A high-precision position sensorless control method for high-speed permanent magnet synchronous motor

Abstract

The application discloses a high-precision position sensorless control method for a high-speed permanent magnet synchronous motor, which comprises the following steps: establishing a ZOH discrete model of the motor in an alpha-beta coordinate system considering line filter parameters, and constructing a sliding mode observer; according to the amplitude and phase relationship between the current observation error and the back electromotive force, adaptively compensating the back electromotive force output by the sliding mode observer, and obtaining the compensated back electromotive force estimation value; injecting a gamma-axis current signal in a gamma-delta coordinate system, constructing an inductance identification model which is not affected by position estimation error through voltage and current difference under different injection states, and feeding back the identification result to the observer to realize the cooperative compensation of the rotor position estimation error caused by Euler discretization, quasi-sliding mode motion and inductance parameter mismatch. The application can improve the estimation accuracy of the motor rotor position, reduce the influence of parameter mismatch on the position sensorless control performance, and improve the stability and dynamic performance of the control system under high-speed running conditions.

Description

Technical Field

[0001] This invention belongs to the field of permanent magnet synchronous motor control technology, specifically relating to a high-precision sensorless control method for a high-speed permanent magnet synchronous motor. Background Technology

[0002] High-speed permanent magnet synchronous motors (HSPMSMs), with their advantages of high speed, small size, and high power density, have become core drive components in fields such as electric spindles, flywheel energy storage, fuel cells, and turbomachinery. Accurate acquisition of position information is a key prerequisite for achieving high-performance vector control in HSPMSMs. Traditional position sensors in HSPMSM applications suffer from installation difficulties and limited system reliability; therefore, sensorless vector control technology has become a current research hotspot in the HSPMSM field.

[0003] Existing sensorless control methods mainly include salient pole tracking-based methods and fundamental wave model-based methods. Salient pole tracking-based methods, such as those in the literature [Z. Lin, X. Li, Z. Wang, T. Shi, and C. Xia. Minimization of additional high-frequency torque ripples in sensorless drive of square wave voltage-injected IPMSM, IEEE Transactions on Power Electronics, Vol. 35, No. 12, pp. 13345-13355, December 2020], typically obtain position information by injecting high-frequency signals. This is suitable for zero-speed and low-speed operating ranges, but high-frequency injection introduces additional losses and noise, making it unsuitable for high-speed scenarios. In contrast, fundamental wave model-based methods achieve position estimation by observing back electromotive force or flux linkage, making them more suitable for medium- and high-speed operating conditions. Therefore, they are widely used in HSPMSM high-speed drive systems.

[0004] In fundamental wave model-based methods, such as those described in the literature [L. Guo, W. Xu, N. Jin, and H. Xiao. Sensorless Control Method for SMO PMSM without DC Bias Based on Improved Pre-filter and Speed-Immune Position Error Compensation Strategy, IEEE Transactions on Power Electronics, Vol. 40, No. 4, pp. 5163-5176, April 2025], sliding mode observers have been widely used due to their advantages such as strong robustness, fast dynamic response, and simple implementation. However, existing sliding mode observers still face significant position estimation errors under high-speed operating conditions. On the one hand, non-ideal factors in the structure of the sliding mode observer itself, such as Euler discretization, quasi-sliding mode motion, and motor parameter mismatch, can lead to errors in estimating the back EMF. On the other hand, non-ideal factors such as inverter nonlinearity, current sampling errors, and rotor magnetic field spatial harmonics can affect the current and voltage input to the sliding mode observer, which in turn can also lead to errors in estimating the back EMF. Under high-speed conditions, due to the significant increase in electrical angular frequency, errors in estimating the back EMF can lead to serious position estimation errors, resulting in increased motor current, increased losses, decreased efficiency, and even control instability.

[0005] Therefore, how to compensate for position estimation errors caused by non-ideal factors under high-speed operating conditions has become a key issue in the current sensorless control of HSPMSM. Summary of the Invention

[0006] This invention focuses on the non-ideal factors (Eulerian discretization, quasi-sliding mode motion, and motor parameter mismatch) inherent in the structure of the sliding mode observer. Aiming to eliminate the phase deviation in back EMF estimation, it reconstructs the observer's discrete model and phase compensation mechanism, proposing an adaptive phase-compensated sliding mode observer based on the ZOH (zero-order hold) discrete model to synergistically reduce the position estimation error caused by Euler discretization and quasi-sliding mode motion. Furthermore, addressing the observer's sensitivity to motor electromagnetic parameters, a self-iterative inductance parameter identification algorithm based on the ZOH discrete model is designed. This algorithm is unaffected by position estimation errors and can obtain accurate identification values, thereby reducing the position estimation error caused by motor electromagnetic parameter mismatch.

[0007] A high-precision sensorless control method for a high-speed permanent magnet synchronous motor includes the following steps: (1) Treat the HSPMSM and its connected line filters as a whole, and establish the ZOH discrete model of the HSPMSM in the α-β two-phase stationary coordinate system; (2) Based on the above ZOH discrete model, a current sliding mode observer is constructed, and adaptive phase compensation is performed on the back EMF estimation formula to form a complete back EMF sliding mode observer; (3) Establish the ZOH discrete model of HSPMSM in the γ-δ synchronous rotating coordinate system. Based on the ZOH discrete model, inject an alternating square wave current signal into the γ-axis current reference value. Eliminate the back electromotive force term in the inductance identification model through differential processing. Construct an inductance identification model that is not affected by the rotor position estimation error. Use the inductance identification model, average value filtering and self-iteration algorithm to identify the stator inductance estimate. (4) By combining the back EMF sliding mode observer and the stator inductance estimate, high-precision sensorless control of HSPMSM is achieved.

[0008] Furthermore, the expression for the ZOH discrete model in step (1) is as follows:

[0009]

[0010] in: i αβ ( k +1) and i αβ ( k ) are respectively the first k +1 moment and the k The stator current of the motor in the stationary coordinate system at time α-β. u αβ ( k ) is the first k The stator voltage of the motor in the stationary coordinate system at time α-β e αβn ( k ) is the first k The nominal value of the back electromotive force of the motor in the stationary coordinate system of the α-β phase at time t. ω e The electric angular velocity of the motor. R sn The nominal value of the equivalent stator resistance, L sn This is the nominal value of the equivalent stator inductance. T s The sampling period is j, where j is the imaginary unit. a n , b n , c n These are the model parameters.

[0011] Furthermore, the expression for the back EMF sliding mode observer in step (2) is as follows:

[0012]

[0013]

[0014] in: and The first k +1 moment and the k Estimated values ​​of motor stator current in the stationary coordinate system at time α-β For current error , For the first k Estimated back electromotive force of the motor in the stationary coordinate system at time α-β For model parameters, B and C For observer parameters, This is an estimated value for the electric angular velocity of the motor. m = k smo / Z 0, k smo For sliding mode gain, Z 0 represents the boundary layer thickness, and sat[ ] represents the saturation function.

[0015] Furthermore, the expression for the ZOH discrete model in step (3) is as follows:

[0016]

[0017] in: i γδ ( k +1) and i γδ ( k ) are respectively the first k +1 moment and the k The stator current of the motor in the synchronous rotating coordinate system at time γ-δ u γδ ( k ) is the first k The stator voltage of the motor in the synchronous rotating coordinate system at time γ-δ. e γδ ( k ) is the first k The back electromotive force of the motor in the synchronous rotating coordinate system at time γ-δ. ω e The electric angular velocity of the motor. R s To account for the equivalent stator resistance of the line filter, L s To account for the equivalent stator inductance of the line filter,T s The sampling period is j, where j is the imaginary unit. a , b , c These are the model parameters.

[0018] Furthermore, the expression for the inductor identification model in step (3) is as follows:

[0019] in: This is an estimated value for the stator inductance. K , M , O For model parameters, K =Δ i γ ( k )Δ u δ ( k )-Δ i δ ( k )Δ u γ ( k ), M =Δ i γ ( k )Δ u γ ( k )-Δ i δ ( k )Δ u δ ( k ), O= Δ i γ ( k ) 2 +Δ i δ ( k ) 2 Δ i γδ ( k ) is the first k The stator current difference Δ before and after alternation in the synchronous rotating coordinate system at time γ-δ. u γδ ( k ) is the first k The stator voltage difference Δ before and after alternation in the synchronous rotating coordinate system at time γ-δ. i γ ( k ) and Δ i δ ( k) are respectively Δ i γδ ( k The γ-axis component and δ-axis component of ) , Δ u γ ( k ) and Δ u δ ( k ) are respectively Δ u γδ ( k The γ-axis and δ-axis components of ).

[0020] Furthermore, in step (3), the stator inductance estimate is identified using an inductance identification model, average value filtering, and a self-iterative algorithm. The specific implementation process is as follows: Step S31: Before the alternation, when the square wave current signal is 0, continuously collect data. s The first stator current and stator voltage are obtained. s 1 set of data , i γ1 and i δ1 These are the γ-axis and δ-axis components of the stator current before alternation. u γ1 and u δ1 These are the γ-axis and δ-axis components of the stator voltage before alternation; Step S32: After alternation, the square wave current signal is -| i γinj |At that time, after the delay reaches steady state, continuous data collection is performed. s The second stator current and stator voltage are obtained. s 2 sets of data , i γ2 and i δ2 These are the γ-axis and δ-axis components of the stator current after alternation. u γ2 and u δ2 These are the γ-axis and δ-axis components of the stator voltage after alternation. i γinj The set current amplitude; Step S33: Calculate the data difference before and after alternation using the following formula. and parameters ;

[0021] , ,

[0022] Step S34: Set parameters Replace the parameters in the inductor identification model respectively K , M , O The stator inductance estimate is calculated using an inductance identification model. And update the back EMF sliding mode observer; Step S35: Perform iterative calculations based on steps S31 to S34, repeating the process. N The estimated value of the output stator inductance .

[0023] Furthermore, the termination condition for the iterative calculation in step S35 is:

[0024] in: This represents the average of the stator inductance estimates obtained from the first three iterations. This is the estimated value of the stator inductance obtained in this iteration. k err This is the set error constant.

[0025] Further, in step (4), the back EMF sliding mode observer and the stator inductance estimate are introduced into the sensorless control system of the HSPMSM. The back EMF of the motor is estimated using the back EMF sliding mode observer, and the rotor position angle estimate and electrical angular velocity estimate are calculated by the phase-locked loop. The amplitude is between 0 and -| i γinj The alternating square wave current signal is injected into the γ-axis current reference value, and the identified stator inductance estimate is fed back to the back EMF sliding mode observer after low-pass filtering.

[0026] A computer device includes a memory and a processor, wherein the memory stores a computer program and the processor executes the computer program to implement the above-described high-precision sensorless control method for a high-speed permanent magnet synchronous motor.

[0027] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned high-precision sensorless control method for a high-speed permanent magnet synchronous motor.

[0028] The inventiveness of this invention lies in its classification of three types of errors—Euler discretization, quasi-sliding mode motion, and inductor mismatch—as back electromotive force (EMF) phase deviation. A collaborative compensation framework of ZOH discretization correction, phase compensation, and injection-type inductor identification is proposed. ZOH correction improves the accuracy of the phase baseline in back EMF estimation, phase compensation uniformly handles the remaining phase deviation, and injection-type inductor identification provides real-time high-precision parameters for compensation. These three elements form a mutually reinforcing collaborative mechanism. Therefore, this invention has the following beneficial technical effects: 1. The back EMF sliding mode observer designed in this invention can effectively compensate for the errors caused by Euler discretization and quasi-sliding mode motion. Experimental results show that for the tested motor, the rotor position estimation error can be controlled within 1°, thereby effectively reducing the load current and acceleration time.

[0029] 2. To reduce the sensitivity of the observer to inductance parameters, this invention further proposes inductance identification, which is unaffected by position estimation errors and improves the accuracy of inductance identification. The combination of inductance identification and back EMF sliding mode observer can effectively improve the accuracy of rotor position estimation without position control. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the topology of the HSPMSM drive system.

[0031] Figure 2 This is a schematic diagram of the position estimation algorithm based on the back EMF sliding mode observer in this invention.

[0032] Figure 3 The ratio of the magnitude of the back electromotive force to the ratio of the argument follows -| i γinj | and L sn The relationship diagram of changes.

[0033] Figure 4 This is a schematic diagram of the inductor identification algorithm in this invention.

[0034] Figure 5 This is a block diagram of the high-precision sensorless control system for HSPMSM.

[0035] Figure 6 This is a schematic diagram of the steady-state experimental waveform under a rotational speed of 30 kr / min and rated load.

[0036] Figure 7 This is a schematic diagram of the dynamic experimental waveform under the condition of step rotation speed of 10-20-30 kr / min and no load.

[0037] Figure 8 The rotational speed is 30 kr / min and the initial parameter mismatch value is 0.5. L s and 1.3R s A schematic diagram of the waveform under rated load conditions.

[0038] Figure 9 The rotational speed is 30 kr / min and the initial parameter mismatch value is 1.3. L s and 0.7 R s A schematic diagram of the waveform under rated load conditions. Detailed Implementation

[0039] To describe the present invention in more detail, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] Table 1

[0041] The parameters of the HSPMSM to be controlled in this embodiment are shown in Table 1, and the specific implementation process is as follows: (1) Figure 1 The HSPMSM and the line filter connected to it are considered as a whole, and a ZOH discrete model of the HSPMSM in the α-β two-phase stationary coordinate system is established.

[0042] In the α-β two-phase stationary coordinate system, the HSPMSM and line filter are modeled as a whole, and the expression is:

[0043] in: u αβ = u α +j u β 、i αβ = i α +j i β , e αβ = e α +j e β These are the complex vector forms of the motor stator voltage, current, and back electromotive force, respectively. R s To account for the equivalent stator resistance of the filter, L s To account for the equivalent stator inductance of the filter, ψ f Let p be the amplitude of the permanent magnet flux linkage, and p be the differential operator. ω e Electric angular velocity, θ e This indicates the rotor position.

[0044] set up R sn The nominal value of the equivalent stator resistance, L sn Given the nominal value of the equivalent stator inductance, and incorporating the time-varying effects of motor parameters caused by temperature rise into the back electromotive force, the above equation can be rewritten as:

[0045] in: e αβn = e αβ -( R sn - R s ) i αβ -( L sn - L s )p i αβ .

[0046] Solve the above equation, and discretize the obtained solution using ZOH, with the sampling period set to... T s The ZOH discrete model of HSPMSM in the α-β two-phase stationary coordinate system can be obtained as follows:

[0047]

[0048] (2) Based on the ZOH discrete model, a current sliding mode observer is constructed. At the same time, the phase relationship between the back EMF and the current error is derived. Based on this, the back EMF estimation formula is adaptively phase compensated to form a complete back EMF sliding mode observer.

[0049] set up , Given the estimated values ​​of the complex vector of current and the complex vector of back electromotive force, respectively, and based on the ZOH discrete model of HSPMSM in the α-β coordinate system, the current sliding mode observer is constructed as follows:

[0050] in: , , This is an estimated value for the electric angular velocity. k smo For sliding mode gain, Z 0 represents the boundary layer thickness. uαβcom ( k To compensate for the voltage after one PWM cycle, the expression is:

[0051] in: u γδ ( k The voltage is in the γ-δ coordinate system; when the inverter nonlinearity is ignored, the actual voltage applied to the motor stator is... u αβ ( k )equal u αβcom ( k -1).

[0052] sat is a saturation function, that is:

[0053] By subtracting the current sliding mode observer from the ZOH discrete model of the HSPMSM in the α-β coordinate system, the current error equation can be obtained as follows:

[0054] Solving the above equation in the linear region of the SAT function, we can obtain:

[0055] in: .

[0056] choose m make And when k When large enough, the above formula can be further written as:

[0057] When the observer converges, there is Substituting the above equation into the solution, we get:

[0058] in: .

[0059] To compensate for this phase shift effect, based on the above equation, the back electromotive force estimation formula for adaptive phase compensation is designed as follows:

[0060] To ensure observer stability, it is necessary to To satisfy the requirement that the current error should converge to within the boundary layer of the saturation function in steady state, it is necessary to... To satisfy the above constraints and balance dynamic steady-state performance, we take... k smo =100 andZ 0 = 10.

[0061] Figure 2 This paper presents a position estimation framework based on a back-EMF sliding mode observer, consisting of two parts: the upper part is the back-EMF sliding mode observer, and the lower part is a phase-locked loop. The combination of the two parts yields the estimated rotor electric angular velocity. Electric angular position estimation value ; Figure 2 middle z -1 PI indicates proportional-integral controller, representing unit delay.

[0062] (3) Establish the ZOH discrete model of HSPMSM in the γ-δ synchronous rotating coordinate system, apply two different states of current injection signals to the γ-axis current reference value, eliminate the back electromotive force term in the inductance identification model through differential processing, construct an inductance identification model that is not affected by rotor position estimation error, and use the identification model and self-iterative algorithm to obtain the inductance estimate.

[0063] The ZOH discrete model of HSPMSM in the γ-δ synchronous rotating coordinate system is as follows:

[0064]

[0065] in: u γδ = u γ +j u δ , i γδ = i γ +j i δ , e γδ = e γ +j e δ = These are the complex vector forms of voltage, current, and back electromotive force in the γ-δ coordinate system, respectively. It is the rotor position estimation error.

[0066] The amplitude is between 0 and -| i γinj Alternating square wave signal Reference value for injected γ-axis current, such as Figure 3 As shown, when the injected current amplitude is less than 0.2 pu and the inductor perturbation range is less than ±0.5 pu, the ratio of the magnitudes of the back electromotive force before and after alternation is | eγδ2 | / | e γδ1 | and the ratio of the phase angle arg( e γδ2 ) / arg( e γδ1 Since all of them are close to 1, we can assume that there are e γδ1 = e γδ2 Established.

[0067] Through difference processing, we can obtain:

[0068]

[0069] Where: subscript 1 represents The variable when =0, subscript 2 represents =-| i γinj | is a variable when |.

[0070] Using the above formula, we can obtain an inductor identification model that is unaffected by position estimation errors, namely:

[0071] in: K =Δ i γ ( k )Δ u δ ( k )-Δ i δ ( k )Δ u γ ( k ), M =Δ i γ ( k )Δ u γ ( k )-Δ i δ ( k )Δ u δ ( k ), O= Δ i γ ( k ) 2 +Δ i δ ( k ) 2 .

[0072] The general identification expression can be written as:

[0073] in: y =[ y 1, y 2] T For output items, y 1= K cos( ω e T s )- M sin( ω e T s ), y 2= O sin( ω e T s ), h=K For input items, = [ a , b ] T For the parameters to be estimated, ξ This is the noise term.

[0074] To reduce the impact of noise, we consider using average value filtering to smooth the identification results. The inductance calculation process is as follows:

[0075] in: ω e In steady state, it is assumed to remain unchanged; considering Δ i γ Δ i δ Δ u γ Δ u δ Both are slowly changing direct currents. , , .

[0076] Figure 4 The algorithm flow for inductor identification is demonstrated, and the execution steps are as follows: Step 1: When =0, continuous data collection s 1 time i γ1 , i δ1, u γ1 , u δ1 .

[0077] Step 2: =-| i γinj |Time, delay τ After reaching steady state, continuous data collection was performed. s 2 times i γ2 , i δ2 , u γ2 , u δ2 .

[0078] Step 3: After collecting the data from Step 1 and Step 2, perform average filtering on the data, and then subtract the values ​​to obtain the result. Specifically:

[0079] Step 4: Substitute the data obtained in the above steps into the inductance identification model to calculate the inductance and update it to the observer, and wait for the next inductance identification loop from Step 1 to Step 4.

[0080] To ensure the accuracy of the inductance identification value, a self-iterative algorithm is designed to repeatedly... N The secondary inductor identification loop terminates based on the following evaluation criteria:

[0081] in: It is the average value of the first three inductor identifications. This is the inductance identification value for this test. k err It is the error constant.

[0082] To prevent current surges caused by abrupt changes in the inductance used by the observer, the identified inductance is low-pass filtered before being input into the observer.

[0083] To achieve a high current sampling signal-to-noise ratio while minimizing the interference of the injected current on the normal operation of the motor, this embodiment sets the injected current amplitude to -| i γinj | Set to -0.8A (less than 8.5% of the delta-axis current under rated load). To balance the speed of inductor identification with maximizing the accuracy of average value filtering, set to... s 1= s 2 = 100 τ It was set to 0.01s. k errIt is set to 2%, and the low-pass filter cutoff frequency is set to 30Hz.

[0084] (4) By combining the back EMF sliding mode observer and the inductance identification results, high-precision sensorless control of HSPMSM can be achieved.

[0085] Figure 5 This demonstrates the framework of the HSPMSM high-precision sensorless control system. The part outside the red box is the motor hardware and modulation section, which uses SVPWM (Space Vector Pulse Width Modulation) for three-phase current. i a , i b , i c The current is converted from α / β to α / β by an abc-to-α / β module and then sampled before entering the controller. The red box shows the discrete sensorless algorithm in the controller; the current loop uses PI control, and the current is compared with the reference current via an α / β-to-γ / δ module. Complete the closed loop; the speed loop uses PI (proportional-integral) control, estimating the electrical angular velocity by dividing by the number of pole pairs. n p The estimated angular velocity can then be obtained. , and reference angular velocity Complete the closed loop; estimate the back EMF using a back EMF sliding mode observer and calculate the rotor position and electrical angular velocity estimates using a phase-locked loop; inject a square wave signal into the γ-axis current reference value using inductor identification. The identified inductance value is fed back to the back EMF sliding mode observer. The inductance identification can be triggered manually by the start signal, periodically, or under specific conditions such as the motor temperature rise exceeding a threshold.

[0086] In the following section, we will conduct experiments by building an experimental platform to verify the performance of the HSPMSM high-precision sensorless control method of the present invention under steady-state, dynamic, and parameter mismatch conditions.

[0087] Figure 6 The steady-state experimental waveforms at a speed of 30 kr / min and rated load are shown. The estimated speed is... The rotor position estimation error is 30 kr / min. The current waveform is stable at -0.04°, and the average value of the delta-axis current is [value missing]. The effective value of the phase a current is 9.54A, RMS ( i a The value is 6.59A.

[0088] Figure 7 The dynamic experimental waveforms under no-load conditions with step speeds of 10-20-30 kr / min are shown, and the waveforms change with the estimated speed. Increase, estimate back electromotive force Phase and actual back electromotive force e α The rotor position is basically coincident. The rotor position estimation errors at 10 kr / min, 20 kr / min, and 30 kr / min are significant. The angles were 0.12°, -0.19°, and 0.09°, with acceleration times of 0.76s and 0.78s, respectively.

[0089] Figure 8 and Figure 9 The waveforms of the rated load experiment are shown under two different initial parameter mismatch conditions, with a rotational speed of 30 kr / min and a rotational speed of 0.5 kr / min. L s 1.3 R s The second is 1.3. L s 0.7 R s Before identification, the rotor position estimation error... The angles are 12.37° and -6.53° respectively, while the average value of the delta-axis current is... Given back electromotive forces of 9.88A and 9.38A respectively, estimate the back electromotive force. Compared with the true value e α There is a phase difference; during the identification process, the delta-axis current... i δ With γ-axis current i γ Overall stability; after identification, identify the inductance value. Converging to the true value Near the target area (with an estimation error within ±3%), estimate the back electromotive force. Compared with the true value e α Basically overlapping, position estimation error Approaching zero, and the average value of the delta-axis current. The values ​​were 9.41A and 9.36A, respectively, showing a significant decrease.

[0090] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. Those skilled in the art can readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. A high-precision sensorless control method for a high-speed permanent magnet synchronous motor, characterized in that, Includes the following steps: Step (1): Treat the HSPMSM and its connected line filters as a whole, and establish the ZOH discrete model of the HSPMSM in the α-β two-phase stationary coordinate system; Step (2): Construct a current sliding mode observer based on the above ZOH discrete model, and perform adaptive phase compensation on the back EMF estimation formula to form a complete back EMF sliding mode observer; Step (3): Establish the ZOH discrete model of HSPMSM in the γ-δ synchronous rotating coordinate system. Based on the ZOH discrete model, inject an alternating square wave current signal into the γ-axis current reference value. Eliminate the back electromotive force term in the inductance identification model through differential processing, construct an inductance identification model that is not affected by rotor position estimation error, and use the inductance identification model, average value filtering and self-iterative algorithm to identify the stator inductance estimate. Step (4): Combine the back EMF sliding mode observer and the stator inductance estimate to achieve high-precision sensorless control of HSPMSM.

2. The high-precision sensorless control method for a high-speed permanent magnet synchronous motor according to claim 1, characterized in that, The expression for the ZOH discrete model in step (1) is as follows: ; ; in: i αβ ( k +1) and i αβ ( k ) are respectively the first k +1 moment and the k The stator current of the motor in the stationary coordinate system at time α-β. u αβ ( k ) is the first k The stator voltage of the motor in the stationary coordinate system at time α-β e αβn ( k ) is the first k The nominal value of the back electromotive force of the motor in the stationary coordinate system of the α-β phase at time t. ω e The electric angular velocity of the motor. R sn The nominal value of the equivalent stator resistance, L sn This is the nominal value of the equivalent stator inductance. T s The sampling period is j, where j is the imaginary unit. a n , b n , c n These are the model parameters.

3. The high-precision sensorless control method for a high-speed permanent magnet synchronous motor according to claim 2, characterized in that, The expression for the back EMF sliding mode observer in step (2) is as follows: ; ; ; in: and The first k +1 moment and the k Estimated values ​​of motor stator current in the stationary coordinate system at time α-β For current error , For the first k Estimated back electromotive force of the motor in the stationary coordinate system at time α-β For model parameters, B and C For observer parameters, This is an estimated value for the electric angular velocity of the motor. m = k smo / Z 0, k smo For sliding mode gain, Z 0 represents the boundary layer thickness, and sat[ ] represents the saturation function.

4. The high-precision sensorless control method for a high-speed permanent magnet synchronous motor according to claim 1, characterized in that, The expression for the ZOH discrete model in step (3) is as follows: ; ; in: i γδ ( k +1) and i γδ ( k ) are respectively the first k +1 moment and the k The stator current of the motor in the synchronous rotating coordinate system at time γ-δ u γδ ( k ) is the first k The stator voltage of the motor in the synchronous rotating coordinate system at time γ-δ. e γδ ( k ) is the first k The back electromotive force of the motor in the synchronous rotating coordinate system at time γ-δ. ω e The electric angular velocity of the motor. R s To account for the equivalent stator resistance of the line filter, L s To account for the equivalent stator inductance of the line filter, T s The sampling period is j, where j is the imaginary unit. a , b , c These are the model parameters.

5. The high-precision sensorless control method for a high-speed permanent magnet synchronous motor according to claim 4, characterized in that, The expression for the inductor identification model in step (3) is as follows: ; in: This is an estimated value for the stator inductance. K , M , O For model parameters, K =Δ i γ ( k )Δ u δ ( k )-Δ i δ ( k )Δ u γ ( k ), M =Δ i γ ( k )Δ u γ ( k )-Δ i δ ( k )Δ u δ ( k ), O= Δ i γ ( k ) 2 +Δ i δ ( k ) 2 Δ i γδ ( k ) is the first k The stator current difference Δ before and after alternation in the synchronous rotating coordinate system at time γ-δ. u γδ ( k ) is the first k The stator voltage difference Δ before and after alternation in the synchronous rotating coordinate system at time γ-δ. i γ ( k ) and Δ i δ ( k ) are respectively Δ i γδ ( k The γ-axis component and δ-axis component of ) , Δ u γ ( k ) and Δ u δ ( k ) are respectively Δ u γδ ( k The γ-axis and δ-axis components of ).

6. The high-precision sensorless control method for a high-speed permanent magnet synchronous motor according to claim 5, characterized in that, In step (3), the stator inductance estimate is identified using an inductance identification model, average value filtering, and a self-iterative algorithm. The specific implementation process is as follows: Step S31: Before the alternation, when the square wave current signal is 0, continuously collect data. s The first stator current and stator voltage are obtained. s 1 set of data , i γ1 and i δ1 These are the γ-axis and δ-axis components of the stator current before alternation. u γ1 and u δ1 These are the γ-axis and δ-axis components of the stator voltage before alternation; Step S32: After alternation, the square wave current signal is -| i γinj |At that time, after the delay reaches steady state, continuous data collection is performed. s The second stator current and stator voltage are obtained. s 2 sets of data , i γ2 and i δ2 These are the γ-axis and δ-axis components of the stator current after alternation. u γ2 and u δ2 These are the γ-axis and δ-axis components of the stator voltage after alternation. i γinj The set current amplitude; Step S33: Calculate the data difference before and after alternation using the following formula. and parameters ; ; ， ， ; Step S34: Set parameters Replace the parameters in the inductor identification model respectively K , M , O The stator inductance estimate is calculated using an inductance identification model. And update the back EMF sliding mode observer; Step S35: Perform iterative calculations based on steps S31 to S34, repeating the process. N The estimated value of the output stator inductance .

7. The high-precision sensorless control method for a high-speed permanent magnet synchronous motor according to claim 6, characterized in that, The termination condition for the iterative calculation in step S35 is: ; in: This represents the average of the stator inductance estimates obtained from the first three iterations. This is the estimated value of the stator inductance obtained in this iteration. k err This is the set error constant.

8. The high-precision sensorless control method for a high-speed permanent magnet synchronous motor according to claim 1, characterized in that, In step (4), the back EMF sliding mode observer and the stator inductance estimate are introduced into the sensorless control system of the HSPMSM. The back EMF of the motor is estimated using the back EMF sliding mode observer, and the rotor position angle estimate and electric angular velocity estimate are calculated by the phase-locked loop. The amplitude is between 0 and -| i γinj An alternating square wave current signal is injected into the γ-axis current reference value. The identified stator inductance estimate is then low-pass filtered and fed back to the back EMF sliding mode observer. i γinj The set current amplitude.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: The processor is used to execute the computer program to implement the high-precision sensorless control method for high-speed permanent magnet synchronous motors as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements the high-precision sensorless control method for high-speed permanent magnet synchronous motors as described in any one of claims 1 to 8.

Images