Efficiency Optimization Control Method for High-Speed ​​Brushless DC Motor Systems Based on Time-Sharing Commutation

Abstract

This invention discloses an efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation. It establishes a time-sharing commutation mode to determine the variation law of the motor phase current; obtains the turn-on offset angle and turn-off offset angle based on the maximum electromagnetic power using offline numerical solutions; designs the input layer, hidden layer, and output layer structure of an RBF neural network; performs offline pre-training of the RBF neural network using prior information and online training; updates the neural network parameters by acquiring real-time motor system efficiency information; and applies the optimal turn-on offset angle and optimal turn-off offset angle output by the neural network to the time-sharing commutation control, thereby improving the motor's operating efficiency. The advantages of this invention are that it is not limited to the 120° conduction mode of the brushless DC motor windings, but establishes a time-sharing commutation mode; and it utilizes the RBF neural network to quickly optimize the turn-on offset angle and turn-off offset angle, thereby improving the motor system efficiency.

Description

Technical Field

[0001] This invention belongs to the field of high-speed brushless DC motor system control technology, specifically relating to an efficiency optimization control method for high-speed brushless DC motor systems based on time-sharing commutation. Background Technology

[0002] High-speed brushless DC motors are widely used in home appliances, medical devices and other fields due to their significant advantages such as high power density, small size and simple control method. In particular, their traditional three-phase six-step commutation control method is simple, that is, the motor winding is 120° conduction mode per phase, which can well meet various application scenarios.

[0003] In applications of high-speed brushless DC motors, due to limitations in sensor size, weight, and reliability, sensorless technology is typically used for commutation control. Among these methods, back-EMF zero-crossing detection is widely adopted due to its low cost and high reliability. However, due to factors such as low-pass filter delay, hardware and software delays, and winding impedance characteristics in the detection circuit, the detected back-EMF zero-crossing point has a certain error compared to the ideal commutation point, and this error increases with increasing speed. This detection error not only increases the fluctuation of brushless motor phase current and torque but also reduces torque-to-current ratio and operating efficiency.

[0004] To address the commutation error in back EMF zero-crossing detection methods, scholars both domestically and internationally have established mathematical models for various detection delay angles to compensate for the commutation error. The literature [Sensorless Drive Technology for High-Speed ​​Brushless DC Motors Based on Virtual Neutral Point Voltage. IEEE Transactions on Power Electronics, Vol. 30, No. 6, pp. 3275-3285, June 2015] models the low-pass filter delay and hardware / software delay, quantitatively analyzes the magnitude of the error, and performs corresponding angle compensation for the commutation signal, improving the operating performance of the high-speed brushless DC motor. The literature [Online Lead Angle Adjustment Method for Sine Wave Brushless DC Motors with Hall Sensor Misalignment. IEEE Transactions on Power Electronics, Vol. 32, No. 11, pp. 8247-8253, November 2017] analyzes the influence of the motor's impedance characteristics on the optimal commutation point under high-speed operation, incorporating it into the compensation angle range. The aim is to align the back EMF with the phase current, thereby increasing electromagnetic power and further improving motor operating efficiency. However, existing compensation methods based on back EMF zero-crossing detection only consider the conventional 120° commutation mode and fail to account for the impact of non-fixed 120° commutation modes on motor operation. Therefore, finding a commutation compensation method that considers non-fixed 120° winding conduction, adjusting both the phase and amplitude of the phase current, could provide the possibility of further improving the operating efficiency of the motor system. Summary of the Invention

[0005] In view of the above, this invention provides an efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation. This method aims to optimize the efficiency of the high-speed brushless DC motor system, using the turn-on offset angle and turn-off offset angle as optimization variables. It provides an RBF (Radial Basis Function) neural network efficiency optimization strategy that integrates prior information. The network is pre-trained offline using theoretically calculated turn-on and turn-off offset angles to accelerate neural network convergence and improve initial control performance. Then, through online learning, the turn-on and turn-off offset angles are corrected in real time, thereby achieving adaptive optimization of the offset angle and improvement of system efficiency under time-sharing commutation control conditions.

[0006] An efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation includes the following steps:

[0007] (1) Establish the time-sharing commutation mode of QCSI (quasi-current source inverter) in the motor system, and define the turn-on offset angle during the commutation process. i 1 and the off offset angle i 2;

[0008] (2) Obtain the switching-off angle based on the maximum electromagnetic power using offline numerical solution. i 1m and the off offset angle i 2m ;

[0009] (3) Construct an RBF neural network for predicting the output turn-on offset angle and turn-off offset angle. Use the prior information of the motor under various operating conditions to perform offline pre-training on the RBF neural network. The prior information includes the mechanical angular velocity of the motor, the load torque and the turn-on offset angle. i 1m and the off offset angle i 2m ;

[0010] (4) Use the real-time efficiency information of the motor to train the pre-trained RBF neural network online. At the same time, input the real-time mechanical angular velocity and load torque of the motor into the RBF neural network. Apply the turn-on offset angle and turn-off offset angle predicted by the RBF neural network based on the optimal efficiency to the QCSI time-sharing commutation control.

[0011] Furthermore, the time-sharing commutation mode in step (1) refers to the two switching transistors Q1 and Q2 involved in commutation in QCSI. When the conduction angle of each phase stator winding is greater than 120°, Q2 is turned on before Q1 is turned off, and then Q1 is turned off; the turn-on offset angle... i 1 and the off offset angle i The expression for 2 is as follows:

[0012]

[0013] in: The electric angular velocity of the motor. Compared to the Q2 launch time Time offset time interval (taking the leading time) Positive, lagging (negative) The Q1 shutdown time compared to Time interval of time offset (with lag) Positive, ahead of time (negative) , The time of intersection of the back electromotive force of phase Q1 and the back electromotive force of phase Q2.

[0014] Furthermore, in step (2), the switching-off angle based on the maximum electromagnetic power is obtained by performing offline numerical fast solution to the following set of equations. i 1m and the off offset angle i 2m ;

[0015]

[0016] in: This represents the average electromagnetic power of the motor during operation.

[0017] Furthermore, the average electromagnetic power The expression is as follows:

[0018]

[0019]

[0020] in: e A , e B , e C These are the three back electromotive forces of the motor. i A , i B , i C These are the three-phase stator currents of the motor. The electric angular velocity of the motor. t Indicates time, E This represents the amplitude of the back electromotive force.

[0021] Furthermore, the RBF neural network in step (3) consists of an input layer, a hidden layer, and an output layer. The input layer uses a two-dimensional vector composed of mechanical angular velocity and load torque. The hidden layer is designed with 5 neurons, and the numerical expression of each neuron is as follows:

[0022]

[0023] in: For the hidden layer n The value of each neuron. x For the input layer vector, For the first n A vector of center point parameters For the first n A width parameter, where e is a natural constant. n =1,2,3,4,5;

[0024] The output layer is designed with two neurons, whose values ​​correspond to the on-off offset angle and the off-off offset angle, as shown in the following expressions:

[0025]

[0026] in: For the output layer m The value of each neuron. v mn For the first m The first neuron corresponding to the first n One weight parameter, m =1,2.

[0027] Furthermore, the offline pre-training process in step (3) employs the following loss function. :

[0028]

[0029] in: and These are the on-off offset angle and off-off offset angle, respectively, of the predicted output of the RBF neural network.

[0030] Furthermore, the online training process in step (4) employs the following loss function. :

[0031]

[0032] in: and These represent the average input power and average output power of the motor system after applying the turn-on offset angle and turn-off offset angle predicted by the RBF neural network to the QCSI time-sharing commutation control.

[0033] Furthermore, in step (4), the turn-on offset angle and turn-off offset angle based on the optimal efficiency are applied to the QCSI time-sharing commutation control. That is, the turn-on offset angle and turn-off offset angle predicted by the RBF neural network are used to determine the turn-on time of switch Q2 and the turn-off time of switch Q1, thereby improving the motor operating efficiency.

[0034] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation.

[0035] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation.

[0036] Based on the above technical solution, the present invention has the following beneficial technical effects:

[0037] 1. In this invention, by establishing a time-sharing commutation mode, the turn-on offset angle and turn-off offset angle under maximum electromagnetic power are calculated. Through precise theoretical calculations and offline numerical solutions, and not limited to the 120° winding conduction mode, the specific time-sharing commutation angle for improving the electromagnetic power of the motor can be quickly obtained, thereby optimizing the operating performance of the high-speed brushless DC motor.

[0038] 2. The offline pre-training method based on RBF neural network in this invention can accelerate the convergence speed of neural network, accelerate the optimization speed of motor efficiency, and make the neural network output less prone to divergence, thus making the control system more stable.

[0039] 3. The online training method for the RBF neural network in this invention has strong real-time performance and can quickly adjust parameters based on the acquired motor efficiency information, which can effectively improve the efficiency of the motor control system. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the QCSI topology and time-sharing commutation control process of the motor system in an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of the opposite electromotive force and commutation signal during the time-division commutation process according to an embodiment of the present invention.

[0042] Figure 3This is a schematic diagram of the current flow direction at each stage of the commutation process of the QCSI equivalent circuit in the embodiment of the present invention. (a) corresponds to the steady-state stage before commutation, (b) corresponds to the early stage of commutation, (c) corresponds to the middle stage of commutation, (d) corresponds to the late stage of commutation under high speed conditions, and (e) corresponds to the late stage of commutation under low speed conditions. The faded parts in the figure represent non-conducting lines.

[0043] Figure 4 The switching-off angle based on the maximum electromagnetic power in this embodiment of the invention. i 1m With the off offset angle i 2m The calculation results are shown in the diagram, where (a) corresponds to the activation offset angle. i 1m The calculation result, (b) corresponds to the turn-off offset angle. i 2m Calculation results.

[0044] Figure 5 This is a schematic diagram of the RBF neural network structure used in an embodiment of the present invention. Detailed Implementation

[0045] To describe the present invention in more detail, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0046] This embodiment uses a unit with a phase resistance of 0.23Ω and a phase inductance of 9.2×10⁻⁶. -5 H, opposite potential coefficient k e For a high-speed brushless DC motor with a speed of 1.906 V / (kr / min), the DC power supply voltage of the control system is... U d The voltage is 150V, and the motor is subjected to dual closed-loop control of speed and current.

[0047] The quasi-current source inverter (QCSI) used in this embodiment has advantages such as direct control of winding current and built-in short-circuit protection. It is a better topology for realizing high-speed motor drive control under PAM mode. Its equivalent circuit is as follows: Figure 1 As shown in the diagram, the QCSI consists of two parts: a front-end chopper circuit and a back-end inverter circuit. The chopper circuit is composed of a power MOSFET S. w1 and S w2 Freewheeling diode D w1 D w2 And D1, constant current inductor L Composed of 0 components, flowing through the inductor L The current of 0 is i L The average value when the current is stable is I L The average current flowing through diode D1 isi D1 , R and L These are the phase resistance and phase inductance of the motor; e A , e B , e C The three opposite potentials, i A , i B , i C For three-phase stator current, i in The measured three-phase line voltage is the DC side output current. u AB , u BC , u CA The voltage at neutral point N u N and three-phase terminal voltage u A , u B , u C All reference points are taken from point O.

[0048] To optimize the control efficiency of the motor system, this embodiment provides a high-speed brushless DC motor system efficiency optimization control method based on time-sharing commutation. The specific implementation process is as follows:

[0049] (1) Establish the time-sharing commutation mode and determine the phase current variation law.

[0050] The motor system employs dual closed-loop control based on speed and current, with the given motor speed being... n ref The actual motor speed was measured to be n In the front-end chopper circuit, the power switch S... w2 Keep it off, by controlling the power switch S w1 Perform chopper control to make the current i L The speed loop PI (proportional-integral) controller outputs the setpoint. i L_ref S w1 When conducting, at voltage U d Current under action i L Gradually increase; when S w1 When turned off, the current i L via Dw2 Freewheeling, controlled by the power switch S w1 The constant current inductor current amplitude can be controlled by adjusting the duty cycle.

[0051] During motor commutation, the sequential conduction of each phase winding can be achieved by controlling the turn-on and turn-off times of the inverter bridge power transistors. It is stipulated that when the upper bridge arm power transistor S of phase A... AH When the circuit is turned on, the A-phase winding is forward-biased, denoted as A+; when the A-phase lower arm power transistor S... AL When conducting, phase A winding conducts negatively, denoted as A-; the definitions for phases B and C are similar. For example... Figure 2 As shown, taking the commutation process from A+B- to A+C- as an example, in the conventional fixed 120° conduction mode, based on the intersection of the back electromotive forces of phase B and phase C, i.e., the zero-crossing point of the line back electromotive force, phase B is simultaneously controlled to disconnect and phase C to open; specifically, in t BC At that moment, the power transistor S is turned off. BL At the same time, S was opened. CL When the conduction angle is less than 120°, the power transistor S... BL The turn-off time will be earlier than that of the power transistor S. CL At the turn-on time, the current in phase A will flow through diode D. BH The freewheeling current gradually decays. This single-phase decaying freewheeling current range reduces the effective value of the winding current, thereby reducing the average electromagnetic power within one electrical cycle; therefore, this situation is not considered. When the conduction angle is greater than 120°, S is turned on. CL The time will be earlier than the shutdown of S BL At that moment, phase C is already activated, but phase B has not yet been deactivated. For example... Figure 2 As shown, let S be... CL The opening time compared to t BC The offset time interval is t 1 (Taking the lead) t BC (Time is positive), S BL The timing of shutdown compared to t BC The offset time interval is t 2 (taking the lag) t BC (When the time is positive), during the commutation process, A+B- first changes to A+BC-, and then to A+C-. This commutation method, which controls the winding's opening and closing in a time-sharing manner, is called time-sharing commutation.

[0052] The offset angle for winding activation is i 1. The offset angle of the winding turn-off is i 2. Satisfies:

[0053]

[0054] in: oh e Let be the electric angular velocity of the motor.

[0055] To ensure that the conduction time of each phase is greater than 120°, t 1 and t 2. Must meet t 1+ t 2 > 0, and t 1. t 2 can be negative.

[0056] By adjusting the offset angle i 1. i Adjusting the offset angle 2 can significantly alter the transient process of phase current variation between stable values, allowing for adjustable range in both phase and amplitude of the fundamental phase current. In other words, by selecting the offset angle, the phase of the fundamental phase current can be aligned with the phase of the back EMF while simultaneously increasing the amplitude of the fundamental current, thereby obtaining greater electromagnetic power and further improving system operating efficiency.

[0057] Next, we will analyze the time-division commutation process of A+B- transforming into A+BC-, and then into A+C-, as an example. When A+B- is in the on state, the power transistor S... AH S BL When the circuit is turned on, the currents in phases A and B are in a stable state, and are respectively the inductor currents. I L - I L ,like Figure 3 As shown in (a). Let's define the moment when phase C begins to conduct as... Figure 2 At time 0 of the mid-time commutation, the power transistor S CL Open to S BL The turn-off period is defined as the early stage of time-sharing commutation, during which all three phase windings are conducting, such as... Figure 3 As shown in (b). In S BL After the circuit is turned off, there is a stage where the currents in both phase B and phase C windings have not reached a steady state, meaning that the current in phase B flows through diode D. BH Freewheeling current has not reached a steady-state value of 0, and the C-phase current has not reached a steady-state value. I L This stage is defined as the mid-term of time-sharing commutation, such as... Figure 3 As shown in (c). The period from when either phase B or C reaches a steady state until both phase B and C reach a steady state is defined as the later stage of time-sharing commutation.

[0058] During the early stage of commutation, the power transistor that is conducting is S. AH S BL S CLThe current in phase C starts to rise from 0, and current flows through all three phase windings. The turn-on time of phase C is defined as time 0. At this time, phases B and C are connected to the negative terminal of the DC side, and the terminal voltage... u B , u C When the voltage is 0 (ignoring the turn-on voltage drop of the MOSFET), the voltage equation satisfies:

[0059]

[0060] During this stage, the current in phase A is equal to the inductor current. I L Therefore, the currents in phases B and C satisfy... i B + i C =- I L .

[0061] To simplify the calculation, a Taylor expansion of the opposite potential at time 0 of the time-sharing commutation, retaining the first term, yields:

[0062]

[0063] in: E This represents the amplitude of the back electromotive force.

[0064] Combining the above equations, we can obtain the B and C phase currents during the early stage of time-sharing commutation (0≤ t < t 1+ t The expression for 2) is:

[0065]

[0066] During the commutation phase, the power transistor that is conducting is S. AH S CL When the power transistor S BL After being turned off, the B-phase current flows through the diode D connected in parallel to the upper bridge arm. BH Freewheeling, because the amplitude of the C-phase current is less than I L Inductor current I L A portion of the current will be shunted through D1, therefore diode D1 conducts. At this time, the voltages at phases A and B are clamped at... U d The voltage at phase C is 0, and the three-phase voltage equations satisfy:

[0067]

[0068] Since the switching of power transistors is not involved in the middle and later stages of commutation and the duration is relatively short, the three-phase back EMF can be considered as a constant during this period.t The two opposite potential values ​​are:

[0069]

[0070] in: i 2 is the turn-off offset angle, which satisfies i 2= oh e t 2.

[0071] Due to the duration of the early commutation phase t 1+ t 2 is relatively long, resistance R The impact on the current change process cannot be ignored. Furthermore, the commutation phases are relatively short in the middle and later stages. R The numerical value is relatively small, and its impact on current dynamics is relatively weak. For ease of analysis, resistance is ignored in the modeling of the mid-to-late stages of commutation. R The effect, combined with the above two equations, yields:

[0072]

[0073] Therefore, it can be seen that the absolute values ​​of the changes required for the B and C phase currents to reach their stable values ​​are equal. Thus, the order in which the B and C phase currents reach their stable values ​​can be classified according to the magnitude of the absolute values ​​of the change rates of the B and C phase currents, specifically into two operating conditions: high speed and low speed.

[0074] When operating at high speed, the B-phase current i B When the current reaches steady state, i.e., drops to 0, the C-phase current... i C Still has not reached a stable value I L ,have:

[0075]

[0076] Combining the above two equations, we can obtain E > U d / 3, since the magnitude of this reverse electromotive force is positively correlated with the motor speed, this operating condition can be called the high-speed operating condition.

[0077] Define Δ t a For power transistor S BL The time required for the current in phase B to drop to zero after the circuit is turned off is:

[0078]

[0079] When the B-phase current drops to 0, the C-phase current at this moment... i cfor:

[0080]

[0081] At this time, only the A-phase and C-phase windings have current flowing through them, and the voltage at the A-phase terminal is clamped. U d The voltage at the C-phase terminal is 0, such as Figure 3 As shown in (d), the voltage and current satisfy:

[0082]

[0083] Define Δ t b This is to ensure that the current in phases A and C reaches a stable value after the current in phase B decreases to a stable value. I L - I L The time required is:

[0084]

[0085] The analysis of the time-sharing commutation process under high-speed conditions is now complete. During the stable operation of the motor, the commutation process of the three-phase current is symmetrical. Combining the changes in the three-phase current during the time-sharing commutation period and the current values ​​when the three-phase current is stable, the specific expression of the three-phase current within one electrical cycle can be obtained.

[0086] Taking the half-cycle waveform of phase A as an example, based on the symmetry of the three phases, the phase A current rises from 0 to... I L During this period, its change process should correspond to the C-phase current decreasing from 0 to - I L The change is symmetrical; the current in phase A changes from... I L During the period of decreasing to 0, its change process should correspond to the above-mentioned B-phase current from - I L The process of rising to 0 is symmetrical.

[0087] Under high-speed operating conditions, the expression for the A-phase current takes the zero-crossing point of the A-phase back electromotive force as the time when it is 0. i A In the expression of the first half of the cycle (0≤ t <π / oh e (Time) is:

[0088]

[0089] in: t m1 =π / 6 oh e - t 1,t m2 =π / 6 oh e + t 2, t m3 =π / 6 oh e + t 2+D t a , t m4 =π / 6 oh e + t 2+D t a +D t b , t m5 =π / 2 oh e + t 2, t m6 =π / 2 oh e + t 2+D t a , t m7 =π / 2 oh e + t 2+D t a +D t b , t m8 =5π / 6 oh e - t 1, t m9 =5π / 6 oh e + t 2, t m10 =5π / 6 oh e + t 2+D t a , t m11 =π / oh e ; k 1~ k 5 meet:

[0090]

[0091] The waveform of phase A current is symmetrical about the time axis in both the first and second halves of the cycle. The currents of phases B and C can be obtained by lagging phase A current by 120° and 240° electrical degrees, respectively.

[0092] The analysis method is similar under low-speed conditions, and will not be elaborated here. The current flow direction in the later stage of commutation is as follows: Figure 3 As shown in (e).

[0093] (2) Obtain the switching-off angle based on the maximum electromagnetic power using offline numerical solution. i 1m and the off offset angle i 2m .

[0094] The three-phase back electromotive force of a high-speed motor is close to a sine wave. e A , e B , e C satisfy:

[0095]

[0096] The average electromagnetic power of the high-speed brushless DC motor during operation is calculated based on the phase current expression and the back electromotive force as follows:

[0097]

[0098] in: e A , e B , e C The three back EMFs of the motor, i A , i B , i C It is a three-phase current.

[0099] Solve the equation as follows:

[0100]

[0101] By utilizing offline numerical methods for rapid solution, the switching-off angle based on the maximum electromagnetic power can be quickly obtained. i 1m and the off offset angle i 2m Calculated value. Figure 4 Give motor speed n The range is 20,000~40,000 rpm, and the inductor current is... I LWhen the range is 4~12A, the turn-on offset angle obtained based on the maximum electromagnetic power i 1m and the off offset angle i 2m . i 1m The calculated results range from 2 to 12°. i 2m The calculation results range from -1.8° to 0°. i 1m The changes are quite significant depending on the operating conditions, while i 2m The variation with operating conditions is relatively small. Since the operating point of the motor's maximum electromagnetic power is close to the operating point of the system's optimal efficiency, a suitable optimization range can be selected based on the characteristics of its variation range during subsequent optimization.

[0102] (3) Design the structure of the input layer, hidden layer and output layer of the RBF neural network.

[0103] RBF neural networks are used to optimize the turn-on offset angle and turn-off offset angle, thereby optimizing the control efficiency of the motor system. Their structure is as follows: Figure 5 As shown. The input layer is designed with the operating conditions of a high-speed brushless DC motor, i.e., the mechanical angular velocity. oh m Load torque T m The hidden layer is designed with 5 neurons and is a Gaussian function, and its expression is as follows:

[0104]

[0105] in: h n For the hidden layer n The value of each neuron. x For the input layer vector, For the first n A vector of center point parameters For the first n A width parameter, where e is a natural constant. n It is a natural number and 1 ≤ n ≤5.

[0106] The output layer is designed for time-division commutation and uses the following turn-on offset angle. i 1 and the off offset angle i 2. The linear relationship with the hidden layer is as follows:

[0107]

[0108] in: y m For the output layer mThe value of each neuron. v mn For the first m The corresponding output neuron of the first n One weight parameter, m It is a natural number and 1 ≤ m ≤2.

[0109] (4) The RBF neural network is pre-trained offline using prior information.

[0110] Prior information refers to the calculated activation offset angle. i 1m and the off offset angle i 2m During offline pre-training, the loss function of the neural network is designed as follows:

[0111]

[0112] in: This represents the loss function during offline pre-training. , These are the output values ​​of the RBF neural network, namely the turn-on offset angle and turn-off offset angle used for time-division commutation control.

[0113] The parameter update method is as follows:

[0114]

[0115] in: In the offline pre-training stage, corresponding to Online training correspondence , For the updated number m The corresponding output neuron of the first n One weight parameter, For the first m The corresponding output neuron of the first n The error value of each weight parameter, For the updated number n A vector of center point parameters For the updated number n The error value of the parameter vector of each center point. For the updated number n One width parameter, For the updated number n The error value of the width parameter, k w , k d , k σ For a given learning rate, this embodiment kw , k d , k σ All are 0.01.

[0116] During pre-training, parameter adjustments reduce the loss function, thus affecting the neural network's output. i 1. i 2 will gradually approach i 1m and i 2m After training, the output of the RBF neural network has been adjusted to... i 1m and i 2m The parameters at this point are used as the initial state for subsequent online optimization, accelerating the convergence speed of the neural network during online training and preventing the neural network output from diverging. This is because the solved on-off offset angle and off-off offset angle... i 1m , i 2m They are based on maximum electromagnetic power, not motor efficiency, so they can only ensure minimum copper loss and cannot take into account the impact of iron loss and other types of loss on overall efficiency.

[0117] Online training updates network parameters by acquiring real-time efficiency information, enabling the network to output the optimal activation offset angle. i 1m_best and the off offset angle i 2m_best This minimizes total losses, including copper losses, iron losses, and other types of losses, effectively improving motor operating efficiency.

[0118] (5) Train the RBF neural network online.

[0119] During online training, the loss function for the neural network is designed as follows:

[0120]

[0121] in: This represents the loss function during online training. Input average power to the system, The neural network parameter update rules are the same as those used during offline pre-training, to determine the system's average output power.

[0122] Figure 4 The calculation results cover actual working conditions, and i 1m , i 2m and i1m_best , i 2m_best Since the values ​​are relatively close, a 150% margin can be considered when optimizing online. i 1m_best The optimization range is defined as [0°, 18°]. i 2m_best The optimization range is defined as [-2.5°, 0.5°], and at the same time... i 1m_best + i 2m_best Conditions greater than 0.

[0123] During online training, the loss function is calculated by acquiring the real-time operating status of the motor system. As the neural network parameters are adjusted, the loss function gradually decreases, indicating improved system efficiency. Gradually increasing the size optimizes motor efficiency.

[0124] Based on the aforementioned RBF neural network optimization model, the optimal activation offset angle can be iteratively calculated. i 1m_best and the off offset angle i 2m_best The optimal offset angle obtained in this embodiment i 1m_best , i 2m_best The system is based on accurately capturing the moment of the intersection of the opposite electromotive forces (i.e., the zero-crossing point of the line back EMF). However, in practice, when using back EMF to detect commutation signals, a phase delay at the zero-crossing point of the line back EMF occurs due to filtering delays and hardware / software delays. The detection circuit based on the zero-crossing point of the line back EMF samples the three-phase terminal voltages, which are then processed by filtering, comparison, and isolation circuits to obtain signals for speed and rotor position estimation. H 1. H 2 and H 3. The sum of the hardware delay time caused by each circuit part and the software delay time caused by the sampling and control program is denoted as... t d The corresponding commutation lag angle caused by hardware and software delays is β = oh e t d The commutation hysteresis angle caused by LPF is α It can be calculated from the motor speed and sampling circuit parameters.

[0125] Taking into account the effects of the above delays, the final lead angle used to control the motor's on / off states is (based on the detection of the zero-crossing point):

[0126]

[0127] After the system stabilizes, an efficiency optimization strategy is implemented for control. The motor operates in a speed-current dual closed-loop state, and the signal is obtained by detecting the zero-crossing point of the back EMF. H 1. H 2. H 3. H 1. H 2. H 3. Delay angle α + β With the optimal offset angle i 1m_best and i 2m_best Combined, as a commutation signal, it is used to control the switching on and off of the power transistors in the three-phase inverter bridge, thereby realizing time-sharing commutation control of the high-speed brushless DC motor.

[0128] We used the above motor parameters for experimental verification: First, based on the current change under the control of the quasi-current source inverter, we calculated the turn-on offset angle and turn-off offset angle based on the maximum electromagnetic power offline; then, we pre-trained the RBF neural network offline to obtain the initial parameters and initial output for online training; finally, by acquiring the input current, input voltage, motor speed, and motor load torque of the motor system online, we established the loss function for online training, adjusted the network parameters online, and thus optimized the motor efficiency. Table 1 compares the efficiency of the motor system at a speed of 40,000 rpm and an average inductor current of 10 A, under the following conditions: 120° fixed conduction mode of the winding, after time-sharing commutation pre-training, and after online training.

[0129] Table 1

[0130]

[0131] As can be seen, compared with the 120° fixed conduction mode of the winding, the efficiency optimization control method of high-speed brushless DC motor system based on time-sharing commutation adopted in this embodiment effectively improves the efficiency of the motor system, from the original 91.77% to 93.91%.

[0132] In summary, the efficiency optimization control of the high-speed brushless DC motor system based on time-sharing commutation of this invention can effectively improve the operating efficiency of the motor drive system, thereby improving the motor's operating performance and meeting the needs of various high-efficiency applications.

[0133] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. Those skilled in the art can readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. A method for efficiency optimization control of a high-speed brushless DC motor system based on time-sharing commutation, characterized in that, Includes the following steps: (1) Establish the time-sharing commutation mode of QCSI in the motor system and define the turn-on offset angle during the commutation process. θ 1 and the off offset angle θ 2. QCSI is a quasi-current source inverter; The time-sharing commutation mode refers to the two switching transistors Q1 and Q2 involved in commutation in QCSI. When the conduction angle of each phase stator winding is greater than 120°, Q2 is turned on before Q1 is turned off, and then Q1 is turned off; the turn-on offset angle... θ 1 and the off offset angle θ The expression for 2 is as follows: in: The electric angular velocity of the motor. Compared to the Q2 launch time The time interval of the time offset The Q1 shutdown time compared to The time interval of the time offset The time of intersection of the back electromotive force of phase Q1 and the back electromotive force of phase Q2; (2) Obtain the switching-off angle based on the maximum electromagnetic power using offline numerical solution. θ 1m and the off offset angle θ 2m ; (3) Construct an RBF neural network for predicting the output turn-on offset angle and turn-off offset angle. Use the prior information of the motor under various operating conditions to perform offline pre-training on the RBF neural network. The prior information includes the mechanical angular velocity of the motor, the load torque and the turn-on offset angle. θ 1m and the off offset angle θ 2m ; (4) Use the real-time efficiency information of the motor to train the pre-trained RBF neural network online. At the same time, input the real-time mechanical angular velocity and load torque of the motor into the RBF neural network. Apply the turn-on offset angle and turn-off offset angle predicted by the RBF neural network based on the optimal efficiency to the QCSI time-sharing commutation control.

2. The efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in claim 1, characterized in that: In step (2), the switching-off offset angle based on the maximum electromagnetic power is obtained by rapidly solving the following set of equations offline numerically. θ 1m and the off offset angle θ 2m ; in: This represents the average electromagnetic power of the motor during operation.

3. The efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in claim 2, characterized in that: The average electromagnetic power The expression is as follows: in: e A , e B , e C These are the three back electromotive forces of the motor. i A , i B , i C These are the three-phase stator currents of the motor. t Indicates time, E This represents the amplitude of the back electromotive force.

4. The efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in claim 1, characterized in that: The RBF neural network in step (3) consists of an input layer, a hidden layer, and an output layer. The input layer uses a two-dimensional vector composed of mechanical angular velocity and load torque. The hidden layer is designed with 5 neurons, and the numerical expression of each neuron is as follows: in: For the hidden layer n The value of each neuron. x For the input layer vector, For the first n A vector of center point parameters For the first n A width parameter, where e is a natural constant. n =1,2,3,4,5; The output layer is designed with two neurons, whose values ​​correspond to the on-off offset angle and the off-off offset angle, as shown in the following expressions: in: For the output layer m The value of each neuron. v mn For the first m The first neuron corresponding to the first n One weight parameter, m =1,2.

5. The efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in claim 1, characterized in that, The offline pre-training process in step (3) uses the following loss function. : in: and These are the on-off offset angle and off-off offset angle, respectively, of the predicted output of the RBF neural network.

6. The efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in claim 1, characterized in that, The online training process in step (4) employs the following loss function. : in: and These represent the average input power and average output power of the motor system after applying the turn-on offset angle and turn-off offset angle predicted by the RBF neural network to the QCSI time-sharing commutation control.

7. The efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in claim 1, characterized in that: In step (4), the turn-on offset angle and turn-off offset angle based on the optimal efficiency are applied to the QCSI time-sharing commutation control. That is, the turn-on offset angle and turn-off offset angle predicted by the RBF neural network are used to determine the turn-on time of switch Q2 and the turn-off time of switch Q1, thereby improving the motor operating efficiency.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: The processor is used to execute the computer program to implement the efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements the efficiency optimization control method for a high-speed brushless DC motor system based on time-sharing commutation as described in any one of claims 1 to 7.

Images