CN107579684B - Sensorless brushless direct current motor commutation error correction method based on neural network controller - Google Patents
Sensorless brushless direct current motor commutation error correction method based on neural network controller Download PDFInfo
- Publication number
- CN107579684B CN107579684B CN201711062666.XA CN201711062666A CN107579684B CN 107579684 B CN107579684 B CN 107579684B CN 201711062666 A CN201711062666 A CN 201711062666A CN 107579684 B CN107579684 B CN 107579684B
- Authority
- CN
- China
- Prior art keywords
- error
- commutation
- neural network
- network controller
- voltage difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The invention relates to a sensorless brushless direct current motor commutation error correction method based on a neural network controller. Firstly, the counter potential zero crossing point of the motor is extracted according to the line voltage difference of the motor, then the self-adaptive neural network controller is adopted to correct the phase error, the error convergence speed is high, and the adaptability to the influence of parameter perturbation and the like is strong. The neural network controller consists of an input layer, a hidden layer and an output layer, wherein the input layer receives the opposite potential voltage difference error and the integral thereof as two input nodes, the input layer to the hidden layer and the hidden layer to the output layer are respectively connected by weighting factors W1 and W2, and the control performance of error quick convergence can be obtained by reasonably selecting the learning rate of the controller. The stability condition of the closed-loop system is obtained by lyapunov stability theory analysis.
Description
Technical Field
The invention relates to the technical field of brushless direct current motors, in particular to a sensorless brushless direct current motor rotor commutation error correction method based on a neural network controller, which is suitable for the fields of industry, aerospace control and the like and realizes the accurate commutation of a sensorless brushless direct current motor.
background
brushless dc motors are widely used in the industrial, robotic, automotive, aerospace and military fields due to their advantages of high power density, high efficiency, high torque-to-inertia ratio, compact structure, etc. The traditional brushless direct current motor needs a position sensor to provide rotor position information so as to realize accurate phase change control, however, the installation of the position sensor not only increases the equipment cost, but also puts higher requirements on the daily maintenance of the motor, and simultaneously reduces the overall stability and reliability of the system. Especially in applications where reliability is a high requirement, such as military and aerospace applications, errors in position sensors can cause serious problems. In order to avoid the above-described adverse effects, the position-sensorless motor driving technique is widely applied to the brushless dc motor. Among the numerous position sensorless driving methods, the back emf zero crossing detection method is most applied because of its advantages of simplicity, easy implementation, high reliability, and the like. However, the commutation signal generated by the back emf zero crossing detection method is inevitably accompanied by commutation errors due to the effects of low pass filtering, armature reaction, device delay, and the like. Although the error can be calculated directly or corrected by the PI controller, it can be affected by back emf harmonic coefficients, rotor speed and ambient temperature due to the strong non-linearity of the motor. If a conventional controller is used, the effect of error correction may be reduced in the event of motor parameter disturbances.
Disclosure of Invention
the technical problem to be solved by the invention is as follows: the commutation signal generated by the counter potential zero crossing point detection method is inevitably accompanied by commutation errors, however, the conventional controller is greatly disturbed by motor parameters.
the technical scheme adopted by the invention for solving the technical problems is as follows: a sensorless brushless direct current motor commutation error correction method based on a neural network controller detects a zero crossing point according to line voltage difference and extracts basic commutation information, and then compensates commutation errors through a self-adaptive neural network controller, so that high-precision high-stability commutation control of sensorless control of a brushless direct current motor is achieved.
The principle of the invention is as follows:
1. Correction principle of phase error
when the motor accurately commutates, the opposite potential signal is symmetrical trapezoidal wave. When the commutation has an error, the opposite potential signal has obvious distortion, and the waveform is not symmetrical any more: during lagging phase conversion, the counter potential voltage before conduction is obviously larger than that at the end of conduction; during leading commutation, the back-emf voltage before conduction is significantly less than that at the end of conduction. It can be found that the reverse potential difference at both points can be used to indicate a commutation error. The invention just utilizes the deviation to calculate the commutation error, so the back electromotive voltage value at two moments needs to be sampled. The reverse potential difference versus commutation error can be expressed as a non-linear function:
△u=f(α) (16)
Where Δ u is the reverse potential difference and α is the commutation error.
2. Correction method based on neural network
Firstly, zero crossing point detection is carried out to provide basic commutation information, then the back electromotive force voltage difference delta u before and after motor commutation is used as feedback quantity to be sent to a neural network correction controller, and deviation compensation quantity output by the controller is fused with the basic commutation information to control motor commutation.
(1) Obtaining basic commutation information:
The phase voltage equation of the brushless DC motor can be written as
wherein the subscript x ═ a, b, c, uxIs the voltage of the DC terminal to ground, unIs neutral to ground voltage ixis phase current, R is phase resistance, L is phase inductance, exare of opposite potential. Taking x-phase and y-phase as conducting phase, and the other phase as non-conducting phase as an example, the line voltage can be obtained according to a formula,
wherein u isxyrepresenting the line voltage between the two phases.
Taking the CB-phase conduction as an example, when the B-phase lower tube is conducted and the C-phase upper tube is conducted, the line voltage u can be obtained according to the equation (18)caand uaband can derive these two line voltagesThe difference of (a) to (b),
at this time, ic=-ib,ia=0,ec=-ebAccordingly, formula (19) can be simplified to
ucaab=-2ea (20)
The same principle is that:
Based on this, back-emf zero-crossings can be obtained by detecting line voltage differences to provide basic commutation information.
(2) neural network controller architecture for error correction:
In the correction scheme based on the neural network controller (as shown in figure 3), the motor speed omega and the winding current iLThe PI controller is used for regulating, basic information required by commutation is extracted from a zero crossing point detected from the line voltage difference, and then the neural network controller is used for outputting error compensation quantity and fusing the error compensation quantity with the basic commutation information to be used for commutation control of the sensorless motor. The neural network controller (as shown in figure 5) is composed of a 2-3-1 structure neural network. The input layer receives the voltage difference error of the opposite potentials before and after the phase change and the integral value thereof; the output layer outputs the compensation quantity of the phase error; the input layer to hidden layer and hidden layer to output layer are connected by weighting factors of W1 and W2 respectively, and the weighting factors are updated by minimizing an error function by adopting a gradient descent method. The neural network controller receives a correction error of e-uref-△u,urefIs an error reference input, and Deltau is the difference value of opposite potential voltages before and after commutation representing the commutation error. The specific implementation of the controller is as follows:
x1(k)=e(k),x2(k)=∑e(k) (22)
wherein x isi(k) For the kth iteration of the ith node of the input layer, hj(k) Is the k-th iteration function of the j-th node of the hidden layer, y (k) is the k-th iteration output of the output layer, e (k) represents the counter potential voltage difference error input of the k-th iteration, W1_ij(k) a weighting factor, W, representing the k-th iteration connecting the ith node of the input layer and the jth node of the hidden layer2_j(k) and representing the weighting factor connecting the jth node of the hidden layer and the output layer of the kth iteration.
the k-th iteration error function is defined as,
Updating the weighting factor by minimizing an error function, and solving the increment of the weighting factor by adopting a gradient descent method,
wherein, Δ W1_ij(k) And Δ W2_j(k) Is the increment of updating the next weighting factor, eta is the learning rate, J (k) is the kth iteration error function, Deltau (k) is the kth voltage difference of the opposite potential characterizing the commutation error, and Deltatheta (k) is the phase error of the commutation point at the kth time.
due to the non-linear relationship and parameter uncertainty between commutation error and the voltage value characterizing the commutation error,cannot be represented explicitly by an analytic function, and the change trend of the analytic function is represented by a symbolic function. Accordingly, formula (26) can be arranged as,
The new weighting factor may be updated as,
(3) analyzing the stability of the neural network controller and setting parameters:
In order to ensure the stability of the correction system, stable conditions are obtained by the Lyapunov stability theory analysis, and a learning rate selection rule is set under the condition of ensuring the stability, so that the learning rate selection rule is adaptively adjusted to adapt to the continuously changing error environment. The concrete implementation is as follows:
The lyapunov function is defined as:
where V (k) is the k-th updated Lyapunov function, the Lyapunov function increment Δ V (k) between the k-th and k + 1-th iterations is,
the input error may be expressed as,
e(k+1)=e(k)+△e(k) (31)
Where Δ e (k) is the back emf voltage difference error input increment for the k +1 th time relative to the k-th time, which can be written as,
By substituting equation (27) into equation (32), it can be obtained,
in the case of substituting equations (31) and (33) into equation (30), the increment of the lyapunov function can be rewritten as,
it can be found that when Δ V is less than or equal to 0, the stability of the closed-loop system can be ensured. Therefore, the learning rate η needs to be satisfied,
the learning rate of the neural network controller has a significant effect on the error convergence rate. Too slow a learning rate may result in slow convergence, but too large η may result in system oscillation and instability. In order to achieve better effect, the method sets the selection rule of the learning rate eta as follows,
Wherein Δ utolerableIs a voltage difference tolerance value caused by commutation errors.
Compared with the prior art, the invention has the advantages that:
(1) Aiming at the problem of correcting the commutation error of the rotor of the brushless direct current motor without the position sensor, the invention adopts the neural network controller to overcome the nonlinear characteristic of the brushless direct current motor on the basis of the traditional voltage difference PI regulation method, accurately compensates the commutation error and improves the commutation accuracy.
(2) The learning rate of the neural network controller is obtained by analyzing the stability of the closed-loop system, and can be adjusted in a self-adaptive manner under the condition of ensuring the stability according to the real-time state, so that the stability of the system is ensured, and a faster convergence effect can be obtained.
Drawings
FIG. 1 is a flow chart of commutation error correction according to the present invention;
FIG. 2 is a waveform diagram of reverse potential in the presence of commutation error, wherein FIG. 2(a) is a diagram of reverse potential in the event of commutation advance, and FIG. 2(b) is a diagram of reverse potential in the event of commutation retard;
FIG. 3 is an overall block diagram of the control system of the present invention;
FIG. 4 is a diagram of a neural network control module of the present invention;
fig. 5 is an internal structure diagram of the neural network controller according to the present invention.
Detailed Description
The invention is further described with reference to the following figures and detailed description.
In the specific implementation process, the specific implementation steps of the invention are as follows:
correction method based on neural network, motor speed omega and winding current iLThe phase-change error correction information is obtained by a neural network controller under the regulation of a PI controller. Firstly, zero crossing point detection is carried out according to the voltage difference of the two lines to provide basic commutation information, then the back electromotive force voltage difference delta u before and after the motor commutation is used as a feedback quantity to be sent to a neural network correction controller, and the deviation compensation quantity output by the controller is fused with the basic commutation information to control the motor commutation. The method specifically comprises the following steps:
(1) detecting line voltage difference and extracting zero crossing points:
as can be seen from the above formula, by detecting the line voltage difference, the counter potential zero crossing point can be obtained, and the counter potential zero crossing point is delayed by 30 electrical angles according to the speed measurement information to generate the actual phase change signal.
(2) Collecting the voltage difference of the phase change point, inputting the voltage difference into a neural network controller, and calculating an error correction compensation quantity:
and the voltage difference of the commutation points is sent to a neural network controller, error correction compensation quantity is calculated in real time by adopting the neural network structure, weight update rate and learning rate regulation mechanism, and the compensation quantity output by the controller is fused with basic commutation information by a system to control motor commutation.
(3) And (3) repeating the steps (1) and (2) by the control system, realizing closed-loop correction of the commutation error and driving the motor to operate.
Those skilled in the art will appreciate that the details of the present invention not described in detail herein are well within the skill of those in the art.
Claims (2)
1. a sensorless brushless direct current motor commutation error correction method based on a neural network controller is characterized in that: firstly, zero crossing point detection is carried out to provide basic commutation information, then the voltage difference of opposite electric potentials before and after motor commutation is used as feedback quantity to be sent to an adaptive neural network correction controller, and the phase commutation error is accurately compensated after the deviation compensation quantity output by the controller is fused with the basic commutation information, so that the high-precision commutation control of the sensorless brushless direct current motor is realized;
The neural network controller for correcting the commutation error consists of a 2-3-1 structure neural network; the input layer receives the voltage difference error of the opposite potentials before and after the phase change and the integral value thereof; the output layer outputs the compensation quantity of the phase error; from input layer to hidden layer and from hidden layer to output layer respectively by W1And W2connecting the weighting factors, and updating the weighting factors by minimizing an error function by adopting a gradient descent method; the error of the voltage difference of the opposite electric potentials received by the neural network controller is e-urefΔ u, where urefIs a voltage difference reference input, and Δ u is an opposite potential voltage difference value representing a commutation error, the specific implementation of the controller is as follows:
x1(k)=e(k),x2(k)=∑e(k) (1)
wherein x isi(k) for the kth iteration of the ith node of the input layer, hj(k) is the k-th iteration function of the j-th node of the hidden layer, y (k) is the k-th iteration output of the output layer, e (k) represents the counter potential voltage difference error input of the k-th iteration, W1_ij(k) A weighting factor, W, representing the k-th iteration connecting the ith node of the input layer and the jth node of the hidden layer2_j(k) representing a weighting factor connecting a jth node of the hidden layer and the output layer of the kth iteration;
the k-th iteration error function is defined as,
Updating the weighting factor by minimizing an error function, and solving the increment of the weighting factor by adopting a gradient descent method,
wherein, Δ W1_ij(k) and Δ W2_j(k) Is the increment of updating the next weighting factor, eta is the learning rate, J (k) is the kth iteration error function, uref(k) is the kth voltage difference reference input, Δ u (k) is the opposite potential kth voltage difference value characterizing the commutation error, and Δ θ (k) is the commutation point phase error at the kth time;
due to the non-linear relationship and parameter uncertainty between commutation error and the voltage value characterizing the commutation error,Cannot be represented explicitly by an analytic function, and the variation trend is represented by a symbolic function, so that the formula can be arranged as,
The new weighting factor may be updated as,
2. The sensorless brushless DC motor commutation error correction method based on neural network controller of claim 1, wherein: the stability condition of the neural network controller is obtained by analyzing the Lyapunov stability theory, and a learning rate selection rule is set under the condition of ensuring stability, so that the neural network controller is adaptively adjusted to adapt to a continuously changing error environment, and the method is specifically realized as follows: the lyapunov function is defined as:
Wherein V (k) is the Lyapunov function after the kth update;
The lyapunov function increment between the kth and the (k + 1) th iterations is,
Where Δ V (k) is the Lyapunov function increment from the k +1 th time to the k-th time;
the input error may be expressed as,
e(k+1)=e(k)+Δe(k) (10)
Where Δ e (k) is the back emf voltage difference error input increment relative to the kth time at (k + 1), the increment of the input error can be written as,
by substituting formula (6) for formula (11),
In the case of substituting equations (10) and (12) into equation (9), the increment of the lyapunov function can be rewritten as,
It can be shown that when Δ V is less than or equal to 0, the stability of the closed-loop system can be ensured, therefore, the learning rate η needs to be satisfied,
the learning rate of the neural network controller has obvious influence on the error convergence rate, slow convergence can be caused by too slow learning rate, but oscillation and instability of the system can be caused by too large eta, in order to achieve better effect, the method sets the selection rule of the learning rate eta as follows,
Wherein, Δ utolerableIs a voltage difference tolerance value caused by commutation errors.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711062666.XA CN107579684B (en) | 2017-11-02 | 2017-11-02 | Sensorless brushless direct current motor commutation error correction method based on neural network controller |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711062666.XA CN107579684B (en) | 2017-11-02 | 2017-11-02 | Sensorless brushless direct current motor commutation error correction method based on neural network controller |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107579684A CN107579684A (en) | 2018-01-12 |
CN107579684B true CN107579684B (en) | 2019-12-13 |
Family
ID=61041331
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711062666.XA Active CN107579684B (en) | 2017-11-02 | 2017-11-02 | Sensorless brushless direct current motor commutation error correction method based on neural network controller |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107579684B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109150033B (en) * | 2018-08-08 | 2022-04-22 | 南京航空航天大学 | Current synchronization control method of doubly salient brushless direct current motor |
JP2022514535A (en) * | 2018-12-27 | 2022-02-14 | ▲広▼▲東▼美的白色家▲電▼技▲術▼▲創▼新中心有限公司 | BLDC back electromotive force zero cross point threshold determination method, device and storage medium |
CN110716577B (en) * | 2019-11-13 | 2020-12-15 | 北京航空航天大学 | Sensorless adaptive commutation error compensation method for brushless direct current motor of magnetic suspension control moment gyroscope |
CN112542967A (en) * | 2020-11-27 | 2021-03-23 | 中国科学院半导体研究所 | Motor phase commutation control method and system based on back electromotive force neural network |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08149884A (en) * | 1994-11-21 | 1996-06-07 | Toyo Electric Mfg Co Ltd | Control method for inverter having constant measuring and setting function |
CN102751920A (en) * | 2012-06-21 | 2012-10-24 | 天津工业大学 | Method for identifying and controlling two-induction-motor synchronous system |
CN103107710A (en) * | 2011-11-14 | 2013-05-15 | 深圳市安邦信电子有限公司 | High-voltage inverter adaptive control system based on neural network and construction method thereof |
CN103532459A (en) * | 2013-10-15 | 2014-01-22 | 东南大学 | Linear servo motor control method for numerically-controlled machine tool driving |
CN106059409A (en) * | 2016-05-27 | 2016-10-26 | 北京航空航天大学 | Position sensor-free brushless direct-current motor rotor phase commutation error correction method and control system |
CN106067077A (en) * | 2016-06-01 | 2016-11-02 | 新奥泛能网络科技股份有限公司 | A kind of load forecasting method based on neutral net and device |
CN106099988A (en) * | 2016-08-20 | 2016-11-09 | 南京理工大学 | LCL type photovoltaic combining inverter adaptive control system and control method |
CN106655918A (en) * | 2016-11-09 | 2017-05-10 | 北京航空航天大学 | Sensorless brushless DC motor commutation deviation fast correction control system |
-
2017
- 2017-11-02 CN CN201711062666.XA patent/CN107579684B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08149884A (en) * | 1994-11-21 | 1996-06-07 | Toyo Electric Mfg Co Ltd | Control method for inverter having constant measuring and setting function |
CN103107710A (en) * | 2011-11-14 | 2013-05-15 | 深圳市安邦信电子有限公司 | High-voltage inverter adaptive control system based on neural network and construction method thereof |
CN102751920A (en) * | 2012-06-21 | 2012-10-24 | 天津工业大学 | Method for identifying and controlling two-induction-motor synchronous system |
CN103532459A (en) * | 2013-10-15 | 2014-01-22 | 东南大学 | Linear servo motor control method for numerically-controlled machine tool driving |
CN106059409A (en) * | 2016-05-27 | 2016-10-26 | 北京航空航天大学 | Position sensor-free brushless direct-current motor rotor phase commutation error correction method and control system |
CN106067077A (en) * | 2016-06-01 | 2016-11-02 | 新奥泛能网络科技股份有限公司 | A kind of load forecasting method based on neutral net and device |
CN106099988A (en) * | 2016-08-20 | 2016-11-09 | 南京理工大学 | LCL type photovoltaic combining inverter adaptive control system and control method |
CN106655918A (en) * | 2016-11-09 | 2017-05-10 | 北京航空航天大学 | Sensorless brushless DC motor commutation deviation fast correction control system |
Also Published As
Publication number | Publication date |
---|---|
CN107579684A (en) | 2018-01-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107579684B (en) | Sensorless brushless direct current motor commutation error correction method based on neural network controller | |
CN106655918B (en) | A kind of quick corrective control of brushless DC motor without position sensor commutation deviation | |
Zhang et al. | Sensorless predictive current controlled DC–DC converter with a self-correction differential current observer | |
CN110429889B (en) | Amplitude-adjustable square wave injection maximum torque current ratio motor control method | |
CN111756238B (en) | Novel modeling and composite anti-interference control method for disturbed direct current buck converter system | |
CN110165924B (en) | Improved dead-beat control method of single-phase PWM rectifier | |
Wang et al. | Current sensorless finite-time control for buck converters with time-varying disturbances | |
CN112310991B (en) | Energy storage converter improved finite set model prediction control method based on disturbance observation | |
CN115149806B (en) | Adaptive model prediction control method for staggered parallel Boost converter | |
KR102189823B1 (en) | Automated torque ripple reduction apparatus of motor | |
Ramos-Paja et al. | Sliding-mode controller for a photovoltaic system based on a Cuk converter | |
CN113285593B (en) | Direct-current buck converter system control method based on composite integral sliding mode control | |
Metry et al. | A variable step-size MPPT for sensorless current model predictive control for photovoltaic systems | |
CN108448986B (en) | Permanent magnet motor current control method based on adjustable bandwidth type predictive control | |
CN111835251B (en) | Permanent magnet synchronous motor high-performance control method based on speed-free sensing | |
CN104993485B (en) | A kind of Shunt Hybrid Active Power Filter system and its control method | |
CN112994553A (en) | Simplified model prediction voltage control method for permanent magnet motor system | |
CN113972870B (en) | Robust predictive torque control method for switched reluctance motor | |
CN116087596A (en) | Current detection method and detection system for electromagnetic compatibility monitoring of airborne power supply characteristics | |
CN115566717A (en) | Inverter control strategy based on improved particle swarm algorithm disturbance observer composite sliding mode | |
CN114301361B (en) | Control method of electrolytic capacitor-free permanent magnet synchronous motor driving system based on bus current control | |
CN115940640A (en) | DC-DC buck converter sliding mode control algorithm based on improved discrete approach law | |
CN115498696A (en) | Three-phase rectifier analog resistance control method for unbalanced power grid | |
Zhang et al. | Composite Fast Terminal Sliding Mode Control of DC-DC Converters for Renewable Energy Systems | |
Sheng et al. | Adaptive fast terminal sliding mode control based on radial basis function neural network for speed tracking of switched reluctance motor |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |