commands. However, the performance of MPDSC heavily relies on accurate mathematical models of the motor. In practical systems, unmodeled dynamics, parameter perturbations (such as resistance and inductance variations with temperature), and unknown external load disturbances can introduce significant model mismatch, leading to decreased prediction accuracy and deteriorated control performance. Furthermore, even in relatively advanced MPDSCs, the control structure still retains some redundancy from the cascade concept. The disturbance compensation of the speed loop and the current loop exhibits cascade characteristics. The disturbance complexity caused by this structure limits the dynamic response speed of the system and weakens its overall anti-interference robustness.

[0003] To reduce or even eliminate reliance on precise mathematical models, model-free predictive control (MPDSC) and various robust MPC methods have become important solutions. However, most current MPDSC strategies inherently lack the ability to identify and compensate for periodic unmodeled disturbances. For example, disturbance components such as the voltage dead-zone effect inherent in inverter operation and periodic errors in current measurement have frequency characteristics related to the system fundamental or carrier frequency, making accurate separation and real-time compensation difficult with traditional observers. More importantly, many advanced control schemes have not completely escaped cascaded or quasi-cascaded control architectures. When dealing with periodic disturbances, the bandwidth differences and delay effects between inner and outer loops cause the disturbances to propagate and amplify across multiple loops, ultimately leading to significant speed pulsations or even mechanical oscillations, severely limiting the performance of PMSMs in high-precision applications.

[0004] Periodic disturbances pose a significant challenge to PMSM drive systems, especially under low-speed conditions. These disturbance sources primarily include inverter nonlinearity (such as voltage dead zone), measurement offsets and proportional errors of current and voltage sensors, shaft alignment errors in the mechanical transmission system, and periodically varying load torque. These disturbances introduce low-order (such as first, second, and sixth) harmonics into the motor current and torque, resulting in periodic torque pulsations, which are further amplified by the mechanical system into speed fluctuations. At low speeds, due to the smaller back electromotive force, the system control voltage decreases, and system nonlinearity increases, making the relative impact of these periodic disturbance components more pronounced and causing more severe effects on the stability and accuracy of the control system. Summary of the Invention

[0005] Purpose of the invention: To address the above problems, this invention provides a model-free predictive control method and system for permanent magnet motors with an embedded generalized integrator observer that dynamically resists and suppresses periodic disturbances.

[0006] Technical Solution: To solve the above problems, this invention employs a model-free predictive control method for permanent magnet motors with an embedded generalized integrator observer, comprising the following steps:

[0007] (1) Collect system parameters of the permanent magnet motor drive system;

[0008] (2) Construct a superlocal model of the permanent magnet motor, including the speed model, q-axis current model and d-axis current model. By introducing new variables, integrate the disturbance variables of the mechanical angular velocity and q-axis current of the permanent magnet motor to form a lumped disturbance. Reconstruct the speed model and q-axis current model in the superlocal model to obtain the reconstructed speed control model.

[0009] (3) The reconstructed speed control model is observed by an improved composite observer, which is a third-order extended state observer with an embedded quasi-second-order generalized integrator, to obtain the speed prediction model after the improved composite observer observes the disturbance.

[0010] The d-axis current model is subjected to perturbation by an extended state observer, resulting in a predicted d-axis current model after perturbation by the extended state observer.

[0011] (4) Based on the speed prediction model and the d-axis current prediction model, establish the control law equations for the d-axis voltage and the q-axis voltage;

[0012] (5) Based on the collected system parameters, the voltages of the d-axis and q-axis are calculated through the control law equations of the d-axis voltage and q-axis voltage, and the motor drive is controlled according to the calculated voltages of the d-axis and q-axis.

[0013] Furthermore, the hyperlocal model of the permanent magnet motor in step (2) is as follows:

[0014]

[0015] in, The mechanical angular velocity of the permanent magnet motor. and These are the d-axis current and q-axis current in the rotating coordinate system of the permanent magnet motor, respectively. and These are the d-axis voltage and q-axis voltage in the rotating coordinate system of the permanent magnet motor, respectively. , , These are all disturbances other than input variables, namely the mechanical angular velocity, d-axis current, and q-axis current of the permanent magnet motor. This represents the scaling factor for the rotational speed model in the hyperlocal model. This is the scaling factor for the d-axis current model in the hyperlocal model. This is the scaling factor for the q-axis current model in the hyperlocal model.

[0016] Furthermore, the reconstructed speed control model is as follows:

[0017]

[0018] in, The mechanical angular velocity of the permanent magnet motor. For the new variable introduced, The proportional coefficient for the reconstructed speed control model. The lumped disturbance for the reconstructed speed control model.

[0019] Furthermore, the improved composite observer is as follows:

[0020]

[0021] in, This is the observation error of the mechanical angular velocity. The mechanical angular velocity of the permanent magnet motor The observed values, For new variables The observed values, Lumped disturbance for the reconstructed speed control model The observed values, , , , , and Configuration parameters related to improving the observation performance of the composite observer, , All of these are the state variables of the nth second-order generalized integral periodic perturbation observer.

[0022] Furthermore, the improved composite observer is discretized using forward Euler to obtain the speed prediction model, which is as follows:

[0023]

[0024] in, For the present The observation error at time, For the present Mechanical angular velocity of permanent magnet motor The observed values, For the present The observed values ​​of the predicted mechanical angular velocity of the permanent magnet motor at any given time. The sampling time for the permanent magnet motor current is . For the present New variables at time The observed values, For the present The observed value of the new variable prediction at time 1. For the current aggregate disturbance The observed values, For the front The q-axis voltage of the permanent magnet motor at any given time. The observed values ​​are the predicted values ​​of the lumped disturbance.

[0025] Furthermore, the reconstructed speed control model and d-axis current model are discretized using forward Euler discretization to obtain:

[0026]

[0027]

[0028] in, for Predicted value of the mechanical angular velocity of the permanent magnet motor at any given time. For the present The predicted value of the new variable at time step. For the present Predicted value of the mechanical angular velocity of the permanent magnet motor at any given time. This is the predicted value of the d-axis current of the permanent magnet motor. For the front The d-axis voltage of the permanent magnet motor at all times. This represents the current disturbance of the d-axis current of the permanent magnet motor. For the front The d-axis current of the permanent magnet motor at any given time.

[0029] Furthermore, the control law equations for the d-axis voltage and the q-axis voltage are as follows:

[0030]

[0031] in, The control law for the q-axis voltage. The control law of the d-axis voltage. This is a reference value for the speed of the permanent magnet motor. These are the observed values ​​for the predicted mechanical angular velocity of the permanent magnet motor. The observed values ​​are the predicted values ​​for the new variable. These are the observed values ​​for the predicted d-axis current of the permanent magnet motor. The observed values ​​are the predicted values ​​of the d-axis current disturbance of the permanent magnet motor.

[0032] Furthermore, after calculating the q-axis voltage in step (5), the q-axis voltage is modified according to the voltage limit, and the modified q-axis voltage controls the motor drive; the voltage limit expression is:

[0033]

[0034] in, This is the maximum allowable current for the permanent magnet motor. For the front The q-axis current of the permanent magnet motor at any given time.

[0035] This invention also employs a direct speed prediction control system for a permanent magnet motor, comprising:

[0036] The acquisition module is used to acquire system parameters of the permanent magnet motor drive system;

[0037] The superlocal model reconstruction module is used to construct a superlocal model of the permanent magnet motor, including the speed model, q-axis current model and d-axis current model. By introducing new variables, the disturbance variables of the mechanical angular velocity and q-axis current of the permanent magnet motor are integrated to form a lumped disturbance. The speed model and q-axis current model in the superlocal model are reconstructed to obtain the reconstructed speed control model.

[0038] The prediction model construction module is used to observe disturbances in the reconstructed speed control model through an improved composite observer, wherein the improved composite observer is a third-order extended state observer with an embedded quasi-second-order generalized integrator, to obtain a speed prediction model after the improved composite observer observes the disturbances; it is also used to observe disturbances in the d-axis current model through an extended state observer, to obtain a d-axis current prediction model after the extended state observer observes the disturbances.

[0039] The control law establishment module is used to establish the control law equations for d-axis voltage and q-axis voltage based on the speed prediction model and the d-axis current prediction model.

[0040] The control module is used to calculate the d-axis and q-axis voltages based on the collected system parameters through the control law equations of the d-axis and q-axis voltages, and control the motor drive according to the calculated d-axis and q-axis voltages.

[0041] Beneficial Effects: Compared with existing technologies, the significant advantage of this invention is the elimination of the traditional MPDSC speed-current loop cascade structure. By integrating the speed outer loop and q-axis current loop disturbances into a single lumped disturbance, a second-order speed hyperlocal model is constructed, reducing the design cost of a disturbance state variable, decoupling speed and current coupling, and improving dynamic response by combining dq-axis voltage control law design. Through a third-order extended state observer (TESO) architecture with an embedded second-order quasi-generalized integrator (QGI), both dynamic disturbance rejection and periodic disturbance suppression are considered. TESO quickly estimates wide-band disturbances, while QGI accurately suppresses specific harmonics, ensuring system stability. Addressing the overcurrent risk of non-cascaded structures, a current limiting method is proposed based on current transient characteristics and the motor's maximum allowable current. This eliminates the need for additional current loop adjustment, controlling the current within the rated range through voltage constraints, effectively avoiding dynamic overload and ensuring safe control of non-cascaded MPDSCs. Attached Figure Description

[0042] Figure 1 This is a structural diagram of the direct speed prediction control method in this invention.

[0043] Figure 2 This is the experimental result of embedding a 6.6Hz QGI into TESO when the motor is not under load torque in this invention.

[0044] Figure 3 This is the experimental result of embedding a 6.6Hz QGI into TESO when the motor is subjected to rated load torque in this invention.

[0045] Figure 4 This is the experimental result of embedding a 13.2Hz QGI into TESO when the motor is not under load torque in this invention.

[0046] Figure 5 This is the experimental result of embedding a 13.2Hz QGI into TESO when the motor is subjected to rated load torque in this invention.

[0047] Figure 6 This is the experimental result of embedding a 40Hz QGI into TESO when the motor is not under load torque in this invention.

[0048] Figure 7 This is the experimental result of embedding a 40Hz QGI into TESO when the motor is subjected to rated load torque in this invention.

[0049] Figure 8 This is the experimental result of embedding 6.6Hz, 13.2Hz and 40Hz QGI into TESO simultaneously when the motor is not under load torque in this invention.

[0050] Figure 9 This is the experimental result of simultaneously embedding 6.6Hz, 13.2Hz and 40Hz QGI into TESO when the motor is subjected to rated load torque in this invention.

[0051] Figure 10 This is the rotational speed test result of the RMPDSC dynamic performance test in this invention.

[0052] Figure 11 The results are experimental results of phase current for dynamic performance testing of RMPDSC in this invention.

[0053] Figure 12 This is the result of the current limiting experiment in this invention. Detailed Implementation

[0054] like Figure 1 As shown in this embodiment, a model-free predictive control method for permanent magnet motors with an embedded generalized integral observer is proposed. Addressing the speed harmonic problem caused by periodic unmodeled disturbances in Model Predictive Direct Speed ​​Control (MPDSC) of permanent magnet synchronous motors (PMSM), and the shortcomings of traditional cascaded control structures such as response hysteresis and weak anti-disturbance robustness, a robust model-predictive direct speed control (RMPDSC) method based on QGI-TESO is proposed. By reconstructing the second-order speed hyperlocal model of the PMSM, the disturbances of the original speed outer loop and q-axis current loop are integrated into a single lumped disturbance, eliminating redundancy in the cascaded structure. The dq-axis voltage control law is derived using a deadbeat-free algorithm, simplifying the system control structure design. A TESO composite observer with embedded multiple second-order QGIs is designed, relying on TESO to achieve fast estimation of wide-band disturbances. The first, second, and sixth harmonics are accurately suppressed through corresponding frequency QGIs, and parameter configuration is provided through Lyapunov functions and root loci to ensure asymptotic stability of the system. A q-axis voltage limiting strategy is set to effectively avoid the overcurrent risk generated during dynamic processes.

[0055] The control method specifically includes the following steps:

[0056] Step 1: Collect system parameters of the permanent magnet motor drive system; in this embodiment, the internal controller data is transmitted to the host computer via serial communication to monitor the speed and current control in real time. The system current sampling frequency is 10kHz, and the inverter dead time is 100ns. The prototype parameters are shown in Table 1:

[0057] Table 1 Prototype Parameters

[0058]

[0059] Step 2: Construct a hyperlocal model of the permanent magnet motor. By introducing new variables, integrate the disturbance variables of the mechanical angular velocity and q-axis current of the permanent magnet motor to form a lumped disturbance. Reconstruct the speed model and q-axis current model in the hyperlocal model to obtain the reconstructed speed control model.

[0060] First, a mathematical model of the permanent magnet motor in a rotating coordinate system is constructed, including the current model, the motion equation model, and the electromagnetic torque equation.

[0061] The mathematical model for the current in a permanent magnet motor is as follows:

[0062] (1)

[0063] in, , These are the d-axis voltage and q-axis voltage in the rotating coordinate system of the permanent magnet motor, respectively. , These are the d-axis current and q-axis current in the rotating coordinate system of the permanent magnet motor, respectively. , These are the d-axis inductance and q-axis inductance in the rotating coordinate system of a permanent magnet motor (for surface-mounted permanent magnet synchronous motors: = ); R s This refers to the stator resistance of the permanent magnet motor. For permanent magnet flux; ω is the rotor electrical angular velocity of the motor.

[0064] The equations of motion and electromagnetic torque of the electric motor are as follows:

[0065] (2)

[0066] In the formula, Let be the mechanical angular velocity of the motor. This represents the torque of the motor. For load torque, The viscous damping coefficient is... For rotational inertia, For extreme logarithms, This is the torque coefficient of the motor.

[0067] By adopting the concept of hyperlocal modeling, the mathematical model of the permanent magnet motor is simplified, and the first-order ideal mathematical model of equations (1) and (2) is transformed. The model is equivalent to a combination of input variables and disturbance components, including a speed model, a q-axis current model, and a d-axis current model. The equivalent model is as follows:

[0068] (3)

[0069] In the formula, , and It includes all perturbations in the equation except for the input variables, including unmodeled components and disturbances caused by parameter changes. , and The scaling factor for the hyperlocal model is denoted as , where , , .

[0070] In equation (3), besides the DC component disturbance, , and There are also periodic unmodeled disturbances caused by current measurement errors, inverter dead zones, and shaft asymmetry.

[0071] Equation (3) shows that there is a cascaded relationship between the rotational speed and the q-axis current, and there are two disturbance variables in each relationship. The disturbance is in the outer speed loop, and the outer speed loop response is relatively slow. Therefore, by constructing a new variable, we can both eliminate a disturbance state variable and... The disturbance is equivalent to being suppressed in the q-axis current loop, and the new variable is:

[0072] (4)

[0073] Furthermore, the expression for the q-axis current can be derived:

[0074] (5)

[0075] Substituting equation (4) into equation (3) and rearranging, we obtain the new speed control model as follows:

[0076] (6)

[0077] In the formula, It is the lumped disturbance in the second-order reconstruction velocity model, and at the same time It includes periodic disturbances of the first, second, and sixth order. :

[0078] (7)

[0079] As shown in equation (7), the frequency of the periodic disturbance remains unchanged after model reconstruction. Although the gain is amplified to some extent due to the differentiation operation, the frequency configuration of the periodic disturbance is consistent before and after reconstruction. Therefore, it can be seen that after model reconstruction, the frequency of the periodic disturbance will be... and The integration is equivalent to a single perturbation state variable. It can effectively improve the understanding of This improves the suppression efficiency and reduces the design cost of a state variable in disturbance suppression design. After simplification, the reconstructed speed control model is obtained:

[0080] (8)

[0081] In equation (3), after reconstructing the q-axis current and speed equations, equation (8) is obtained. The control objective of the reconstructed speed control model is the speed, so the control objective of the entire system is divided into two parts: speed and d-axis current. For the speed, it should follow its reference value. Since the d-axis current does not participate in generating torque, the d-axis current should be 0, so the cost function can be determined as:

[0082] (9)

[0083] In the formula, k1 and k2 represent the weighting coefficients of the two constraint terms. Since the rotational speed and d-axis current are completely decoupled through the observer, a deadbeat-free approach can be used to solve the problem in order to achieve a more accurate solution and reduce computational burden. First, the rotational speed and d-axis current are discretized using the forward Euler method, and the following can be obtained:

[0084] (10)

[0085] (11)

[0086] In equation (9), the most ideal cost function solution is the d-axis voltage and q-axis voltage that make the reference value consistent with the predicted value, so we can directly let:

[0087] (12)

[0088] Therefore, by combining equations (10)-(12), the control law of the dq axis voltage can be obtained:

[0089] (13)

[0090] In equation (13), and Apart from variables that can be obtained through the system's sensors, other variables, especially lumped disturbances, cannot be directly acquired. and Since it cannot be obtained directly from the system, a corresponding disturbance observer needs to be established to obtain it, while also taking into account the disturbance frequency characteristics present in the lumped disturbance.

[0091] Step 3: The reconstructed speed control model is subjected to disturbances by an improved composite observer. The improved composite observer is a third-order extended state observer with an embedded quasi-second-order generalized integrator, resulting in a speed prediction model after the improved composite observer observes the disturbances. The d-axis current model is subjected to disturbances by an extended state observer, resulting in a d-axis current prediction model after the extended state observer observes the disturbances.

[0092] The improved composite observer is as follows:

[0093] (14)

[0094] in, This is the observation error of the mechanical angular velocity. The mechanical angular velocity of the permanent magnet motor The observed values, For new variables The observed values, Lumped disturbance for the reconstructed speed control model The observed values, , , , , and Configuration parameters related to improving the observation performance of the composite observer, , All of these are the state variables of the nth second-order generalized integral periodic perturbation observer.

[0095] To ensure the stability of equation (14), a Lyapunov function is constructed:

[0096] (15)

[0097] After substituting each differential term and using Young's inequality to handle the overlapping terms, the negative definite standard form can be obtained as follows:

[0098] (16)

[0099] In the formula, the observer bandwidth is defined as... The configuration parameters of a third-order perturbation observer are designed using the bandwidth definition method of a third-order low-pass filter: To ensure its stability, periodic perturbation parameters are defined. Let be the perturbation frequency, and let = (0.1~0.25) , = (0.5~2), thus ensuring that in the standard formula , , , and Since all values ​​are greater than zero, equation (16) holds, meaning that equation (15) is negative definite. Therefore, the observer satisfies Lyapunov stability and is asymptotically stable.

[0100] Discretize the improved composite observer using forward Euler:

[0101] (17)

[0102] in, For the present The observation error at time, For the present Mechanical angular velocity of permanent magnet motor The observed values, For the present The observed values ​​of the predicted mechanical angular velocity of the permanent magnet motor at any given time. The sampling time for the permanent magnet motor current is . For the present New variables at time The observed values, For the present The observed value of the new variable prediction at time 1. For the current aggregate disturbance The observed values, For the front The q-axis voltage of the permanent magnet motor at any given time. The observed values ​​are the predicted values ​​of the lumped disturbance.

[0103] The d-axis current model is subjected to perturbations using an extended state observer, resulting in a predicted d-axis current model after the perturbations are observed. and It can be extracted using a common second-order extended state observer (ESO).

[0104] Step 4: Based on the speed prediction model and the d-axis current prediction model, establish the control law equations for the d-axis voltage and the q-axis voltage;

[0105] Considering that the response time of rotational speed is slower than that of current variable, the estimated values ​​of reconstructed current and disturbance at time k+1 in equation (17) will be used to compensate for the control delay in the actual system. Finally, by combining the rotational speed prediction equations, the output rate of the dq-axis voltage equation can be obtained:

[0106] (18)

[0107] in, The control law for the q-axis voltage. The control law of the d-axis voltage. This is a reference value for the speed of the permanent magnet motor. These are the observed values ​​for the predicted mechanical angular velocity of the permanent magnet motor. The observed values ​​are the predicted values ​​for the new variable. These are the observed values ​​for the predicted d-axis current of the permanent magnet motor. The observed values ​​are the predicted values ​​of the d-axis current disturbance of the permanent magnet motor.

[0108] Because the system design reconfigures the current, and overcurrent protection is also a crucial aspect of the system design during dynamic processes, current limiting is also necessary. Since the control process directly outputs the q-axis voltage through predictive equations and observers, the q-axis current can only be indirectly constrained by controlling the q-axis voltage. Given that the transient change time of the current is much shorter than the change time of the mechanical disturbance, the mechanical disturbance can usually be ignored during the transient process of electrical parameters. Therefore...

[0109] (19)

[0110] Then, the relationship between the reconstructed current and the q-axis current can be restated as follows:

[0111] (20)

[0112] Assume the maximum allowable current of the motor is expressed as Substituting into the discretized equation (20), we can obtain

[0113] (twenty one)

[0114] Then, the q-axis voltage limiting under current constraint conditions can be obtained:

[0115] (twenty two)

[0116] Thus, equations (17), (18), and (22) complete the robust model predictive control method based on TESO-QGI, which enhances the suppression effect on AC unmodeled disturbances on the basis of DC unmodeled disturbances.

[0117] Step 5: Based on the collected system parameters, the d-axis and q-axis voltages are calculated using the control law equations for the d-axis and q-axis voltages. The motor drive is then controlled according to the calculated d-axis and q-axis voltages. Furthermore, the q-axis voltage limiting formula is derived based on the maximum allowable current of the motor, and the q-axis voltage is corrected. The corrected voltage command is then converted into an inverter signal using SVPWM to drive the motor, forming a control closed loop and achieving high-precision PMSM control.

[0118] This embodiment's control method uses a core logic chain of "signal acquisition - disturbance observation - model prediction - voltage output - constraint protection": The bottom-level signal acquisition module acquires the d / q-axis current and mechanical angular velocity of the PMSM at a frequency of 10kHz, filters them, and transmits them to subsequent modules; the core disturbance observation module uses a TESO with an embedded second-order QGI to receive feedback signals and voltage control rates. The TESO quickly estimates wide-band disturbances, and the QGI accurately extracts periodic unmodeled disturbances, then outputs the lumped disturbance quantity. The model prediction module, based on the reconstructed second-order speed model and the observed lumped disturbance, combines speed reference values ​​and d-axis current reference values, and calculates the d / q-axis voltage control rate through forward Euler discretization and a deadbeat-free algorithm. The voltage limiting and drive module derives the q-axis voltage limiting formula based on the motor's maximum allowable current, converts the corrected voltage command into an inverter signal using SVPWM to drive the motor, forming a control closed loop and achieving high-precision control of the PMSM.

[0119] The control performance of the motor was tested, taking a motor speed of 100 rpm as an example. The experimental results are shown below. Figure 2-9 As shown, 6.6Hz bandwidth (1) was embedded respectively. st ), 13.2Hz (2 nd ) and 40Hz (6 th The test results of the motor with no load torque and with rated load torque, obtained by comparing the single generalized integral of the motor, show that... Figure 2 and Figure 3 In the figure, based on the comparison of Total Harmonic Distortion (THD), before applying the load, the first harmonic content in the system was low, and the effect of QGI was not obvious. After applying the load, the misalignment problem between the load and the motor shaft system worsened, and a more obvious first harmonic appeared. However, when QGI was embedded, the suppression effect of the first harmonic was significant, and the peak-to-peak speed (the difference between the maximum value and the minimum value, which reflects the fluctuation of the speed) decreased from 35 rpm to 25 rpm. Figure 4 and Figure 5 To incorporate the experimental results of 13.2Hz QGI, the second harmonic was effectively suppressed under both no-load and rated load conditions, and the peak-to-peak speed was reduced from 33rpm to 23rpm and from 35rpm to 32rpm, respectively. Figure 6 and Figure 7 The experimental results for embedding 40Hz QGI show that, similar to the suppression effect of the second harmonic, the sixth harmonic under both no-load and rated load conditions was effectively suppressed, with the peak-to-peak value decreasing from 33 rpm to 18 rpm and from 35 rpm to 25 rpm, respectively.

[0120] pass Figure 2-7The experimental results show that the addition of QGI can effectively suppress rotational speed harmonics at specific frequency points. Therefore, integrating multiple QGIs can achieve rotational speed suppression at multiple frequency points. The experimental results are as follows: Figure 8 and Figure 9 As shown, when multiple QGIs are integrated into a third-order extended state observer, in no-load harmonic suppression, the second and sixth harmonics are reduced from 3.1% and 3.3% to 0.4% and 0.6%, respectively, and the peak-to-peak speed is reduced from 33 rpm to 10 rpm. Under rated operating conditions, the first, second, and sixth harmonics are reduced from 2.6%, 1.8%, and 3.5% to 0.4%, 0.3%, and 0.6%, respectively, and the peak-to-peak speed is reduced from 35 rpm to 10 rpm, resulting in a significant improvement in steady-state speed performance.

[0121] Figure 10 and Figure 11 The experimental results comparing the dynamic performance of RMPDSC with and without QGI embedding show that embedding QGI improves the system's dynamic performance. This is because QGI increases gain within a certain bandwidth, suppressing transient speed changes. Therefore, during loading, speed drops are smaller and settling times are shorter. However, due to the increased bandwidth, the overshoot is slightly higher during transients compared to RMPDSC without QGI embedding.

[0122] Finally, to ensure the current overload problem of the non-cascaded system during dynamic processes, Figure 12 The current limiting of the system was tested, and the experimental results show that the current limiting proposed in this method plays a key role in the transient process, and can effectively achieve the current limiting within the set range to ensure that the motor does not overload.

Claims

1. A method for model-free predictive control of permanent magnet motor with embedded generalized integral observer, characterized in that, The method comprises the following steps: (1) collecting system parameters of the permanent magnet motor driving system; (2) constructing a hyper-local model of the permanent magnet motor, including a speed model, a q-axis current model and a d-axis current model, integrating disturbance variables of mechanical angular velocity and q-axis current of the permanent magnet motor into a lumped disturbance by introducing new variables, reconstructing the speed model and the q-axis current model in the hyper-local model to obtain a reconstructed speed control model; (3) observing the disturbance of the reconstructed speed control model through an improved composite observer, the improved composite observer being a third-order extended state observer with a quasi-second-order generalized integrator embedded therein, to obtain a speed prediction model after the disturbance is observed by the improved composite observer; the improved composite observer being: ； wherein, is the observation error of the mechanical angular velocity, is the observation value of the mechanical angular velocity of the permanent magnet motor , is the observation value of the new variable , is the observation value of the lumped disturbance of the reconstructed speed control model , , , , , and are configuration parameters related to the observation performance of the improved compound observer, , are the state variables of the nth second-order generalized integral periodic disturbance observer constructed; observing the disturbance of the d-axis current model through an extended state observer to obtain a d-axis current prediction model after the disturbance is observed by the extended state observer; (4) establishing control rate equations of d-axis voltage and q-axis voltage based on the speed prediction model and the d-axis current prediction model; (5) calculating the d-axis voltage and the q-axis voltage through the control rate equations of d-axis voltage and q-axis voltage based on the collected system parameters, and controlling the motor driving according to the calculated d-axis voltage and q-axis voltage.

2. The method of model-free predictive control of permanent magnet motor with embedded generalized integral observer according to claim 1, characterized in that, The hyper-local model of the permanent magnet motor in the step (2) is: ； wherein, is the mechanical angular velocity of the permanent magnet motor, and are the d-axis current and q-axis current of the permanent magnet motor in the rotating coordinate system, respectively; and are the d-axis voltage and q-axis voltage of the permanent magnet motor in the rotating coordinate system, respectively; , , are all the disturbances of the mechanical angular velocity, d-axis current, q-axis current of the permanent magnet motor except for the input variables, respectively; is the proportional coefficient of the speed model in the superlocal model, is the proportional coefficient of the d-axis current model in the superlocal model, is the proportional coefficient of the q-axis current model in the superlocal model.

3. The method of model-free predictive control of permanent magnet motor with embedded generalized integral observer according to claim 2, characterized in that, The reconstructed speed control model is: ； wherein is the mechanical angular velocity of the permanent magnet motor, is a new variable introduced, is a proportional coefficient of the reconstructed speed control model, is a lumped disturbance of the reconstructed speed control model.

4. The method of model-free predictive control of permanent magnet motor with embedded generalized integral observer according to claim 3, characterized in that, The speed prediction model is obtained by discretizing the improved composite observer using forward Euler, and the speed prediction model is: ； wherein, is the observation of the current time instant, is the observation of the current time instant mechanical angular speed of the permanent magnet motor, is the observation of the current time instant mechanical angular speed of the permanent magnet motor prediction, is the observation of the current time instant mechanical angular speed of the permanent magnet motor prediction, is the sampling time of the permanent magnet motor current, is the observation of the current time instant new variable, is the observation of the current time instant new variable prediction, is the observation of the current time instant lumped disturbance, is the observation of the previous time instant q-axis voltage of the permanent magnet motor, is the observation of the lumped disturbance prediction.

5. The method of model-free predictive control of permanent magnet motor with embedded generalized integral observer according to claim 4, characterized in that, The reconstructed speed control model and the d-axis current model are discretized by forward Euler to obtain: ； ； in, for Predicted value of the mechanical angular velocity of the permanent magnet motor at any given time. For the present The predicted value of the new variable at time step. For the present Predicted value of the mechanical angular velocity of the permanent magnet motor at any given time. This is the predicted value of the d-axis current of the permanent magnet motor. For the front The d-axis voltage of the permanent magnet motor at all times. This represents the current disturbance of the d-axis current of the permanent magnet motor. For the front The d-axis current of the permanent magnet motor at any given time.

6. The method of model-free predictive control of permanent magnet motor with embedded generalized integral observer according to claim 5, characterized in that, The control rate equations of d-axis voltage and q-axis voltage are: ； wherein, is a control rate of the q-axis voltage, is a control rate of the d-axis voltage, is a reference value of the permanent magnet motor speed, is an observation value of the permanent magnet motor mechanical angular velocity prediction value, is an observation value of the new variable prediction value, is an observation value of the permanent magnet motor d-axis current prediction value, is an observation value of the permanent magnet motor d-axis current disturbance prediction value.

7. The method of model-free predictive control of permanent magnet motor with embedded generalized integral observer according to claim 6, characterized in that, After the q-axis voltage is calculated in the step (5), the q-axis voltage is modified according to voltage limiting, and the motor driving is controlled by the modified q-axis voltage; the voltage limiting expression is: ； wherein, is the maximum allowed current of the permanent magnet motor, is the previous is the q-axis current of the permanent magnet motor at the instant.

8. A permanent magnet motor model-free predictive control system embedded with a generalized integral observer, characterized in that, It comprises: a collection module for collecting system parameters of the permanent magnet motor driving system; a hyper-local model reconstruction module for constructing a hyper-local model of the permanent magnet motor, including a speed model, a q-axis current model and a d-axis current model, integrating disturbance variables of mechanical angular velocity and q-axis current of the permanent magnet motor into a lumped disturbance by introducing new variables, reconstructing the speed model and the q-axis current model in the hyper-local model to obtain a reconstructed speed control model; a prediction model construction module for observing the disturbance of the reconstructed speed control model through an improved composite observer, the improved composite observer being a third-order extended state observer with a quasi-second-order generalized integrator embedded therein, to obtain a speed prediction model after the disturbance is observed by the improved composite observer; and for observing the disturbance of the d-axis current model through an extended state observer to obtain a d-axis current prediction model after the disturbance is observed by the extended state observer; the improved composite observer being: ； wherein, is the observation error of the mechanical angular velocity, is the mechanical angular velocity of the permanent magnet motor is the observation value of the new variable is the observation value of the new variable is the observation value of the lumped disturbance of the reconstructed speed control model is the observation value of the lumped disturbance of the reconstructed speed control model , , , , , and are configuration parameters related to the improved compound observer observation performance, , are the state variables of the constructed nth second-order generalized integral periodic disturbance observer a control rate establishment module for establishing control rate equations of d-axis voltage and q-axis voltage based on the speed prediction model and the d-axis current prediction model; a control module for calculating the d-axis voltage and the q-axis voltage through the control rate equations of d-axis voltage and q-axis voltage based on the collected system parameters, and controlling the motor driving according to the calculated d-axis voltage and q-axis voltage.

Images