A variable pitch control method, device and system based on a BP neural network

Abstract

The present disclosure relates to a variable pitch control method, device and system based on BP neural network. The method comprises: obtaining a first target speed and an actual speed of a wind turbine, determining a first error value between the first target speed and the actual speed, inputting the first target speed, the actual speed and the first error value into a trained BP neural network, obtaining a proportional coefficient and an integral coefficient output by the BP neural network, updating configuration parameters of a PI controller using the proportional coefficient and the integral coefficient, obtaining an updated PI controller, inputting the first error into the updated PI controller, obtaining a first blade angle output by the updated PI controller, and adjusting a blade angle of the wind turbine based on the first blade angle. The present scheme can improve the accuracy of the control parameters of the variable pitch controller.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of wind power generators, and particularly relates to a variable pitch control method, device and system based on a BP neural network. BACKGROUND

[0002] In the related art, when a wind turbine reaches a rated speed, a torque is given constant, and a pitch angle is controlled by a pitch controller to make the absorbed wind energy constant, so as to make the power and the speed constant. In order to achieve good control effect of the pitch PI control, the proportional and integral control effects must be adjusted. The common pitch controller in the wind power field is a classic incremental digital PI controller. In this way, the proportion and the integral in the PID control are in a linear relationship. However, the relationship between the proportion and the integral is not necessarily a simple linear combination, which leads to that the control parameters of the pitch controller are not accurate based on the linear relationship between the proportion and the integral, and further leads to that it is difficult to accurately control the pitch. SUMMARY

[0003] In order to overcome the problems in the related art, the present disclosure provides a variable pitch control method, device and system based on a BP neural network.

[0004] According to a first aspect of the embodiments of the present disclosure, a variable pitch control method based on a BP neural network is provided, comprising:

[0005] obtaining a first target speed and an actual speed of a wind power generator, and determining a first error value between the first target speed and the actual speed;

[0006] inputting the first target speed, the actual speed and the first error value into a trained BP neural network to obtain a proportional coefficient and an integral coefficient output by the BP neural network;

[0007] updating configuration parameters of a PI controller by using the proportional coefficient and the integral coefficient to obtain an updated PI controller;

[0008] inputting the first error into the updated PI controller to obtain a first pitch angle output by the updated PI controller;

[0009] adjusting a pitch angle of the wind power generator based on the first pitch angle, and returning to the operation step of obtaining the first target speed and the actual speed of the wind power generator.

[0010] In some embodiments of the present disclosure, before the operation step of obtaining the first target speed and the actual speed of the wind power generator, the method further comprises:

[0011] In a case where the rotating speed of the wind driven generator reaches a rated rotating speed, a second target rotating speed and an actual rotating speed of the wind driven generator are acquired;

[0012] A difference between the second target rotating speed and the actual rotating speed is calculated to obtain a second error value corresponding to a current parameter optimization round;

[0013] A weighting coefficient in a BP neural network to be trained is corrected based on the second error value to obtain a corrected BP neural network;

[0014] Whether the BP neural network to be trained converges is determined based on the second target rotating speed and the actual rotating speed;

[0015] In a case where it is determined that the BP neural network to be trained does not converge, the step of acquiring the second target rotating speed and the actual rotating speed of the wind driven generator in a case where the rotating speed of the wind driven generator reaches the rated rotating speed is executed until it is determined that the BP neural network to be trained converges, and the trained BP neural network is obtained.

[0016] In some embodiments of the present disclosure, the BP neural network comprises an input layer, a hidden layer and an output layer;

[0017] The correcting the weighting coefficient in the BP neural network to be trained based on the second error value to obtain the corrected BP neural network comprises:

[0018] The first change amount is calculated by using the following formula:

[0019]

[0020] wherein, Δw li (k) is the first change amount of the kth parameter optimization round, Δw li (k-1) is the first change amount of the (k-1)th parameter optimization round, η is a learning rate, and α is an inertia coefficient, is the i-th output value of the hidden layer, e(k) is the second error value, G(β) is a nonlinear function of the pitch angle, y(k) is the actual rotating speed, and u(k) is the first blade angle, is an output value of the network layer, and g(x) is an activation function of the output layer;

[0021] The first change amount is added to an output layer weighting coefficient corresponding to a previous parameter optimization round to obtain a first weighting coefficient;

[0022] The second change amount of the hidden layer is calculated by using the following formula:

[0023]

[0024] wherein, Δwij (k) is the second change amount of the kth parameter optimization round, Aw ij (k-1) is the second change amount of the k-1th parameter optimization round, is the jth input value of the input layer, and f(x) is an activation function of the output layer.

[0025] The second change amount is added to the hidden layer weight coefficient corresponding to the previous parameter optimization round to obtain a second weight coefficient.

[0026] The first weight coefficient is used to correct the weight coefficient of the output layer, and the second weight coefficient is used to correct the weight coefficient of the hidden layer, to obtain the corrected BP neural network.

[0027] In some embodiments of the present disclosure, the determination of whether the to-be-trained BP neural network converges based on the second target rotating speed and the actual rotating speed comprises:

[0028] The performance index function value of the to-be-trained BP neural network is calculated using the following formula:

[0029] J = 0.5(r(k)-y(k)) 2

[0030] wherein J is the performance index function value, r(k) is the second target rotating speed, and y(k) is the actual rotating speed.

[0031] In the case where the performance index function value meets the preset condition, it is determined that the to-be-trained BP neural network converges.

[0032] In the case where the performance index function value does not meet the preset condition, it is determined that the to-be-trained BP neural network does not converge.

[0033] In some embodiments of the present disclosure, the adjustment of the blade angle of the wind turbine based on the first blade angle comprises:

[0034] The current pitch angle of the wind turbine is obtained.

[0035] The second pitch angle is determined based on the current pitch angle and the first pitch angle by using a variable pitch gain table; the variable pitch gain table is obtained based on a nonlinear function of the pitch angle.

[0036] The current pitch angle of the wind turbine is adjusted to the second pitch angle.

[0037] According to a second aspect of the embodiments of the present disclosure, a variable pitch control device based on a BP neural network is provided, comprising:

[0038] The first determining unit is configured to obtain a first target rotating speed and an actual rotating speed of the wind turbine, and determine a first error value between the first target rotating speed and the actual rotating speed.

[0039] The input unit is configured to input the first target rotating speed, the actual rotating speed and the first error value into the trained BP neural network, and obtain a proportional coefficient and an integral coefficient output by the BP neural network.

[0040] The updating unit is configured to update configuration parameters of a PI controller by using the proportional coefficient and the integral coefficient, and obtain an updated PI controller.

[0041] The output unit is configured to input the first error into the updated PI controller, and obtain a first blade angle output by the updated PI controller.

[0042] The adjusting unit is configured to adjust a blade angle of the wind turbine based on the first blade angle, and return to the operation of obtaining the first target rotating speed and the actual rotating speed of the wind turbine.

[0043] In some embodiments of the present disclosure, the apparatus further includes:

[0044] The obtaining unit is configured to obtain a second target rotating speed and an actual rotating speed of the wind turbine when a rotating speed of the wind turbine reaches a rated rotating speed.

[0045] The calculating unit is configured to calculate a difference between the second target rotating speed and the actual rotating speed, and obtain a second error value corresponding to a current parameter optimization round.

[0046] The correcting unit is configured to correct a weighting coefficient in the BP neural network to be trained based on the second error value, and obtain a corrected BP neural network.

[0047] The second determining unit is configured to determine whether the BP neural network to be trained converges based on the second target rotating speed and the actual rotating speed.

[0048] The executing unit is configured to return to the step of obtaining the second target rotating speed and the actual rotating speed of the wind turbine when the rotating speed of the wind turbine reaches the rated rotating speed until it is determined that the BP neural network to be trained converges, and obtain the trained BP neural network, when it is determined that the BP neural network to be trained does not converge.

[0049] According to a third aspect of an embodiment of the present disclosure, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method according to any one of the first aspect when executing the computer program.

[0050] According to a fourth aspect of the embodiments of the present disclosure, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the method according to any one of the first aspect.

[0051] According to a fifth aspect of the embodiments of the present disclosure, a computer program product is provided, and the computer program product comprises a computer program. The computer program is executed by a processor to implement the method according to any one of the first aspect.

[0052] The technical scheme provided by the embodiments of the present disclosure can have the following beneficial effects: by obtaining the first target rotating speed and the actual rotating speed of the wind turbine, determining the first error value between the first target rotating speed and the actual rotating speed, inputting the first target rotating speed, the actual rotating speed and the first error value into the BP neural network which is trained, obtaining the proportional coefficient and the integral coefficient output by the BP neural network, updating the configuration parameters of the PI controller by using the proportional coefficient and the integral coefficient, obtaining the updated PI controller, inputting the first error into the updated PI controller, obtaining the first blade angle output by the updated PI controller, and adjusting the blade angle of the wind turbine based on the first blade angle. The proportional coefficient and the integral coefficient in the BP neural network are optimized in a non-linear manner by using the BP neural network, so as to improve the accuracy of the control parameters of the variable pitch controller, and further improve the accuracy of the control of the blades.

[0053] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0054] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0055] Figure 1 FIG. 1 is a flowchart of a variable pitch control method based on a BP neural network according to an exemplary embodiment.

[0056] Figure 2 FIG. 2 is a schematic diagram of a BP neural network structure according to an exemplary embodiment.

[0057] Figure 3 FIG. 3 is a variable pitch control flowchart based on a BP neural network according to an exemplary embodiment.

[0058] Figure 4 FIG. 4 is a block diagram of a variable pitch control device based on a BP neural network according to an exemplary embodiment.

[0059] Figure 5is a block diagram of an apparatus for a BP neural network based pitch control method according to an exemplary embodiment. DETAILED DESCRIPTION

[0060] The exemplary embodiments will be described in detail herein with reference to the attached drawings. In the following description, like reference numerals refer to like elements, unless the context clearly dictates otherwise. The following exemplary embodiments are not intended to represent all embodiments in accordance with the present disclosure. Rather, they are merely examples of apparatuses and methods in accordance with some aspects of the present disclosure as detailed in the appended claims.

[0061] The terminology used in the present disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the present disclosure. As used in the present disclosure and the appended claims, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0062] It should be understood that although the terms first, second, third, etc. can be used herein to describe various information, these terms are not intended to denote a particular order of the information. These terms are used merely to distinguish one type of information from another. For example, a first information can be termed a second information, and similarly, a second information can be termed a first information, without departing from the scope of the present disclosure. Depending on the context, the word "if' and "when' as used herein can be interpreted to mean "upon determining" or "in response to determining."

[0063] In addition, the steps of various forms shown in the present disclosure can be reordered, added or deleted. For example, the steps described in the present disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, which are not limited herein.

[0064] In the related art, after the wind turbine reaches the rated speed, the torque is constant, and the pitch angle is controlled by the pitch controller to make the absorbed wind energy constant, thereby making the power and speed constant. In order to achieve good control effect, the proportional and integral control effects must be adjusted. The common pitch controller in the wind power field is a classic incremental digital PI controller. In this way, the proportion and integral in the PID control are in a linear relationship. However, the relationship between the proportion and the integral is not necessarily a simple linear combination, which leads to inaccurate determination of the control parameters of the pitch controller based on the linear relationship between the proportion and the integral, and further leads to difficulty in precise control of the pitch.

[0065] To solve the above problems, the present disclosure provides a variable pitch control method, device and system based on BP neural network. The first target speed and the actual speed of the wind turbine are obtained, the first error value between the first target speed and the actual speed is determined, the first target speed, the actual speed and the first error value are input into the trained BP neural network, the proportional coefficient and the integral coefficient output by the BP neural network are obtained, the configuration parameters of the PI controller are updated using the proportional coefficient and the integral coefficient, the updated PI controller is obtained, the first error is input into the updated PI controller, the first blade angle output by the updated PI controller is obtained, and the blade angle of the wind turbine is adjusted based on the first blade angle. The proportional coefficient and the integral coefficient in the BP neural network are optimized in a nonlinear manner through the BP neural network, thereby improving the accuracy of the control parameters of the variable pitch controller, and further improving the accuracy of controlling the blades.

[0066] Figure 1 is a flow chart of a variable pitch control method based on BP neural network according to an exemplary embodiment. As shown in Figure 1 It should be noted that the variable pitch control method based on BP neural network in the embodiments of the present disclosure is applied to a variable pitch control device based on BP neural network. As shown in Figure 1 The method can include the following steps:

[0067] Step 101, obtaining the first target speed and the actual speed of the wind turbine, and determining the first error value between the first target speed and the actual speed.

[0068] In one embodiment, the first target speed can be a target speed preset according to actual demand.

[0069] As an example, the first target speed can be subtracted from the actual speed to obtain the first error value

[0070] Step 102, inputting the first target speed, the actual speed and the first error value into the trained BP neural network to obtain the proportional coefficient and the integral coefficient output by the BP neural network.

[0071] In one embodiment, the first target speed, the actual speed and the first error value can be input into the trained BP neural network, and the proportional coefficient and the integral coefficient can be calculated by the BP neural network based on the first target speed, the actual speed and the first error value.

[0072] In some embodiments of the present application, as Figure 2As shown, the BP neural network can be divided into network input layer, network hidden layer and network output layer. In the present application, the number of network hidden layers is 1. Assuming that the input layer has M input nodes, Q hidden layer nodes and 2 output nodes. The input nodes correspond to the selected system operating state quantities, which can generally be the input and output quantities of the system at different times, and r(k), y(k), e(k) and 1 are selected here. The output nodes correspond to the 2 parameters of the PI controller. The hidden layer neuron activation function f(x) is taken as the positive and negative symmetric Sigmoid function, and the output layer neuron activation function g(x) is taken as the non-negative Sigmoid function. It is agreed that the upper subscripts (1), (2) and (3) represent the input layer, the hidden layer and the output layer respectively.

[0073] In one embodiment, the network input layer can be represented by the following formula:

[0074]

[0075] wherein j = 1, 2, …, M, x(j) is the system input node, is the network input layer.

[0076] The input of the network hidden layer is:

[0077]

[0078] wherein w ij is the network hidden layer weighting coefficient.

[0079] The output of the network hidden layer is:

[0080]

[0081] wherein i = 1, 2, …, Q, and f(x) is the activation function of the network hidden layer.

[0082] The input of the network output layer is:

[0083]

[0084] wherein w li is the weighting coefficient of the network output layer.

[0085] The output of the network output layer is:

[0086]

[0087] wherein l = 1, 2, g(x) is the activation function of the network output layer, Kp is the updated proportional coefficient of the PI controller, and Ki is the updated integral coefficient of the PI controller.

[0088] Step 103, updating the configuration parameters of the PI controller by using the proportional coefficient and the integral coefficient to obtain an updated PI controller.

[0089] In one embodiment, the current proportional coefficient of the PI controller is updated as the proportional coefficient output by the BP neural network, and the current integral coefficient of the PI controller is updated as the integral coefficient output by the BP neural network.

[0090] Step 104, inputting the first error into the updated PI controller to obtain a first blade angle output by the updated PI controller.

[0091] In one embodiment, the PI controller can calculate the first blade angle by using the following formula:

[0092] u(k) = u(k-1) + Au(k)

[0093] Au(k) = Kp*(e(k) - e(k-1)) + Ki*(e(k))

[0094] wherein u(k) is the output of the PI controller at the current kth control round (i.e., the first blade angle), u(k-1) is the output of the PI controller at the last control round, e(k) is the system error (i.e., the first error value), i.e., the input of the PI controller, Kp is the updated proportional coefficient of the PI controller, and Ki is the updated integral coefficient of the PI controller.

[0095] It should be noted that before the wind turbine is controlled for the first time, the proportional coefficient and the integral coefficient can be initialized.

[0096] Step 105, adjusting the blade angle of the wind turbine based on the first blade angle, and returning to the operation of obtaining the first target speed and the actual speed of the wind turbine.

[0097] In one embodiment, steps 101 to 105 can be executed cyclically during the operation of the wind turbine until the wind turbine stops operating.

[0098] In some embodiments of the present application, step 105 can specifically include the following steps:

[0099] obtaining a current pitch angle of the wind turbine;

[0100] determining a second pitch angle based on the current pitch angle and the first blade angle by using a variable pitch gain table; the variable pitch gain table is obtained based on a nonlinear function of the pitch angle;

[0101] adjusting the current pitch angle of the wind turbine to the second pitch angle.

[0102] In one embodiment, asFigure 3 As shown, after the PI controller calculates the first blade angle u(k), the second blade angle u'(k) is determined based on the current blade angle and the first blade angle using the pitch gain table (i.e., the nonlinear function G(β) based on the blade pitch angle), and the current blade pitch angle of the wind turbine is adjusted to the second blade pitch angle u'(k).

[0103] In some embodiments of this application, such as Figure 3 As shown, the first target speed r(k) and actual speed y(k) of the wind turbine are obtained, and the first error value e(k) between r(k) and y(k) is determined. r(k), y(k) and e(k) are input into the trained BP neural network to obtain the proportional coefficient Kp and integral coefficient Ki of the BP neural network output. The configuration parameters of the PI controller are updated using Kp and Ki to obtain the updated PI controller. e(k) is input into the updated PI controller to obtain the first blade angle u(k) output by the updated PI controller. The blade angle of the wind turbine is adjusted based on the first blade angle. Using the pitch gain table (i.e., the nonlinear function G(β) based on the pitch angle), the second pitch angle u'(k) is determined based on the current pitch angle and the first pitch angle. The current pitch angle of the wind turbine is adjusted to the second pitch angle u'(k).

[0104] In some embodiments of this application, prior to step 101, the method may further include:

[0105] Step a1: When the wind turbine reaches its rated speed, obtain the second target speed and the actual speed of the wind turbine.

[0106] Step a2: Calculate the difference between the second target speed and the actual speed to obtain the second error value corresponding to the current parameter optimization cycle.

[0107] Step a3: Based on the second error value, the weighting coefficients in the BP neural network to be trained are corrected to obtain the corrected BP neural network.

[0108] In some embodiments of this application, the BP neural network includes an input layer, a hidden layer, and an output layer, and step a3 may specifically include the following steps:

[0109] The first change is calculated using the following formula:

[0110]

[0111]

[0112] Where, Δw li (k) represents the first change in the k-th parameter optimization round, Δw li(k-1) is the first change of the (k-1)th parameter optimization round, η is a learning rate, and α is an inertia coefficient, is the i-th output value of the hidden layer, e(k) is a second error value, G(β) is a nonlinear function of the pitch angle, y(k) is an actual rotating speed, and u(k) is a first blade angle, is an output value of the network layer, and g(x) is an activation function of the output layer;

[0113] The first change is added to the output layer weighting coefficient corresponding to the last parameter optimization round to obtain a first weighting coefficient;

[0114] The second change of the hidden layer is calculated by using the following formula:

[0115]

[0116] where Δw ij (k) is the second change of the kth parameter optimization round, Δw ij (k-1) is the second change of the (k-1)th parameter optimization round, is the j-th input value of the input layer, and f(x) is an activation function of the output layer;

[0117] The second change is added to the hidden layer weighting coefficient corresponding to the last parameter optimization round to obtain a second weighting coefficient;

[0118] The first weighting coefficient is used to correct the weighting coefficient of the output layer, and the second weighting coefficient is used to correct the weighting coefficient of the hidden layer, to obtain a corrected BP neural network.

[0119] In an embodiment, the network weighting coefficient can be corrected according to the gradient descent method and additionally an inertia term for quickly converging to a global minimum.

[0120] As an example, the change Δw li (k) of the output layer network weighting coefficient in the kth parameter optimization round can be represented by the following formula:

[0121]

[0122] where η is a learning rate, and α is an inertia coefficient.

[0123] The partial derivative term in the above formula can be derived as follows:

[0124]

[0125] where

[0126] Unknown, the symbol function substitute; The nonlinear gain scheduling function of the variable pitch controller is a function of the pitch angle β, and is only related to the actual pitch angle fed back by the wind turbine;

[0127] For When l = 1,

[0128] When l = 2, Wherein Derivation of the output layer activation function:

[0129]

[0130] Wherein, The hidden layer output.

[0131] Therefore, the output layer network weighting coefficient learning algorithm can be simplified as:

[0132]

[0133] Similarly, the hidden layer network weighting coefficient learning algorithm is:

[0134]

[0135] Wherein,

[0136] Step a4, determining whether the to-be-trained BP neural network converges based on the second target rotating speed and the actual rotating speed.

[0137] In some embodiments of the present application, step a4 can specifically include the following steps:

[0138] The performance index function value of the to-be-trained BP neural network is calculated by using the following formula:

[0139] J = 0.5(r(k)-y(k)) 2

[0140] Wherein, J is the performance index function value, r(k) is the second target rotating speed, and y(k) is the actual rotating speed;

[0141] In the case that the performance index function value meets the preset condition, it is determined that the to-be-trained BP neural network converges;

[0142] In the case that the performance index function value does not meet the preset condition, it is determined that the to-be-trained BP neural network does not converge.

[0143] Step a5, in the case that it is determined that the to-be-trained BP neural network does not converge, returning to execute step a1 until it is determined that the to-be-trained BP neural network converges, and obtaining the BP neural network after training.

[0144] In one embodiment, the variable pitch nonlinear function G(β) can be initialized in advance according to the blade parameters of the wind turbine before step a1 is performed. In addition, the network weighting coefficient, learning rate and inertia coefficient can be initialized in advance.

[0145] According to the variable pitch control method based on the BP neural network provided in the embodiments of the present disclosure, by obtaining the first target rotating speed and the actual rotating speed of the wind turbine, a first error value between the first target rotating speed and the actual rotating speed is determined, the first target rotating speed, the actual rotating speed and the first error value are input into the trained BP neural network to obtain the proportional coefficient and the integral coefficient output by the BP neural network, the configuration parameters of the PI controller are updated by using the proportional coefficient and the integral coefficient to obtain an updated PI controller, the first error is input into the updated PI controller to obtain the first blade angle output by the updated PI controller, and the blade angle of the wind turbine is adjusted based on the first blade angle. The proportional coefficient and the integral coefficient in the BP neural network are optimized in a nonlinear manner by the BP neural network, so that the accuracy of the control parameters of the variable pitch controller can be improved, and the accuracy of the control of the blades is further improved.

[0146] Figure 4 is a variable pitch control device block diagram based on a BP neural network according to an exemplary embodiment. Referring to Figure 4 The device includes a first determination unit 401, an input unit 402, an update unit 403, an output unit 404 and an adjustment unit 405.

[0147] The first determination unit 401 is configured to obtain the first target rotating speed and the actual rotating speed of the wind turbine, and determine a first error value between the first target rotating speed and the actual rotating speed.

[0148] The input unit 402 is configured to input the first target rotating speed, the actual rotating speed and the first error value into the trained BP neural network to obtain the proportional coefficient and the integral coefficient output by the BP neural network.

[0149] The update unit 403 is configured to update the configuration parameters of the PI controller by using the proportional coefficient and the integral coefficient to obtain an updated PI controller.

[0150] The output unit 404 is configured to input the first error into the updated PI controller to obtain the first blade angle output by the updated PI controller.

[0151] The adjustment unit 405 is configured to adjust the blade angle of the wind turbine based on the first blade angle, and return to the operation step of obtaining the first target rotating speed and the actual rotating speed of the wind turbine.

[0152] In some embodiments of the present application, the device further comprises:

[0153] The acquisition unit is configured to acquire the second target rotating speed and the actual rotating speed of the wind power generator when the rotating speed of the wind power generator reaches the rated rotating speed.

[0154] The calculation unit is configured to calculate a difference between the second target rotating speed and the actual rotating speed to obtain a second error value corresponding to the current parameter optimization round.

[0155] The correction unit is configured to correct the weighting coefficients in the BP neural network to be trained based on the second error value to obtain a corrected BP neural network.

[0156] The second determination unit is configured to determine whether the BP neural network to be trained converges based on the second target rotating speed and the actual rotating speed.

[0157] The execution unit is configured to return to execute the step of acquiring the second target rotating speed and the actual rotating speed of the wind power generator when the rotating speed of the wind power generator reaches the rated rotating speed until it is determined that the BP neural network to be trained converges to obtain the trained BP neural network, in a case where it is determined that the BP neural network to be trained does not converge.

[0158] In some embodiments of the present application, the BP neural network comprises an input layer, a hidden layer and an output layer, and the correction unit can be specifically configured to execute the following steps:

[0159] The first change amount is calculated by using the following formula:

[0160]

[0161] wherein, Δw li (k) is the first change amount of the kth parameter optimization round, Δw li (k-1) is the first change amount of the (k-1)th parameter optimization round, η is a learning rate, and α is an inertia coefficient, is the i-th output value of the hidden layer, e(k) is the second error value, G(β) is a nonlinear function of the pitch angle, y(k) is the actual rotating speed, and u(k) is the first blade angle, is the output value of the network layer, and g(x) is an activation function of the output layer.

[0162] The first change amount is added to the output layer weighting coefficient corresponding to the previous parameter optimization round to obtain a first weighting coefficient.

[0163] The second change amount of the hidden layer is calculated by using the following formula:

[0164]

[0165] wherein, Δw ij(k) is the second change amount of the kth parameter optimization round, Aw ij (k-1) is the second change amount of the k-1th parameter optimization round, is the jth input value of the input layer, and f(x) is an activation function of the output layer.

[0166] The second change amount is added to the hidden layer weight coefficient corresponding to the last parameter optimization round to obtain a second weight coefficient.

[0167] The first weight coefficient is used to correct the weight coefficient of the output layer, and the second weight coefficient is used to correct the weight coefficient of the hidden layer, to obtain a corrected BP neural network.

[0168] In some embodiments of the present application, the second determining unit can be specifically configured to perform the following steps:

[0169] The performance index function value of the BP neural network to be trained is calculated using the following formula:

[0170] J = 0.5(r(k)-y(k)) 2

[0171] Wherein, J is the performance index function value, r(k) is the second target speed, and y(k) is the actual speed.

[0172] In the case where the performance index function value meets the preset condition, it is determined that the BP neural network to be trained converges;

[0173] In the case where the performance index function value does not meet the preset condition, it is determined that the BP neural network to be trained does not converge.

[0174] In some embodiments of the present application, the execution unit can be specifically configured to perform the following steps:

[0175] Obtaining the current pitch angle of the wind turbine;

[0176] Determining the second pitch angle based on the current pitch angle and the first pitch angle using the variable pitch gain table; the variable pitch gain table is obtained based on a nonlinear function of the pitch angle;

[0177] Adjusting the current pitch angle of the wind turbine to the second pitch angle. As for the device in the above embodiments, the specific manner in which each module performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

[0178] The variable pitch control device based on the BP neural network according to the embodiment of the present disclosure obtains a first target rotating speed and an actual rotating speed of the wind driven generator, determines a first error value between the first target rotating speed and the actual rotating speed, inputs the first target rotating speed, the actual rotating speed and the first error value into the trained BP neural network, obtains a proportional coefficient and an integral coefficient output by the BP neural network, updates the configuration parameters of the PI controller by using the proportional coefficient and the integral coefficient, obtains an updated PI controller, inputs the first error into the updated PI controller, obtains a first blade angle output by the updated PI controller, and adjusts the blade angle of the wind driven generator based on the first blade angle. The proportional coefficient and the integral coefficient in the BP neural network are optimized in a non-linear manner by using the BP neural network, so that the accuracy of the control parameters of the variable pitch controller can be improved, and the accuracy of the control on the blade is improved.

[0179] Figure 5 is a block diagram of an apparatus for a variable pitch control method based on a BP neural network according to an example embodiment. For example, the apparatus 500 can be an electronic device, such as a programmable logic controller (PLC), a computer, etc.

[0180] Referring to Figure 5 , the apparatus 500 can include one or more of the following components: a processing component 502, a memory component 504, a power supply component 506, a multimedia component 508, a sensor component 514, and a communication component 516.

[0181] The processing component 502 usually controls overall operations of the apparatus 500, such as display, data communications. The processing component 502 can include one or more processors 520 to execute instructions, to complete all or part of the steps of the methods described above. In addition, the processing component 502 can include one or more modules to facilitate the interaction between the processing component 502 and other components. For example, the processing component 502 can include a multimedia module to facilitate the interaction between the multimedia component 508 and the processing component 502.

[0182] The memory 504 is configured to store various types of data to support the operation of the apparatus 500. Examples of these data include instructions for operating any application or method on the apparatus 500. The memory 504 can be realized by any type of volatile or non-volatile storage devices or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage or flash memory.

[0183] The power component 506 provides power to the various components of the device 500. The power component 506 can include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 500.

[0184] The multimedia component 508 includes a screen providing an output interface between the device 500 and a user. In some embodiments, the screen can include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, swiping, and gestures on the touch panel. The touch sensors can not only sense a boundary of a touching or swiping action, but also detect duration and pressure related to the touching or swiping action.

[0185] The sensor component 514 includes one or more sensors to provide the device 500 with state assessments in various aspects. For example, the sensor component 514 can include an absolute encoder to assess a position of a rotor, an incremental encoder to assess a rotational speed of the rotor, an anemometer to assess a wind speed, and other sensors for a wind turbine, such as a vibration sensor, a temperature sensor, a pressure sensor, and a wind direction vane.

[0186] The communication component 516 is configured to facilitate wired or wireless communication between the device 500 and other devices. For example, an industrial bus, such as modbus-tcp, modbus-rtu, canopen, profibus, and the like, can be used to implement network communication for a wind turbine.

[0187] In example embodiments, the device 500 can be implemented using one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, or other electronic elements to perform the methods described above.

[0188] In example embodiments, a non-transitory computer-readable storage medium including instructions, such as the memory 504 including instructions, is also provided. The instructions can be executable by the processor 520 of the device 500 to perform the methods described above. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disc, and an optical data storage device, and the like.

[0189] In example embodiments, a computer program product including a computer program is also provided. The computer program can be executed by the processor 520 of the device 500 to implement the methods described above.

[0190] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0191] It is to be understood that the application is not limited to the precise structures hereinabove described and shown in the drawings, for purposes of illustration and education only, and that variations and changes can be made by persons skilled in the art in the implementation of the application without departing from the scope thereof. The scope of the application is limited only by the claims that follow.

Claims

1. A method for pitch control based on BP neural network, characterized in that, The method comprises the following steps: obtaining a first target speed and an actual speed of a wind turbine, and determining a first error value between the first target speed and the actual speed; inputting the first target speed, the actual speed and the first error value into a trained BP neural network to obtain a proportional coefficient and an integral coefficient output by the BP neural network; updating configuration parameters of a PI controller using the proportional coefficient and the integral coefficient to obtain an updated PI controller; inputting the first error value into the updated PI controller to obtain a first blade angle output by the updated PI controller; adjusting a blade angle of the wind turbine based on the first blade angle, and returning to the step of obtaining the first target speed and the actual speed of the wind turbine; wherein, before the step of obtaining the first target speed and the actual speed of the wind turbine, the method further comprises: obtaining a second target speed and an actual speed of the wind turbine when the speed of the wind turbine reaches a rated speed; calculating a difference between the second target speed and the actual speed to obtain a second error value corresponding to a current parameter optimization round; correcting a weighting coefficient in a to-be-trained BP neural network based on the second error value to obtain a corrected BP neural network; determining whether the to-be-trained BP neural network converges based on the second target speed and the actual speed; when it is determined that the to-be-trained BP neural network does not converge, returning to the step of obtaining the second target speed and the actual speed of the wind turbine when the speed of the wind turbine reaches the rated speed until it is determined that the to-be-trained BP neural network converges, and obtaining the trained BP neural network; wherein, the BP neural network comprises an input layer, a hidden layer and an output layer; the step of correcting the weighting coefficient in the to-be-trained BP neural network based on the second error value to obtain the corrected BP neural network comprises: calculating a first change amount using the following formula: where Δw li (k) is the first change of the kth parameter optimization round, Δw li (k-1) is the first change of the k-1th parameter optimization round, η is the learning rate; α is the inertia coefficient, is the i-th output value of the hidden layer, e(k) is the second error value, G(β) is a nonlinear function of the pitch angle, y(k) is the actual speed, u(k) is the first blade angle, is the output value of the network layer, g(x) is the activation function of the output layer; adding the first change amount to an output layer weighting coefficient corresponding to a previous parameter optimization round to obtain a first weighting coefficient; calculating a second change amount of the hidden layer using the following formula: where Δw ij (k) is the second change amount of the kth parameter optimization round, Δw ij (k-1) is the second change amount of the k-1th parameter optimization round, is the jth input value of the input layer, and f(x) is the activation function of the output layer. adding the second change amount to a hidden layer weighting coefficient corresponding to the previous parameter optimization round to obtain a second weighting coefficient; correcting the weighting coefficient of the output layer using the first weighting coefficient and correcting the weighting coefficient of the hidden layer using the second weighting coefficient to obtain the corrected BP neural network.

2. The BP neural network based pitch control method of claim 1, wherein, the step of determining whether the to-be-trained BP neural network converges based on the second target speed and the actual speed comprises: calculating a performance index function value of the to-be-trained BP neural network using the following formula: J = 0.5(r(k) - y(k)) 2 wherein, J is the performance index function value, r(k) is the second target speed, and y(k) is the actual speed; when the performance index function value meets a preset condition, it is determined that the to-be-trained BP neural network converges; when the performance index function value does not meet the preset condition, it is determined that the to-be-trained BP neural network does not converge.

3. The BP neural network based pitch control method of claim 1, wherein, The adjusting the blade angle of the wind turbine based on the first blade angle comprises: obtaining a current pitch angle of the wind turbine; determining a second pitch angle based on the current pitch angle and the first blade angle by using a pitch gain table; the pitch gain table is obtained based on a nonlinear function of the pitch angle; adjusting the current pitch angle of the wind turbine to the second pitch angle.

4. A variable pitch control device based on BP neural network, characterized in that, The method of any one of claims 1-3, comprising: a first determining unit configured to obtain a first target rotating speed and an actual rotating speed of the wind turbine, and determine a first error value between the first target rotating speed and the actual rotating speed; an input unit configured to input the first target rotating speed, the actual rotating speed and the first error value into a BP neural network which is trained, to obtain a proportional coefficient and an integral coefficient output by the BP neural network; an updating unit configured to update configuration parameters of a PI controller by using the proportional coefficient and the integral coefficient, to obtain an updated PI controller; an output unit configured to input the first error value into the updated PI controller, to obtain a first blade angle output by the updated PI controller; an adjusting unit configured to adjust a blade angle of the wind turbine based on the first blade angle, and return to the operation of obtaining the first target rotating speed and the actual rotating speed of the wind turbine; wherein the device further comprises: an obtaining unit configured to obtain a second target rotating speed and an actual rotating speed of the wind turbine in a case that a rotating speed of the wind turbine reaches a rated rotating speed; a calculating unit configured to calculate a difference between the second target rotating speed and the actual rotating speed, to obtain a second error value corresponding to a current parameter optimization round; a correcting unit configured to correct a weighting coefficient in a BP neural network to be trained based on the second error value, to obtain a corrected BP neural network; a second determining unit configured to determine whether the BP neural network to be trained converges based on the second target rotating speed and the actual rotating speed; an executing unit configured to return to the step of obtaining the second target rotating speed and the actual rotating speed of the wind turbine in the case that the rotating speed of the wind turbine reaches the rated rotating speed until it is determined that the BP neural network to be trained converges, to obtain the BP neural network which is trained, in a case that it is determined that the BP neural network to be trained does not converge.

5. An electronic device, comprising: comprise: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method of any one of claims 1-3 when executing the computer program.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method of any one of claims 1-3.

7. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the method of any one of claims 1-3.

Images