A control system and method for online correction of reference torque of a switched reluctance motor

Abstract

The present application relates to the technical field of switched reluctance motor control, and particularly relates to a control system and method for online correction of reference torque of a switched reluctance motor, which directly adjusts the output of a neural network with a torque error by online correction of a reference torque value, thereby omitting a proportional differential controller and reducing calculation amount. Compared with the prior art, the present application has the following advantages: according to the periodic relationship between the reference torque value of the SRM and the phase current, a function conforming to the variation law of the phase current of the SRM is used as an activation function of the hidden layer of the neural network, and a neural network for online correction of the reference torque is designed to conform to the strong nonlinear characteristics of the SRM; the difference between the reference torque value and the instantaneous torque value is used as the basis for adjustment, and the weight coefficients and the hidden layer coefficients of the neural network are adjusted online to correct the reference torque value, thereby effectively suppressing torque ripple; and the calculation amount is low, the system is embedded, and is suitable for online control of the SRM.

Description

Technical Field

[0001] This invention relates to the field of switched reluctance motor control technology, and specifically to a control system and method for online correction of the reference torque of a switched reluctance motor. Background Technology

[0002] Switched reluctance motors (SRMs) do not require rare metals for manufacturing, and their rotors have no windings or permanent magnets. They have advantages such as simple structure, low manufacturing cost, and wide speed range. Their main switching devices are directly connected in series with the phase windings, avoiding the possibility of short circuits and breakdowns in the drive circuit, thus providing extremely high safety. As a result, they have received great attention in today's society, which is more focused on green environmental protection, and are expected to become the preferred power source for the next generation of new energy vehicles.

[0003] The high torque ripple of the SRM limits its application. To reduce torque ripple, common control methods include Direct Instantaneous Torque Control (DITC) and Torque Sharing Function (TSF). DITC uses torque as the control target, achieving torque hysteresis control. However, this method is not easy to control the current shape, easily resulting in large current spikes and an unsmooth current flow. TSF is another commonly used method. By distributing a reference torque to each phase of the SRM, and during the SRM's commutation process, comparing the real-time current with the distributed current, a current hysteresis loop is controlled. This achieves the goal of controlling electromagnetic torque and speed by controlling the current, thus reducing torque ripple during current smoothing.

[0004] The invention patent with publication number CN110022109A discloses a torque-current neural network model SRM torque ripple control method and system. It designs a method that combines feedforward and error compensation. The feedforward path obtains the reference current by adjusting the reference torque through the neural network. The error compensation calculates the error using the reference torque value and the instantaneous torque value. The obtained error is sent to the proportional-derivative controller. At the same time, the output of the proportional-derivative controller serves as the basis for neural network adjustment.

[0005] This method reduces torque ripple by adjusting the reference current, but it still suffers from the drawback of high computational complexity. Summary of the Invention

[0006] The purpose of this invention is to provide a control system and method for online correction of reference torque of a switched reluctance motor. By adjusting the output of a torque error correction neural network, the reference torque value is ultimately adjusted to reduce torque ripple.

[0007] To achieve the above objectives, the present invention provides a control system for online correction of the reference torque of a switched reluctance motor, comprising a PI controller, a neural network regulator, a torque distribution module, a torque-to-current conversion module, a current hysteresis module, a power conversion circuit, a switched reluctance motor, an encoder, a torque meter, and a speed calculation module. The PI controller, the neural network regulator, the torque distribution module, and the torque-to-current conversion module are connected sequentially. The output terminal of the torque-to-current conversion module is connected to the input terminal of the current hysteresis module via the feedback current of the switched reluctance motor. The output terminal of the current hysteresis module is connected to the input terminal of the power conversion circuit. The output terminal of the power conversion circuit is connected to the input terminal of the switched reluctance motor. The switched reluctance motor, the encoder, and the speed calculation module are connected sequentially.

[0008] In addition to the PI controller's output being directly connected to the neural network regulator's input, the torque deviation, obtained by subtracting the PI controller's output from the torque meter's output, is also connected to the neural network regulator's input.

[0009] The PI controller is used to process the speed error and its rate of change, and to adjust the output reference torque value through PI adjustment.

[0010] The neural network regulator is used to correct the reference torque value online based on the current rotor angle and torque error, so as to obtain the corrected reference torque value.

[0011] The torque distribution module is used to distribute the corrected reference torque value to each phase to obtain the reference torque value of each phase, and the distribution is based on the current rotor angle.

[0012] The torque-to-current conversion module is used to convert the reference torque value of each phase into a reference phase current.

[0013] The current hysteresis module is used to compare the given phase current with the feedback phase current to calculate the drive signal sent to the power conversion circuit.

[0014] The power conversion circuit is used to control the motor operation by applying voltage to each phase winding or disconnecting the voltage on the winding;

[0015] The encoder is used to detect the position of the motor rotor;

[0016] The torque table is used to look up the current actual output torque value, i.e., the instantaneous torque value, based on the input current motor rotor position and current phase current.

[0017] The aforementioned speed calculation module is used to calculate the rotor speed based on the motor rotor position measured by the encoder.

[0018] The neural network regulator has three layers: an input layer, a hidden layer, and an output layer. The inputs are the rotor angle θ and the reference torque value T. ref The output is the corrected reference torque value T. mod The value of the j-th hidden function of the hidden layer at time k is

[0019]

[0020] Among them, c j (k) is the value of the j-th center point at time k, b j (k) represents the width of the j-th implicit function at time k, ||θ-c j (k)|| is the rotor angle and c j (k) center distance.

[0021] The neural network learning method employs gradient descent, and the performance index function is selected as follows:

[0022]

[0023] In the formula, T ref For reference torque value, T ins This represents the instantaneous torque value; the goal of the neural network is to approximate and track T. ins ,

[0024]

[0025] The weight iteration algorithm is as follows:

[0026]

[0027] In the formula, δ is the learning rate;

[0028]

[0029] In the formula, α is the momentum factor, and w j (k), w j (k-1), w j (k-2) represent the weighted value at time k of the j-th hidden node, the weighted value at the time before time k of the j-th hidden node, and the weighted value at the time before time k-1 of the j-th hidden node, respectively, Δw j (k) is w j The increment of (k);

[0030] c j (k), c j (k-1), c j(k-2) represent the center value of the j-th hidden node at time k, the center value of the j-th hidden node before time k, and the center value of the j-th hidden node before time k-1, respectively, Δc j (k) is c j The increment of (k);

[0031] b j (k), b j (k-1), b j (k-2) represent the width value of the j-th hidden node at time k, the width value of the j-th hidden node at time k before time k, and the width value of the j-th hidden node at time k-1 before time k, respectively. Δb j (k) is b j The increment of (k);

[0032] Corrected reference torque value T mod for

[0033]

[0034] In the formula, N is the number of hidden layer nodes in the neural network, and its value ranges from 5 to 9.

[0035] In this case, the current of the switched reluctance motor changes periodically and is successively phased by a step angle θ. step The input layer first preprocesses θ to make it m*θ step The period is in the interval [0, m*θ] step [Internal changes, where m is the phase number of the SRM, initializing the implicit function center value]

[0036] c j (0)=θ step *floor(θ / θ step )+θ step / 2

[0037] Where c j (0) represents the initial time c. j The value, floor() means rounding down, θ step Let θ be the step angle. step Defined by the following formula: θ step =360 / (m*N) r ), N r This represents the number of rotors in a switched reluctance motor.

[0038] The torque-to-current conversion module uses a linear calculation method to convert torque to current, and the formula is as follows:

[0039]

[0040] Where m is the number of phases of the switched reluctance motor; when the switched reluctance motor has 3 phases, m = 1, 2, 3; i m Let T be the reference current for phase m. m is the reference torque value for phase m, and k is the linearization coefficient of the inductor partial conductance, with a value ranging from 0.05 to 1.

[0041] This invention also proposes a control method for online correction of the reference torque of a switched reluctance motor. The control system using the aforementioned online correction of the reference torque of the switched reluctance motor includes the following steps:

[0042] The speed error is obtained by subtracting the given speed from the real-time speed.

[0043] The speed error is processed by PI calculation to obtain the reference torque value;

[0044] The reference torque value and rotor angle are input into the neural network regulator to obtain the corrected reference torque value;

[0045] The rotor position angle is obtained from the output of the switched reluctance motor, and the instantaneous torque value is obtained by looking up a table in combination with the feedback phase current.

[0046] The torque difference is obtained by subtracting the reference torque value from the instantaneous torque value.

[0047] Adjust the implicit function coefficients and weight coefficients of the neural network based on the torque difference;

[0048] Based on the rotor angle, the corrected reference torque value is distributed to obtain the phase torque value;

[0049] The phase torque value is converted into current to obtain the reference phase current value;

[0050] The difference between the reference phase current and the feedback phase current is calculated, and the on / off state or freewheeling of the drive circuit is controlled based on the calculation result.

[0051] The rotor position angle is calculated to obtain the real-time speed of the motor.

[0052] This invention provides a control system and method for online correction of the reference torque of a switched reluctance motor (SRM). The system includes a PI controller, a neural network regulator, a torque distribution module, a torque-to-current conversion module, a current hysteresis module, a power conversion circuit, a switched reluctance motor, an encoder, a torque meter, and a speed calculation module. By directly adjusting the output of the neural network using the torque error through online correction of the reference torque value, the proportional-derivative controller is eliminated, reducing computational load. Compared with existing technologies, the advantages of this invention are: based on the periodic relationship between the SRM reference torque value and the phase current, a function conforming to the SRM phase current variation law is used as the activation function of the hidden layer of the neural network, designing a neural network for online correction of the reference torque that conforms to the strong nonlinear characteristics of the SRM; the difference between the reference torque value and the instantaneous torque value is used as the adjustment basis to adjust the weight coefficients and hidden layer coefficients of the neural network online, correcting the reference torque value and effectively suppressing torque ripple; simultaneously, the low computational load allows for the construction of an embedded system suitable for online control of SRMs. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a schematic diagram of the control system structure for online correction of reference torque of a switched reluctance motor according to the present invention.

[0055] Figure 2 This is a schematic diagram of the neural network regulator principle of the present invention.

[0056] Figure 3 This is a schematic diagram of the power conversion circuit of the present invention.

[0057] Figure 4 This is a flowchart of the control method for online correction of reference torque of a switched reluctance motor according to the present invention.

[0058] Figure 5 This is a schematic diagram of the control system structure for the switched reluctance motor reference torque of the present invention, where no online correction is performed.

[0059] Figure 6 This is a schematic diagram of the phase current and output torque pulsation of the present invention when the switched reluctance motor speed is 300 rpm and the load torque is 2 Nm in a specific embodiment of the present invention.

[0060] Figure 7This is a schematic diagram of phase current and output torque pulsation in a specific embodiment of the present invention when the switched reluctance motor speed is 300 rpm and the load torque is 2 Nm, and the reference torque is not adjusted and controlled online.

[0061] Figure 8 This is a schematic diagram of the phase current and output torque pulsation of the present invention when the switched reluctance motor speed is 700 rpm and the load torque is 5 Nm in a specific embodiment of the present invention.

[0062] Figure 9 This is a schematic diagram of the phase current and output torque pulsation when the switched reluctance motor speed is 700 rpm and the load torque is 5 Nm in a specific embodiment of the present invention, without online adjustment control of the reference torque. Detailed Implementation

[0063] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0064] Please see Figures 1 to 3 This invention provides a control system for online correction of the reference torque of a switched reluctance motor, comprising a PI controller, a neural network regulator, a torque distribution module, a torque-to-current conversion module, a current hysteresis module, a power conversion circuit, a switched reluctance motor, an encoder, a torque meter, and a speed calculation module. The PI controller, the neural network regulator, the torque distribution module, and the torque-to-current conversion module are connected sequentially. The output of the torque-to-current conversion module is connected to the input of the current hysteresis module via the feedback current of the switched reluctance motor. The output of the current hysteresis module is connected to the input of the power conversion circuit. The output of the power conversion circuit is connected to the input of the switched reluctance motor. The output of the switched reluctance motor is connected to the input of the encoder. The encoder outputs a rotor angle θ, which is connected to the torque distribution module, the torque meter, and the speed calculation module, respectively.

[0065] In addition to the PI controller's output being directly connected to the neural network regulator's input, the torque deviation, obtained by subtracting the PI controller's output from the torque meter's output, is also connected to the neural network regulator's input.

[0066] The PI controller is used to process the speed error and its rate of change, and to adjust the output reference torque value through PI adjustment.

[0067] The neural network regulator is used to correct the reference torque value online based on the current rotor angle and torque error, so as to obtain the corrected reference torque value.

[0068] The torque distribution module is used to distribute the corrected reference torque value to each phase to obtain the reference torque value of each phase, and the distribution is based on the current rotor angle.

[0069] The torque-to-current conversion module is used to convert the reference torque value of each phase into a reference phase current.

[0070] The current hysteresis module is used to compare the given phase current with the feedback phase current to calculate the drive signal sent to the power conversion circuit.

[0071] The power conversion circuit is used to control the motor operation by applying voltage to each phase winding or disconnecting the voltage on the winding;

[0072] The encoder is used to detect the position of the motor rotor;

[0073] The torque table is used to look up the current actual output torque value, i.e., the instantaneous torque value, based on the input current motor rotor position and current phase current.

[0074] The aforementioned speed calculation module is used to calculate the rotor speed based on the motor rotor position measured by the encoder.

[0075] like Figure 2 As shown, the neural network regulator has three layers: an input layer, a hidden layer, and an output layer. The inputs are the rotor angle θ and the reference torque value T. ref The output is the corrected reference torque value T. mod The value of the j-th hidden function of the hidden layer at time k is

[0076]

[0077] Among them, c j (k) is the value of the j-th center point at time k, b j (k) represents the width of the j-th implicit function at time k, ||θ-c j (k)|| is the rotor angle and c j (k) center distance.

[0078] The current of a switched reluctance motor (SRM) changes periodically, with each phase differing by a step angle θ. step The input layer first preprocesses θ to make it m*θ step The period is in the interval [0, m*θ] step [Internal changes, where m is the phase number of the SRM, initializing the implicit function center value]

[0079] c j (0)=θ step *floor(θ / θ step)+θ step / 2

[0080] Where c j (0) represents the initial time c. j The value, floor() means rounding down, θ step Let θ be the step angle. step Defined by the following formula: θ step =360 / (m*N) r ), N r This represents the number of rotors in a switched reluctance motor.

[0081] Neural network learning uses the gradient descent method, and the performance index function is selected as follows:

[0082]

[0083] In the formula, T ref For reference torque value, T ins This represents the instantaneous torque value; the goal of the neural network is to approximate and track T. ins When the deviation approaches 0, the output of the neural network equals the instantaneous torque value T. ins ,

[0084]

[0085] The weight iteration algorithm is as follows:

[0086]

[0087] In the formula, δ is the learning rate;

[0088]

[0089] In the formula, α is the momentum factor, and w j (k), w j (k-1), w j (k-2) represent the weighted value at time k of the j-th hidden node, the weighted value at the time before time k of the j-th hidden node, and the weighted value at the time before time k-1 of the j-th hidden node, respectively, Δw j (k) is w j The increment of (k);

[0090] c j (k), c j (k-1), c j (k-2) represent the center value of the j-th hidden node at time k, the center value of the j-th hidden node before time k, and the center value of the j-th hidden node before time k-1, respectively, Δc j (k) is c j The increment of (k);

[0091] b j (k), b j (k-1), b j (k-2) represent the width value of the j-th hidden node at time k, the width value of the j-th hidden node at time k before time k, and the width value of the j-th hidden node at time k-1 before time k, respectively. Δb j (k) is b j The increment of (k).

[0092] Corrected reference torque value T mod for

[0093]

[0094] In the formula, N is the number of hidden layer nodes in the neural network, and its value ranges from 5 to 9.

[0095] Furthermore, the current distribution module uses a cubic distribution method, with the specific formula as follows:

[0096] θ on Let θ be the conduction angle. off For the shut-off angle, θ ov τ is the overlap angle. t It is a periodicity.

[0097]

[0098] The cubic allocation method has a smoother transition effect compared to other allocation methods, which helps to reduce torque pulsation during commutation.

[0099] Furthermore, in the torque-to-current conversion module, the torque-to-current calculation adopts a linear calculation method, and the formula is as follows:

[0100]

[0101] Where m is the number of phases of the switched reluctance motor; when the switched reluctance motor has 3 phases, m = 1, 2, 3; i m Let T be the reference current for phase m. m is the reference torque value for phase m, and k is the linearization coefficient of the inductor partial conductance, with a value ranging from 0.05 to 1.

[0102] The advantages of the linear calculation method are: based on the torque-current conversion formula, the coefficient k is obtained by linearizing the inductance bias line, which simplifies the calculation, and the resulting error is corrected by adjusting the reference torque value through a neural network.

[0103] Please see Figure 3The power conversion circuit adopts an asymmetrical half-bridge drive circuit. The current hysteresis module compares the instantaneous phase current of the SRM with the reference phase current and calculates the phase current error. After limiting calculation, it determines whether each phase winding of the switched reluctance motor is working in the single-phase conduction region or the commutation conduction region based on the SRM rotor angle. Then, it determines the "+1", "0", and "-1" switching states of the asymmetrical half-bridge drive circuit. By controlling the voltage of each phase winding, it controls the phase current of the SRM.

[0104] Hysteresis control can reduce the frequency of current switching control and avoid power loss caused by frequent switching of power devices.

[0105] like Figure 4 As shown, this invention also proposes a control method for online correction of the reference torque of a switched reluctance motor. The control system using the aforementioned online correction of the reference torque of the switched reluctance motor includes the following steps:

[0106] The speed error is obtained by subtracting the given speed from the real-time speed.

[0107] The speed error is processed by PI calculation to obtain the reference torque value;

[0108] The reference torque value and rotor angle are input into the neural network regulator to obtain the corrected reference torque value;

[0109] The rotor position angle is obtained from the output of the switched reluctance motor, and the instantaneous torque value is obtained by looking up a table in combination with the feedback phase current.

[0110] The torque difference is obtained by subtracting the reference torque value from the instantaneous torque value.

[0111] Adjust the implicit function coefficients and weight coefficients of the neural network based on the torque difference;

[0112] Based on the rotor angle, the corrected reference torque value is distributed to obtain the phase torque value;

[0113] The phase torque value is converted into current to obtain the reference phase current value;

[0114] The difference between the reference phase current and the feedback phase current is calculated, and the on / off state or freewheeling of the drive circuit is controlled based on the calculation result.

[0115] The rotor position angle is calculated to obtain the real-time speed of the motor.

[0116] Furthermore, the present invention also proposes a specific embodiment, which is illustrated through simulation experiments and comparison with a reference torque control system for a switched reluctance motor without online correction:

[0117] Please see Figures 5 to 9This specific embodiment is based on simulation experiments on the Matlab / Simulink platform. The parameters of the switched reluctance motor are as follows: number of phases of SRM is 3, number of stator poles is 12, number of rotor poles is 8, DC bus voltage is 60V, moment of inertia is 0.0082kg·mm, and friction coefficient is 0.01N·ms. The simulation window time is set to 0.025s. Since this method only improves the inner loop of current compensation, it will not affect the speed control effect. The simulation results retain the last 0.025s graph.

[0118] A neural network regulator is used to correct the reference torque value online. The adjusted reference torque value is then input into the torque distribution module. The resulting graph is shown below: Figure 6 This is a schematic diagram of the phase current and output torque pulsation of the present invention when the SRM speed is 300 rpm and the load torque is 2 Nm. Figure 8 This is a schematic diagram of the phase current and output torque pulsation of the present invention when the SRM speed is 700 rpm and the load torque is 5 Nm.

[0119] like Figure 5 As shown, no correction is applied to the reference torque value, which is equivalent to removing the neural network regulator. The output of the PI controller is directly connected to the input of the torque distribution module, and the resulting effect is illustrated in the diagram below: Figure 7 This is a schematic diagram of phase current and output torque ripple when the SRM speed is 300 rpm and the load torque is 2 Nm. Figure 9 This is a schematic diagram of phase current and output torque pulsation when the SRM speed is 700 rpm and the load torque is 5 Nm.

[0120] By comparing the torque ripple curves under the same conditions, it can be seen that the method of the present invention can effectively suppress the torque ripple of the switched reluctance motor by making online corrections to the reference torque value through a neural network regulator.

[0121] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A control system for online correction of reference torque of a switched reluctance motor, characterized in that, The system includes a PI controller, a neural network regulator, a torque distribution module, a torque-to-current conversion module, a current hysteresis module, a power conversion circuit, a switched reluctance motor, an encoder, a torque meter, and a speed calculation module. The PI controller, the neural network regulator, the torque distribution module, and the torque-to-current conversion module are connected in sequence. The output of the torque-to-current conversion module is connected to the input of the current hysteresis module via the feedback current of the switched reluctance motor. The output of the current hysteresis module is connected to the input of the power conversion circuit. The output of the power conversion circuit is connected to the input of the switched reluctance motor. The output of the switched reluctance motor is connected to the input of the encoder. The encoder outputs a rotor angle θ, which is connected to the torque distribution module, the torque meter, and the speed calculation module, respectively. The PI controller is used to process the speed error and its rate of change, and to adjust the output reference torque value through PI regulation. The neural network regulator is used to correct the reference torque value online based on the current rotor angle and torque error, so as to obtain the corrected reference torque value. The torque distribution module is used to distribute the corrected reference torque value to each phase to obtain the reference torque value of each phase, and the distribution is based on the current rotor angle. The torque-to-current conversion module is used to convert the reference torque value of each phase into a reference phase current. The current hysteresis module is used to compare the given phase current with the feedback phase current to calculate the drive signal sent to the power conversion circuit. The power conversion circuit is used to control the motor operation by applying voltage to each phase winding or disconnecting the voltage on the winding; The encoder is used to detect the position of the motor rotor; The torque table is used to look up the current actual output torque value, i.e., the instantaneous torque value, based on the input current motor rotor position and current phase current. The aforementioned speed calculation module is used to calculate the rotor speed based on the motor rotor position measured by the encoder. The neural network regulator has three layers: an input layer, a hidden layer, and an output layer. The input is the rotor angle. and reference torque value The output is the corrected reference torque value. The value of the j-th hidden function of the hidden layer at time k is in, Let j be the value of the j-th center point at time k. Let the width of the j-th implicit function at time k be . For rotor angle and The center distance.

2. The control system for online correction of reference torque of a switched reluctance motor as described in claim 1, characterized in that, Neural network learning uses the gradient descent method, and the performance index function is selected as follows: In the formula, For reference torque value, The instantaneous torque value; the goal of the neural network is to approximate and track it. , The weight iteration algorithm is as follows: In the formula, For learning rate; In the formula, Momentum factor , , Let represent the weighted value at time k of the j-th hidden node, the weighted value at the time preceding time k of the j-th hidden node, and the weighted value at the time preceding time k-1 of the j-th hidden node, respectively. yes The increment; , , Let represent the center value of the j-th hidden node at time k, the center value of the j-th hidden node at time k before time k, and the center value of the j-th hidden node at time k-1 before time k, respectively. yes The increment; , , Let represent the width value of the j-th hidden node at time k, the width value of the j-th hidden node at time k before time k, and the width value of the j-th hidden node at time k-1 before time k, respectively. yes The increment; Corrected reference torque value for In the formula, N is the number of hidden layer nodes in the neural network, and its value ranges from 5 to 9.

3. The control system for online correction of reference torque of a switched reluctance motor as described in claim 2, characterized in that, The current of the switched reluctance motor changes periodically, with each phase differing by a step angle. The input layer first preprocesses θ to make it... The period is in the interval [0, ] [Internal changes, where m is the phase number of the SRM, initializing the implicit function center value] in Representing the initial time The value, floor() means rounding down. Defined by the following formula: , This represents the number of rotors in a switched reluctance motor.

4. The control system for online correction of reference torque of a switched reluctance motor as described in claim 3, characterized in that, The torque-to-current conversion module uses a linear calculation method to convert torque to current, and the formula is as follows: Where m is the number of phases of the switched reluctance motor. When the switched reluctance motor has 3 phases, m = 1, 2, 3. Let m be the reference current for phase m. is the reference torque value for phase m, and k is the linearization coefficient of the inductor partial conductance, with a value ranging from 0.05 to 1.

5. A control method for online correction of reference torque of a switched reluctance motor, employing the control system for online correction of reference torque of a switched reluctance motor as described in claim 1, characterized in that, Includes the following steps: The speed error is obtained by subtracting the given speed from the real-time speed. The speed error is processed by PI calculation to obtain the reference torque value; The reference torque value and rotor angle are input into the neural network regulator to obtain the corrected reference torque value; The rotor position angle is obtained from the output of the switched reluctance motor, and the instantaneous torque value is obtained by looking up a table in combination with the feedback phase current. The torque difference is obtained by subtracting the reference torque value from the instantaneous torque value. Adjust the implicit function coefficients and weight coefficients of the neural network based on the torque difference; Based on the rotor angle, the corrected reference torque value is distributed to obtain the phase torque value; The phase torque value is converted into current to obtain the reference phase current value; The difference between the reference phase current and the feedback phase current is calculated, and the on / off state or freewheeling of the drive circuit is controlled based on the calculation result. The rotor position angle is calculated to obtain the real-time speed of the motor.

Images