A permanent magnet synchronous motor control method based on load torque compensation

Abstract

The present invention discloses a permanent magnet synchronous motor control method based on load torque compensation, wherein the speed of the permanent magnet synchronous motor is controlled by a PI speed controller, a load torque observer observes the load torque, and the output of the load torque observer is used to compensate the output of the PI speed controller for the load torque. The load torque observation uses an algorithm that automatically adjusts the feedback gain according to the change in the load torque given value and the change in the load torque observation value, and can quickly reduce the load torque observation error and feed-forward the load torque observation value to the current regulator when the system speed changes, or the parameters change, the load is disturbed, and the load torque given value changes or / and the load torque observation value changes, thereby improving the control accuracy of the permanent magnet synchronous motor.

Description

[0001] This invention patent application is a divisional application. The original application number is 202010918629.X. The application date is September 4, 2020. The name of the invention is elevator permanent magnet synchronous motor drive control system. Technical Field

[0002] The present invention relates to the technical field of permanent magnet synchronous motors, and more specifically, to a permanent magnet synchronous motor control method based on load torque compensation. Background Art

[0003] Permanent magnet synchronous motors have the advantages of high efficiency, high torque, and good speed performance, and are widely used in manufacturing, electric vehicles, industrial production and other fields. When using permanent magnet synchronous motors as elevator traction motors, their low speed, rapidity, hard mechanical characteristics, etc. are better than AC asynchronous motors. Their life, energy consumption, and maintainability are better than DC motors. It is easy to achieve an ideal elevator drive with low speed and high torque, and meet the requirements of strong overload capacity and wide speed regulation range for elevator drives. The control method based on load torque compensation can reduce the impact of load disturbances, but when the load torque changes, the given speed changes, or the internal parameters of the control system are perturbed, the load torque observation deviation will increase, affecting the performance indicators of the permanent magnet synchronous motor speed control, such as steady-state, dynamic, and speed regulation range. Summary of the invention

[0004] The purpose of the present invention is to provide a permanent magnet synchronous motor control method based on load torque compensation, which feeds the observed load torque into the current regulator, improves the load torque observation response speed and reduces the torque observation volatility. The speed of the permanent magnet synchronous motor is controlled by a PI speed controller, and the load torque observer observes the load torque. The output of the load torque observer is used to compensate the output of the PI speed controller for load torque. PI speed controller PI speed controller output load torque given value and torque current given component i′ q for

[0005]

[0006] Where p is the number of motor pole pairs, ψ f is the permanent magnet flux; K p is the proportional coefficient of the PI speed controller, T i is the integral time constant of the PI speed controller; the motor's rotor angular velocity error e = ω * -ω,ω * is the given rotor angular velocity, ω is the rotor angular velocity, and e(t) is the instantaneous value of the rotor angular velocity error.

[0007] The load torque observer is

[0008]

[0009] Where J is the moment of inertia, is the estimated value of the rotor angular velocity, g is the feedback gain of the load torque observer and g<0; k g is the sliding mode gain of the load torque observer and k g ≤-|e2 / J|, is the load torque observation error, T L is the load torque, is the observed value of load torque; i q is the q-axis current.

[0010] The load torque observer is based on the load torque given value and load torque observation The feedback gain g is adjusted by:

[0011] Step S1: The load torque observer performs T on the load torque. L Observe and obtain the load torque observation value The PI speed controller performs control calculation to obtain the load torque given value

[0012] Step S2: Calculation

[0013] Step S3, determine whether ΔT is greater than ε2; ​​when ΔT is greater than ε2, set the feedback gain g equal to g min And exit; when ΔT is less than or equal to ε2, enter step S4;

[0014] Step S4, determine whether ΔT is less than ε1; when ΔT is less than ε1, set the feedback gain g equal to g max And exit; when ΔT is greater than or equal to ε1, enter step S5;

[0015] Step S5: Feedback gain g is calculated according to

[0016]

[0017] Calculate. Among them, ε1 is the torque change lower limit comparison threshold, ε2 is the torque change upper limit comparison threshold, and 0<ε1<ε2; g max is the high value of feedback gain, g min is the low value of feedback gain, and g min <g max <0.

[0018] The method of load torque observer to compensate the output of PI speed controller is to use the observed value of load torque Converted into torque current compensation component Feedforward compensation to the input of the q-axis current PI controller; q-axis torque current setpoint for

[0019]

[0020] The PI speed controller parameters and the load torque observer parameters are optimized and adjusted using the particle swarm algorithm. When the PI speed controller parameters are optimized and adjusted using the particle swarm algorithm, the parameter vector to be optimized is θ = [K p , T i ], the number of parameter vectors to be optimized is 2, that is, the dimension N of the search space is equal to 2. When the PI speed controller parameters are optimized using the particle swarm algorithm, the objective function Q1 for calculating the fitness value is

[0021]

[0022] Where, e(t) is the instantaneous value of the rotor angular velocity error, t m is the transition time of the motor rotor angular velocity step response, t=0 is the starting time of the motor step response; Q 11 γ m1 (1-sgn(e(t)+ω δ )) is the angular velocity overshoot penalty function, γ m1 is a sufficiently large positive number, ω δ is the rotor angular velocity overshoot limit; Q 12 is the steady-state error penalty function, ω Δ is the steady-state error limit of the rotor angular velocity; γ m2 ≥2.

[0023] When the load torque observer parameters are optimized using the particle swarm algorithm, the parameter vector to be optimized is θ1 = [G max , G min , ε1, ε2, α], the number of parameter vectors to be optimized is 2, that is, the dimension N of the search space is equal to 5. max With G max The relationship is g min With G min The relationship is k g The relationship with α is

[0024]

[0025] Among them, α≥1; further, the value of α is selected in the range of 1 to 5. When the load torque observer parameters are optimized and adjusted using the particle swarm algorithm, the objective function Q2 for calculating the fitness value is:

[0026]

[0027] in, is the load torque observation error, e2(t) is the instantaneous value of the load torque observation error, t p is the tracking adjustment time of the motor load torque observed step response, t = 0 is the load mutation moment of the load torque observed step response; Q 21 γ p1 (1-sgn(e2(t)+T δ )) is the torque observation tracking overshoot penalty function, T δ is the torque observation tracking overshoot limit, γ p1 is a sufficiently large positive number; max(|e2(t)|) is the absolute value of the maximum torque observation steady-state error, γ p2 is the fitness balance emphasis coefficient, γ p2 >0;Q 22 γ p1 (1-sgn(e2(t)+T Δ )) is the torque observation steady-state error penalty function, T Δ is the load torque observed steady-state error limit; γ p3 ≥2.

[0028] The particle swarm algorithm is:

[0029] Step 101, initialize the particle swarm; the initial position of the particle is Where M is the number of particles;

[0030] Step 102, taking the initial position of each particle as the initial optimal value of each particle, calculating the fitness value of each particle and saving it as the optimal particle fitness value of each particle; comparing the fitness values ​​of each particle to obtain the initial particle swarm optimal solution and the particle swarm optimal fitness value and saving them;

[0031] Step 103, according to formula

[0032]

[0033] m n+1 =m n +v n+1

[0034] Update the speed and position of each particle; n is the current number of iterations, v n and m nis the velocity vector and position of the particle; c0 is the inertia weight, ranging from 0 to 1.4; c1 and c2 are learning factors, ranging from 1 to 2; is a random number ranging from 0 to 1; is the optimal solution found by the particle itself so far, Represents the optimal solution of the particle swarm of the entire population so far;

[0035] Step 104, calculating the fitness value of each particle;

[0036] Step 105, And the corresponding optimal particle fitness value is updated. And the corresponding particle swarm optimal fitness value is updated;

[0037] Step 106, judging whether the loop termination condition is satisfied, if yes, then ending the particle swarm algorithm, and the final particle swarm optimal solution is the optimal parameter of the optimized parameter vector; otherwise, returning to step 103.

[0038] The permanent magnet synchronous motor control method based on load torque compensation detects the rotor position θ of the permanent magnet synchronous motor and the three-phase current i a 、i b and i c ; Based on the three-phase current i a 、i b and i c Clark transformation is performed on the permanent magnet synchronous motor to obtain the current i in the α-β axis coordinate system: α 、Current i β , according to the current i α 、Current i β Perform Park transformation on the rotor position θ to obtain the current i in the dq axis coordinate system d 、Current i q .

[0039] The permanent magnet synchronous motor control method based on load torque compensation is implemented by a permanent magnet synchronous motor drive control system including a PI speed controller, a load torque observer, a q-axis current controller, a d-axis current controller, a Clarke transformation module, a position and speed detection module, a Park transformation module, a Park inverse transformation module, an SVPWM module and a three-phase inverter.

[0040] The periodic control process of the permanent magnet synchronous motor drive control system specifically includes the following steps:

[0041] Step 1: Detect the rotor position θ, rotor angular velocity ω and three-phase current i of the permanent magnet synchronous motor a 、i b and ic ;

[0042] Step 2: Based on the three-phase current i a 、i b and i c Clark transformation is performed on the permanent magnet synchronous motor to obtain the current i in the α-β axis coordinate system: α 、Current i β , according to the current i α 、Current i β Perform Park transformation on the rotor position θ to obtain the current i in the dq axis coordinate system d 、Current i q ;

[0043] Step 3: Feedback gain g is based on the load torque given value T L * and load torque observation to adjust for changes in

[0044] Step 4: Load torque observer is based on rotor angular velocity ω and current i q Observe the load torque and obtain the load torque observation value and torque current compensation component

[0045] Step 5: PI speed controller is based on the input rotor given angular velocity ω * And the rotor angular velocity ω is used for control calculation to obtain the load torque given value and torque current given component i′ q ;

[0046] Step 6: Given component i' based on torque current q and torque current compensation component i″ q Calculate the q-axis torque current given value i q * ; The d-axis current controller is based on the d-axis torque current given value i d * and the current i in the d-axis coordinate system d The difference between them is used for PI control operation to obtain the control voltage U in the d-axis coordinate system d ; The q-axis current controller is based on the q-axis torque current given value i q * and the current i in the q-axis coordinate system q The difference between them is used for PI control operation to obtain the control voltage U in the q-axis coordinate system q ; Based on the control voltage U in the dq axis coordinate system d , U qPerform Park inverse transformation to obtain the control voltage U in the α-β axis coordinate system α , U β ; d-axis torque current given value i d * is equal to 0;

[0047] Step 7: Set the control voltage U in the α-β axis coordinate system α , U β As the input of the SVPWM module, the SVPWM module controls the three-phase inverter to generate a three-phase AC power supply U a , U b , U c , thereby driving the permanent magnet synchronous motor to operate.

[0048] In the above steps, the order of step 3 and steps 4 and 5 can be interchanged, that is, steps 4 and 5 can be performed first, and then step 3.

[0049] The beneficial effect of the present invention is that the load torque observation value is feed-forward compensated to the given value of the current regulator, and the relevant influence caused by the load disturbance or the change of the system parameters can be offset without a large adjustment of the given current part of the PI speed controller output, and the jitter of the system is effectively weakened. The load torque observation adopts an algorithm that automatically adjusts the feedback gain according to the change of the load torque given value and the change of the load torque observation value, avoiding the problem that the load torque observer selects a fixed small feedback gain to cause large torque observation fluctuations and selects a fixed large feedback gain to cause a long convergence time. When the control parameters, model parameters, etc. of the system change or the load is disturbed, resulting in a change in the load torque given value or / and a change in the load torque observation value, the observation error of the load torque can be quickly reduced, and the observation effect and the rapidity and accuracy of the motor speed control are improved. The feedback gain g is automatically adjusted when the load torque set value changes. When the load torque observation value has not changed significantly, but the load torque set value changes due to changes in the rotor angular velocity set value and / or the actual value of the rotor angular velocity, or when the load torque set value changes due to changes in system model parameters, which will cause the load torque observation value to fluctuate significantly, the feedback gain g can be adjusted in advance. When the load torque observation value actually produces an observation error, the response speed of the observer is accelerated, the observation error of the load torque observation value is quickly reduced, and the speed and accuracy of the motor speed control are further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a block diagram of an embodiment of a permanent magnet synchronous motor drive control system for an elevator;

[0051] Figure 2 A flow chart of an embodiment of a method for automatically adjusting feedback gain;

[0052] Figure 3 Effect diagram of adjusting the relative decay rate of siege step length for siege step length decay factor ξ;

[0053] Figure 4 is the given rotor angular velocity signal and load torque signal. DETAILED DESCRIPTION

[0054] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0055] Figure 1 The present invention is a block diagram of an embodiment of a permanent magnet synchronous motor drive control system for an elevator. Figure 1 In the Clarke transformation module, the three-phase current i of the permanent magnet synchronous motor (PMSM) is input a 、i b and i c , output the current i in the two-phase stationary α-β axis coordinate system α 、i β ; The position sensor in the position and speed detection module detects the rotor position θ of the permanent magnet synchronous motor and converts it into the rotor angular velocity ω output; the Park conversion module inputs the current i α 、i β and the rotor position θ, the output current i in the rotating dq axis coordinate system d 、i q ;PI speed controller input rotor given angular velocity ω * and rotor angular velocity ω, output load torque given value T L * and torque current given component i′ q ; Load torque observer inputs load torque given value T L * , rotor angular velocity ω and current i q , output torque current compensation component i″ q ; Torque current given component i′ q and torque current compensation component i″ q After adding, it is used as the q-axis torque current given value i * q ; q-axis current PI controller input q-axis torque current given value i * q and current i d , output control voltage U in q-axis coordinate system q ; d-axis current PI controller input q-axis torque current given value i * d and current i d , output control voltage U in d-axis coordinate system d , d-axis torque current given value i* d =0; the control voltage U in the dq axis coordinate system is input to the Park inverse transformation module d , U q , output the control voltage U in the α-β axis coordinate system α , U β SVPWM module (i.e. space vector pulse width modulation module) input control voltage U α , U β , output pulse signal to the three-phase inverter, the three-phase inverter converts the DC voltage U dc Convert to three-phase AC power supply U a , U b , U c , thereby driving the permanent magnet synchronous motor to operate.

[0056] Ignoring the influence of core eddy current and hysteresis loss, the i d =0, the PMSM rotor magnetic field oriented control is established, and the mathematical model of PMSM in the dq axis rotating coordinate system is established. The voltage equation is:

[0057]

[0058] For the use of i d =0 control mode of salient pole PMSM vector control system, the electromagnetic torque equation is:

[0059] T e =1.5pψ f i q (2)

[0060] The PMSM motion equation is:

[0061]

[0062] In formula (1), (2), and (3), u d 、u q are the voltages of the dq axes respectively; i d 、i q are the currents of d and q axes respectively; L d , L q are the inductance of d and q axes respectively; T e is the electromagnetic torque; T L is the load torque; R is the stator resistance; p is the number of motor poles; ω e is the rotor electrical angular velocity, i.e. angular frequency; ω is the rotor angular velocity, i.e. the motor rotor mechanical angular velocity; ψ f is the permanent magnet flux; J is the moment of inertia; B is the friction coefficient; t is time.

[0063] From the mathematical model of permanent magnet synchronous motor (1), (2), (3) in dq coordinates, it can be seen that i d =0 control strategy, the motor torque output can be controlled by formula (2), and the motor operation can be controlled by combining the speed and current double closed loop. Let the motor rotor angular velocity error e = ω * -ω, ω* is the given rotor angular velocity of the motor. The PI speed controller regulates and controls the rotor angular velocity error e, and outputs the given torque current component i′ q , direct torque control of the motor is performed. Formula (4) is the transfer function of the PI speed controller.

[0064]

[0065] In formula (4), K p is the proportional coefficient of the PI speed controller, T i is the integral time constant of the PI speed controller. The load torque given value T output by the PI speed controller L * and torque current given component i′ q for:

[0066]

[0067] The load torque given value T L * is sent to the load torque observer, and e(t) is the instantaneous value of the rotor angular velocity error.

[0068] The parameters of the speed loop controller can be manually adjusted according to the conventional method of PID parameter adjustment; based on experience, the proportional coefficient K p The adjustment range is greater than 0 and less than 10I N , I N is the rated current of the permanent magnet synchronous motor; the integral time constant T i The adjustment range is 0.001~0.5s.

[0069] According to the PMSM electromagnetic torque and motion equation, for a constant step load, it can be considered as a constant value within the change cycle, that is, The motor rotor angular velocity and load torque are used as state variables to form the PMSM state equation:

[0070]

[0071] Based on equation (6), with load torque and motor rotor angular velocity as observation objects, load torque observer embodiment 1 is established as:

[0072]

[0073] In formula (7), is the observed value of the load torque, is the estimated value of the rotor angular velocity, g is the feedback gain of the load torque observer, k g is the sliding mode gain of the load torque observer embodiment 1, and the load torque observer embodiment 1 is a sliding mode observer. Compared with the load torque, the motor friction accounts for a small proportion, so let B = 0, ignoring the influence of friction, then the load torque observer embodiment 1 of formula (7) becomes:

[0074]

[0075] According to (6) and (8) when B=0, the error equation of the load torque observer embodiment 1 is obtained as follows:

[0076]

[0077] In formula (9), is the rotor angular velocity estimation error, is the load torque observation error, and the observer sliding surface is defined as According to the sliding mode reachability condition, the stability condition of the observer system in equation (8) is k g ≤-|e2 / J|, and g<0.

[0078] Based on equation (6), with the load torque and the motor rotor angular velocity as the observation objects, a load torque observer embodiment 2 can also be established as follows:

[0079]

[0080] Compared with the load torque, the motor friction accounts for a small proportion. Let B = 0 and ignore the influence of friction. Then the load torque observer embodiment 2 of formula (10) becomes:

[0081]

[0082] In formula (10) and (11), is the observed value of the load torque, is the estimated value of the rotor angular velocity, g is the feedback gain of the load torque observer, k W is the proportional gain of the load torque observer embodiment 2, and the load torque observer embodiment 2 is a state observer. According to equation (6) and equation (11) when B=0, the error equation of the load torque observer embodiment 2 is obtained as follows:

[0083]

[0084] In formula (12), is the rotor angular velocity estimation error, is the load torque observation error. The state observer of equation (11) is an autonomous linear system. W <0, and g<0, the observer is asymptotically stable. Formula (7) of load torque observer embodiment 1 and formula (10) of load torque observer embodiment 2 both take into account the friction factor of the motor. The addition of small friction damping will have an adverse effect on the rapidity of system response, but can increase the stability on the basis of formula (8) and formula (11), respectively.

[0085] When the observer embodiment 1 of equations (7) and (8) is selected, the sliding mode gain k g The setting method is to follow

[0086]

[0087] In formula (13), α≥1; generally, the value of α is selected in the range of 1 to 5, for example, α is selected to be equal to 1.5. Load torque observer embodiment 1 In the process of observing the load torque, k g The absolute value of |e2| is too small. When |e2| is large, the observer cannot enter the sliding mode state. g The absolute value of is large enough to ensure that the observer enters the sliding mode state, but the steady-state observation fluctuation of the load torque becomes larger; k g The value of changes with the load torque observation error, which can simultaneously increase the stability of the observer and reduce the steady-state observation fluctuation of the load torque.

[0088] When the observer embodiment 2 of equations (10) and (11) is selected, the proportional gain k W The setting method is to follow

[0089]

[0090] In formula (14), T N is the rated torque of the motor, β>0; generally, the β value is selected in the range of 1 to 20, for example, β=10. When β is selected to increase, the steady-state fluctuation of the load torque observation becomes larger, but the torque observation tracking overshoot becomes smaller; when β is selected to decrease, the steady-state fluctuation of the load torque observation becomes smaller, but the torque observation tracking overshoot becomes larger.

[0091] In the observers represented by equations (7), (8) or (10), (11), the value of the feedback gain g has a great influence on the load torque observation result. The larger the feedback gain g, the smaller the observed torque fluctuation, but the slower the observed torque identification speed; the smaller the feedback gain g, the faster the observed torque speed, but the greater the observed torque fluctuation. In consideration of this problem, in the traditional load torque observer, the observation speed and fluctuation of the load torque are comprehensively considered, and the feedback gain g is taken as a compromise value, but this will abandon the advantages of small fluctuation when the feedback gain is large and fast observation speed when the feedback gain is small.

[0092] When the motor uses a PI controller for speed control, the influence of parameter changes and external load disturbances on the system is mainly suppressed by increasing the proportional coefficient in the controller, but too large a proportional coefficient will cause the system stability to decrease. In order to solve the contradiction between the rapid response and anti-interference of the PI speed controller and the stability, the observer is used to observe the load disturbance changes in real time, and the load torque observation value is feedforward compensated to the current regulator. Under the premise of reducing the proportional coefficient of the PI controller, the anti-interference performance of the system can be not reduced. In order to make full use of the advantages of the feedback gain g at high and low values, according to the load torque observation value and the load torque given value change at two adjacent moments, when the load torque given value changes little and the load torque observation value changes little, a larger feedback gain g is given to make the observation result less volatile and more stable; when the load torque given value changes greatly or the load torque observation value changes greatly, a smaller feedback gain g is given to speed up the observation speed. Finally, by adjusting the feedback gain g, a comprehensive result of fast observation speed, small fluctuation and stronger stability is obtained.

[0093] Figure 2 The flow chart of the feedback gain automatic adjustment method embodiment is used in the load torque observer embodiment 1 or the load torque observer embodiment 2. Figure 1 In the elevator permanent magnet synchronous motor drive control system embodiment, the feedback gain is automatically adjusted. Figure 2 In the figure, ε1 is the torque change lower limit comparison threshold, ε2 is the torque change upper limit comparison threshold; the sum of the change in the load torque given value and the change in the load torque observed value for the last two times Where, ΔT L * is the difference between the two most recent load torque given values, is the difference between the two most recent load torque observations. In a periodic control process of a permanent magnet synchronous motor drive control system, Figure 2 The feedback gain g shown in (b) is adjusted later than the load torque observation and the output calculation of the PI speed controller, and is:

[0094] Step S1: The load torque observer performs T on the load torque.L Observe and obtain the load torque observation value The PI speed controller performs control calculation to obtain the load torque given value At this time for for Until the next feedback gain g adjustment, the becomes becomes

[0095] Step S2: Calculation calculate

[0096] Step S3, determine whether ΔT is greater than ε2; ​​when ΔT is greater than ε2, set the feedback gain g equal to g min And exit; when ΔT is less than or equal to ε2, enter step S4;

[0097] Step S4, determine whether ΔT is less than ε1; when ΔT is less than ε1, set the feedback gain g equal to g max And exit; when ΔT is greater than or equal to ε1, enter step S5;

[0098] Step S5: Feedback gain g is calculated according to

[0099]

[0100] Perform calculations.

[0101] Among them, ε1 is the torque change lower limit comparison threshold, ε2 is the torque change upper limit comparison threshold, and 0<ε1<ε2; g max is the high value of feedback gain, g min is the low value of feedback gain, and g min <g max <0.

[0102] In the process of periodic control of motor speed, Figure 2 The feedback gain g shown in (a) is adjusted prior to the load torque observation and the output calculation of the PI speed controller. The feedback gain g adjustment method changes the above step S1 to step S5, steps S2-S5 to steps S1-S4, and the exit in each step is changed to entering step S5, and

[0103] When ΔT is greater than ε2, it indicates that the load torque observation value fluctuates greatly, or the load torque given value changes greatly due to changes in system model parameters, rotor angular velocity given value, and rotor angular velocity actual value, which will cause the load torque observation value to fluctuate greatly. The feedback gain g is selected to be equal to gmin Rapid identification and observation of load torque; when ΔT is less than ε1, it indicates that the load torque set value changes little and the state load torque observation value fluctuates little, and the feedback gain g is selected to be equal to g max The load torque identification and observation with stability as the main focus are performed; when ΔT is greater than or equal to ε1 and less than or equal to ε2, the feedback gain g is calculated according to formula (15), so that the feedback gain g decreases with the increase of ΔT in this interval, avoiding the adverse effect on the working stability of the torque observer caused by the drastic change of the feedback gain g due to the small change of ΔT. Figure 2 In the given torque change comparison threshold 0<ε1<ε2, the specific values ​​of ε1 and ε2 are related to the sampling control period (cycle time) of the PI speed controller, the permanent magnet synchronous motor and its load conditions. ε2 is generally within the range of less than 5% of the rated torque. For example, if the rated torque is 22N·m, ε1=0.1N·m and ε2=0.6N·m. The value of the feedback gain g satisfies g min <g max <0, generally, g min ≥-5000. g min The value should be selected so that when the load torque changes suddenly, the torque observation tracking overshoot of the load torque observer output observation value is within the torque observation tracking overshoot limit; g max The value should be selected when the load torque is constant, the load torque observer and the PI speed controller are both in steady state, and the sum of the change in the load torque given value and the change in the load torque observation value for the last two times ΔT is less than ε1; for example, select the feedback gain g max =-0.5, g min =-10. Select g min , g max The specific method for determining the values ​​of ε1, ε2 is as follows: first, when the load torque remains unchanged and the load torque observer and the PI speed controller are both in a steady state, let the feedback gain g start from a larger value. For example, let the feedback gain g gradually decrease from -0.01. The steady-state error of the load torque observation will gradually increase. When the steady-state error of the load torque observation reaches the steady-state error limit of the load torque observation, the feedback gain g value at this time is determined to be g max ; Keep the load torque unchanged and set the feedback gain g equal to g max , and continuously measure the F1-time ΔT value, and take the average value of the maximum F2 ΔT measurement values ​​in the F1-time ΔT measurement as the torque change lower limit comparison threshold ε1; then, when the load torque observer and the PI speed controller are both in a steady state, the load torque is suddenly changed, and under the premise of ensuring that the torque observation tracking overshoot of the load torque observer output observation value is within the torque observation tracking overshoot limit, the tracking adjustment time of the load torque observer output observation value is as short as possible to adjust and determine g minvalue; then, keep the load torque unchanged and set the feedback gain g equal to g min At the same time, the ΔT value is measured continuously for F1 times, and the average value of the maximum F2 ΔT measurement values ​​in the F1 ΔT measurement at this time is used as the torque change upper limit comparison threshold ε.

[0104] When designing the PI speed controller and load torque observer of the elevator permanent magnet synchronous motor drive control system embodiment, the PI speed controller parameters and load torque observer can also be adjusted using optimization algorithms such as particle swarm algorithm, wolf pack algorithm, genetic algorithm, etc. When the particle swarm algorithm is used to adjust the PI speed controller parameters, the objective function for comprehensively evaluating the various performance indicators of the PI speed controller is established as follows:

[0105]

[0106] In formula (16), Q 11 t in m is the transition time of the motor rotor angular velocity step response, t=0 is the starting time of the motor step response; Q 11 γ m1 (1-sgn(e(t)+ω δ )) is the angular velocity overshoot penalty function, where γ m1 Take a sufficiently large (t m 5 times or more of the reasonable value), ω δ is the rotor angular velocity overshoot limit (i.e. the maximum rotor angular velocity overshoot allowed by the system); when the overshoot of the motor rotor angular velocity step response does not exceed the rotor angular velocity overshoot limit ω δ When , the overshoot penalty function term is equal to 0, otherwise it is equal to γ m1 ;Q 12 is the steady-state error penalty function, ω Δ is the rotor angular velocity steady-state error limit; when the steady-state error of the motor rotor angular velocity step response does not exceed the rotor angular velocity steady-state error limit ω Δ When , the steady-state error penalty function term is equal to 0, otherwise it is equal to γ m1 ; Q1 is the objective function value, that is, the fitness value of the particle swarm algorithm for optimizing the parameters of the PI speed controller; the smaller the fitness value of the particle, the better the corresponding position. γ m1 When taking the value, first estimate t m For example, if the rated speed of the motor is 1500r / min (corresponding to the rated rotor angular velocity of 157rad / s), the rated load starting time is about 0.2s, t m The reasonable value of is no more than 0.5s, then γ m1 Relative to 0.5, it is large enough, 5 times or more. For example, take γ m1 =2.5.m2 Generally, the value is greater than or equal to 2. Its value determines the time interval for measuring the steady-state error of the rotor angular velocity and judging whether the steady-state error of the rotor angular velocity exceeds the steady-state error limit of the rotor angular velocity ω. Δ ; For example, γ m2 When the value is equal to 6, then at 5 times the transition time t m The steady-state error of the rotor angular velocity is measured in the interval. If other index factors need to be considered in the PI speed controller parameter optimization, for example, whether the overshoot is small enough, whether the steady-state error is small enough, etc., other objective functions other than equation (16) can be established.

[0107] When the particle swarm algorithm is used to adjust the parameters in the load torque observer embodiment 1 or the load torque observer embodiment 2, the objective function for comprehensively evaluating the performance indicators of the load torque observer is established as follows:

[0108]

[0109] In formula (17), is the load torque observation error, e2(t) is the instantaneous value of the load torque observation error; Q 21 t in p is the tracking adjustment time of the motor load torque observed step response, t = 0 is the load mutation moment of the load torque observed step response; Q 21 γ p1 (1-sgn(e2(t)+T δ )) is the torque observation tracking overshoot penalty function, where γ p1 Take a sufficiently large (t p +γ p2 max(5 times or more of the reasonable value of |e2(t)|), T δ is the torque observation tracking overshoot limit. When the torque observation tracking overshoot does not exceed the torque observation tracking overshoot limit T δ When the torque observation tracking overshoot penalty function term is equal to 0, otherwise it is equal to γ p1 .Q 22 In the first term, max(|e2(t)|) is the absolute value of the maximum torque observation steady-state error, γ p2 is the fitness balance emphasis coefficient, which is a constant greater than 0; Q 22 γ p1 (1-sgn(e2(t)+T Δ )) is the torque observation steady-state error penalty function, T Δ is the load torque observation steady-state error limit; when the torque observation steady-state error does not exceed the load torque observation steady-state error limit T Δ When the torque observation steady-state error penalty function term is equal to 0, otherwise it is equal to γp1 Q2 is the objective function value, that is, the fitness value of the load torque observer parameters using the particle swarm algorithm; the smaller the fitness value of the particle, the better the corresponding position. p1 When taking the value, first estimate t p +γ p2 ·The reasonable value (upper limit) of max(|e2(t)|); for example, assuming the rated torque of the motor is 22N·m, the maximum torque observation tracking adjustment time t is expected p At about 0.1s, the load torque observed steady-state error limit T Δ is 1N·m, γ p2 When it is 0.1, the torque observation tracking adjustment time term and the maximum torque observation steady-state error absolute value term are relatively balanced, or the objective function values ​​Q2 of the two play a similar role; at this time, t p +γ p2 The reasonable value of max(|e2(t)|) is no more than 0.2, γ p1 The value can be equal to 1, or a constant greater than 1. p2 value, the weight of the torque observation tracking adjustment time term in the objective function value Q2 becomes larger, and it is more inclined to the rapidity of torque observation; increasing γ p2 value, the weight of the absolute value of the maximum torque observation steady-state error in the objective function value Q2 becomes larger, and it is more inclined to the steady-state performance of torque observation. p3 Generally, the value is greater than or equal to 2, and its value determines the time interval for measuring the steady-state error of load torque observation. For example, γ p3 When the value is equal to 6, the tracking adjustment time (i.e. transition process time) t p The load torque observation steady-state error is measured in the interval.

[0110] The specific steps of the particle swarm algorithm are:

[0111] Step 101, initialize the particle swarm. Assume that the initial position of the particle in the particle swarm is Where M is the number of particles in the particle swarm, which is generally selected between 20 and 150, and the initial position is required to obey random distribution.

[0112] (1) When optimizing the parameters of the PI speed controller, the parameter vector to be optimized is θ = [K p , T i ], at this time, the search space dimension N of the particle swarm algorithm is equal to 2. In the final optimization result, the particle position value with the smallest fitness value is the optimal parameter of the PI speed controller. The position of the i-th particle is expressed as Corresponding to the parameter vector θ to be optimized; the position value interval is [m imin m imax], the range interval can be given based on existing knowledge or experience, for example, parameter K p The value interval [m 1min m 1max ] is [0 10I N ], parameter T i The value interval [m 2min m 2max ] is [0.001 0.5].

[0113] (2) When optimizing the parameters of the load torque observer embodiment 1, the parameter vector to be optimized is θ1=[G max , G min , ε1, ε2, α], at this time, the search space dimension N of the particle swarm algorithm is equal to 5. After optimizing the optimal position of the particle, g max , g min according to Sliding mode gain k g According to the parameter α, they are calculated according to formula (13). The initial position of the i-th particle is expressed as Corresponding to the parameter vector θ1 to be optimized; the position value interval is [m imin m imax ], the range interval can be given based on existing knowledge or experience, for example, parameter G max The value interval [m 1min m 1max ] is [-4 4]; parameter G min The value interval [m 2min m 2max ] is [-4 4]; the value range of parameter ε1 is [m 3min m 3max ] and the value range of parameter ε2 [m 4min m 4max ] are [0 0.05T N ], T N is the rated torque of the motor; the value range of parameter α is [m 5min m 5max ] is [1 5].

[0114] (3) When optimizing the parameters of the load torque observer embodiment 2, the parameter vector to be optimized is θ2 = [G max , G min , ε1, ε2, β], at this time, the search space dimension N of the particle swarm algorithm is equal to 5. After optimizing the optimal position of the particle, g max , g min according to Proportional gain k W According to the parameter β, they are calculated according to formula (14). The initial position of the i-th particle is expressed as Corresponding to the parameter vector θ2 to be optimized; the position value interval is [m imin m imax ], the range interval can be given based on existing knowledge or experience. For example, the value intervals of the first four parameters in θ2 are the same as the value intervals of the first four parameters in θ1; the value interval of parameter β is [m 5min m 5max ] is [1 20].

[0115] In each optimization object of step 101, since the initial position of the particle is required to obey the random distribution and the spatial search algorithm of the particle swarm is a linear method, the feedback gain high value g is directly max and feedback gain low value g min When searching and optimizing in the interval [-50000], the absolute value of the feedback gain with a low value has a great influence on the working state of the observer. For example, in the interval [-10 0], the probability of particles entering is small, and it is difficult to optimize the feedback gain with a high value g. max and feedback gain low value g min The optimal position of the feedback gain g is not directly adjusted during the optimization process. max and feedback gain low value g min Instead of searching and optimizing, it is optimized in a similar way to gain. The parameter interval [-10 0] is expanded in the search interval, making it easier to obtain a high value of feedback gain g. max and feedback gain low value g min The optimal position; at this time, the feedback gain is high value g max and feedback gain low value g min The parameter range is [-10000 -0.0001]; the feedback gain high value g max Under normal circumstances, the interval [-0.0001 0] will not be selected, otherwise it will cause the observer to respond too slowly; in addition, the upper limit of the interval is -0.0001, which also avoids high feedback gain values ​​g max The observer cannot work properly if the value is 0.

[0116] Step 102: Set the initial position m of each particle (0) As the initial optimal value m of each particle b (0) , calculate the fitness function value of each particle (i.e., particle fitness value) and save it as the particle optimal fitness value of each particle; compare the fitness values ​​of each particle to obtain the optimal solution m of the initial particle swarm g (0) And the optimal fitness value of the particle swarm is saved. Assume that the initial velocity of the particle is Also obeying the random distribution, the initial velocity of the i-th particle is expressed as The speed change extreme value of the parameter [v imin v imax ] is generally set to 5% to 20% of the parameter value range; for example, when the optimized parameter vector is θ, the parameter T i The value interval [m 2min m 2max ] is [0.001 0.5], the interval range is 0.499, then the second dimension variable of each particle (parameter T i )'s velocity change extreme value [v 2min v 2max ] The value of 5% is [-0.02495 0.02495], and the value of 20% is [-0.0998 0.0998]; when the optimized parameter vector is θ1 or θ2, the parameter G max The value interval [m 1min m 1max ] is [-4 4], the interval range is 8, then the first dimension variable of each particle (parameter G max )'s velocity change extreme value [v 1min v 1max ]According to 5%, the value is [-0.4 0.4], and according to 20%, the value is [-1.6 1.6].

[0117] Step 103: Location update.

[0118]

[0119] m n+1 =m n +v n+1

[0120] Update the speed and position of each particle; the speed change of each dimensional variable cannot exceed the corresponding speed change extreme value of each dimensional variable, and the updated position of each dimensional variable cannot exceed the corresponding value interval of each dimensional variable. In formula (18), n is the current number of iterations, v n and m n is the velocity vector and position of the particle; c0 is the inertia weight, ranging from 0 to 1.4. Adjusting its value can change the search range and search speed. Furthermore, adaptively reducing the c0 value as the number of iterations increases is conducive to achieving a balance between search capability and convergence speed; c1 and c2 are learning factors, ranging from 1 to 2, and it is recommended that both be equal to 2; is a random number ranging from 0 to 1; is the optimal solution (optimal position) found by the particle itself so far, It represents the optimal solution (optimal position) of the particle swarm of the entire population so far.

[0121] Step 104, fitness value update: calculate the fitness value of each particle.

[0122] Step 105: Update the optimal fitness value of the particle and the optimal fitness value of the particle swarm. And the corresponding particle optimal fitness value is updated. And the corresponding particle swarm optimal fitness value is updated.

[0123] Step 106, determine whether the loop termination condition is met, if yes, then end the particle swarm algorithm, the final particle swarm optimal solution is the optimal parameter of the optimized parameter vector, that is, the PI speed controller, or the optimal parameter of the load torque observer; otherwise, return to step 103.

[0124] The loop termination condition is generally to reach the maximum iteration step limit or the optimal particle fitness value is less than a certain threshold. The maximum iteration step limit is used as the loop termination condition. The maximum iteration step is usually selected between 100 and 2000. The particle swarm algorithm is used to adjust the parameter K p 、T i When optimizing and setting the threshold conditions of the particle swarm's optimal fitness value, if the motor's startup time is required to be within 0.2s, the threshold of the optimal particle fitness value can be set to 0.2. When the requirements are raised to ensure a certain margin, the threshold of the optimal particle fitness value can be smaller than 0.2, for example, set to 0.15. When the particle swarm algorithm is used to tune the load torque observer parameters and set the threshold conditions of the particle swarm's optimal fitness value, it is necessary to refer to the expected torque observation tracking adjustment time t p , load torque observation steady-state error limit T Δ , fitness balance emphasis coefficient γ p2 etc. to determine the threshold value; let T Δ is 1N·m, γ p2 = 1.5, the desired torque observation tracking adjustment time t p Less than 0.04s, γ p2 If it is equal to 0.05, the threshold condition of the optimal fitness value of the particle swarm can be set at 0.06.

[0125] The particle swarm algorithm is used to optimize the parameter K p 、T i When the particle fitness value of each particle is calculated according to formula (16), each particle is required to be used as the corresponding controller parameter in turn to control the motor start (or start in the simulation system) to obtain the motor rotor angular velocity step response e(t) required in formula (16). The transition process time t is determined based on e(t). m , calculate the particle fitness value Q1.

[0126] The particle swarm algorithm is used to tune the load torque observer parameters. The parameters of the PI speed controller in the elevator permanent magnet synchronous motor drive control system embodiment have been tuned, and the load torque compensation control is realized. When the particle fitness value of each particle is calculated according to formula (17), it is required to convert each particle position into the corresponding load torque observer parameter in turn. When the motor given speed remains unchanged and the PI speed controller is in a steady state, the load torque is suddenly changed to control the motor operation (or run in the motor simulation system) to obtain the motor load torque observation step response e2(t) required in formula (17). The transition process time t is determined based on e2(t). p The particle fitness value Q2 is calculated by observing the steady-state error of the load torque.

[0127] Among the load torque observer parameters to be optimized, g max , g min There is a constraint g between min <g max , corresponding to the constraint G min >G max When initializing or updating the particle position, first, particle m i1 According to the value interval [m 1min m 1max ] to initialize or update the position, then the particle m i2 According to the value interval [m i1 m 2max ] to initialize or update the position so that each particle m in each iteration i All satisfy the constraint G min >G max , that is, satisfying the constraint g min <g max There is a constraint between ε1 and ε2. ε1<ε2. When initializing or updating the particle position, first, particle m i The parameter m in i3 (i.e. ε1) according to the value interval [m 3min m 3max ] Randomly generate positions or update positions, then particle m i The parameter m i4 (i.e. ε2) according to the value interval [m i3 m 4max ] Randomly generate positions or update positions so that particle m i The constraint ε1<ε2 is satisfied.

[0128] Observe the load torque observation value Then, the observed value of the load torque Converted into torque current compensation component i″ q Feedforward compensation to the input of the q-axis current PI controller, the torque current given component i' output by the PI speed controller q The q-axis torque current given value i of the q-axis current PI controller is * q for:

[0129]

[0130] In formula (19), k q =1 / (1.5pψ f ) is the torque observation compensation coefficient. Comparing equation (5) with equation (19), it can be obtained that when the load is disturbed or the system parameters change, the load torque compensation is not added in equation (5), and a larger K needs to be selected. p The value is used to provide a sufficiently large given current change to offset the impact of load disturbance or system parameter changes, so as to ensure that the motor speed can be quickly constant; while formula (19) feeds forward the load torque observation value to the current regulator, without the need for a large K p When the load is disturbed or the system parameters are changed, a sufficiently large given current change can be provided to offset the related impact of the load disturbance or the change of system parameters, thereby reducing the output pressure of the PI speed controller.

[0131] When the feedback gain value is fixed, the smaller the feedback gain g, the larger the oscillation amplitude of the load torque observation and the stronger the volatility; the larger the feedback gain g, the smaller the oscillation amplitude of the load torque observation and the higher the observation accuracy. The automatic gain adjustment algorithm solves the problem of large torque observation fluctuation caused by small feedback gain in the load torque observer and the problem of long convergence time of large feedback gain. The convergence time and fluctuation amplitude indicators are better than the compromise gain algorithm. It can quickly track the change value of the load torque and quickly reduce the observation error caused by the given change or parameter change. It has a small oscillation amplitude and high observation accuracy, achieving a good observation effect.

[0132] When the given speed is changed under the rated load torque, although the actual load torque does not change, it can be seen from the load torque observer constructed by equations (7), (8) or (10), (11) that when the rotor angular velocity ω changes, even if the load torque does not change, the observed torque observation value will change, resulting in observation error. When the given speed is changed under the rated load torque, the control and adjustment process of the permanent magnet synchronous motor control system is that the PI speed controller first changes the output load torque given value T according to the given speed. L * Change the torque current given value i *q changes, thus making the electromagnetic torque T of the permanent magnet synchronous motor e The feedback gain g is only based on the change of the load torque observation value. Automatic adjustment is performed. At this time, only when the rotor angular velocity ω changes, the load torque observation value After the change occurs, the feedback gain g is adjusted; the feedback gain g is adjusted according to the change ΔT of the load torque given value. L * and the change in the observed value of load torque The absolute value of the load torque is automatically adjusted when the given speed changes. L * Change, load torque observation value If the load torque has not changed, adjust the feedback gain g in advance. When an observation error actually occurs, the observer response speed can be accelerated to eliminate (reduce) the load torque observation value as quickly as possible. The observation error is reduced, thereby improving the rapidity and accuracy of motor speed control. Similarly, when the system model parameters change, the load torque given value T L * Prior to load torque observation When the load torque is changed, the feedback gain g is also based on the change in the load torque given value ΔT L * and the change in the observed value of load torque Automatic adjustment can be performed to adjust the feedback gain g in advance, speed up the observer response speed, and eliminate (reduce) the load torque observation value as soon as possible The observation error can be reduced and the speed control of the motor can be further improved. Of course, if the load is disturbed and the observation value is When changes occur, When a major change occurs, Figure 2 It can be seen that the feedback gain g can also be automatically adjusted to eliminate (reduce) the load torque observation value as quickly as possible. The observation error makes the load torque observation value Keep up with the load torque T as soon as possible L changes.

[0133] Furthermore, in the embodiment of the elevator permanent magnet synchronous motor drive control system, after the parameters of the PI speed controller and the load torque observer are adjusted manually or by optimization, the parameters of the PI speed controller can be manually fine-tuned while realizing load torque compensation control, or the parameters of the PI speed controller can be re-optimized using a particle swarm algorithm according to steps 101-106.

[0134] In each of the above, select g min , g max In the specific methods of determining the value and comparison threshold, the parameters in the PI speed controller have been adjusted and are implemented under the condition of load torque compensation control; when the parameter value is determined by the manual method, it is recommended that F1 is an integer greater than or equal to 20, and F2 is an integer greater than or equal to 5 and less than or equal to 0.5F1.

[0135] In the periodic control process of the permanent magnet synchronous motor speed in the drive control system, the load torque given value T calculated at the kth moment (or the kth step) is L * T L * (k), load torque observation value Recorded as The k-1 moment is the moment of the periodic control process before the k moment, and the load torque given value T L * T L * (k-1), load torque observation value Recorded as The k-2 moment is the moment of the periodic control process before the k-1 moment, and the load torque given value T L * T L * (k-2), load torque observation value Recorded as The steps of permanent magnet synchronous motor speed control in the elevator permanent magnet synchronous motor drive control system include:

[0136] Step 1: Detect the rotor position θ, rotor angular velocity ω and three-phase current i of the permanent magnet synchronous motor a 、i b and i c ;

[0137] Step 2: Based on the three-phase current i a 、i b and i c Clark transformation is performed on the permanent magnet synchronous motor to obtain the current i in the α-β axis coordinate system: α 、Current i β , according to the current i α 、Current i β Perform Park transformation on the rotor position θ to obtain the current i in the dq axis coordinate system d 、Current i q ;

[0138] Step 3: Load torque observer is based on rotor angular velocity ω and current i q Observe the load torque and obtain the load torque observation value and torque current compensation component i″ q ;

[0139] Step 4: PI speed controller is based on the input rotor given angular velocity ω * And the rotor angular velocity ω is used for control calculation to obtain the load torque given value and torque current given component i′ q ;

[0140] Step 5: Feedback gain g is based on the load torque given value T L * and load torque observation to adjust for changes in

[0141] Step 6: Given component i' based on torque current q and torque current compensation component i″ q Calculate the q-axis torque current given value i q * ; The d-axis current controller is based on the d-axis torque current given value i d * and the current i in the d-axis coordinate system d The difference between them is used for PI control operation to obtain the control voltage U in the d-axis coordinate system d ; The q-axis current controller is based on the q-axis torque current given value i q * and the current i in the q-axis coordinate system q The difference between them is used for PI control operation to obtain the control voltage U in the q-axis coordinate system q ; Based on the control voltage U in the dq axis coordinate system d , U q Perform Park inverse transformation to obtain the control voltage U in the α-β axis coordinate system α , U β ; d-axis torque current given value i d * is equal to 0;

[0142] Step 7: Set the control voltage U in the α-β axis coordinate system α , U β As the input of the SVPWM module, the SVPWM module controls the three-phase inverter to generate a three-phase AC power supply U a , U b , U c , thereby driving the permanent magnet synchronous motor to operate.

[0143] In the above steps, the order of steps 3, 4 and 5 can be interchanged, that is, step 5 can be performed first, and then steps 3 and 4. Figure 2 (b) In the above steps 3, 4, and 5, first perform load torque observation and speed control, and then automatically adjust the feedback gain. Figure 2 (b) in the figure first automatically adjusts the feedback gain, then observes the load torque and controls the speed. At this time, among the above steps, step five is performed first, and then steps three and four are performed.

[0144] The parameters of the PI speed controller and the load torque observer can also be optimized uniformly using particle swarm optimization, wolf pack algorithm, genetic algorithm and other optimization algorithms. Taking the wolf pack algorithm as an example, at this time, the system motor is given a rotor angular velocity ω * The step signal is Figure 4 As shown in (a) of Figure 4 (b) Change the load torque T L Given the rotor angular velocity ω * The signal is a step signal, and the rotor angular speed is given by Not greater than the rated angular velocity of the motor, and not less than 80% of the rated angular velocity of the motor. Figure 4 In (b), when the motor starts (t = 0), the load torque is the load torque high value T Lmax ; When the motor enters the stable rotor angular velocity state (t>t z ), the load torque changes from the high value T Lmax Mutation reduced to low value T Lmin ; The load torque is maintained at a low value T Lmin Run time After that, from the low value T Lmin Mutation increases to high values ​​of T Lmax ; Among them, the load torque high value T Lmax Not greater than the rated load torque of the motor, the load torque low value T Lmin Not less than 10% of the rated load torque of the motor, the load torque high value T Lmax With low load torque T Lmin The difference between them shall not be less than 50% of the rated load torque of the motor; 2~5t zA step signal is used as the given rotor angular velocity signal of the motor. The load torque mutation is controlled when the motor rotor angular velocity is in steady-state operation. The parameters of the PI speed controller and the load torque observer are optimized simultaneously by constructing the comprehensive performance index of the motor startup phase and the steady-state operation phase. The influence of the load torque observer's performance is unified to the rotor angular velocity performance index, which simplifies the parameter optimization process and improves the rotor angular velocity performance index to the greatest extent.

[0145] Considering that the main function of the load torque observer is to improve the anti-load interference ability of the PI speed controller, the objective function for comprehensively evaluating the performance indicators of the PI speed controller and the load torque observer is established as follows:

[0146]

[0147] In formula (20), Q 31 The integral term in is the IAE criterion for the startup phase of the motor rotor angular velocity step response, t z is the transition time of the motor rotor angular velocity step response, t=0 is the starting time of the motor step response; Q 31 The second term γ z1 (1-sgn(e(t)+ω δ )) is the angular velocity overshoot penalty function, where γ z1 Take a sufficiently large ( 5 times or more of the reasonable value), ω δ is the rotor angular velocity overshoot limit (i.e. the maximum rotor angular velocity overshoot allowed by the system); when the overshoot of the motor rotor angular velocity step response does not exceed the rotor angular velocity overshoot limit ω δ When , the overshoot penalty function term is equal to 0, otherwise it is equal to γ z1 ;Q 32 The integral term in is the IAE criterion of the steady-state working stage of the motor rotor angular velocity, which comprehensively considers the steady-state error and the anti-interference performance of the problem; Q 32 The second term γ z1 (1sgnet+ωΔ is the steady-state error penalty function, ωΔ is the steady-state error limit of the rotor angular velocity; when the steady-state error of the motor rotor angular velocity step response does not exceed the steady-state error limit of the rotor angular velocity ω Δ When , the steady-state error penalty function term is equal to 0, otherwise it is equal to γ z1 ; Q3 is the objective function value, that is, the fitness value of the wolf pack algorithm for parameter optimization of the PI speed controller and the load torque observer; the smaller the fitness function value of the individual wolf, the better the corresponding position. γ z2 The value is generally greater than or equal to 6, and its value determines the time interval for measuring the steady-state error of the rotor angular velocity. For example, γ z2When the value is equal to 10, then at 9 times the transition time t z The steady-state error of the rotor angular velocity is measured in the interval; t is equal to γ z2 The time t2 should be later than Figure 4 Medium load torque from low value T Lmin Mutation increases to high values ​​of T Lmax moment. γ z is the fitness balance adjustment coefficient, which is a constant greater than 0 and is used to balance the starting performance and steady-state performance (including steady-state error and anti-interference ability) of the motor; reduce γ z value, the weight of the startup performance item in the objective function value Q3 becomes larger, and the system performance is more inclined to the rapidity of motor startup; increasing γ z value, the weight of the steady-state performance item in the objective function value Q3 becomes larger, and the system performance is more inclined to the steady-state performance and anti-interference ability of speed control. z1 When taking values, first estimate For example, if the rated speed of the motor is 1500r / min (corresponding to the rated rotor angular velocity of 157rad / s) and the starting time is about 0.2s, then The value of the item does not exceed 40; let γ z2 The value is equal to 10, γ z is equal to 2, and the rotor angular velocity steady-state error limit ω Δ Equal to 2rad / s, at this time The reasonable value of γ is no more than 60. z1 It can be taken as a constant equal to 300 or greater than 300. If other index factors need to be considered in the optimization of PI speed controller parameters, for example, whether the transition process time is short enough, whether the steady-state error is small enough, etc., other objective functions other than formula (20) can be established.

[0148] The specific steps of using the wolf pack algorithm to optimize the parameters of the PI speed controller and the load torque observer in the embodiment of the elevator permanent magnet synchronous motor drive control system are:

[0149] Step 301, initialize the wolf pack. Assume that the initial position of each individual wolf in the wolf pack is Where M is the number of individuals, which is generally selected between 20 and 150, and the initial position is required to obey random distribution. For different optimization objects, there are:

[0150] (1) For the PI speed controller and load torque observer embodiment 1, the parameter vector to be optimized is μ1=[K p , T i , G max , G min, ε1, ε2, α], at this time, the search space dimension N of the wolf pack algorithm is equal to 7. After optimizing the final position of the alpha wolf, g max , g min according to Sliding mode gain k g According to the parameter α, the initial position of the i-th wolf is expressed as Corresponding to the parameter vector μ1 to be optimized; the position value interval is [z imin z imax ], the range interval can be given based on existing knowledge or experience, for example, parameter K p The value interval [z 1min z 1max ] is [0 10I N ], parameter T i The value interval [z 2min z 2max ] is [0.001 0.5], parameter G max The value interval [z 3min z 3max ] is [-4 4]; parameter G min The value interval [z 4min z 4max ] is [-4 4]; the value range of parameter ε1 is [z 5min z 5max ] and the value range of parameter ε2[z 6min z 6max ] are [0 0.05T N ], T N is the rated torque of the motor; the value range of parameter α is [z 7min z 7max ] is

[15] .

[0151] (2) For the PI speed controller and load torque observer embodiment 2, the parameter vector to be optimized is μ2 = [K p , T i , G max , G min , ε1, ε2, β], at this time, the search space dimension N of the wolf pack algorithm is equal to 7. After optimizing the final position of the leader, g max , g min according to Proportional gain k W According to the parameter β, the initial position of the i-th wolf is expressed as Corresponding to the parameter vector μ2 to be optimized; the position value interval is [z imin z imax], the range interval can be given based on existing knowledge or experience. For example, the value intervals of the first 6 parameters in vector μ2 are the same as those of the first 6 parameters in vector μ1, and the value interval of parameter β is [z 7min z 7max ] is

[120] .

[0152] Step 302, hunting competition. The fitness value of each wolf in the wolf pack is calculated according to formula (20). The smaller the fitness value, the better the position of the wolf. The wolf with the best position is selected as the alpha wolf, and the R1 wolves with the best positions other than the alpha wolf are the candidate wolves. The R1 candidate wolves start hunting search behavior according to formula (21), and each candidate wolves compete for the alpha wolf according to the size of the fitness value, specifically:

[0153] Step 3021, the candidate wolf randomly selects h1 directions, searches for parameters along each direction according to formula (21), moves forward one step, and then moves back; calculates the fitness value after moving forward according to formula (20), selects the minimum fitness value in all directions, and if the minimum fitness value is smaller than the fitness value of the original position of the candidate wolf, replaces the original position of the candidate wolf with the position where the minimum fitness value is located; when the fitness value of the candidate wolf is smaller than the fitness value of the leader wolf, it is regarded as the leader wolf and withdraws from the hunting competition, otherwise, enters step 3022;

[0154] Step 3022, each candidate wolf repeats step 3021 h2 times;

[0155] Step 3023: After all R1 wolves complete step 3022, they exit the hunting competition.

[0156]

[0157] In formula (21), i = 1, 2, ..., R1; j = 1, 2, ..., N; rand (-1, 1) is a random number uniformly distributed in [-1 1]; l = 1, 2, ..., h1; stepa is the hunting step length, and the recommended value range of stepa is [0.1 0.9]; z i =[z i1 z i2 … z iN ] is the position of the ith candidate wolf. R1 is recommended to be a fixed value or a random value in the interval [0.1M 0.25M]; the number of directions h1 is recommended to be in the interval [3 8], and the number of repetitions h2 is recommended to be in the interval [3 10].

[0158] Step 303, summoning and charging. The alpha wolf and the other wolves other than the candidate wolf start charging and searching according to formula (22), charging towards the alpha wolf. The fitness value of the new position is calculated according to formula (20). When the new position searched by the i-th wolf is better than its current position, its position is changed, otherwise it remains unchanged; if the new position searched by the i-th wolf is better than the alpha wolf's position, the i-th wolf is converted to the alpha wolf and re-initiates the call.

[0159] z′ ij =z ij +rand(-1,1)·stepb·(z bj -z ij ) (twenty two)

[0160] In formula (22), i=1, 2,…,M-R1-1; j=1, 2,…,N; z′ i =[z′ i1 z′ i2 … z′ i8 ] indicates the position of the i-th wolf’s search update; z i =[z i1 z i2 … z iN ] represents the current position of the i-th wolf; z b =[z b1 z b2 … z bN ] indicates the current position of the leader wolf; stepb is the running step length, and the recommended value range of stepb is [1.3 2.5].

[0161] Step 304, siege the prey. At the call of the alpha wolf, the other wolves siege the prey according to formula (23). The fitness value of the new position is calculated according to formula (20). When the new position searched by the i-th wolf during the siege is better than the current position, the position of the wolf is changed, otherwise it remains unchanged; if the new position searched by the i-th wolf during the siege is better than the alpha wolf's position, the i-th wolf is converted to the alpha wolf.

[0162]

[0163] In formula (23), i = 1, 2, ..., M-1; j = 1, 2, ..., N; represents the current position of the i-th wolf (i.e., the position after n iterations), The new position searched by the i-th wolf; z b =[z b1 z b2 …z bN ] represents the current position of the leader; δ is a pre-set threshold, and the recommended value range of δ is [0.1 0.4]; z jmaxand z jmin are the maximum and minimum values ​​of the j-th dimension parameter value interval respectively; stepc is the siege step length, which is calculated according to formula (24).

[0164]

[0165] In formula (24), n is the current number of iterations, n max is the maximum number of iterations set; stepc max 、stepc min are the maximum siege step length and the minimum siege step length respectively, and ξ is the siege step length attenuation factor. min The value range of stepc is [0.3 1.3]. max The value is stepc min 5 to 100 times. n increases from 1 to n max When stepc increases exponentially from stepc max Decay to stepc min , adjusting the size of the siege step length attenuation factor ξ can adjust the relative attenuation speed of the siege step length in the early and late stages; the larger the value of ξ, the faster the attenuation speed of stepc in the early stage relative to the late stage, and the later stage tends to be equal to stepc for a long time. min The smaller the value of ξ, the slower the decay speed of stepc in the early stage relative to the later stage, and the later stage still decays and approaches stepc. min ; The value range of ξ is [1.5 10]. Figure 3 The effect diagram of adjusting the relative decay speed of the siege step length stepc for the siege step length decay factor ξ; Figure 3 In the figure, the horizontal axis is the number of iterations n, n max =1000; the vertical axis is the siege step length stepc, stepc max Equal to 10, stepc min equal to 1; curve ① is the attenuation curve of stepc when ξ is equal to 2, and curve ② is the attenuation curve of stepc when ξ is equal to 8.

[0166] When the value of ξ is large, the proportion of the process of small-step precise search in the later stage increases, which is more conducive to finding the precise optimal solution; when the value of ξ is small, the diversity of the later stage of the search is enhanced, which is more conducive to avoiding the local optimum near the optimal solution; for different optimization objects such as finding the precise optimal solution and avoiding the local optimum near the optimal solution, the value of the siege step attenuation factor ξ can be reasonably selected to adapt to the greatest extent to achieve different optimization effects.

[0167] Step 305, judging the termination condition. If the number of loop iterations reaches or the fitness value of the leader is less than a certain threshold, the optimization process is terminated and the leader position parameter is the optimal parameter of the optimized parameter vector. Otherwise, set n=n+1 and go to step 306.

[0168] Step 306, competitive update. According to the principle of survival of the fittest, randomly generate R2 wolves to replace the eliminated wolves with the worst R2 fitness value, and compete to update the wolf pack, and go to step 302. R2 is recommended to take a fixed value or a random value in the interval [0.05M 0.15M].

[0169] In the above steps, when the new positions of individual wolves are randomly generated, or when individual wolves search for new positions, the variables of each dimension of the new position of each wolf cannot exceed the corresponding value range of the variable. max , g min There is a constraint g between min <g max , corresponding to the constraint G min >G max When a new position of an individual wolf is randomly generated, or when an individual wolf searches for a new position, first, the individual wolf matches z i The parameter z in i3 According to the value interval [z 3min z 3max ] Randomly generate a position or update the position, then the individual wolf z i The parameter z i4 According to the value interval [z i3 z 4max ] Randomly generate positions or update positions so that individual wolves i Satisfy the constraint G min >G max , that is, satisfying the constraint g min <g max There is a constraint between ε1 and ε2. ε1<ε2. When a new position of an individual wolf is randomly generated, or when an individual wolf searches for a new position, first, the individual wolf z i The parameter z in i5 (i.e., ε1) according to the value interval [z 5min z 5max ] Randomly generate a position or update the position, then the individual wolf z i The parameter z i6 (i.e., ε2) according to the value interval [z i5 z 6max ] Randomly generate positions or update positions so that individual wolves i The constraint ε1<ε2 is satisfied.

[0170] In step 305, the termination condition adopts the maximum number of iterations limit method, and the maximum number of iterations n max It is recommended to take a fixed value in the interval [20 500]. When setting the condition that the head wolf adaptation value is less than a certain threshold, the threshold termination condition needs to comprehensively consider the starting performance and anti-interference performance of the motor speed control. For example, if the rated speed of the motor is 1500r / min (corresponding to the rated rotor angular velocity of 157rad / s), the starting time is required to be within 0.2s, and the anti-interference performance and starting performance are basically balanced, then the threshold of the termination condition can be set to 30.

[0171] In the above steps, when calculating the fitness value of the new position according to formula (20), it is required to convert the individual wolf positions into corresponding PI speed controller parameters and load torque observer parameters in turn to control the motor operation (or operation in the simulation system). * is a step signal, load torque T L When the rotor angular velocity changes suddenly in a steady state, the speed response e(t) of the motor is obtained, and the transition process time t is determined based on e(t). z At the same time, the wolf’s fitness value Q3 is calculated based on e(t).

[0172] In the above embodiments, the rotor angular velocity steady-state error refers to the difference between the instantaneous value and the steady-state value of the motor rotor angular velocity in steady state, and the rotor angular velocity steady-state error limit is the maximum absolute value of the rotor angular velocity steady-state error allowed by the system; the rotor angular velocity steady-state error limit is usually the same as the maximum value of the rotor angular velocity steady-state error allowed by the system. The torque observation tracking overshoot limit is usually 1% to 10% of the rated torque of the motor. Specifically, the torque observation tracking overshoot limit is 2% of the rated torque, or 5% of the rated torque, or 10% of the rated torque, and so on. Tracking adjustment time refers to the process time from when the load torque suddenly changes from one fixed value to another fixed value, from the moment when the sudden change starts to when the load torque observer outputs the observed value and stably enters the range of the load torque observation steady-state error limit; the load torque observation steady-state error refers to the error between the instantaneous value of the observed torque and the load torque when the load torque is constant and the load torque observer is in a steady state, which includes the observation error caused by the sliding mode observer jitter (or the steady-state fluctuation of the state observer) and the observation error caused by interference other than load fluctuation; the load torque observation steady-state error limit is the maximum absolute value of the load torque observation steady-state error allowed by the load torque observer; the load torque observation steady-state error limit is usually the same as the maximum value of the load torque observation steady-state error allowed by the system; the load torque observation steady-state error limit is usually 1% to 5% of the rated torque of the motor. Specifically, the load torque observation steady-state error limit is 1% of the rated torque, or 2% of the rated torque, or 5% of the rated torque, etc. The torque observation tracking overshoot refers to the maximum deviation of the load torque after the load torque suddenly changes from one fixed value to another fixed value, and the observed value output by the load torque observer exceeds the maximum deviation of the load torque after the sudden change. When the steady-state error of the load torque observation is within the vicinity of the steady-state error limit of the load torque observation, for example, within the range of 95% to 105%, or within the range of 98% to 102%, it is considered that the steady-state error of the load torque observation has increased to the steady-state error limit of the load torque observation.

[0173] In the present invention, the elevator permanent magnet synchronous motor drive control system is a permanent magnet synchronous motor speed control system. The permanent magnet synchronous motor speed control system and the permanent magnet synchronous motor control method based on load torque compensation provided by the present invention can be used in other permanent magnet synchronous motor application occasions in addition to being used in the elevator permanent magnet synchronous motor drive control system.

[0174] In addition to the technical features described in the specification, other technologies involved in the present invention are conventional technologies mastered by those skilled in the art. For example, the q-axis current controller and the d-axis current controller use PI controllers for control and controller parameter selection, the PI speed controller control parameter selection, the position and speed detection module uses a rotary transformer or a photoelectric encoder to detect the rotation angle and rotation speed of the permanent magnet synchronous motor rotor, and the Clarke transformation module, Park transformation module, Park inverse transformation module and SVPWM module, three-phase inverter transformation method and application method, etc., are all conventional technologies mastered by those skilled in the art.

Claims

1. A permanent magnet synchronous motor control method based on load torque compensation, wherein the speed of the permanent magnet synchronous motor is controlled by a PI speed controller, a load torque observer observes the load torque, and the output of the load torque observer is used to perform load torque compensation on the output of the PI speed controller; characterized in that: PI speed controller PI speed controller output load torque given value and torque current given component i′ q for Where p is the number of motor pole pairs, ψ f is the permanent magnet flux; K p is the proportional coefficient of the PI speed controller, T i is the integral time constant of the PI speed controller; the motor's rotor angular velocity error e = ω * -ω,ω * is the given rotor angular velocity, ω is the rotor angular velocity, and e(t) is the instantaneous value of the rotor angular velocity error; The load torque observer is Where J is the moment of inertia, is the estimated value of the rotor angular velocity, g is the feedback gain of the load torque observer and g<0; k g is the sliding mode gain of the load torque observer and k g ≤-|e2 / J|, is the load torque observation error, T L is the load torque, is the observed value of load torque; i q is the q-axis current; The load torque observer is based on the load torque given value and load torque observation The feedback gain g is adjusted by: Step S1: The load torque observer performs T on the load torque. L Observe and obtain the load torque observation value The PI speed controller performs control calculation to obtain the load torque given value Step S2: Calculation Step S3, determine whether ΔT is greater than ε2; ​​when ΔT is greater than ε2, set the feedback gain g equal to g min And exit; when ΔT is less than or equal to ε2, enter step S4; Step S4, determine whether ΔT is less than ε1; when ΔT is less than ε1, set the feedback gain g equal to g max And exit; when ΔT is greater than or equal to ε1, enter step S5; Step S5: Feedback gain g is calculated according to Calculate; where ε1 is the torque change lower limit comparison threshold, ε2 is the torque change upper limit comparison threshold, and 0<ε1<ε2; g max is the high value of feedback gain, g min is the low value of feedback gain, and g min <g max <0; The parameters of PI speed controller and load torque observer are optimized and adjusted using particle swarm algorithm; The method of load torque compensation for the output of PI speed controller is to use the observed value of load torque Converted into torque current compensation component i″ q Feedforward compensation to the input of the q-axis current PI controller; q-axis torque current setpoint for:

2. The permanent magnet synchronous motor control method based on load torque compensation according to claim 1, characterized in that: When the PI speed controller parameters are optimized using the particle swarm algorithm, the parameter vector to be optimized is θ = [K p , T i ], the search space dimension N is equal to 2.

3. The permanent magnet synchronous motor control method based on load torque compensation according to claim 2, characterized in that: When the PI speed controller parameters are optimized using the particle swarm algorithm, the objective function Q1 for calculating the fitness value is: Where, e(t) is the instantaneous value of the rotor angular velocity error, t m is the transition time of the motor rotor angular velocity step response, t=0 is the starting time of the motor step response; Q 11 γ m1 (1-sgn(e(t)+ω δ )) is the angular velocity overshoot penalty function, γ m1 is a sufficiently large positive number, ω δ is the rotor angular velocity overshoot limit; Q 12 is the steady-state error penalty function, ω Δ is the steady-state error limit of the rotor angular velocity; γ m2 ≥2.

4. The permanent magnet synchronous motor control method based on load torque compensation according to claim 1, characterized in that: When the load torque observer parameters are optimized using the particle swarm algorithm, the parameter vector to be optimized is θ1 = [G max , G min , ε1, ε2, α], the search space dimension N is equal to 5; g max With G max The relationship is g min With G min The relationship is k g The relationship with α is Among them, α≥1.

5. The permanent magnet synchronous motor control method based on load torque compensation according to claim 4, characterized in that: When the load torque observer parameters are optimized using the particle swarm algorithm, the objective function Q2 for calculating the fitness value is: in, is the load torque observation error, e2(t) is the instantaneous value of the load torque observation error, t p is the tracking adjustment time of the motor load torque observed step response, t = 0 is the load mutation moment of the load torque observed step response; Q 21 γ p1 (1-sgn(e2(t)+T δ )) is the torque observation tracking overshoot penalty function, T δ is the torque observation tracking overshoot limit, γ p1 is a sufficiently large positive number; max(|e2(t)|) is the absolute value of the maximum torque observation steady-state error, γ p2 is the fitness balance emphasis coefficient, γ p2 >0;Q 22 γ p1 (1-sgn(e2(t)+T Δ )) is the torque observation steady-state error penalty function, T Δ is the load torque observed steady-state error limit; γ p3 ≥2.

6. The permanent magnet synchronous motor control method based on load torque compensation according to claim 1, characterized in that:Detect the rotor position θ and three-phase current i of the permanent magnet synchronous motor a 、i b and i c ; Based on the three-phase current i a 、i b and i c Clark transformation is performed on the permanent magnet synchronous motor to obtain the current i in the α-β axis coordinate system: α 、Current i β , according to the current i α 、Current i β Perform Park transformation on the rotor position θ to obtain the current i in the dq axis coordinate system d 、Current i q .

Images