Low speed sensorless control method based on observer considering convergence effect

Abstract

The application provides a low-speed position sensorless control method based on an observer considering convergence effect q Firstly, a permanent magnet motor current model is used to construct q a shaft current sliding surface; then, a sliding mode observer based on the shaft current model of the sliding surface is designed to accurately estimate the rotor speed and integrate the rotor position. Affected by the convergence effect, when the observer is used for low-speed operation, large load mutation can cause motor stalling, and then cause the loss of signals for the observer, and the output speed and position disorder. Therefore, a table compensation strategy based on deep symbolic regression and data filling is provided, large load mutation data is generated based on small load mutation data, so that the position error can be compensated in real time, and the stalling problem caused by the convergence effect is eliminated. The application not only realizes noiseless low-speed position sensorless control, but also solves the stalling problem caused by the convergence effect under the condition of low-speed large load mutation.

Description

Technical Field

[0001] This invention belongs to the field of permanent magnet motor control technology, specifically relating to a low-speed sensorless control method based on an observer that takes into account convergence effects. Background Technology

[0002] Currently, permanent magnet synchronous motors (PMSMs) are renowned for their compact size, high power density, and excellent stability, and are widely used in robot joint modules (RJMs). In traditional permanent magnet motor vector control, the use of mechanical position sensors increases system size and cost, while also increasing control complexity and reducing system reliability. In contrast, sensorless control methods offer advantages such as small size, low cost, and high reliability, and have a broader prospect for long-term application in robot joints. Currently, sensorless control methods for RJM permanent magnet motors can be broadly divided into two categories based on speed range: low speed (below 10% of the motor's rated speed) using high-frequency injection (HFI) methods, and high speed using back electromotive force (EMF) or flux linkage detection methods. Among these, the back EMF detection method in the high-speed range has a clear theoretical basis and is relatively mature. However, the HFI method used in the low-speed range has some problems: First, high-frequency injection generates noise, which is unacceptable in some RJM applications; second, HFI is only applicable to PMSMs with salient poles or saturated salient poles, and its applicability to commonly used surface-mounted PMSMs, whether inner or outer rotors, is insufficient. A potential method for achieving low-speed position observation is to reference back EMF estimation, that is, to directly identify relevant parameters in the motor model using an observer to achieve position detection. Although position cannot be directly obtained from the motor model, speed information can be acquired and the position calculated through integration, which makes observer-based position-free control methods possible. The foundation for this method is speed observation.

[0003] Although the theoretical basis of the above methods is relatively simple, their practical application is limited by computational delay, hysteresis introduced by filters, and the convergence effect (CE). These factors restrict their widespread application and further development. Compared to the former two, CE, as an inherent characteristic of the observer, exhibits hysteresis in the response, and its impact analysis and improvement research are relatively limited. Especially under conditions of low speed and large load abrupt changes, CE may cause motor stall, resulting in signal loss and position estimation distortion. Solving the CE problem would lay a solid foundation for the widespread application of observer-based methods. Summary of the Invention

[0004] This invention addresses the limitations of traditional low-speed sensorless control methods by proposing an observer-based low-speed sensorless control method that considers convergence effects. This method is suitable for the low-speed operation of permanent magnet synchronous motors used in robot joint modules. First, a... Axial current sliding surface; subsequently, a design based on this sliding surface is developed. A sliding mode observer based on the shaft current model is used to accurately estimate the rotor speed, and its integration yields the rotor position. Due to convergence effects, large load surges during low-speed operation using this observer can cause motor stall, leading to signal loss for the observer and disordered output speed and position. To address this, a lookup table compensation strategy based on Deep Symbolic Regression (DSR) and Data Filling (DF) is proposed. Data for large load surges is generated based on small load surge data to achieve real-time compensation for position errors, thereby eliminating the stall problem caused by convergence effects. This invention not only achieves noiseless, low-speed, sensorless control but also solves the stall problem caused by convergence effects under large load surges at low speeds. Its noiseless characteristic makes it more suitable for robot joint module motors with high precision and stability requirements, and the solution to the stall problem ensures the method's anti-interference performance in the face of load surges.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] In a first aspect, the present invention provides an observer-based low-speed sensorless control method considering convergence effects, the method comprising the following steps:

[0007] S1: Obtain the three-phase current of the permanent magnet motor and transform it according to the current rotor position, converting the three-phase current into... shaft current and ;

[0008] S2: Construction Shaft current sliding surface , ,in, for Shaft estimation current, for Actual shaft current, for Shaft estimation current error;

[0009] S3: Design based on sliding surface Axis sliding mode observer, calculation The differential value of the shaft estimated current is used to obtain the result through integration. Axis current estimation, and based on Shaft estimation current and Obtaining the actual shaft current Shaft current estimation error, through The estimated rotor speed of the motor is obtained from the shaft current estimation error;

[0010] S4: Integrate the estimated rotor speed of the motor to obtain the estimated rotor position;

[0011] S5: Based on the load mutation value output by the torque sensor in real time and the estimated speed of the motor rotor, a lookup table compensation strategy based on deep symbolic regression and data filling is adopted to obtain the estimated position error for compensation. The estimated rotor position is corrected in real time for the dynamic instantaneous process under low-speed load mutation, so as to realize noiseless low-speed sensorless control.

[0012] Furthermore, in S1, the three-phase current is converted into... shaft current and The calculation formula is:

[0013]

[0014] in, For permanent magnet motors Three-phase current, This indicates the rotor position.

[0015] Furthermore, in S3, the calculation The differential value of the shaft estimated current is calculated using the following formula:

[0016]

[0017] in, For motor phase resistance, and They are motors Shaft inductor, express Shaft control voltage, For the sliding mode gain of the observer, For symbolic functions, For the magnetic flux linkage of the motor;

[0018] pass The estimated rotor speed of the motor is obtained from the shaft current error, and the calculation formula is as follows:

[0019]

[0020] in, This indicates the estimated speed of the motor rotor.

[0021] Furthermore, sliding mode gain Based on the Lyapunov stability principle, the following conditions are required to ensure the stability of the sliding mode observer:

[0022]

[0023] in, Set the speed for the motor.

[0024] Furthermore, in step S4, the estimated rotor speed is integrated to obtain the estimated rotor position. The calculation formula is:

[0025]

[0026] in, This represents the initial position of the motor rotor.

[0027] Furthermore, in step S5, a deep symbolic regression method is used to generate compensation data under load mutations that cause motor stall. Specifically, this includes: firstly, measuring the estimated position error under small load mutations at several selected low-speed points as training data, where low speed is below 10% of the motor's rated speed, small load is 70% or less of the motor's rated load, and above 70% of the motor's rated load is considered a large load; then, using deep symbolic regression to construct a functional relationship between the estimated position error and the load mutation value, the compensation value under large load mutations that cause motor stall is obtained.

[0028] Furthermore, in step S5, when there are unsampled points within the velocity measurement range and load range, a data filling method is used to complete the data, specifically including:

[0029] Linear interpolation is performed on the known compensation value on a two-dimensional grid consisting of the load mutation value and the rotational speed;

[0030] The interpolated compensation surface is smoothed by two-dimensional Gaussian filtering and non-negative constraints are applied to generate a complete compensation surface.

[0031] Secondly, embodiments of the present invention also provide an electronic device, including a processor and a memory, wherein the memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above-described method.

[0032] Compared with the prior art, the present invention has at least the following beneficial technical effects:

[0033] This invention proposes an observer-based low-speed sensorless control method considering convergence effects, applicable to the low-speed operation control of permanent magnet synchronous motors used in robot joint modules. First, a... Axial current sliding surface; subsequently, a design based on this sliding surface is developed. A sliding mode observer based on the shaft current model is used to accurately estimate the rotor speed, and its integration yields the rotor position. Due to the convergence effect, when using this observer for low-speed operation, large load changes can cause the motor to stall, resulting in signal loss for the observer and disordered output speed and position. To address this, a lookup table compensation strategy based on deep symbolic regression (DSR) and data imputation (DF) is proposed. This strategy can generate data for large load changes based on small load change data, thereby achieving real-time compensation for position errors and eliminating the stall problem caused by the convergence effect.

[0034] This invention can not only achieve noiseless low-speed sensorless control, but also solve the stalling problem caused by convergence effect under sudden changes in low speed and large load. Its noiseless characteristic makes it more suitable for robot joint module motors with high requirements for accuracy and stability. The solution to the stalling problem ensures the anti-interference performance of this method in the face of sudden load changes.

[0035] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0036] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0037] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0038] Figure 1 A schematic flowchart of the observer-based low-speed sensorless control method considering convergence effects provided by the present invention.

[0039] Figure 2 Provided by the present invention Block diagram of the principle of the sliding mode observer.

[0040] Figure 3 A schematic diagram illustrating the compensation table generation process for Deep Symbolic Regression (DSR) and Data Imputation (DF) provided by this invention.

[0041] Figure 4 A schematic diagram of the electronic device provided by the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0043] In the description of this invention, it should be noted that some processes described in this application specification and drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may be performed in any order or in parallel. Furthermore, various numbers are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0044] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0045] Traditional low-speed sensorless control methods for permanent magnet motors in robots typically employ high-frequency injection, which not only introduces noise but also consumes some energy in generating high-frequency signals, affecting the motor's maximum output torque. This invention proposes an observer-based low-speed sensorless control method considering convergence effects, suitable for the low-speed operation of permanent magnet synchronous motors used in robot joint modules. First, a... Axial current sliding surface; secondly, design based on this sliding surface. A sliding mode observer based on the shaft current model is used to accurately estimate the rotor speed. Subsequently, the estimated speed is integrated to obtain the rotor position. However, due to convergence effects, large load surges during low-speed operation using this observer can cause motor stall, leading to signal loss for the observer and erratic output speed and position. To address this, a lookup table compensation strategy based on deep symbolic regression and data filling is proposed. Data on large load surges is generated from small load surge data to achieve real-time compensation for position errors, thereby eliminating the stall problem caused by convergence effects. Finally, real-time torque change detection and real-time speed estimation are input into the compensation table generated by the aforementioned method. The required position error compensation value is output and added to the estimated position, then output to the dual-loop vector control system to achieve low-speed, stall-free closed-loop control.

[0046] The flow chart of the observer-based low-speed sensorless control method considering convergence effect proposed in this invention is as follows: Figure 1 and Figure 2 As shown. In this embodiment, the electrical parameters of the permanent magnet motor are as follows: inductance is... The resistance is The rotor flux is ， The permanent magnet pair number is 4, the low-speed test rotation speed is 100 r / min, and the control cycle is... Moment of inertia of motor coefficient of friction The implementation process for achieving noiseless, low-speed, sensorless control and solving the stall problem caused by convergence effect under sudden changes in low speed and large load is as follows:

[0047] 1) Condition measurement and recording:

[0048] Measuring permanent magnet motors Three-phase current, denoted as And record the previous switching cycle. The estimated rotor speed of the motor obtained by the shaft sliding mode observer and rotor estimated position ,as well as Shaft control voltage and Shaft estimation current ;

[0049] 2) Coordinate transformation:

[0050] According to rotor position The Clark and Park transformations were performed sequentially, and the phase current was transformed using the following formula. Convert to shaft current and :

[0051]

[0052] 3) Selection of sliding surface:

[0053] Based on the design principle of the sliding mode observer and the design concept of the method proposed in this invention, the first step is to select the sliding surface used in the sliding mode observer. :

[0054]

[0055] in, for Shaft estimation current, for Actual shaft current, for Shaft estimation current error.

[0056] 4) Shaft current estimation calculation:

[0057] exist The shaft current model includes information about the motor rotor speed, as shown in the following equation:

[0058]

[0059] in, For motor phase resistance, and They are motors Shaft inductor, For the magnetic flux of the motor, This refers to the rotor speed of the motor.

[0060] At the same time, according to Shaft sliding mode observer selection Since the shaft current estimation error is used as the sliding surface, it is first calculated using the following formula. Differential value of shaft-estimated current:

[0061]

[0062] in, This is the sliding mode gain of the observer. For symbolic functions, through input Shaft current estimation error, output changes rapidly. To match the estimated rotational speed. The above formula is used to calculate... After estimating the differential value of the shaft current, the result can be obtained through integration. Axis current estimation.

[0063] 5) Estimated rotational speed acquisition:

[0064] Based on the setup in step 3 Shaft sliding mode observer, and the one obtained in step 3 The shaft current estimation error can be used to obtain the estimated rotor speed of the motor using the following formula:

[0065]

[0066] in, This indicates the estimated speed of the motor rotor.

[0067] In this embodiment of the invention, sliding mode gain According to the Lyapunov stability principle, the selection of the sliding mode observer must meet the following conditions to ensure its stability, as shown in the following equation:

[0068]

[0069] in, Set the speed for the motor.

[0070] 6) Estimated location acquisition:

[0071] Based on the estimated rotational speed obtained in step 5, and the relationship between rotational speed and position, the estimated position is obtained using the integration process shown in the following formula. :

[0072]

[0073] in, This represents the initial position of the motor rotor.

[0074] The above steps can be used to obtain the initial estimated speed and estimated position required by the dual closed-loop vector control system. The following section addresses the problem of stalling caused by convergence effect (CE) under sudden changes in low speed and large load.

[0075] 7) Selection of the table lookup compensation method:

[0076] To address the motor stalling problem caused by convergence effect under sudden changes in low speed and high load, this invention chooses to directly compensate for the estimated position (adding the estimated position error obtained through DSR and DF to the estimated position), rather than speed compensation, because position is the ultimate factor affecting motor performance. Considering the extremely short influence time of CE, the compensation value can be considered a constant. This also places high demands on the response speed of the compensation method; therefore, this invention employs a lookup table method. The input to the lookup table is the real-time torque from the RJM torque sensor and... The shaft sliding mode observer estimates the rotational speed and outputs it for compensation. In the event of a sudden load change, this method can quickly output a response. .

[0077] 8) Low-speed, high-load compensation data acquisition based on DSR:

[0078] During the actual data collection and construction of the aforementioned compensation table, when there is a significant sudden change in load, the motor may stall, resulting in the inability to obtain data. Furthermore, the workload required to collect such a large amount of data is infeasible. Inspired by AI-based data generation methods, this invention innovatively employs Deep Symbolic Regression (DSR) to generate data under large load mutations. Specifically, it first measures data under small load (70% or less load is considered small load, and above is considered large load) mutations at several selected low-speed points. As training data, then using DSR to construct The functional relationship between the load mutation value and the model is used to obtain the compensation value under large load mutation.

[0079] In DSR, mathematical expressions can be represented as expression trees, where internal nodes correspond to operators and leaf nodes correspond to input variables or constants. Each expression tree is converted into a token sequence through preorder traversal. For a length of... expression , No. tokens From token store Choose from. The expression is generated sequentially using a recurrent neural network (RNN), where For neural network parameters, before... tokens Under the condition of selection The conditional probability is:

[0080]

[0081] in, Represents the first RNN. The output vectors correspond to the token The probability of.

[0082] After preorder sampling of the expression, the corresponding symbolic expression is instantiated and evaluated using a reward function. Its fitness is typically measured by the normalized root mean square error (NRMSE), calculated by dividing the RMSE by the standard deviation of the target value. :

[0083]

[0084] in, Indicates the measured , This represents the input load mutation value, while Candidate expressions, This represents the total number of variables participating in the computation. After the expression is generated, the generation process is further optimized using a recurrent neural network (RNN) and policy gradient to improve the fitness of the expression. The RNN operates on a scale of [size missing]. The process involves training on a batch of data and returning the gradient of a defined loss function to generate the optimized reward, as follows:

[0085]

[0086] in, For loss function, Parameters for determining the degree of risk sought. It is a reward function. It is a set of expressions experience Reward percentile, The policy gradient guides the direction of parameter updates. As an indicator function, it is updated only for high-reward samples. This method allows us to obtain the corresponding value at the selected speed. The expression for the functional relationship between the load mutation value and the load mutation value. Then, by substituting the large load mutation value, the compensation value corresponding to the speed can be obtained.

[0087] 9) Data acquisition for complete low-speed load compensation based on DF coverage:

[0088] Considering that sampling and generating large load mutation data for each speed point using DSR is not feasible, this paper introduces the DF method to fill in missing data in two dimensions: load mutation value and speed. First, missing values ​​for unselected speed and unselected load mutation points are generated by interpolation, and then Gaussian filtering is used to suppress interpolation errors and noise.

[0089] In two-dimensional interpolation, assuming the compensation function value... It is known on a regular grid. Based on this, its four neighboring points can be used. Calculate the new query point The interpolated values ​​at the location, where First along Linear interpolation is performed in the direction to obtain two intermediate values. and :

[0090]

[0091] Then, using along Two intermediate values ​​obtained from the direction and along Interpolation is performed in the direction to obtain the final interpolation result. :

[0092]

[0093] After interpolation, the originally sparse data in the lookup table becomes denser, which can be viewed as a surface in three-dimensional space. Subsequently, a two-dimensional Gaussian convolution is performed on it to suppress interpolation errors and high-frequency noise. The Gaussian kernel is defined as follows:

[0094]

[0095] in, and Indicates the horizontal and vertical distances relative to the center of the nucleus. The value of the Gaussian kernel. The standard deviation is used to control the width of the Gaussian distribution, thus determining the degree of smoothing. Applying Gaussian filtering... The process can be represented as two-dimensional convolution:

[0096]

[0097] in, This represents a surface that has undergone smoothing. This indicates that the function will be used. With nuclear Perform convolution. This represents a dummy variable in the convolution integral. During this process, Convolution is performed with a Gaussian kernel. To prevent negative values ​​from being obtained, a non-negativity constraint needs to be applied. Therefore, the final expression is:

[0098]

[0099] in, This represents the approximate surface obtained through interpolation. The standard deviation is expressed as Two-dimensional Gaussian kernel, This is a non-negative lower bound. Using the DF method, a range covering the entire low-speed range and all load conditions can be generated. Compensated surface.

[0100] In this embodiment, the compensation table generation process for Deep Symbolic Regression (DSR) and Data Imputation (DF) is as follows: Figure 3 As shown, firstly, the estimated position error under small load changes at a selected speed in the low-speed range is sampled; then, DSR is used to generate the load change value and estimated position error at the selected speed. The functional relationship between them is shown in Table 1 below:

[0101] Table 1. Relationships of DSR Generation Expression Functions

[0102]

[0103] By substituting the large load mutation value, the corresponding estimated position error can be obtained; then, DR is used to generate the estimated position error covering the entire low-speed range, thereby generating a compensation table covering all operating conditions in the low-speed range, which is used to compensate for the estimated position error caused by CE.

[0104] The above process enables noiseless, low-speed, sensorless control and solves the stall problem caused by convergence effect under sudden changes in low speed and high load.

[0105] As described in the above embodiments, in order to achieve noiseless low-speed sensorless control and solve the stalling problem caused by convergence effect under sudden changes in low speed and large load, this invention provides an observer-based low-speed sensorless control method considering convergence effect, applicable to the low-speed operation of permanent magnet synchronous motors used in robot joint modules. First, a current model of the permanent magnet motor is constructed... Axial current sliding surface; subsequently, a design based on this sliding surface is developed. A sliding mode observer based on the shaft current model is used to accurately estimate the rotor speed, and its integration yields the rotor position. Due to convergence effects, large load surges during low-speed operation using this observer can cause motor stall, leading to signal loss for the observer and disordered output speed and position. To address this, a lookup table compensation strategy based on Deep Symbolic Regression (DSR) and Data Imputation (DF) is proposed. Data for large load surges is generated based on small load surge data to achieve real-time compensation for position errors, thereby eliminating the stall problem caused by convergence effects. This invention not only achieves noiseless, low-speed, sensorless control but also solves the stall problem caused by convergence effects under large load surges at low speeds. Its noiseless characteristic makes it more suitable for robot joint module motors with high precision and stability requirements, and the solution to the stall problem ensures the method's anti-interference performance in the face of load surges.

[0106] Furthermore, refer to Figure 4 As shown, this embodiment of the invention also provides an electronic device that can perform the above-described method. The electronic device may include a processor 10, a memory 11, a communication bus 12, and a communication interface 13, and may also include a computer program stored in the memory 11 and capable of running on the processor 10.

[0107] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 and calls data stored in the memory 11 to perform various functions of the electronic device and process data.

[0108] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, electronic devices, or computer program products, etc. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The words "a" or "an" preceding a component do not exclude the presence of a plurality of such components. This invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer.

[0110] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0111] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A low-speed sensorless control method based on an observer, considering convergence effects, characterized in that, The method includes the following steps: S1: Obtain the three-phase current of the permanent magnet motor and transform it according to the current rotor position, converting the three-phase current into... shaft current and ; S2: Construction Shaft current sliding surface , ,in, for Shaft estimation current, for Actual shaft current, for Shaft estimation current error; S3: Design based on sliding surface Axis sliding mode observer, calculation The differential value of the shaft estimated current is used to obtain the result through integration. Axis current estimation, and based on Shaft estimation current and Obtaining the actual shaft current Shaft current estimation error, through The estimated rotor speed of the motor is obtained from the shaft current estimation error; S4: Integrate the estimated rotor speed of the motor to obtain the estimated rotor position; S5: Based on the load mutation value output by the torque sensor in real time and the estimated speed of the motor rotor, a lookup table compensation strategy based on deep symbolic regression and data filling is adopted to obtain the estimated position error for compensation. The estimated position of the rotor in the dynamic instantaneous process under low-speed load mutation is corrected in real time to achieve noiseless low-speed sensorless control. In step S5, deep symbolic regression is used to generate compensation data under load mutations that cause motor stall. Specifically, this includes: first, measuring the estimated position error under small load mutations at several selected low-speed points as training data, where low speed is below 10% of the motor's rated speed, small load is 70% or less of the motor's rated load, and above 70% of the motor's rated load is considered a large load; then, deep symbolic regression is used to construct a functional relationship between the estimated position error and the load mutation value to obtain the compensation value under large load mutations that cause motor stall. In step S5, when there are unsampled points within the speed measurement range and load range, a data filling method is used to complete the data, specifically including: Linear interpolation is performed on the known compensation value on a two-dimensional grid consisting of the load mutation value and the rotational speed; The interpolated compensation surface is smoothed by a two-dimensional Gaussian filter and a non-negative constraint is applied to generate a complete compensation surface.

2. The method according to claim 1, characterized in that, In S1, the three-phase current is converted into... shaft current and The calculation formula is: in, For permanent magnet motors Three-phase current, This indicates the rotor position.

3. The method according to claim 2, characterized in that, In S3, calculation The differential value of the shaft estimated current is calculated using the following formula: in, For motor phase resistance, and They are motors Shaft inductor, express Shaft control voltage, For the sliding mode gain of the observer, For symbolic functions, For the magnetic flux of the motor; pass The estimated rotor speed of the motor is obtained from the shaft current error, and the calculation formula is as follows: in, This indicates the estimated speed of the motor rotor.

4. The method according to claim 3, characterized in that, Sliding mode gain Based on the Lyapunov stability principle, the following conditions must be followed to ensure the stability of the sliding mode observer: in, Set the speed for the motor.

5. The method according to claim 3, characterized in that, In step S4, the estimated rotor speed is integrated to obtain the estimated rotor position. The calculation formula is: in, This represents the initial position of the motor rotor.

6. An electronic device, characterized in that, It includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor, the processor executing the machine-executable instructions to perform the method as described in any one of claims 1-5.

Images