CN102594244A

CN102594244A - Joint control method of primary frequency modulation for doubly-fed wind power generation set

Info

Publication number: CN102594244A
Application number: CN2012100377634A
Authority: CN
Inventors: 文乐斌; 李群; 孙蓉; 李强; 刘建坤; 顾伟; 顾天畏; 柳伟
Original assignee: Southeast University; Electric Power Research Institute of State Grid Jiangsu Electric Power Co Ltd
Current assignee: State Grid Corp of China SGCC; Southeast University; State Grid Jiangsu Electric Power Co Ltd; Electric Power Research Institute of State Grid Jiangsu Electric Power Co Ltd
Priority date: 2012-02-20
Filing date: 2012-02-20
Publication date: 2012-07-18
Anticipated expiration: 2032-02-20
Also published as: CN102594244B

Abstract

The invention discloses a combined control method for primary frequency modulation of a doubly-fed wind turbine. The continuous Hopfield neural network is used to optimize the PD controller parameters online, and a neural controller with strong self-adaptive ability is established, which can realize rotor kinetic energy and standby power. The joint control of the joint control, the minimum frequency change is used as the objective function, the network weight corresponds to the system state variable, and the output of the neuron is used as the parameter of the PD controller, and the parameter is obtained by combining the expression of the objective function and the expression of the energy function Change rules, and then look for steady-state output according to the rules. The present invention uses the PSCAD/EMTDC simulation platform to carry out detailed simulation research on the neural joint control strategy, and compares it with the traditional frequency control strategy. The result shows that the Hopfield neural joint control has better primary frequency modulation control effect.

Description

Combined control method of primary frequency modulation for doubly-fed wind turbines

技术领域 technical field

本发明涉及双馈风电机组一次调频控制方法，用以研究利用风力发电机的控制提高电力系统频率稳定性，属于风力发电技术领域。The invention relates to a primary frequency modulation control method of a doubly-fed wind power generating set, which is used for researching and improving the frequency stability of a power system by using the control of a wind power generator, and belongs to the technical field of wind power generation.

背景技术 Background technique

近年来，可再生能源的开发利用越来越受到世界各国的广泛重视，风能作为一种可持续发展的新能源，以其无污染性和可再生性，成为一种很有发展前途的绿色能源。风能和潮汐能、太阳能等能源相比，其利用率最高，具有可与常规发电方式比拟的竞争力。因此近几年来我国的风力发电产业得到了迅速的发展，风电装机容量不断提高。In recent years, the development and utilization of renewable energy has attracted more and more attention from all over the world. Wind energy, as a sustainable new energy, has become a promising green energy due to its non-pollution and renewability. . Compared with tidal energy, solar energy and other energy sources, wind energy has the highest utilization rate and is competitive with conventional power generation methods. Therefore, in recent years, my country's wind power industry has developed rapidly, and the installed capacity of wind power has continued to increase.

在风力发电技术中，基于变速恒频双馈感应电机(DFIG)的风力发电系统与传统的基于普通异步发电机的恒速恒频风力发电系统相比具有明显的优势，因此已逐渐成为风电市场的主流机型。由于双馈风电机组控制系统使其机械功率与系统电磁功率的解耦、转速与系统频率的解耦，风力机转子机械部分无法对系统频率变化做出快速有效的响应，因此其旋转动能对系统惯量的贡献几乎没有。大量的双馈风电机组接入电网替代部分常规发电机组，整个系统的惯量必然会受到影响而相对减少。已知系统的惯量与频率降低的变化率有关，在电网发生严重的频率降低事故时，惯量越低，系统频率降低得越快，因此系统的频率将更难控制。In wind power generation technology, the wind power generation system based on the variable speed constant frequency double-fed induction motor (DFIG) has obvious advantages compared with the traditional constant speed constant frequency wind power generation system based on the ordinary asynchronous generator, so it has gradually become the wind power market mainstream models. Because the doubly-fed wind turbine control system decouples the mechanical power from the electromagnetic power of the system, and decouples the rotational speed from the system frequency, the mechanical part of the wind turbine rotor cannot respond quickly and effectively to changes in the system frequency, so its rotational kinetic energy has a great impact on the system. The contribution of inertia is next to nothing. When a large number of double-fed wind turbines are connected to the grid to replace some conventional generators, the inertia of the entire system will inevitably be affected and relatively reduced. It is known that the inertia of the system is related to the change rate of frequency reduction. When a serious frequency reduction accident occurs in the power grid, the lower the inertia, the faster the system frequency will decrease, so the system frequency will be more difficult to control.

如今，越来越多的电力公司提出了严格的风电场并网技术导则，频率控制能力是其中重要的技术要求之一。因此，要求风力发电能像常规发电厂一样具有参与电网一次频率的能力已成为一项重要而迫切的任务。Nowadays, more and more power companies have put forward strict technical guidelines for grid connection of wind farms, and frequency control capability is one of the important technical requirements. Therefore, it has become an important and urgent task to require wind power generation to have the ability to participate in the primary frequency of the power grid like conventional power plants.

国内外对双馈风电机组频率控制方式的研究主要包括三种方法：1)转子动能控制。双馈风电机组的转子中储有大量的旋转动能，通过增加附加的频率控制环节控制转子转矩参考值，可实现在频率下降时降低转速，从而释放叶片中的动能提供频率支撑。2)备用功率控制。类似于同步发电机组运行时需要具有一定的备用容量，为了保证变速风机的功率储备，风机必须运行在不是最大风能追踪的工作点。一般可通过控制桨距角或调节功率转速曲线，来使风机卸载运行，使有功功率参考值低于最佳功率曲线，从而保证一次调频备用容量。在系统频率大幅降低时，减小桨距角或运行至最优功率转速曲线上从而增加有功输出，参与一次调频。3)联合控制。同时考虑转子动能和备用功率控制。The domestic and foreign studies on the frequency control methods of doubly-fed wind turbines mainly include three methods: 1) Rotor kinetic energy control. A large amount of rotational kinetic energy is stored in the rotor of a double-fed wind turbine. By adding an additional frequency control link to control the rotor torque reference value, the speed can be reduced when the frequency drops, thereby releasing the kinetic energy in the blades to provide frequency support. 2) Standby power control. Similar to the need for a certain reserve capacity when the synchronous generator set is running, in order to ensure the power reserve of the variable speed wind turbine, the wind turbine must operate at a working point that is not the maximum wind energy tracking. Generally, the wind turbine can be operated unloaded by controlling the pitch angle or adjusting the power speed curve, so that the active power reference value is lower than the optimal power curve, so as to ensure the reserve capacity of primary frequency modulation. When the system frequency is greatly reduced, reduce the pitch angle or run to the optimal power speed curve to increase the active power output and participate in a frequency modulation. 3) Joint control. Both rotor kinetic energy and reserve power control are considered.

传统的控制模式，包括转子动能控制、备用功率控制、以及普通联合控制，均需建立一个有效地系统数学模型，而对于DFIG风电机组，由于风速的不确定性和电力电子模型的复杂性，模型趋向于非线性和时变性，建立一套详细完整的数学模型十分困难；另一方面，传统控制方法的控制参数一般是依据经验人为确定的，具有很大的主观性，而控制参数的变化对系统控制特性影响较大，因此所得控制参数鲁棒性较差。Traditional control modes, including rotor kinetic energy control, standby power control, and common joint control, all need to establish an effective system mathematical model, and for DFIG wind turbines, due to the uncertainty of wind speed and the complexity of power electronic models, the model It tends to be nonlinear and time-varying, and it is very difficult to establish a set of detailed and complete mathematical models; on the other hand, the control parameters of traditional control methods are generally determined artificially based on experience, which is highly subjective, and the change of control parameters has great influence on The control characteristics of the system have great influence, so the robustness of the obtained control parameters is poor.

发明内容 Contents of the invention

本发明所要解决的技术问题是提供一种双馈风电机组频率控制方法，无需精确的数学模型即可执行控制功能，且其控制参数是根据系统的运行结构实时计算得出的，具有强的适应性和鲁棒性。The technical problem to be solved by the present invention is to provide a doubly-fed wind turbine frequency control method, which can perform the control function without an accurate mathematical model, and its control parameters are calculated in real time according to the operating structure of the system, which has strong adaptability and robustness.

为解决上述技术问题，本发明提供一种双馈风电机组一次调频联合控制方法，其特征在于，包括以下步骤：In order to solve the above technical problems, the present invention provides a combined control method for primary frequency modulation of doubly-fed wind turbines, which is characterized in that it includes the following steps:

(1)在双馈风机控制系统中分别建立转子动能控制模块和备用功率控制模块，转子动能控制模块和备用功率控制模块的输入变量均为频率偏差，转子动能控制模块的输出变量为转矩参考值的修改量，备用功率控制模块的输入变量为桨距角参考值的修改量；(1) The rotor kinetic energy control module and the backup power control module are respectively established in the double-fed fan control system. The input variables of the rotor kinetic energy control module and the backup power control module are frequency deviation, and the output variables of the rotor kinetic energy control module are torque reference The modification amount of the value, the input variable of the standby power control module is the modification amount of the pitch angle reference value;

(2)建立Hopfield神经网络，确定控制对象系统模型，且目标函数是电网发生频率偏移时系统频率变化最小，使控制对象的状态变量对应于网络权值，并将神经元的输出作为PD控制器的参数；(2) Establish the Hopfield neural network, determine the system model of the control object, and the objective function is to minimize the system frequency change when the frequency offset of the power grid occurs, so that the state variable of the control object corresponds to the network weight, and the output of the neuron is used as the PD control parameters of the device;

(3)将目标函数表达式与标准能量函数表达式对应起来，推导得到连接权矩阵和网络的偏值表达式；(3) Corresponding the expression of the objective function and the expression of the standard energy function, and deriving the partial value expression of the connection weight matrix and the network;

(4)将得到的连接权重矩阵和网络输入偏值矩阵代入Hopfield网络的动态方程中，并取神经元输出的非线性特性为对称型S非线性作用函数，推导得到控制器参数的变化规律。(4) Substitute the obtained connection weight matrix and network input bias matrix into the dynamic equation of the Hopfield network, and take the nonlinear characteristics of neuron output as the symmetric S nonlinear action function, and derive the change rule of the controller parameters.

本发明所达到的有益效果：发明的基于Hopfield神经网络的双馈风电机组一次调频联合控制方法，根据频率下降最小为目标采用Hopfield神经网络对控制器的参数进行了优化设计，可以更合理的安排转子动能控制和备用功率控制的联合调频。当系统发生频率偏移时，风机即可实现释放储存在转子叶片中的旋转动能，并增加有功出力提供频率支撑，能进一步缓解系统的频率下降。采用的Hopfield神经网络控制器实现了参数自适应的功能，不易受外界环境变化的影响，能很快适应系统参数等发生的变化。Beneficial effects achieved by the present invention: the invented combined control method for primary frequency modulation of double-fed wind turbines based on the Hopfield neural network uses the Hopfield neural network to optimize the design of the parameters of the controller according to the goal of minimizing the frequency drop, which can be arranged more reasonably Joint frequency modulation of rotor kinetic energy control and reserve power control. When the system frequency shifts, the fan can release the rotational kinetic energy stored in the rotor blades, and increase the active output to provide frequency support, which can further alleviate the frequency drop of the system. The Hopfield neural network controller adopted realizes the function of parameter self-adaptation, is not easily affected by changes in the external environment, and can quickly adapt to changes in system parameters.

附图说明 Description of drawings

图1为Hopfield神经联合控制示意图；Figure 1 is a schematic diagram of Hopfield neural joint control;

图2为四机两区域仿真系统；Figure 2 is a four-machine two-area simulation system;

图3为PSCAD中的神经控制器A和神经控制器B示意图；Fig. 3 is the schematic diagram of neural controller A and neural controller B in PSCAD;

图4为神经控制器A中的Hopfield神经网络示意图；Fig. 4 is the Hopfield neural network schematic diagram in neural controller A;

图5为功率扰动各种策略下调频曲线比较示意图；Figure 5 is a schematic diagram of comparison of frequency modulation curves under various power disturbance strategies;

图6为风速扰动各种策略下调频曲线比较示意图。Figure 6 is a schematic diagram of the comparison of frequency modulation curves under various strategies of wind speed disturbance.

具体实施方式 Detailed ways

下面结合附图对本发明具体叙述如下：Below in conjunction with accompanying drawing, the present invention is specifically described as follows:

(1)技术方案的第一部分：综合考虑转子动能控制和备用功率控制，在双馈风机控制系统中分别建立转子动能控制模块和备用功率控制模块(见附图1)。控制设计的核心是Hopfield神经控制器A和Hopfield神经控制器B，转子动能控制模块和备用功率控制模块输入变量均为频率偏差Δf，输出变量分别为转矩参考值的修改量ΔT_ref和桨距角参考值的修改量Δβ_ref，目的是优化联合控制的频率控制特性。(1) The first part of the technical solution: considering rotor kinetic energy control and backup power control comprehensively, a rotor kinetic energy control module and a backup power control module are respectively established in the doubly-fed fan control system (see Figure 1). The core of the control design is the Hopfield neural controller A and the Hopfield neural controller B. The input variables of the rotor kinetic energy control module and the reserve power control module are the frequency deviation Δf, and the output variables are the modified amount of the torque reference value ΔT _ref and the pitch The modification amount Δβ _ref of the angle reference value is aimed at optimizing the frequency control characteristic of the joint control.

(2)技术方案的第二部分：建立Hopfield神经网络，确定控制对象系统模型，并使网络权值已知，对应于控制对象的状态变量，将神经元的输出作为PD控制器参数，确定目标函数是频率变化最小。(2) The second part of the technical solution: establish the Hopfield neural network, determine the system model of the control object, and make the network weight known, corresponding to the state variable of the control object, use the output of the neuron as the parameter of the PD controller, and determine the target The function is that the frequency change is minimal.

由于电力电子变换器对电功率快速的控制，可以认为风电机组中频率控制环节整体所调节的转矩参考值修改量和桨距角参考值修改量与实际输出之间没有动态：Due to the rapid control of the electric power by the power electronic converter, it can be considered that there is no dynamic between the torque reference value modification amount and the pitch angle reference value modification amount adjusted by the frequency control link of the wind turbine and the actual output:

$Δ Δ {T T}_{ref ref} = = - - {K K}_{f f 11} Δf Δ f - - {K K}_{in in 11} \frac{df df}{dt dt} - - - - - - ((11))$

$Δ Δ {β β}_{ref ref} = = - - {K K}_{f f 22} Δf Δf - - {K K}_{in in 22} \frac{df df}{dt dt}$

控制器A控制的对象可认为是 $\{\begin{matrix} \frac{df}{dt} = K_{A} Δ T_{ref} \\ y = f \end{matrix},$ 控制器B控制的对象可认为是 $\{\begin{matrix} \frac{df}{dt} = K_{B} Δ β_{ref} \\ y = f \end{matrix},$ 其中神经控制器A的参数A和神经控制器B的参数B可通过对简单的转子动能控制和简单的备用功率控制的运行参数进行简单辨识得到。其中f为系统频率，Δf为频率偏差，ΔT_ref为转矩参考值修改量，Δβ_ref为桨距角参考值修改量，K_f1、K_in1分别为神经控制器A的比例调节系数与微分调节系数，K_f2、K_in2分别为神经控制器B的比例调节系数与微分调节系数，y为控制器的输出，K_A为控制器A的控制对象的参数，K_B为控制器B的控制对象的参数。The object controlled by controller A can be considered as $\{\begin{matrix} \frac{df}{dt} = K_{A} Δ T_{ref} \\ the y = f \end{matrix},$ The object controlled by controller B can be considered as $\{\begin{matrix} \frac{df}{dt} = K_{B} Δ β_{ref} \\ the y = f \end{matrix},$ The parameter A of the neural controller A and the parameter B of the neural controller B can be obtained by simply identifying the operating parameters of the simple rotor kinetic energy control and the simple standby power control. Where f is the system frequency, Δf is the frequency deviation, ΔT _ref is the modification amount of the torque reference value, Δβ _ref is the modification amount of the pitch angle reference value, K _f1 and K _in1 are the proportional adjustment coefficient and differential adjustment of neural controller A respectively Coefficients, K _f2 and K _in2 are the proportional adjustment coefficient and differential adjustment coefficient of the neural controller B respectively, y is the output of the controller, K _A is the parameter of the control object of the controller A, and K _B is the control object of the controller B parameters.

对控制器A、B的一般情况的系统模型y(t)＝C[Ax(t)+Bu(t)]进行讨论，这样推导的结论具有一般性，模型如下：Discuss the system model y(t)=C[Ax(t)+Bu(t)] of the general situation of controllers A and B, the conclusions derived in this way are general, and the model is as follows:

$\{\begin{matrix} \frac{dx dx ((t t))}{dt dt} = = Bu Bu ((t t)) \\ y the y ((t t)) = = C C [[Ax Ax ((t t)) + + \frac{dx dx ((t t))}{dt dt}]] \end{matrix} - - - - - - ((22))$

式中，y(t)为系统输出变量，x(t)为系统状态变量，u(t)为系统输入变量，A、B、C均为描述系统模型的待定系数。In the formula, y(t) is the system output variable, x(t) is the system state variable, u(t) is the system input variable, and A, B, C are undetermined coefficients describing the system model.

控制器采用PD控制器，其输出为：The controller adopts PD controller, and its output is:

$u u ((t t)) = = {k k}_{p p} e e ((t t)) + + {k k}_{d d} \frac{de de ((t t))}{dt dt} - - - - - - ((33))$

式中，k_p、k_d为控制器比例调节系数和微分调节系数，e(t)为控制系统误差，r(t)为额定频率，取恒值；y(t)为当前频率)，即：In the formula, k _p and k _d are the proportional adjustment coefficient and differential adjustment coefficient of the controller, e(t) is the error of the control system, r(t) is the rated frequency and takes a constant value; y(t) is the current frequency), that is :

e(t)＝r(t)-y(t)(4)e(t)=r(t)-y(t)(4)

控制系统的目标函数取：The objective function of the control system is:

$E E. = = \frac{11}{22} {e e}^{22} ((t t)) - - - - - - ((55))$

将式(2)、(3)、(4)代入式(5)展开得：Substituting equations (2), (3), and (4) into equation (5), we get:

$E E. ((t t)) = = \frac{11}{22} {e e}^{22} ((t t)) = = \frac{11}{22} {{r r ((t t)) - - C C [[Ax Ax ((t t)) + + Bu Bu ((t t))]]}}^{22}$

$= = \frac{11}{22} {{r r ((t t)) - - C C [[Ax Ax ((t t)) + + B B (({k k}_{p p} ((t t)) e e ((t t)) + + {k k}_{d d} ((t t)) \frac{de de ((t t))}{dt dt}))]]}}^{22} - - - - - - ((66))$

由于有两个控制器的参数即P、D参数，取Hopfield网络输出神经元数为2，即CHNN由两个神经元组成，CHNN在t时刻的输出为：Since there are two controller parameters, that is, P and D parameters, the number of Hopfield network output neurons is 2, that is, CHNN is composed of two neurons, and the output of CHNN at time t is:

V(t)＝[V₁(t)，V₂(t)]^T＝[k_p(t)，k_d(t)]^T(7)V(t) = [V ₁ (t), V ₂ (t)] ^T = [k _p (t), k _d (t)] ^T (7)

技术方案的第三部分：将目标函数表达式与标准能量函数表达式对应起来，推导得到连接权矩阵和网络偏值表达式。The third part of the technical solution: Corresponding the expression of the objective function with the expression of the standard energy function, deriving the connection weight matrix and the expression of the network partial value.

令V₁＝k_p，V₂＝k_d，将E_N(t)展开得：Let V ₁ =k _p , V ₂ =k _d , expand E _N (t) to get:

${E E.}_{N N} ((t t)) = = - - \frac{11}{22} [[{w w}_{1111} ((t t)) {k k}_{p p} ((t t)) {k k}_{p p} ((t t)) + + {w w}_{1212} ((t t)) {k k}_{p p} ((t t)) {k k}_{d d} ((t t)) + + {w w}_{21 twenty one} ((t t)) {k k}_{d d} ((t t)) {k k}_{p p} ((t t)) - - - - - - ((88))$

$+ + {w w}_{22 twenty two} ((t t)) {k k}_{d d} ((t t)) {k k}_{d d} ((t t))]] - - {k k}_{p p} ((t t)) {I I}_{11} ((t t)) - - {k k}_{d d} ((t t)) {I I}_{22} ((t t))$

其中V₁、V₂分别为神经元1和神经元2的输出，w_ij为神经元i和神经元j的连接权值，I₁(t)、I₂(t)分别为神经元1和神经元2的阈值。Among them, V ₁ and V ₂ are the output of neuron 1 and neuron 2 respectively, w _ij is the connection weight of neuron i and neuron j, I ₁ (t), I ₂ (t) are neuron 1 and neuron Threshold of neuron 2.

当Hopfield网络处于平衡状态时，能量函数最小，w₁₂＝w₂₁，此时：When the Hopfield network is in a balanced state, the energy function is the smallest, w ₁₂ =w ₂₁ , at this time:

$\frac{&PartialD; &PartialD; {E E.}_{N N}}{&PartialD; &PartialD; {k k}_{p p}} = = \frac{&PartialD; &PartialD; E E.}{&PartialD; &PartialD; {k k}_{p p}} = = 00 - - - - - - ((99))$

$\frac{{&PartialD; &PartialD; E E.}_{N N}}{&PartialD; &PartialD; {k k}_{d d}} = = \frac{&PartialD; &PartialD; E E.}{&PartialD; &PartialD; {k k}_{d d}} = = 00$

其中E_N为网络的标准能量函数，E为目标函数，e为系统偏差，x系统状态，r为恒定输入。Where E _N is the standard energy function of the network, E is the objective function, e is the system deviation, x is the system state, and r is the constant input.

由 $\frac{{&PartialD; E}_{N}}{&PartialD; k_{p}} = \frac{&PartialD; E}{&PartialD; k_{p}} = 0$ 得：Depend on $\frac{{&PartialD; E.}_{N}}{&PartialD; k_{p}} = \frac{&PartialD; E.}{&PartialD; k_{p}} = 0$ have to:

$\frac{{&PartialD; &PartialD; E E.}_{N N}}{&PartialD; &PartialD; {k k}_{p p}} = = \frac{11}{22} (({- - 22 w w}_{1111} {k k}_{p p} - - 22 {w w}_{1212} {k k}_{d d})) - - {I I}_{11} = = 00 - - - - - - ((1010))$

$\frac{&PartialD; &PartialD; E E.}{&PartialD; &PartialD; {k k}_{p p}} = = \frac{11}{22} (({22 B B}^{22} {C C}^{22} {e e}^{22} {k k}_{p p} + + {22 B B}^{22} C C {k k}_{d d} e e \frac{de de}{dt dt} + + 22 {ABC ABC}^{22} xe xe - - 22 BCre BCre)) = = 00 - - - - - - ((1111))$

由上面两式得：From the above two formulas:

ω₁₁＝-B²C²e²， $ω_{12} = ω_{21} = - 2 B^{2} C^{2} e \frac{de}{dt}, - - - (12)$ ω ₁₁ =-B ² C ² e ² , $ω_{12} = ω_{twenty one} = - 2 B^{2} C^{2} e \frac{de}{dt}, - - - (12)$

I₁＝-2ABC²ex+2BCreI ₁ ＝-2ABC ² ex+2BCre

同理由 $\frac{{&PartialD; E}_{N}}{&PartialD; k_{d}} = \frac{&PartialD; E}{&PartialD; k_{d}} = 0$ 得：same reason $\frac{{&PartialD; E.}_{N}}{&PartialD; k_{d}} = \frac{&PartialD; E.}{&PartialD; k_{d}} = 0$ have to:

$ω_{22} = - B^{2} C^{2} {(\frac{de}{dt})}^{2},$ $ω_{12} = ω_{21} = - 2 B^{2} C^{2} e \frac{de}{dt},$ (13) $ω_{twenty two} = - B^{2} C^{2} {(\frac{de}{dt})}^{2},$ $ω_{12} = ω_{twenty one} = - 2 B^{2} C^{2} e \frac{de}{dt},$ (13)

${I I}_{22} = = - - 22 {ABC ABC}^{22} \frac{de de}{dt dt} x x + + 22 BCr BCr \frac{de de}{dt dt}$

通过上面推导得到连接权矩阵W和网络的偏值(阈值)I如下：Through the above derivation, the connection weight matrix W and the bias (threshold) I of the network are obtained as follows:

$W W = = - - [\begin{matrix} {[[CBe CBe ((t t))]]}^{22} & 22 {B B}^{22} {C C}^{22} e e ((t t)) \frac{de de ((t t))}{dt dt} \\ 22 {B B}^{22} {C C}^{22} e e ((t t)) \frac{de de ((t t))}{dt dt} & {[[CB CB \frac{de de ((t t))}{dt dt}]]}^{22} \end{matrix}] - - - - - - ((1414))$

$I I = = {[\begin{matrix} - - 22 A A {BC BC}^{22} e e ((t t)) x x ((t t)) + + 22 BCr BCr ((t t)) e e ((t t)) \\ - - 22 {ABC ABC}^{22} \frac{de de ((t t))}{dt dt} x x ((t t)) + + 22 BCr BCr ((t t)) \frac{de de ((t t))}{dt dt} \end{matrix}]}^{T T} - - - - - - ((1515))$

(3)技术方案的第四部分：将得到连接权重矩阵和网络输入偏值矩阵代入Hopfield网络的动态方程中，并取神经元输出的非线性特性为对称型S非线性作用函数，推导得到控制器参数的变化规律。(3) The fourth part of the technical solution: Substituting the obtained connection weight matrix and the network input bias matrix into the dynamic equation of the Hopfield network, and taking the nonlinear characteristics of the neuron output as the symmetric S nonlinear action function, deriving the control Variations of device parameters.

标准Hopfield网络的动态方程为：The dynamic equation of the standard Hopfield network is:

$\{\begin{matrix} {C C}_{i i} \frac{{du du}_{i i}}{dt dt} = = \underset{j j}{Σ Σ} {w w}_{ij ij} {V V}_{j j} + + {I I}_{i i} \\ {V V}_{i i} = = f f (({u u}_{i i})) \end{matrix} - - - - - - ((1616))$

取C_i＝1.0，将所求的W和I代入上式得：Take C _i =1.0, and substitute the obtained W and I into the above formula to get:

$\frac{{du du}_{11}}{dt dt} = = {w w}_{1111} {V V}_{11} + + {w w}_{1212} {V V}_{22} + + {I I}_{11}$

$= = - - {B B}^{22} {C C}^{22} {e e}^{22} g g (({u u}_{11})) + + 22 {B B}^{22} {C C}^{22} e e \frac{de de}{dt dt} g g (({u u}_{22})) - - 22 {ABC ABC}^{22} ex ex + + 22 BCre BCre - - - - - - ((1717))$

$\frac{{du du}_{22}}{dt dt} = = {w w}_{21 twenty one} {V V}_{11} + + {w w}_{22 twenty two} {V V}_{22} + + {I I}_{22}$

$= = 22 {B B}^{22} {C C}^{22} e e \frac{de de}{dt dt} g g (({u u}_{11})) - - {B B}^{22} {C C}^{22} {((\frac{de de}{dt dt}))}^{22} g g (({u u}_{22})) - - 22 {ABC ABC}^{22} \frac{de de}{dt dt} x x + + 22 BCr BCr \frac{de de}{dt dt}$

其中w_ij为第i个神经元与第j个神经元的连接权值，I_i、u_i、V_i分别为第i个神经元的偏值、状态量和输出，f(u_i)为作用函数。Where w _ij is the connection weight between the i-th neuron and the j-th neuron, I _i , u _i , V _i are the bias value, state quantity and output of the i-th neuron respectively, and f(u _i ) is Action function.

取神经元输出的非线性特性为对称型S非线性作用函数(增益K)：The nonlinear characteristic of neuron output is taken as the symmetric S nonlinear action function (gain K):

$g g (({u u}_{i i})) = = {K K}_{i i} \frac{{11 - - e e}^{- - {β β}_{i i} {u u}_{i i}}}{11 + + {e e}^{- - {β β}_{i i} {u u}_{i i}}} = = \frac{{22 K K}_{i i}}{11 + + {e e}^{- - {β β}_{i i} {u u}_{i i}}} - - {K K}_{i i},, i i = = 1,2 1,2 - - - - - - ((1818))$

其中g(u_i)为作用函数，K_i为增益，β_i为参数。Where g(u _i ) is the action function, K _i is the gain, and β _i is the parameter.

网络的实际输出为：The actual output of the network is:

k_p＝g(u₁)k _p =g(u ₁ )

(19)(19)

k_d＝g(u₂)k _d =g(u ₂ )

由于 $1 + e^{- β_{1} u_{1}} = \frac{{2 K}_{1}}{k_{p} + K_{1}},$ $1 + e^{- β_{2} u_{2}} = \frac{{2 K}_{2}}{k_{d} + K_{2}},$ 则有：because $1 + e^{- β_{1} u_{1}} = \frac{{2 K}_{1}}{k_{p} + K_{1}},$ $1 + e^{- β_{2} u_{2}} = \frac{{2 K}_{2}}{k_{d} + K_{2}},$ Then there are:

$\frac{{dk dk}_{p p}}{{du du}_{11}} = = \frac{- - 22 {K K}_{11} {e e}^{- - {β β}_{11} {u u}_{11}} (({- - β β}_{11}))}{{((11 + + {e e}^{- - {β β}_{11} {u u}_{11}}))}^{22}} = = 22 {β β}_{11} {K K}_{11} \frac{{K K}_{11} - - {k k}_{p p}}{{k k}_{p p} + + {K K}_{11}} \frac{{(({k k}_{p p} + + {K K}_{11}))}^{22}}{{(({22 K K}_{11}))}^{22}} = = \frac{{β β}_{11} (({K K}_{11}^{22} - - {k k}_{p p}^{22}))}{22 {K K}_{11}}$

$\frac{{dk dk}_{p p}}{dt dt} = = \frac{{dk dk}_{p p}}{d d} \frac{{du du}_{11}}{dt dt}$

(20)(20)

$= = \frac{{β β}_{11} (({K K}_{11}^{22} - - {k k}_{p p}^{22}))}{22 {K K}_{11}} ((- - {B B}^{22} {C C}^{22} {e e}^{22} g g (({u u}_{11})) + + 22 {B B}^{22} {C C}^{22} e e \frac{de de}{dt dt} g g (({u u}_{22})) - - 22 {ABC ABC}^{22} ex ex + + 22 BCre BCre))$

同理可得：In the same way:

$\frac{{dk dk}_{d d}}{{du du}_{22}} = = \frac{{β β}_{22} (({K K}_{22}^{22} - - {k k}_{d d}^{22}))}{22 {K K}_{22}} - - - - - - ((21 twenty one))$

$\frac{{dk dk}_{d d}}{dt dt} = = \frac{{dk dk}_{d d}}{d d {u u}_{22}} \frac{{du du}_{22}}{dt dt}$

$= = \frac{{β β}_{22} (({K K}_{22}^{22} - - {k k}_{d d}^{22}))}{22 {K K}_{22}} (({22 B B}^{22} {C C}^{22} e e \frac{de de}{dt dt} g g {(({u u}_{11})) - - {B B}^{22} {C C}^{22} ((\frac{de de}{dt dt}))}^{22} g g (({u u}_{22})) - - 22 {ABC ABC}^{22} \frac{de de}{dt dt} x x + + 22 BCr BCr \frac{de de}{dt dt})) - - - - - - ((22 twenty two))$

求解微分方程式(20)和式(22)，可得到优化后的k_p、k_d，从而实现PD参数的整定。By solving differential equation (20) and equation (22), the optimized k _p and k _d can be obtained, so as to realize the tuning of PD parameters.

例如采用KUNDUR的4机2区域模型，改造形成本发明的仿真系统对文中的控制策略进行仿真验证。如图2所示，仿真系统包括同步电机4台，输出功率分别为120MW、120MW、124MW、120MW；双馈风电场忽略了其中各台风电机组的差异性，用一个双馈风机等效，总输出功率为80MW；系统负荷L1、L2分别为206MW、342MW。系统在5s时，同步电机G2由于失步故障退出运行，或者仿真区域的风速出现扰动，由额定风速12m/s变为10m/s，分析此时系统频率变化情况。在PSCAD平台里搭建仿真系统模型，分别对转子动能控制、备用功率控制、联合控制、Hopfield神经联合控制进行仿真分析，比较各控制策略下系统调频特性差异。For example, the 4-machine 2-area model of KUNDUR is used to transform and form the simulation system of the present invention to simulate and verify the control strategy in the text. As shown in Figure 2, the simulation system includes 4 synchronous motors with output powers of 120MW, 120MW, 124MW, and 120MW respectively; the DFIG wind farm ignores the differences among the wind turbines, and uses a DFIG equivalent, the total The output power is 80MW; the system loads L1 and L2 are 206MW and 342MW respectively. When the system is in 5s, the synchronous motor G2 stops running due to out-of-step failure, or the wind speed in the simulation area is disturbed, changing from the rated wind speed of 12m/s to 10m/s, and the system frequency change at this time is analyzed. The simulation system model is built on the PSCAD platform, and the simulation analysis is carried out on the rotor kinetic energy control, standby power control, joint control, and Hopfield neural joint control, and the difference of system frequency modulation characteristics under each control strategy is compared.

在PSCAD中建立Hopfield神经控制器A和Hopfield神经控制器B模块如图3所示。两个模块内部均包含一个两神经元的Hopfield神经网络，以神经控制器A为例，如图4所示，将系统的状态输入对应于网络的权值和阈值，神经元的输出作为控制器的参数，非线性微分方程的求解过程是自动完成的，其中sfunction模块为S型函数模块。Establish Hopfield neural controller A and Hopfield neural controller B modules in PSCAD as shown in Figure 3. Both modules contain a two-neuron Hopfield neural network. Taking neural controller A as an example, as shown in Figure 4, the state input of the system corresponds to the weight and threshold of the network, and the output of the neuron is used as the controller parameters, the solution process of the nonlinear differential equation is completed automatically, and the sfunction module is an S-type function module.

通过仿真结果如图5、6比较可以看出，本发明所提出的基于Hopfield神经网络的双馈风电机组一次调频联合控制方法可以实现参数自适应的功能，因而更能够合理安排转子动能控制和备用功率控制的调频比例，比之于恒参数的联合控制具有自适应的优势，所以能进一步缓解系统的频率下降，提高最低频率对改善双馈风电机组一次调频特性具有更理想的控制效果。From the comparison of the simulation results shown in Figures 5 and 6, it can be seen that the Hopfield neural network-based DFIG combined control method for primary frequency modulation proposed by the present invention can realize the function of parameter self-adaptation, so it can more reasonably arrange rotor kinetic energy control and backup Compared with the joint control of constant parameters, the frequency modulation ratio of power control has the advantage of self-adaptation, so it can further alleviate the frequency drop of the system, and increasing the minimum frequency has a more ideal control effect on improving the primary frequency modulation characteristics of double-fed wind turbines.

Claims

1. a double-fed fan motor unit primary frequency modulation combination control method is characterized in that, may further comprise the steps:

(1) in the double-fed blower fan control system, sets up rotor kinetic energy control module and non-firm power control module respectively; The input variable of rotor kinetic energy control module and non-firm power control module is frequency departure; The output variable of rotor kinetic energy control module is the index word of torque reference value, and the input variable of non-firm power control module is the index word of propeller pitch angle reference value;

(2) set up the Hopfield neural net; Confirm the controlling object system model; And it is minimum that target function is that electrical network occurrence frequency when skew system frequency changes, and the state variable that makes controlling object is corresponding to network weight, and with the parameter of neuronic output as the PD controller;

(3) target function expression formula and standard energy function expression are mapped, deriving obtains the inclined to one side value expression of connection weight matrix and network;

(4) connection weight matrix that obtains and network are imported in the dynamical equation of inclined to one side value matrix substitution Hopfield network, and the nonlinear characteristic of getting neuron output is symmetric form S nonlinear interaction function the Changing Pattern of the controlled device parameter of deriving.