SMO-EKF brushless direct current motor position-free control method based on SBA-GWO double-layer optimization

Abstract

The invention discloses an SMO-EKF brushless direct current motor position-free control method based on SBA-GWO double-layer optimization, SMO adopts sliding mode surface gradient-rotating speed double-factor segmented gain, and an improved SBA-GWO algorithm is combined with a global exploration mechanism of a stingless bee algorithm and a group cooperation characteristic of a grey wolf optimization algorithm. An elite initialization strategy and a dynamic search mode switching mechanism based on a grey wolf level are introduced, and an SBA-GWO algorithm is adopted for parameter optimization: in the first layer, SMO counter electromotive force THD is taken as fitness, and optimal gain is searched; in the second layer, the EKF estimation error quadratic sum is used as the fitness, and the Q / R adaptive residual noise time-varying property is optimized. And constructing an SMO-EKF cooperative control closed loop according to the optimized parameters to realize brushless direct current motor position-free control. Compared with the prior art, in the BLDC full-rotating-speed-domain position-free control, the dynamic response speed and the steady-state buffeting suppression effect are both considered, the robustness and the engineering realizability are remarkably improved, a physical sensor is not needed, and the engineering cost is low.

Description

Technical Field

[0001] This invention relates to the field of sensorless motor control, specifically to a sensorless control method for SMO-EKF brushless DC motors based on SBA-GWO dual-layer optimization. Background Technology

[0002] In the field of brushless DC motor (BLDC) control, sensorless control technology has become a core research direction in industrial automation, new energy vehicles, and smart homes because it avoids the problems of increased cost, limited installation space, and reduced reliability associated with physical sensors (such as encoders and Hall sensors). Among them, control schemes based on sliding mode observers (SMO) and extended Kalman filters (EKF) have become the mainstream choice for mid-to-high-end motor control systems due to their effective estimation capabilities of motor states. However, current technologies still have several key shortcomings, making it difficult to meet the high-precision control requirements under all operating conditions.

[0003] Currently, traditional SMOs employ fixed or single-factor adaptive gain, which cannot balance the requirements of "rapid approach" and "chatter suppression." At low speeds, back EMF waveform distortion is prone to occur, while at high speeds, current observation errors converge lagging, and position estimation accuracy significantly decreases during speed switching. EKF's Q / R matrix is ​​mostly based on empirical fixed values, failing to adapt to noise differences at different motor speeds. At high speeds, system noise increases, and fixed Q / R results in decreased filtering accuracy and slower dynamic response, making it difficult to cover wide-speed-range control requirements.

[0004] Therefore, current technologies have shortcomings in gain adaptation, noise dynamics, and observer coordination. There is an urgent need for a sensorless control method that can achieve high precision and low jitter across the entire speed range to meet the needs of practical applications. Summary of the Invention

[0005] Purpose of the invention: In view of the problems pointed out in the background art, the present invention provides a positionless control method for SMO-EKF brushless DC motor based on SBA-GWO dual-layer optimization, which realizes high-precision, low-jitter, and fast-response sensorless control of brushless DC motor across the entire speed range, meets the motor control performance requirements of industrial drive, smart home and other scenarios, and reduces reliance on physical sensors, balancing cost and reliability.

[0006] Technical solution: This invention discloses a positionless control method for SMO-EKF brushless DC motors based on SBA-GWO dual-layer optimization, comprising the following steps:

[0007] S1: Constructing a brushless DC motor Coordinate coefficient mathematical model and SMO infrastructure, wherein the SMO infrastructure design is a "sliding surface gradient-rotation speed" dual-factor piecewise adaptive gain;

[0008] S2: Establish the EKF state estimation model, clarify the state variables and control variables, and bring the prior state after discretization by the extended Kalman filter of the brushless DC motor state estimation into the extended Kalman filter recursive process for processing.

[0009] S3: Using the SBA-GWO algorithm with the SMO back EMF THD as the fitness function. Optimize SMO gain parameters to find the optimal gain;

[0010] S4: Reuse the SBA-GWO algorithm, using the sum of squared errors estimated by EKF as the fitness function. Optimize the EKF noise covariance matrix parameters;

[0011] S5: Construct an SMO-EKF collaborative control closed loop based on the optimized parameters of S3 and S4 to achieve full-speed-domain positionless control of the brushless DC motor.

[0012] Furthermore, the SMO infrastructure includes a current observation equation and a sliding surface function. The sliding surface function is defined as the current observation error. The designed "sliding surface gradient-speed rotation" dual-factor piecewise adaptive gain cancels out the error term through the back EMF observation value, so that the observation error converges to 0.

[0013] Furthermore, the "sliding surface gradient-rotation speed" dual-factor piecewise adaptive gain in S1 is specifically as follows:

[0014]

[0015] in, , , The gain parameter to be optimized increases in value with increasing rotational speed. , This is the rated speed of the motor. =1000rpm is the medium speed threshold , The gradient threshold and minimum gradient value of the sliding surface. This represents the error in current observation.

[0016] Furthermore, the back electromotive force observation is calculated by combining a two-factor piecewise adaptive gain with a sigmoid function for smoothing, specifically as follows:

[0017]

[0018]

[0019] in, , The back electromotive force observation value, This is the gain coefficient.

[0020] Furthermore, the state variables of the EKF state estimation model are The control variables are The output variable is , , They are respectively , Shaft stator current, Electric angular velocity, For rotor electrical angle, , They are respectively , Shaft stator voltage.

[0021] Furthermore, in S3, the fitness is defined by the SMO back electromotive force THD, i.e., the objective function is to minimize the total harmonic distortion rate (THD) of the SMO output back electromotive force.

[0022]

[0023] in, , The back electromotive force observation value, Calculated via Fourier transform This represents the amplitude of the fundamental back electromotive force. for Second harmonic amplitude.

[0024] Furthermore, step S4 considers that EKF needs to simultaneously optimize the accuracy of rotational speed and position estimation, and constructs a weighted fitness function:

[0025]

[0026] in, , Let λ be the estimated EKF rotational speed and position values ​​for the k-th iteration, and λ be the position error weighting coefficient. Electric angular velocity, For rotor electrical angle, This represents the total number of iterations.

[0027] Furthermore, when the SBA-GWO algorithm uses the SMO back electromotive force THD as the fitness to optimize the SMO gain parameter and find the optimal gain, the optimization variable, i.e., each particle in the population, is the SMO gain parameter. , , , Let f be the gain parameter to be optimized, and f be the fitness function. The SBA-GWO algorithm is reused, with the sum of squared errors of the EKF estimation as the fitness. When optimizing the parameters of the EKF noise covariance matrix, the optimization variable is each particle in the population. , The process noise covariance matrix is... The process noise covariance matrix is ​​4×4. Representation matrix The four different parameters to be optimized on the main diagonal represent the process noise covariance of current, speed, and position, respectively. To measure the noise covariance matrix, where Let be the parameters to be optimized on the main diagonal of matrix R, and be the process noise covariance of the sensor. Let f be the population size, and let f be the fitness function. .

[0028] Furthermore, the SBA-GWO algorithm introduces GWO elite direction guidance during the SBA algorithm's bee-hiring phase, guiding each particle in the population... Fitness is calculated, and the three particles with the best fitness are selected as the guide individuals for the GWO. ,in, , , For α wolf, β wolf, and δ wolf, respectively, the improved formula is as follows:

[0029]

[0030] in, The candidate position vector of the i-th particle at the t-th iteration. The current position vector of the i-th particle at the t-th iteration. The current position vector of the i-th particle at the (t+1)-th iteration. The particle with the highest fitness in the population at the t-th iteration. For the search radius, Generate a random decimal number between 0 and 1. The term refers to the elite guidance direction introduced by GWO; in the reconnaissance bee phase, when the particle satisfies T represents multiple consecutive iterations, and these are labeled as stationary particles. An adaptive reset is performed on continuously stationary particles, as shown in the following formula:

[0031]

[0032] in, The maximum value of the j-th dimension parameter, The minimum value of the j-th dimension parameter, The value of the j-th dimension parameter of the i-th particle at the (t+1)-th iteration. The value of the j-th dimension parameter of the i-th particle at the t-th iteration. Let be the Euclidean distance between the particle and the optimal particle.

[0033] Furthermore, in S5, the SMO-EKF cooperative control closed loop combines the sliding mode observer (SMO) and the extended Kalman filter (EKF) in a cooperative closed-loop structure, forming a two-layer observation mechanism of "SMO coarse estimation - EKF fine refinement":

[0034] Substituting the S3-optimized SMO gain parameters into the SMO infrastructure, based on... Shaft measured current , With voltage , Rough estimate of output rotor speed Rough estimation of location Using this as the initial state of EKF, and combining the noise covariance matrix parameters Q and R optimized by S4, a high-precision rotational speed is output through EKF discretization recursive calculation. With position ,Will Feedback is sent to the speed loop PI controller, compared with the speed reference value to generate the d-axis current reference value idref, and then... Used for Park transformation to achieve complete control.

[0035] Beneficial effects:

[0036] 1. This invention integrates SBA global exploration and GWO local development, and avoids the local convergence defects of a single algorithm through elite guidance, neighborhood perturbation and adaptive reset.

[0037] 2. The optimization effect of this invention on SMO is that after optimizing the dual-factor piecewise gain, the back EMF harmonic distortion rate is significantly reduced, the fluctuation of current observation error is reduced, the convergence speed is accelerated at high speed, and the chattering is significantly reduced at low speed, thus achieving stable position coarse estimation across the entire speed range.

[0038] 3. The process noise matrix of the SBA-GWO optimization of EKF in this invention. With measurement noise matrix This allows noise parameters to dynamically match the motor's noise characteristics across the entire speed range, resulting in a smaller steady-state error in EKF speed estimation compared to traditional fixed Q / R; enhanced anti-interference capability: the optimized... / The Kalman gain can balance the prediction and measurement weights of the EKF in real time, and adjust the Kalman gain when there are sudden load changes or speed switching. Adaptive adjustment, position estimation error The fluctuation range is small, avoiding the estimation drift caused by noise mismatch in traditional EKF.

[0039] 4. The coarse estimation result of the SMO in this invention provides the EKF with an initial state that closely matches the actual working conditions, avoiding convergence lag caused by initial value deviation of the EKF; at the same time, the fine-tuning capability of the EKF compensates for the small errors of the SMO in the steady-state stage, forming a complementary closed loop of "coarse estimation-fine-tuning", solving the problems of residual chattering of a single SMO or insufficient dynamic response of a single EKF. The collaborative architecture does not require additional sensor hardware, and performance upgrades can be achieved only through algorithm optimization, reducing hardware costs; at the same time, the high-precision speed and position signals output by the collaboration can directly adapt to the control requirements of the motor speed loop and current loop, reducing the adjustment pressure of subsequent controllers and improving the robustness and reliability of the entire control system. Attached Figure Description

[0040] Figure 1 This is a flowchart of the brushless motor control strategy;

[0041] Figure 2 A diagram illustrating the parameter optimization process for the SBA-GWO algorithm;

[0042] Figure 3 The graph shows a performance comparison of the SBA-GWO algorithm under different fitness functions.

[0043] Figure 4 Diagram showing the control effect of a traditional SMO;

[0044] Figure 5 This is a diagram showing the optimized SMO-EKF control effect of the present invention. Detailed Implementation

[0045] 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 only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] This embodiment discloses a positionless control method for an SMO-EKF brushless DC motor based on SBA-GWO dual-layer optimization. The method steps are as follows:

[0047] S1: Constructing a brushless DC motor Based on the limitations of traditional gain design and the requirements of full-speed-domain operation, a coordinate coefficient mathematical model and SMO basic architecture are designed to implement a two-factor segmented adaptive gain.

[0048] S1.1: Establishing a brushless DC motor at rest on two phases State-space equations in coordinate system and sliding mode brushless DC motor The equations for shaft voltage and current are:

[0049]

[0050]

[0051] in, , They are respectively , Shaft stator voltage, , They are respectively , Shaft stator current, For stator resistance, For stator equivalent inductance, Electric angular velocity, It is a permanent magnet flux chain. The rotor electrical angle is given.

[0052] S1.2: Constructing the current observation equations and sliding surface functions of the SMO:

[0053] The equation for current observation is:

[0054]

[0055] The sliding surface function is defined as the current observation error:

[0056]

[0057] in, , They are respectively , Shaft stator current observations , The back electromotive force observation value, The observed rotor electrical angle values ​​are summarized as follows:

[0058]

[0059] The motor parameters simulated in this embodiment (adapted for brushless DC motors in the transportation field) are shown in Table 1 below:

[0060] Table 1 BLDC Parameter Table

[0061] parameter numerical values parameter numerical values Stator resistance 7.2Ω Stator inductor 5.5mh Extreme logarithm 4 Damping coefficient 1.2Nms Moment of inertia <![CDATA[0.8kg.m 2 ]]> Rated speed 1700rpm

[0062] To make the observation error converge to 0 (i.e. →0), requires back electromotive force observation. To offset the error term, the designed two-factor piecewise adaptive gain formula is as follows:

[0063]

[0064] in, , , For the gain parameters to be optimized, This is the rated speed of the motor. =1000rpm is the medium speed threshold , The gradient threshold and minimum gradient value of the sliding mode surface are given; the back electromotive force observation is combined with piecewise gain and sigmoid function smoothing calculation, as shown in the following formula:

[0065]

[0066]

[0067] in, , The back electromotive force observation value, For sliding surface (current observation error). is the gain coefficient (determining the rate of change of the function). By dynamically adapting the gain using a two-factor method, the error equation is made to converge the observation error to zero across the entire rotational speed domain.

[0068] S2: Establish the EKF state estimation model, clarifying the state variables, control variables, and discretized recursive calculation process. The core logic of EKF is to deduce the unmeasurable state (speed, position) in reverse from the known input (voltage) and measurable output (current).

[0069] S2.1: Establish the EKF state estimation model and the state-space equations of the nonlinear system of the brushless DC motor:

[0070]

[0071] Among them, state variables Control variables Output variables .

[0072] , .

[0073] right After linearization, the corresponding Jacobian matrices are obtained as follows:

[0074]

[0075]

[0076] Based on the above equations, the prior state equation after discretization using the extended Kalman filter for brushless DC motor state estimation can be obtained as follows:

[0077]

[0078] Where k represents the current iteration number, , , , These represent the input state variables in the k-th iteration. Shaft stator current, Shaft stator current, electric speed, and electric angle; Represents the shaft-stator voltage and shaft-stator voltage under k-1 iterations.

[0079] Substitute into the extended Kalman filter recursive process:

[0080]

[0081] in, These are prior estimates. This is the prior estimate from the previous time step. Sampling time.

[0082]

[0083] in, Here is the state transition matrix. It is a 4th order identity matrix. Let Jacobi be the state transition matrix. At that time, among them The Jacobian matrix is ​​equivalent to a constant that is unaffected by time t.

[0084]

[0085] in, Let be the prior error covariance matrix. Let Q be the optimal covariance matrix at the previous time step, and let Q represent the covariance matrix of the system process noise.

[0086]

[0087] in, R represents the Kalman filter gain, and R represents the covariance matrix of the system measurement noise.

[0088]

[0089] in, for The inverse matrix, This is the output state variable during the k-th iteration.

[0090] The optimal covariance matrix is ​​updated as follows:

[0091]

[0092] in, The process noise covariance matrix is... To measure the noise covariance matrix.

[0093] S3: Optimize SMO gain parameters using the SBA-GWO algorithm to achieve source chattering suppression.

[0094] S3.1: Optimize the SMO gain parameters using the SBA-GWO algorithm. Initialize the core parameters of the SBA-GWO algorithm: population size N (balancing optimization accuracy and computational cost) is set to 30, maximum number of iterations tmax (adapting to real-time control requirements) is set to 50, and the stingless wasp search radius (controlling the global exploration range) is set to 0.5. Adapt to a reasonable SMO gain range of [5, 200] to ensure that the global exploration does not miss the optimal solution while avoiding inefficient searching. The gray wolf population has α-wolf (globally optimal), β-wolf (second best), and δ-wolf (third best) hierarchical factors. The optimization variable is the SMO gain parameter designed in S1. The fitness function design for SMO gain optimization aims to minimize the total harmonic distortion (THD) of the SMO output back electromotive force, and constructs the fitness objective function as follows:

[0095]

[0096] in, Calculated via Fourier transform This represents the amplitude of the fundamental back electromotive force. for Second harmonic amplitude.

[0097] This invention introduces GWO elite orientation guidance during the hired bee phase, guiding each particle in the population. Calculate fitness and select the three particles with the best fitness as the guide individuals for GWO. , The objective function is to minimize the total harmonic distortion (THD) of the SMO output back electromotive force, where , , The α wolf (optimal), β wolf (second best), and δ wolf (third best) are respectively represented by the following position update formulas:

[0098]

[0099] in, For the search radius, Generate a random decimal number between 0 and 1 (inclusive of 0, exclusive of 1). This is an elite-guided direction introduced by GWO.

[0100] During the reconnaissance bee phase, when the particles satisfy... These are marked as stationary particles. An adaptive reset is performed on continuously stationary particles, as shown in the following formula:

[0101]

[0102] in, The j-th dimension parameter of the best particle in the current population Let be the Euclidean distance between the particle and the optimal particle.

[0103] S4: Reuse the SBA-GWO algorithm to optimize the EKF noise covariance matrix parameters and improve the back-end filtering accuracy.

[0104] Considering that EKF needs to simultaneously optimize the accuracy of rotational speed and position estimation, a weighted fitness function is constructed:

[0105]

[0106] in, , The EKF speed and position estimates are for the kth iteration, and λ=0.1 is the position error weighting coefficient (since the speed error has a more significant impact on control performance, a lower weight is set).

[0107] At this point, the SBA-GWO algorithm is reused, and the fitness function... The weighted fitness function of EKF The optimization variable is : .

[0108] Generate N initial particles, each particle corresponding to a set of Q / R parameter combinations. The process noise covariance matrix is... To measure the noise covariance matrix, the position update formula and the adaptive formula are designed as described in S3 above, outputting the globally optimal particle. The corresponding optimal noise covariance matrix is . Optimal / Substitute into the EKF recursive process and update the Kalman gain. With error covariance matrix .

[0109] S5: Construct an SMO-EKF collaborative control closed loop to achieve full-speed-domain positionless control of the brushless DC motor.

[0110] Construct a SMO-EKF cooperative control closed loop, substituting the optimized SMO gain parameters from S3 into the SMO module, based on... Shaft measured current , With voltage , Rough estimate of output rotor speed Rough estimation of location Using this as the initial state of EKF, and combining it with the optimized Q and R matrices from S4, a high-precision rotational speed is output through EKF discretization recursive calculation. With position Will Feedback is sent to the speed loop PI controller, compared with the speed reference value to generate the d-axis current reference value idref (for brushless DC motors, idref is typically set to 0 to achieve maximum torque control). Used for Park transformation to achieve complete control.

[0111] See Figure 3 This invention utilizes the SBA-GWO algorithm, traditional spikeless algorithm, gray wolf algorithm, and particle swarm optimization algorithm in terms of fitness function. Simulation experiments were conducted under the given conditions, and the results show that the SBA-GWO algorithm has good convergence speed and convergence performance.

[0112] See Figure 4 and Figure 5 , Figure 4 This is a diagram illustrating the control effect of a traditional SMO. Figure 5 The image shows the effect of the optimized SMO-EKF control. During the experiment, the reference speed was changed from 0 rpm to 600 rpm, 600 rpm to 1000 rpm, and 1000 rpm to 1700 rpm. It can be clearly seen that the optimized SMO-EKF control effect in the low-speed range is almost without error compared to the reference speed. Compared with the traditional SMO control, which has distortion in the low-speed range, the improvement is significant.

[0113] The foregoing description of the embodiments enables those skilled in the art to make or use the present invention. Various modifications to the embodiments will be readily apparent to those skilled in the art. The general principles of the invention may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention should not be limited to the embodiments shown herein, but should cover the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A SMO-EKF brushless DC motor positionless control method based on SBA-GWO double-layer optimization, characterized in that, The method comprises the following steps: S1: Constructing a brushless DC motor The coordinate system mathematical model and the SMO infrastructure are designed with "sliding mode surface gradient-speed" double factor segmented adaptive gain. S2: Establishing an EKF state estimation model, defining state variables and control variables, and bringing the prior state of the discrete extended Kalman filter for brushless direct current motor state estimation into the recursive process of the extended Kalman filter; S3: SMO back-EMF THD as fitness function by SBA-GWO algorithm , and the optimal gain is searched by optimizing SMO gain parameter S4: multiplex SBA-GWO algorithm, EKF estimation error sum of squares as fitness function , optimization of EKF noise covariance matrix parameters; S5: Constructing an SMO-EKF collaborative control closed loop according to the optimized parameters in S3 and S4, and completing full-speed-range positionless control of the brushless direct current motor.

2. The SMO-EKF sensorless control method of a brushless DC motor based on SBA- GWO double-layer optimization according to claim 1, characterized in that, The SMO infrastructure comprises a current observation equation and a sliding mode surface function, the sliding mode surface function is defined as a current observation error, and the designed "sliding mode surface gradient-speed" double-factor segmented adaptive gain offsets the error term through the back electromotive force observation value, so that the observation error converges to 0.

3. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 2, characterized in that, The "sliding mode surface gradient-speed" double-factor segmented adaptive gain in S1 is specifically: ； wherein, , , is a gain parameter to be optimized, whose value increases with the increase of the rotational speed, , is the rated rotational speed of the electric machine, = 1000 rpm is a medium speed threshold value, , is a sliding mode surface gradient threshold value and a minimum gradient value, is the current observation error.

4. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 3, characterized in that, The back electromotive force observation value is combined with the double-factor segmented adaptive gain and sigmoid function smoothing calculation, and is specifically: ； ； wherein , is the back electromotive force observation value, is the gain coefficient.

5. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 1, wherein, The state variables of the EKF state estimation model are: The control variables are The output variable is , , They are respectively , Shaft stator current, Electric angular velocity, For rotor electrical angle, , They are respectively , Shaft stator voltage.

6. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 1, wherein, In S3, the SMO back electromotive force THD is used as the fitness, that is, the minimization of the total harmonic distortion rate THD of the SMO output back electromotive force is used as the target, and a fitness target function is constructed: ； wherein , is the back EMF observation value, is calculated by Fourier transformation, is the back EMF fundamental amplitude, is the sub-harmonic amplitude.

7. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 1, wherein, Step S4 considers that the EKF needs to optimize the speed and position estimation accuracy at the same time, and a fitness function containing a weight is constructed: ； wherein, , is the EKF speed, position estimate for the kth iteration, and λ is a position error weight coefficient, is the electrical angular velocity, is the rotor electrical angle, is the total number of iterations.

8. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 6 or 7, characterized in that, In the SBA-GWO algorithm, the SMO gain parameter is optimized to find the optimal gain, and the optimization variable, i.e. each particle in the population, is the SMO gain parameter , , , is the gain parameter to be optimized, and the fitness function f is ; The SBA-GWO algorithm is multiplexed to optimize the EKF noise covariance matrix parameters, with the EKF estimation error sum of squares as the fitness, and the optimization variable is each particle in the population , is the process noise covariance matrix, is a 4 × 4 process noise covariance matrix, represents the matrix The four different to-be-optimized parameters on the main diagonal line in the matrix are current, speed, and position process noise covariance; is the measurement noise covariance matrix, wherein is the to-be-optimized parameter on the main diagonal line in the matrix R, which is the process noise covariance of the sensor; is the population size, and the fitness function f is .

9. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 8, wherein, The SBA-GWO algorithm introduces the GWO elite direction guidance in the SBA algorithm's hiring bee stage, and the GWO elite direction guidance is used to guide each particle in the population The fitness is calculated, and the three particles with the optimal fitness are selected as the guide individuals of the GWO, wherein, , , are alpha wolf, beta wolf and delta wolf respectively, and the improved formula is as follows: ； wherein, the candidate position vector of the i-th particle at the t-th iteration, the current position vector of the i-th particle at the t-th iteration, the current position vector of the i-th particle at the t+1-th iteration, the particle with the highest fitness of the population at the t-th iteration, is the search radius, generating a random number between 0 and 1, is the elite-guided direction introduced by GWO; at the scout bee stage, when a particle meets T means that the iteration is performed continuously for many times, and the particle is marked as a stagnation particle. The adaptive reset is performed on the particle that stagnates continuously, and the formula is as follows: ； wherein, a maximum value of the jth dimension parameter, a minimum value of the jth dimension parameter, a value of the jth dimension parameter of the ith particle at the t+1th iteration, a value of the jth dimension parameter of the ith particle at the tth iteration, is the Euclidean distance of the particle from the best particle.

10. The SBA-GWO double-layer optimized SMO-EKF positionless control method of a brushless DC motor according to claim 1, wherein, In S5, the SMO-EKF collaborative control closed loop combines the sliding mode observer SMO and the extended Kalman filter EKF in a collaborative closed loop structure, forming a "SMO rough estimation-EKF fine tuning" double-layer observation mechanism: The SMO gain parameter optimized by S3 is substituted into the SMO infrastructure, and based on Shaft measured current , Voltage , , the rough estimated value of the output rotor speed and the rough estimated value of the position , which are used as the initial state of the EKF, combined with the noise covariance matrix parameters Q and R optimized by S4, through the discrete recursive calculation of the EKF, the high-precision speed and position are output, and is fed back to the speed loop PI controller, compared with the speed reference value to generate the d-axis current reference value idref, and is used for Park transformation to complete the complete control.

Images