阅读说明：本技术 一种基于神经网络直接转矩控制的永磁直驱风电机组控制方法 (Control method of permanent magnet direct-drive wind turbine generator based on neural network direct torque control ) 是由 田玉蓉 蔡彬 邱雅兰 谌义喜 褚晓广 于 2021-08-10 设计创作，主要内容包括：本发明涉及一种基于神经网络直接转矩控制的永磁直驱风电机组控制方法,属电气工程技术领域。该方法采用神经网络直接转矩控制策略,通过离线训练定子磁链神经网络,逼近定子磁链参考值与电磁转矩的关系曲线,实现变定子磁链参考值控制,控制永磁直驱型风力发电机的定子电流d轴分量i-(ds)保持为零；通过将永磁直驱型风力发电机的定子电压计算式进行离散化处理,计算得到定子电压的α轴、β轴分量的参考值u～(*)-(sα)和u～(*)-(sβ),将此参考值经SVPWM模块调制后产生驱动信号,驱动所述机侧变流器的功率开关管,控制所述永磁直驱型风力发电机工作。本发明可降低发电机定子侧电流谐波畸变和机侧变流器能耗,确保整个风电机组系统性能最优。(The invention relates to a control method of a permanent magnet direct-drive wind turbine generator based on neural network direct torque control, and belongs to the technical field of electrical engineering. The method adopts a neural network direct torque control strategy, approaches a relation curve of a stator flux linkage reference value and electromagnetic torque through off-line training of a stator flux linkage neural network, realizes variable stator flux linkage reference value control, and controls a stator current d-axis component i of a permanent magnet direct-drive wind driven generator ds Remains zero; discretizing a stator voltage calculation formula of the permanent magnet direct drive type wind driven generator to calculate and obtain reference values u of alpha-axis and beta-axis components of the stator voltage * sα And u * sβ And modulating the reference value by an SVPWM module to generate a driving signal to drive a power switch tube of the machine side converter and control the permanent magnet direct drive type wind driven generator to work. The invention can reduce the current harmonic distortion at the stator side of the generator and the energy consumption of the machine side converter, and ensure the optimal performance of the whole wind turbine system.)

1. A permanent magnet direct-drive wind turbine generator control method based on neural network direct torque control comprises the following steps:

step 1, designing a rotating speed tracking controller by adopting a BP neural network-PID control strategy:

11) determining the number of layers of the BP neural network: the neural network comprises 1 input layer, 1 hidden layer and 1 output layer, wherein the input layer has 3 input vectors x_j(k) J is 1,2 and 3, which are respectively the reference value of the rotating speed of the permanent magnet direct drive type wind driven generator at the current momentCurrent time rotational speed measurement value omega_r(k) And the deviation value between them Instant gamex₂(k)＝ω_r(k)，x₃(k) K is the current time; the hidden layer has 5 neurons; the output layer has 3 neurons;

12) performing feedforward calculation of the BP neural network to obtain the output of the hidden layer and the output of the output layer:

input s of the ith neuron (i ═ 1,2,3,4,5) of the hidden layer at the current time_iOutput O_i ²(k) Respectively as follows:

in the formula, w_ij ²(k) Is a connection weight coefficient between the j-th neuron of the input layer and the i-th neuron of the hidden layer at the current moment, f₁(. -) is an excitation function of the hidden layer, and a hyperbolic tangent function tanh is adopted;

input s of the current time of the ith neuron (l ═ 1,2,3) of the output layer_lOutput O_l ³(k) Respectively as follows:

in the formula, w_li ³(k) Is a connection weight coefficient between the ith neuron of the hidden layer and the ith neuron of the output layer at the current time, f₂() is the excitation function of the output layer, and a Sigmoid function is adopted;

order:

13) three outputs K obtained in the step 12)_p、K_i、K_dInputting into a PID controller, and obtaining the output y (k) of the rotating speed tracking controller at the current moment as:

step 2, the stator voltage u of the permanent magnet direct-drive wind driven generator_sStator current i_sRespectively obtaining stator voltage alpha axis component u under alpha beta static coordinate system through abc/alpha beta coordinate transformation_sαBeta axis component u_sβAnd stator current alpha axis component i_sαBeta axis component i_sβ(ii) a Simultaneously, converting the abc/dq coordinate of the stator current of the permanent-magnet direct-drive wind driven generator to obtain a stator current d-axis component i under a dq rotation coordinate system_sdAnd q-axis component i_sq；

The stator flux linkage vector psi of the permanent magnet direct drive type wind driven generator is obtained through calculation_sAlpha-axis component psi_sαAnd the beta-axis component psi_sβ：

In the formula, R_sThe equivalent resistance of the stator winding of the permanent magnet direct-drive wind driven generator is shown, and t is time;

simultaneously calculating to obtain the electromagnetic torque T of the permanent magnet direct-drive wind driven generator_e：

In the formula, P is the pole pair number of the permanent magnet direct-drive wind driven generator;

step 3, taking the y (k) obtained in the step 1 as the electromagnetic torque reference value of the permanent magnet direct drive type wind driven generator at the current momentThen put this inAnd the electromagnetic torque T obtained in the step 2_eMaking a difference, comparing the difference and inputting the difference into a PI controller to obtain a torque angle delta, namely a stator flux linkage vector psi of the permanent magnet direct drive type wind driven generator_sFlux linkage vector with rotorThe included angle between them;

step 4, constructing a stator flux linkage neural network model, and enabling the stator flux linkage neural network model to approach a stator flux linkage amplitude reference value of the permanent magnet direct drive type wind driven generator through off-line trainingAnd electromagnetic torque T_eThe relation curve of (2) realizes zero d-axis current control, and the specific steps are as follows:

41) enabling the stator current d-axis component i of the permanent magnet direct drive type wind driven generator obtained in the step 2_sdWhen the value is equal to 0, thenAnd electromagnetic torque T_eThe relation of (A) is as follows:

in the formula, L_qIs the stator side inductance q-axis component of the permanent magnet direct drive type wind driven generator, and P is the pole pair number psi of the permanent magnet direct drive type wind driven generator_rThe magnetic flux is a rotor flux linkage of the permanent magnet direct drive type wind driven generator;

42) constructing a stator flux linkage neural network model, wherein the model consists of an input layer, 1 hidden layer and an output layer;

the input layer has 1 neuron: electromagnetic torque T at the present time_e(k) Let x be T_e(k)；

The hidden layer has 10 neurons, the jth neuron inputs s_jComprises the following steps:

s_j＝ω_1jx-θ_j，j＝1,2,…,10

in the formula, ω_1jIs the connection weight coefficient between the input layer neuron and the jth neuron of the hidden layer at the current moment, theta_jA threshold value for the jth neuron of the hidden layer;

hidden layer jth neuron output y_jComprises the following steps:

y_j＝f₁(s_j)＝f₁(ω_1jx-θ_j)＝tanh(ω_1jx-θ_j)

in the formula (f)₁(. -) is an excitation function of the hidden layer, and a hyperbolic tangent function tanh is adopted;

the output layer has 1 neuron with inputs s:

in the formula, ω_jIs a connection weight coefficient between the jth neuron of the hidden layer and the neuron of the output layer at the current moment, and theta is a threshold value of the output layer;

the output layer neuron output y is:

wherein f is₂(. -) is the excitation function of the output layer, using a linear function;

order: the output y of the neuron in the output layer is the reference value of the amplitude of the stator flux linkage at the current momentNamely:

43) taking several electromagnetic torques T_eThe value of (A) is in the range of0～T_eNWherein, T_eNFor electromagnetic torque T_eOff-line training the stator flux linkage neural network model according to the formula (3) to enable the stator flux linkage neural network model to approach the formula (3) with sufficient precision;

step 5, calculating stator flux linkage vector psi_sAmplitude of (phi)_sAnd its position angle theta_s：

In the formula, #_sαAnd psi_sβCalculated in step 2;

step 6, calculating the electromagnetic torque T obtained in the step 2_eObtaining a stator flux amplitude reference value of the permanent magnet direct drive type wind driven generator at the current moment by the stator flux neural network model which is constructed and trained in the step 4Combining the equations (4) and (5), the stator flux linkage deviation vector delta psi is calculated according to equation (6)_sAlpha axis component of (delta phi)_sαAnd the beta axis component delta phi_sβ：

Wherein δ is the torque angle obtained in step 3;

step 7, the stator voltage u of the permanent magnet direct-drive wind driven generator is measured_sAlpha-axis component u of_sαAnd a beta axis component u_sβFormula for calculation

Discretizing, combining with the formula (6), and calculating to obtain u_sα、u_sβReference value u of^* _sαAnd u^* _sβ：

In the formula, T is a sampling period;

step 8, mixing u^* _sαAnd u^* _sβAnd after modulation by the SVPWM module, a driving signal is generated to drive a power switch tube of the machine side converter, so as to control the permanent magnet direct drive type wind driven generator to work.

Technical Field

The invention relates to a control method, in particular to a control method of a permanent magnet direct-drive wind turbine generator based on neural network direct torque control, and belongs to the technical field of wind power.

Background

The permanent magnet direct-drive wind turbine generator generally adopts a converter with a back-to-back structure, and comprises a machine side converter and a grid side converter, wherein the machine side converter generally adopts a Voltage Source Converter (VSC). In order to improve the rapidity of the system response, the machine side converter of the permanent magnet direct drive type wind turbine generator generally adopts a Direct Torque Control (DTC) method to control the electromagnetic torque of the generator. The traditional direct torque control adopts a ping-pong control scheme, is simple and easy to implement, but the method uses a hysteresis controller, so that an error band always exists in the control of stator flux linkage and electromagnetic torque, and a large amount of harmonic waves are generated at the stator side of the permanent magnet synchronous generator. Harmonic waves in the stator current enable the electromagnetic torque pulsation of the generator to be large, and the wind turbine generator can generate additional mechanical vibration and torsional vibration. In addition, the d-axis component of the stator-side alternating current cannot be maintained to be zero by the traditional direct torque control, which may cause the demagnetization phenomenon of the permanent magnet and increase the power loss of the wind turbine generator side converter.

Disclosure of Invention

The main purposes of the invention are as follows: aiming at the defects in the prior art, the invention provides a control method of a permanent magnet direct-drive wind turbine generator based on neural network direct torque control, which reduces the harmonic content of the stator side current and the power loss of a machine side converter by adopting a direct torque control strategy based on a BP neural network.

In order to achieve the aim, the permanent magnet direct-drive wind turbine generator set comprises a permanent magnet direct-drive wind driven generator, a machine side converter, a direct current link and a grid side converter, wherein one end of the machine side converter is connected with a stator of the permanent magnet direct-drive wind driven generator, and the other end of the machine side converter is connected with the direct current link; the other end of the direct current link is connected with the grid-side converter; and the grid-side converter is connected with an alternating current power grid through a power frequency transformer.

The invention relates to a control method of a permanent magnet direct-drive wind turbine generator based on neural network direct torque control, which comprises the following steps of:

step 1, designing a rotating speed tracking controller by adopting a BP neural network-PID control strategy:

11) determining the number of layers of the BP neural network: the neural network comprises 1 input layer, 1 hidden layer and 1 output layer, wherein the input layer has 3 input vectors x_j(k) J is 1,2 and 3, which are respectively the reference value of the rotating speed of the permanent magnet direct drive type wind driven generator at the current momentCurrent time rotational speed measurement value omega_r(k) And the deviation value between them Instant gamex₂(k)＝ω_r(k)，x₃(k) K is the current time; the hidden layer has 5 neurons; the output layer has 3 neurons;

12) performing feedforward calculation of the BP neural network to obtain the output of the hidden layer and the output of the output layer:

input s of the ith neuron (i ═ 1,2,3,4,5) of the hidden layer at the current time_iOutput O_i ²(k) Respectively as follows:

in the formula, w_ij ²(k) Is thatA connection weight coefficient between the j-th neuron of the input layer and the i-th neuron of the hidden layer at the current moment, f₁(. -) is an excitation function of the hidden layer, and a hyperbolic tangent function tanh is adopted;

input s of the current time of the ith neuron (l ═ 1,2,3) of the output layer_lOutput O_l ³(k) Respectively as follows:

in the formula, w_li ³(k) Is a connection weight coefficient between the ith neuron of the hidden layer and the ith neuron of the output layer at the current time, f₂() is the excitation function of the output layer, and a Sigmoid function is adopted;

order:

13) three outputs K obtained in the step 12)_p、K_i、K_dInputting into a PID controller, and obtaining the output y (k) of the rotating speed tracking controller at the current moment as:

step 2, the stator voltage u of the permanent magnet direct-drive wind driven generator_sAnd stator current i_sRespectively obtaining stator voltage alpha axis component u under alpha beta static coordinate system through abc/alpha beta coordinate transformation_sαBeta axis component u_sβAnd stator current alpha axis component i_sαBeta axis component i_sβ(ii) a Simultaneously carrying out direct drive on the stator current i of the permanent magnet wind driven generator_sObtaining a stator current d-axis component i under a dq rotation coordinate system through abc/dq coordinate transformation_sdAnd q-axis component i_sq；

The stator flux linkage vector psi of the permanent magnet direct drive type wind driven generator is obtained through calculation_sAlpha-axis component psi_sαAnd the beta-axis component psi_sβ：

In the formula, R_sThe equivalent resistance of the stator winding of the permanent magnet direct-drive wind driven generator is shown, and t is time;

simultaneously calculating to obtain the electromagnetic torque T of the permanent magnet direct-drive wind driven generator_e：

In the formula, P is the pole pair number of the permanent magnet direct-drive wind driven generator;

step 3, taking the y (k) obtained in the step 1 as the electromagnetic torque reference value of the permanent magnet direct drive type wind driven generator at the current momentThen put this inAnd the electromagnetic torque T obtained in the step 2_eMaking a difference, comparing the difference and inputting the difference into a PI controller to obtain a torque angle delta, namely a stator flux linkage vector psi of the permanent magnet direct drive type wind driven generator_sFlux linkage vector with rotorψ_rThe included angle between them; r is

Step 4, constructing a stator flux linkage neural network model, and enabling the stator flux linkage neural network model to approach a stator flux linkage amplitude reference value of the permanent magnet direct drive type wind driven generator through off-line trainingAnd electromagnetic torque T_eThe relation curve of (2) realizes zero d-axis current control, and the specific steps are as follows:

41) enabling the stator current d-axis component i of the permanent magnet direct drive type wind driven generator obtained in the step 2_sdWhen the value is equal to 0, thenAnd electromagnetic torque T_eThe relation of (A) is as follows:

in the formula, L_qIs the stator side inductance q-axis component of the permanent magnet direct drive type wind driven generator, and P is the pole pair number psi of the permanent magnet direct drive type wind driven generator_rThe magnetic flux is a rotor flux linkage of the permanent magnet direct drive type wind driven generator;

42) constructing a stator flux linkage neural network model, wherein the model consists of an input layer, 1 hidden layer and an output layer;

the input layer has 1 neuron: electromagnetic torque T at the present time_e(k) Let x be T_e(k)；

The hidden layer has 10 neurons, the jth neuron inputs s_jComprises the following steps:

s_j＝ω_1jx-θ_j，j＝1,2,…,10

in the formula, ω_1jIs the connection weight coefficient between the input layer neuron and the jth neuron of the hidden layer at the current moment, theta_jA threshold value for the jth neuron of the hidden layer;

hidden layer jth neuron output y_jComprises the following steps:

y_j＝f₁(s_j)＝f₁(ω_1jx-θ_j)＝tanh(ω_1jx-θ_j)

in the formula (f)₁(. -) is an excitation function of the hidden layer, and a hyperbolic tangent function tanh is adopted;

the output layer has 1 neuron with inputs s:

in the formula, ω_jIs a connection weight coefficient between the jth neuron of the hidden layer and the neuron of the output layer at the current moment, and theta is a threshold value of the output layer;

the output layer neuron output y is:

wherein f is₂(. -) is the excitation function of the output layer, using a linear function;

order: the output y of the neuron in the output layer is the reference value of the amplitude of the stator flux linkage at the current momentNamely:

43) taking several electromagnetic torques T_eThe value of (A) ranges from 0 to T_eNWherein, T_eNFor electromagnetic torque T_eOff-line training the stator flux linkage neural network model according to the formula (3) to enable the stator flux linkage neural network model to approach the formula (3) with sufficient precision;

step 5, calculating stator flux linkage vector psi_sAmplitude of (phi)_sAnd its position angle theta_s：

In the formula, #_sαAnd psi_sβCalculated in step 2;

step 6, calculating the electromagnetic torque T obtained in the step 2_eObtaining a stator flux amplitude reference value of the permanent magnet direct drive type wind driven generator at the current moment by the stator flux neural network model which is constructed and trained in the step 4Combining the equations (4) and (5), the stator flux linkage deviation vector delta psi is calculated according to equation (6)_sAlpha axis component of (delta phi)_sαAnd the beta axis component delta phi_sβ：

Wherein δ is the torque angle obtained in step 3;

step 7, the stator voltage u of the permanent magnet direct-drive wind driven generator is measured_sAlpha-axis component u of_sαAnd a beta axis component u_sβFormula for calculation

Discretizing, combining with the formula (6), and calculating to obtain u_sα、u_sβReference value u of^* _sαAnd u^* _sβ：

In the formula, T is a sampling period;

step 8, mixing u^* _sαAnd u^* _sβAnd after modulation by the SVPWM module, a driving signal is generated to drive a power switch tube of the machine side converter, so as to control the permanent magnet direct drive type wind driven generator to work.

Compared with the prior art, the invention has the beneficial effects that: the invention adopts a direct torque control strategy based on a BP neural network, approaches a relation curve of a stator flux linkage reference value and electromagnetic torque through off-line training of a stator flux linkage neural network, realizes variable stator flux linkage reference value control, and controls a d-axis component i of a stator current of a permanent magnet direct-drive type wind driven generator_sdThe zero value is kept, so that the power loss of the machine side converter can be reduced, and the demagnetization phenomenon of the permanent magnet direct drive type wind driven generator is avoided; meanwhile, the space vector modulation (SVPWM) scheme is used to improve the traditional direct torque control effect and inhibit the harmonic distortion of the alternating current at the stator side of the generator. In addition, by adopting the BP-PID controller, the PID control parameter self-tuning is realized, and the overshoot of the electromagnetic torque and the rotating speed is reduced.

Drawings

FIG. 1 is a schematic diagram of a topological structure of a permanent magnet direct-drive wind turbine generator set.

FIG. 2 is a block diagram of the machine side converter control employing a neural network based direct torque control strategy in accordance with the present invention.

FIG. 3 is a diagram of a rotational speed neural network model structure according to the present invention.

FIG. 4 is a diagram of a stator flux neural network model according to the present invention.

FIG. 5 is a vector diagram of a direct torque control system.

FIG. 6 is a block diagram of the machine side converter control employing a conventional direct torque control strategy.

FIG. 7 is a graph comparing stator side AC current harmonics for the present invention and a conventional direct torque control strategy.

FIG. 8 is a graph of electromagnetic torque versus simulation for the present invention versus a conventional direct torque control strategy.

FIG. 9 is a graph of speed versus simulation for the present invention and a conventional direct torque control strategy.

FIG. 10 shows the output power P of the permanent-magnet direct-drive wind power generator_outGraph is shown.

FIG. 11 shows generator output power P under the present invention and conventional direct torque control strategy_outWhen the same, the DC side power P of the machine side converter_dcAnd comparing simulation graphs.

The system comprises a permanent magnet direct drive type wind driven generator 1, a machine side converter 2 and a direct current link 3; 4-network side converter, 5-rotating speed neural network model and 6-stator flux linkage neural network model.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings.

As shown in fig. 1, the permanent magnet direct-drive wind turbine generator set comprises a permanent magnet direct-drive wind driven generator 1, a machine side converter 2, a direct current link 3 and a grid side converter 4, wherein one end of the machine side converter 2 is connected with a stator of the permanent magnet direct-drive wind driven generator 1, and the other end of the machine side converter is connected with the direct current link 3; the other end of the direct current link 3 is connected with a network side converter 4; and the grid-side converter 4 is connected with an alternating current power grid through a power frequency transformer.

The invention relates to a control method of a permanent magnet direct-drive wind turbine generator based on neural network direct torque control, which comprises the following steps of:

step 1, constructing a rotating speed neural network model 5, and designing an outer ring rotating speed tracking controller by adopting a BP neural network-PID control strategy. The specific method comprises the following steps:

as shown in fig. 2, the outer ring rotational speed tracking controller is composed of a rotational speed neural network model 5 and a PID controller, the rotational speed neural network model 5 outputs three parameters PID controller of the PID controller, and the output of the outer ring rotational speed tracking controller is the electromagnetic torque reference value of the permanent magnet direct drive type wind driven generator 1

11) As shown in fig. 3, first, the number of BP neural network layers of the rotational speed neural network model 5 is determined: the neural network comprises 1 input layer, 1 hidden layer and 1 output layer, wherein the input layer has 3 neurons, namely: 3 input vectors x_j(k) J is 1,2,3, which are the reference values of the rotation speed of the permanent magnet direct drive wind generator 1 at the current momentCurrent time rotational speed measurement value omega_r(k) And the deviation value between themInstant gamex₂(k)＝ω_r(k)，x₃(k) K is the current time; the hidden layer has 5 neurons; the output layer has 3 neurons, and their outputs are K at the current time_p、K_i、K_d。

As shown in fig. 2, the reference value of the rotation speed at the current momentMay be determined in one of the following ways:

according to the current wind speed v_wDetermining the current rotation speed reference value according to the wind speed-power curve, the rotation speed-power curve or the optimal tip speed ratio of the permanent magnet direct-drive wind driven generator 1Maximum Power Point Tracking (MPPT) control is realized;

determining the output power of the permanent magnet direct-drive wind driven generator 1 according to the output dispatching instruction of the power grid company, and determining the current rotating speed reference value according to the rotating speed-power curve of the permanent magnet direct-drive wind driven generator 1

12) And performing feedforward calculation to obtain the output of the hidden layer and the output layer of the BP neural network:

input s at the current time of the ith neuron (i ═ 1,2,3,4,5) of the hidden layer_iOutput O_i ²(k) Respectively as follows:

in the formula, w_ij ²(k) Is the connection weight coefficient between the jth neuron of the input layer and the ith neuron of the hidden layer at the current moment, f₁(. -) is an excitation function of the hidden layer, and a hyperbolic tangent function tanh is adopted;

input s of the current time of the ith neuron (1, 2,3) of the output layer_lOutput O_l ³(k) Respectively as follows: :

in the formula, w_li ³(k) Is the connection weight coefficient between the ith neuron of the hidden layer and the ith neuron of the output layer at the current moment, f₂() is the excitation function of the output layer, and a Sigmoid function is adopted;

then there are:

13) and (k) calculating the output y (k) of the outer ring rotating speed tracking controller at the current moment as follows:

step 2, the stator voltage u of the permanent magnet direct-drive wind driven generator 1 is measured_sStator current i_sRespectively obtaining u through abc/alpha beta coordinate transformation_sα、u_sβAnd i_sαAnd i_sβ(ii) a Simultaneously, the stator current i of the permanent magnet direct drive type wind driven generator 1 is converted into the stator current_sObtaining stator current d-axis and q-axis components i under dq coordinate system through abc/dq coordinate transformation_sd、i_sq；

From this calculation the stator flux linkage vector psi_sα, β component ψ_sαAnd psi_sβ：

In the formula, R_sThe equivalent resistance of a stator winding of the permanent magnet direct-drive wind driven generator 1 is shown, and t is time;

the electromagnetic torque T of the permanent magnet direct drive type wind driven generator 1 at the current moment is calculated and obtained_e：

In the formula, P is the pole pair number of the permanent magnet direct drive type wind driven generator 1.

Step 3, taking the y (k) obtained in the step 1 as the electromagnetic torque reference value of the permanent magnet direct drive type wind driven generator 1 at the current momentThen put this inAnd the electromagnetic torque T at the current moment obtained in the step 2_eMaking difference, comparing, inputting into PI controller to obtain torque angle delta, i.e. stator flux linkage vector psi_sVector psi of flux linkage with rotor_rThe angle therebetween (see fig. 5). r is

Step 4, constructing a stator flux linkage neural network model, and enabling the stator flux linkage neural network model to approach a stator flux linkage amplitude reference value of the permanent magnet direct drive type wind driven generator 1 through off-line trainingAnd electromagnetic torque T_eThe relation curve of (2) realizes zero d-axis current control, and the specific steps are as follows:

41) enabling the stator current d-axis component i of the permanent magnet direct drive type wind driven generator 1 obtained in the step 2_sdWhen the value is equal to 0, thenAnd electromagnetic torque T_eThe relation of (A) is as follows:

in the formula, L_qIs a stator side inductance q-axis component, psi, of a permanent magnet direct drive type wind power generator 1_rIs a rotor flux linkage of a permanent-magnet direct-drive wind driven generator 1.

The derivation process of equation (3) is as follows:

taking the non-salient pole type permanent magnet synchronous generator as an example, the relationship between the stator flux linkage and the current is as follows:

electromagnetic torque T_eIt can also be calculated as follows:

let i_sdSubstituting 0 into formula (9) or formula (10),and the | ψ in the formula (9)_sI is changed intoFinishing to obtain:

therefore, the actual electromagnetic torque T at the present time is compared_eBy substituting formula (3), the calculation of i_sdStator flux linkage amplitude | ψ of 0_sThe value of | is given. It can be seen that if the stator flux linkage amplitude | ψ can be measured_sI is controlled to be the value calculated by the formula (3), i at the current moment_sdMust be 0 and thus the corresponding stator flux linkage amplitude | ψ is now present_sL is used as the reference value of the stator flux linkage amplitude at the current moment

42) And constructing a stator flux linkage neural network model. As shown in fig. 4, the model consists of 1 input layer, 1 hidden layer, and 1 output layer;

the input layer has 1 neuron, i.e. 1 input vector: electromagnetic torque T at the present time_e(k) Let x be T_e(k)；

The hidden layer has 10 neurons, the jth of which inputs s_jComprises the following steps:

s_j＝ω_1jx-θ_j，j＝1,2，…，10

in the formula, ω_1jIs the connection weight coefficient between the input layer neuron and the jth neuron of the hidden layer at the current moment, theta_jA threshold for the jth neuron of the hidden layer;

hidden layer jth neuron output y_jComprises the following steps:

y_j＝f₁(s_j)＝f₁(ω_1jx-θ_j)＝tanh(ω_1jx-θ_j)

in the formula (f)₁(. to) as the excitation function of hidden layer, hyperbolic tangent function is adoptedtanh；

The output layer has 1 neuron with inputs s:

in the formula, ω_jIs a connection weight coefficient between the jth neuron of the hidden layer and the neuron of the output layer at the current moment, and theta is a threshold value of the output layer;

the output layer neuron output y is:

wherein f is₂(. -) is the excitation function of the output layer, using a linear function;

order: the output y of the neuron in the output layer is the reference value of the amplitude of the stator flux linkage at the current momentNamely:

43) taking several electromagnetic torques T_eThe value of (A) ranges from 0 to T_eNWherein, T_eNFor electromagnetic torque T_eThe stator flux linkage neural network model is trained off-line according to the formula (3) to approximate the formula (3) with sufficient accuracy, and the minimum value of the expected error is set to be 2e^-10。

Step 5, calculating stator flux linkage vector psi_sAmplitude of (phi)_sAnd its position angle theta_s：

In the formula, #_sαAnd psi_sβCalculated in step 2;

step 6, as shown in fig. 2, the real-time electromagnetic torque T calculated in step 2 is used_eObtaining a stator flux amplitude reference value of the permanent magnet direct drive type wind driven generator 1 at the current moment by the stator flux neural network model which is constructed and trained in the step 4

FIG. 5 is a schematic diagram of vectors in a direct torque control system in a two-phase stationary frame at any time. As shown in fig. 5, θ_sThe angle of the current position of the stator flux linkage is obtained; theta_rAs angle of permanent magnet flux linkage position, i.e. reference stator flux linkage vectorThe size of the position angle of (1); delta is a torque angle; delta psi_sIs the deviation vector of the stator flux linkage; u. of₁、u₂、u₃、u₄、u₅、u₆Is a non-zero basis space voltage vector. At this time, application u is selected₂And u₃Vector, the effect of which is to obtain stator flux linkage error vector delta psi_sAnd superposing the current stator flux linkage on the deviation vector so as to enable the amplitude of the stator flux linkage to reach a reference valueTo further control the electromagnetic torque.

According to FIG. 5, the stator flux linkage deviation vector Δ ψ can be calculated by combining equations (4) and (5), respectively_sOf the alpha and beta components Δ ψ_sαAnd delta phi_sβ：

Wherein δ is the torque angle obtained in step 3;

step 7, the stator voltage u of the permanent magnet direct-drive wind driven generator 1 is measured_sAlpha-axis component u of_sαAnd a beta axis component u_sβCalculating the formula:

the discretization treatment is carried out by the following steps:

in the formula, T is a sampling period.

The stator voltage u can be calculated by combining formula (6)_sα、u_sβReference value ofAnd

step 8, as shown in FIG. 2, add u^* _sαAnd u^* _sβAnd the SVPWM module generates a driving signal after modulation, drives a power switch tube of the machine side converter 2 and controls the permanent magnet direct drive type wind driven generator 1 to work.

In order to verify the effectiveness of the direct torque control strategy based on the BP neural network, the machine side converter 2 of the permanent magnet direct drive type wind driven generator 1 is subjected to comparative simulation analysis by adopting the direct torque control strategy based on the BP neural network and the traditional direct torque control strategy.

As shown in fig. 6, the main ideas of the conventional direct torque control strategy are: the stator voltage u of the permanent magnet direct drive type wind driven generator 1_sa、u_sb、u_scAnd stator current i_sa、i_sb、i_scRespectively obtaining u through abc/alpha beta coordinate transformation_sα、u_sβAnd i_sα、i_sβThe stator flux linkage vector psi is obtained by using the formula (1)_sComponent ψ in the alpha and beta axes_sαAnd psi_sβThen, the stator flux linkage vector psi is obtained by using the formula (4)_sAmplitude of (phi)_sAnd will refer to the magnitude of the stator flux linkage vectorAnd | ψ_sMaking a difference I, and sending the difference I into a flux linkage hysteresis comparator;

according to the measured real-time wind speed v_wObtaining a reference speed by maximum power point tracking control (MPPT)Will be provided withWith the actual speed omega_rMaking difference, obtaining electromagnetic torque reference value by PID controllerAnd the real-time electromagnetic torque T calculated by the formula (2)_eMaking difference, and sending the difference into a torque hysteresis comparator; synthesizing the output of the flux linkage hysteresis comparator, the output of the torque hysteresis comparator, and the position angle θ of the stator flux linkage calculated by equation (5)_sAnd selecting a corresponding voltage vector after the query of the switch table, generating a driving signal to drive a power switch tube of the machine side converter 2, and controlling the permanent magnet direct drive type wind driven generator 1 to work.

Specific simulation parameters are shown in tables 1 and 2.

TABLE 1 wind turbine-related parameters

Parameter(s)	Numerical value
		Rated power PN	1kW
Rated wind speed vN	10m/s
		Starting wind speed v0	2.5m/s
Radius of wind wheel R	2m
		Air density ρ	1.225kg/m3
Pitch angle beta	0°

TABLE 2 permanent magnet synchronous generator-related parameters

Parameter(s)	Numerical value
		Rated power PN	1kW
Rated speed nN	150r/min
		Rated mechanical torque Tm	-63.67N*m
Number of pole pairs P	20
		Stator side d-axis inductance Ld	9mH
Stator side q-axis inductance Lq	9mH
		Permanent magnet flux linkage amplitude | ψr|	0.7Wb
DC side voltage Udc	650V
		Stator side rated flux linkage amplitude | psi* s|	0.7Wb

Setting the simulation time to be 0.4s, setting the initial wind speed to be 6m/s, and increasing the wind speed to 10m/s when the wind speed is 0.2 s.

As shown in fig. 7, fig. 7a and 7b are respectively the harmonic distortion rate of the ac current at the stator side of the permanent-magnet direct-drive wind power generator 1 when the conventional direct torque control method and the direct torque control method based on the BP neural network of the present invention are adopted. As can be seen from fig. 7, the stator side alternating current mainly has odd harmonics, and the direct torque control based on the BP neural network of the present invention reduces the stator side alternating current harmonic distortion rate of the generator by 2.7% compared with the conventional direct torque control.

As shown in fig. 8, fig. 8a and 8b are respectively the electromagnetic torque variation of the permanent magnet direct drive type wind power generator 1 when the conventional direct torque control method and the direct torque control method based on the BP neural network of the present invention are adopted. As can be seen from fig. 8, when the conventional direct torque control method is adopted, the electromagnetic torque fluctuation is about 10N · m when the system is stabilized, whereas when the direct torque control based on the BP neural network of the present invention is adopted, the electromagnetic torque fluctuation is less than 5N · m, and the electromagnetic torque ripple is reduced by at least 50%.

As shown in fig. 9, fig. 9a and 9b are respectively the rotation speed variation of the permanent magnet direct drive type wind power generator 1 when the conventional direct torque control method and the direct torque control method based on the BP neural network of the present invention are adopted. As can be seen from fig. 9, when the conventional direct torque control method is adopted, the rotational speed has a large overshoot at the system start-up and wind speed change stages, whereas when the direct torque control based on the BP neural network of the present invention is adopted, the rotational speed has a faster dynamic response and can reach a steady state in a short time.

FIG. 10 shows the variation of the output power of the permanent magnet direct drive type wind power generator 1, and it can be seen from FIG. 10 that when the wind speed reaches the rated wind speed of 10m/s, the output power P of the generator 1_outReaching the rated power of 1000W. As shown in fig. 11, fig. 11a and fig. 11b are the dc-side power variation of the time-side converter 2 by the conventional direct torque control method and the direct torque control method based on the BP neural network according to the present invention, respectively. As can be seen from fig. 11, when the conventional direct torque control is adopted, the dc-side output power of the machine-side converter 2 is 800W, and when the direct torque control based on the BP neural network of the present invention is adopted, the dc-side output power of the machine-side converter 2 is 910W. Therefore, the direct torque control method based on the BP neural network can reduce the power loss of the machine-side converter 2.

In a word, the direct torque control method based on the BP neural network can inhibit the harmonic distortion of the alternating current at the stator side of the generator, reduce the pulsation of electromagnetic torque and rotating speed and reduce the power loss of the machine side converter.