A control method for wind turbine hub cooling system based on neural network PID

Abstract

This invention discloses a neural network PID control method for a wind turbine hub cooling system. The method includes establishing a control system, including obtaining key parameters and establishing a thermal balance equation and a transfer function for the temperature control process; formulating a control strategy, including setting a target temperature, a centrifugal fan airflow range, and employing a neural network PID control method; and verifying the stability and reliability of the system through simulation and actual machine operation. The advantages of this invention include improving the responsiveness and control accuracy of the cooling system and enabling the cooling system to be dynamically adjusted.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation and energy system control, and in particular to a wind turbine hub heat dissipation system control method based on neural network PID. Background Art

[0002] To protect offshore wind turbines from erosion by salt spray, rain, and other factors, the rotor section often uses a sealed structure. The rotor section primarily consists of the hub, blades, a shroud, and internal components. The heat load within the rotor primarily originates from internal operating components during operation, including the pitch controller, pitch motor, and main shaft. When the unit is operating, components generate heat and transfer heat through their surfaces to the internal hub environment. When both the ambient temperature and solar radiation are high, the air temperature inside the hub can become excessively high, potentially affecting the normal operation of the generator, reducing power generation efficiency, and even causing the unit to shut down.

[0003] The Chinese invention patent "Temperature Control System for Wind Turbine Generators" (Publication No. CN103184984A) discloses a heat dissipation system for cooling the hub. This system uses a ventilator to discharge hot air from the hub into ducts in the blades. The hot air in the ducts transfers heat from the blades to the outside, forming cool air that is then injected into the hub, repeating the cycle to cool the hub. While this technology can cool the hub, it does not explain how to optimally configure the heat dissipation parameters of the heat dissipation system based on the wind turbine model.

[0004] The Chinese invention patent "Coupled Verification and Design Method for the Cooling System of the Wind Rotor of a Wind Turbine Generator" (Publication No. CN111695255A) discloses a triple iterative coupling algorithm for verifying and designing the cooling parameters of the cooling system for wind turbines whose wind rotors use an air-cooled cooling system. This technology verifies the cooling capacity of the wind rotor cooling system through iterative calculation using a dichotomy method by collecting the equipment parameters, environmental parameters, cooling parameters of the wind rotor cooling system, and setting the temperature parameters of the wind turbine. This technology solves the problem of how to configure the cooling parameters based on various parameters at a certain moment, but the external environmental parameters will change over time, and most centrifugal fans have the characteristic of adjustable air volume. This technology does not reflect the characteristics of dynamically adjusting the cooling system under different external environments. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method for controlling a wind turbine hub heat dissipation system based on a neural network PID.

[0006] In order to achieve the above object of the invention, the technical solution adopted by the present invention is as follows:

[0007] A method for controlling a wind turbine hub heat dissipation system based on a neural network PID comprises the following steps:

[0008] Step 1: Establish a mathematical model of the control system, including obtaining important parameters, establishing the thermal balance equation and the transfer function of the temperature control process

[0009] Obtain important parameters, including: obtaining external environment parameters and heat transfer parameters;

[0010] Obtain external environmental parameters, including: ambient temperature T air , ambient solar radiation Φ sun and ambient wind speed u s3 ;

[0011] Calculate heat transfer parameters, including: Calculate the air velocity u on the inner surface of the hub inside the wind wheel c1 , air velocity u on the outer surface of the hub s1 and the air velocity u on the inner surface of the shroud s2 ; and further calculate the convection heat transfer coefficient h inside the hub c1 , the convection heat transfer coefficient h of the hub outer surface s1 and the convection heat transfer coefficient h on the inner surface of the shroud s2 , the convection heat transfer coefficient h of the outer surface of the shroud s3 ; Calculate the thermal resistance r between the hub and the air shroud, and finally use the above parameters to calculate the total heat transfer coefficient K1 from the inside of the hub to the outer surface of the air shroud;

[0012] Establish the thermal balance equation, including: calculating the reference temperature T of the outer surface of the shroud sur1* , calculate the equivalent thermal resistance R; and based on the heat load Φ0 transferred from the hub to the outer surface of the air shroud when the cooling system is not turned on, calculate the temperature difference ΔT0 between the internal temperature of the hub and the reference temperature in this state;

[0013] Establish the transfer function of the temperature control process: Based on the system parameters and heat transfer parameters of the wind turbine itself, the transfer function G(s) of the hub internal temperature control process is constructed, and a certain sampling period is set to update the environmental parameters, heat transfer parameters and transfer function G(s);

[0014] Step 2: Develop a control strategy, including:

[0015] Determine the control target, including setting the target temperature range and centrifugal fan air volume range to ensure the hub is at the appropriate temperature;

[0016] Step 3: Simulation operation of the control system

[0017] A closed-loop control system is constructed based on the mathematical model of the control system, and the particle swarm algorithm is used in the simulation environment to provide the optimal initial PID parameters for the neural network PID controller.

[0018] Monitor the parameter values ​​of the neural network input variables, including the internal temperature of the wheel hub. The neural network PID controller will further adjust the PID parameters based on the parameter values ​​of the input variables.

[0019] The neural network PID controller will output control instructions based on the current PID parameters and dynamically adjust the ventilation volume of the centrifugal fan to maintain the temperature within the target range;

[0020] Step 4: Conduct actual machine verification of the control system to ensure its stability and reliability under different working conditions; when the wind turbine is running, synchronously start the air cooling system control device, monitor the temperature data in real time and judge the preset cooling demand. If the unit meets the cooling demand, further start the cooling centrifugal fan to test the cooling system.

[0021] Furthermore, the air flow rate of each part inside the wind wheel is calculated, wherein the calculation formula of the air flow rate inside the hub is:

[0022]

[0023] Where D1 is the diameter of the inner surface of the hub, and n is the rated rotation speed of the wind wheel;

[0024] Air velocity u on the outer surface of the hub s1 and the air velocity u on the inner surface of the shroud s2 The calculation method refers to the air flow velocity u on the inner surface of the hub c1 ;

[0025] The heat transfer coefficient h of the outer surface of the guide cover is calculated in the heat transfer parameter calculation. s3 , the formula is:

[0026]

[0027] Where x s3 is the length of the air flow section on the outer surface of the air guide cover, λ s3 is the thermal conductivity of the air on the outer surface of the shroud, Nu s3 is the Nusselt number of the outer surface of the shroud;

[0028] The calculation of the convection heat transfer coefficient h inside the hub c1 , the formula is:

[0029]

[0030] Where, ρ cand ρ0 are the air densities inside the hub and on the sea, respectively, u c1 is the air velocity inside the hub;

[0031] Convection heat transfer coefficient h on the outer surface of the hub s1 and the convection heat transfer coefficient h on the inner surface of the shroud s2 The calculation method refers to the convection heat transfer coefficient h inside the hub c1 ;

[0032] The thermal resistance r between the hub and the fairing is calculated using the following formula:

[0033]

[0034] Where δ1 and δ2 are the thickness of the hub and the fairing, respectively, and λ1 and λ2 are the thermal conductivities of the corresponding materials;

[0035] Use the parameters calculated above to calculate the inverse of the total heat transfer resistance, that is, the total heat transfer coefficient K1. The formula is as follows:

[0036]

[0037] Furthermore, the heat balance equation is established as follows:

[0038]

[0039] Where, Φ in is the heat load transferred from the hub to the fairing in thermal equilibrium, and ΔT is the internal temperature of the hub, T in The reference temperature T of the outer surface of the shroud sur1* The temperature difference between the two; R is the equivalent thermal resistance, which represents the heat transfer capacity from the inside of the hub to the external environment and is defined by the following formula:

[0040]

[0041] Where A1 is the heat transfer area between the shroud and the hub, K1 is the total heat transfer coefficient from the hub interior to the shroud exterior; A2 is the shroud exterior area, h s3 is the convection heat transfer coefficient of the outer surface of the guide cover, ε1 is the emissivity of the outer surface of the guide cover, σ is the Stefan-Boltzmann constant, T sur1* is the reference temperature of the outer surface of the shroud.

[0042] Furthermore, the transfer function of the temperature control process is as follows:

[0043]

[0044] Where Δμ represents the change in heat load transferred from the hub to the air shroud, and ΔT represents the temperature difference between the hub internal temperature and the reference temperature of the air shroud outer surface. The transfer function G1(s) represents the response of the cooling centrifugal fan from the initial air volume to the new target air volume that matches the heat load change Δμ. The transfer function G2(s) represents the response of the temperature difference ΔT between the hub internal temperature and the reference temperature from one initial stable state to another stable state when the ventilation volume changes. s is the Laplace transform variable, which represents the complex frequency.

[0045] Furthermore, the updating steps of the environmental parameters, heat transfer parameters and transfer function G(s) are as follows:

[0046] a. Sampling period setting: Set a suitable sampling update period according to the environment in which the wind turbine is located;

[0047] b. Get the latest environmental data: Whenever the predetermined sampling period is reached, get the latest ambient temperature T air , ambient wind speed u s3 and solar radiation Φ sun ;

[0048] c. Calculate and update heat transfer parameters: recalculate heat transfer parameters using new environmental parameters;

[0049] d. Update transfer function: Update the transfer function G2(s) according to the new parameters;

[0050] e. Adjust control strategy: Use the updated parameters and integrate them into the system’s control loop to ensure that the system can adjust the control strategy according to changes in the external environment.

[0051] Furthermore, the specific steps of constructing the neural network PID controller method are as follows:

[0052] 1) Initialize the neural network parameters, including:

[0053] Initialize the connection weights between the input layer and the hidden layer and the connection weights between the hidden layer and the output layer

[0054] Initialize momentum factor α and learning rate η;

[0055] 2) Obtain input data, including: current hub temperature T in , target temperature T set , system error e(k) and control quantity U(k);

[0056] 3) According to the input data, use the formula to calculate the input of the hidden layer node and output and the input of the output layer nodes and output

[0057] 4) Calculate the control quantity increment ΔU(k) ​​based on the current system error e(k) and historical errors;

[0058] 5) Update the control variable U(k) according to the control variable increment ΔU(k);

[0059] 6) According to the current output layer error Use gradient descent method to update the connection weights between the output layer and the hidden layer

[0060] According to the hidden layer error Use gradient descent method to update the connection weights between the input layer and the hidden layer

[0061] 7) Repeat steps 2) to 6) and continuously iteratively update the neural network parameters until the set stopping condition is reached;

[0062] 8) Finally, the ventilation volume of the centrifugal fan is adjusted according to the control quantity U(k) output by the neural network PID to control the hub temperature.

[0063] Furthermore, the particle swarm optimization method for the neural network PID controller parameters includes the following steps:

[0064] 1) Set the particle swarm size and dimension to 3, corresponding to the three parameters Kp, Ki, and Kd of the PID controller; randomly initialize the position and velocity of each particle, and the particle position represents the parameter value of the PID controller;

[0065] 2) Based on the current particle position, run the temperature control model in the simulation environment to generate response data;

[0066] 3) In each iteration, the absolute error integral value is used to calculate the fitness value of each particle, and the initial individual optimal position and the global optimal position are set;

[0067] 4) Iterative optimization process:

[0068] The inertia factor is calculated using a dynamic adjustment formula to balance the exploration between global search and local search, as follows:

[0069] ω=μ+δ·N(0,1)

[0070] Where ω is the inertia factor, μ is the basic weight, δ is the coefficient that controls the degree of deviation between the inertia factor and its expected value; N(0,1) is a value randomly drawn from a normal distribution with mean 0 and standard deviation 1;

[0071] For each particle, the next speed and position are calculated based on the current speed and position using the particle swarm algorithm update formula. The specific formula is as follows:

[0072]

[0073] Where, is the velocity of particle i in the next iteration, is the velocity of particle i in the current iteration, is the individual best position of particle i in the current iteration; is the position of particle i in the next iteration, is the position of particle i in the current iteration, gBest (t) is the global best position of the entire particle swarm in the current iteration, c1 and c2 are learning factors used to control the influence of individual historical best position and global best position on particle velocity, r1 and r2 are random numbers in the interval [0,1];

[0074] Update the individual best position and global best position of each particle according to the current fitness value of each particle;

[0075] 5) Check whether the maximum number of iterations has been reached or whether the fitness value has converged; if the termination condition is met, the iteration ends;

[0076] 6) Neural network controller initialization and parameter adjustment:

[0077] The optimal PID parameters obtained by the particle swarm optimization algorithm are used as the initial parameters of the neural network PID controller, and the input layer and hidden layer weights of the neural network are initialized;

[0078] According to the input variables of the neural network, the PID control parameters are calculated through the neural network, and the PID controller output is dynamically adjusted to optimize the control effect.

[0079] Compared with the prior art, the advantages of the present invention are:

[0080] 1. Efficient and simple heat dissipation design: Based on heat transfer principles, the reference temperature of the outer surface of the shroud is introduced to intuitively represent the relationship between the heat load transferred from the hub to the shroud and the internal temperature of the hub. This avoids the iterative calculation of higher-order nonlinear terms in the heat balance equation, reducing the computational burden of the system.

[0081] 2. Improve the responsiveness and control accuracy of the system: Optimizing the neural network PID controller through the particle swarm algorithm can enable the control system to quickly find the optimal PID parameters, shorten the response time to changes in temperature control targets, and reduce control errors.

[0082] 3. Dynamically adjust the mathematical model of the control object: The present invention can update various parameters in the heat dissipation system according to the set sampling interval to reflect the heat transfer characteristics of the wind wheel in different time periods and different working conditions.

[0083] 4. Enhanced intelligence and automation of the cooling system: By adjusting the control strategy using a neural network PID algorithm, the cooling system's adaptive adjustment capabilities are enhanced, avoiding the "single-wind" phenomenon caused by long-term constant-speed operation of the centrifugal fan. This intelligent control strategy not only saves energy and maintenance costs, but also improves the overall reliability of the cooling system. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 This is a schematic diagram of a heat dissipation control system according to an embodiment of the present invention;

[0085] Figure 2 is a schematic diagram of heat transfer parameters of a heat dissipation system according to an embodiment of the present invention;

[0086] Figure 3 is a flow chart of a method for constructing a transfer function G2(s) according to an embodiment of the present invention;

[0087] Figure 4 1 is a flow chart of optimizing BP neural network PID by particle swarm algorithm according to an embodiment of the present invention;

[0088] Figure 5 This is a control framework diagram constructed using Simulink in an embodiment of the present invention;

[0089] Figure 6 It is a PID parameter iteration diagram of the particle swarm algorithm according to an embodiment of the present invention;

[0090] Figure 7 This is a flow chart of the system operation of an embodiment of the present invention;

[0091] Figure 8 It is a schematic diagram of the control process for different target temperatures in a simulation environment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0092] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0093] 1. Air cooling system

[0094] The air cooling control system includes a centrifugal fan, temperature sensor, control components, etc. placed inside the wind turbine rotor. The control system principle is as follows: Figure 1As shown in the figure, the centrifugal fan operates according to the control element's command, drawing the high-temperature air in the hub into the blades. This allows the low-temperature air in the blades to enter the hub and mix with the high-temperature air in the hub, and this cycle continues. Because the blade surface has a large heat transfer area, the heat load introduced into the blades can be effectively transferred to the external environment through forced convection heat transfer on the blade surface. This allows the high-temperature air entering the blades to be cooled when it returns to the hub, thereby cooling the interior of the wind turbine hub.

[0095] 2. Establishment of control system

[0096] 2.1 Heat transfer model of wind wheel hub

[0097] 2.1.1 Obtaining external environment parameters

[0098] The control process of this control system depends on external environmental parameters, hereinafter referred to as environmental parameters.

[0099] For this control system, the external environmental parameters are: ambient temperature T air , ambient wind speed u s3 , ambient solar radiation Φ sun .

[0100] For ambient temperature T air : In this system, the temperature sensor data installed on the wind turbine can be directly selected.

[0101] For ambient wind speed u s3 : In this system, the measurement data of the anemometer installed on the wind turbine can be directly selected.

[0102] For ambient solar radiation Φ sun Wind turbines are typically not equipped with sensors to measure solar radiation, while offshore weather stations typically record solar radiation intensity, including shortwave radiation (direct and diffuse), daily total radiation, and annual total radiation. Meteorological data from sources close to the turbine installation location or the installation of separate sensors can be used to ensure representative and accurate data.

[0103] 2.1.2 Calculation of important parameters of heat transfer model

[0104] After obtaining the external environmental parameters, the control system uses the obtained environmental parameters to calculate the important parameters of the heat transfer model, hereinafter referred to as heat transfer parameters.

[0105] Heat transfer parameters in the cooling system such as Figure 2 shown.

[0106] In order to calculate the heat transfer parameters more accurately, it is necessary to use appropriate standard formulas for the heat transfer process of each part.

[0107] Since wind speeds in wind turbines are typically between 10 and 20 m / s, the air flow on the outer surface of the shroud can be considered turbulent. The thermophysical properties of the air inside the wind turbine are determined according to the specific operating conditions.

[0108] Convection heat transfer coefficient h on the outer surface of the shroud s3 Calculation method:

[0109] 1. Calculate the Reynolds number of the outer surface of the shroud:

[0110]

[0111] Where: u s3 is the air velocity on the outer surface of the shroud, x s3 is the length of the air circulation section, v s3 is the kinematic viscosity coefficient of air.

[0112] 2. Calculate the Nusselt number on the outer surface of the shroud:

[0113] The external air flow is turbulent, so the Nusselt number is calculated according to the following formula:

[0114] Nu=0.037(Re 0.8 -871)Pr 0.33

[0115] Where: P r is the Prandtl number of air.

[0116] 3. Calculate the convective heat transfer coefficient h on the outer surface of the shroud s3 :

[0117]

[0118] Where: s3 is the thermal conductivity of air.

[0119] Calculation method of air velocity inside the wind wheel:

[0120] According to the rotation speed n of the wind turbine rotor when the wind turbine is running, the air flow rate of each part of the wind turbine rotor when the wind turbine is running is calculated by the diameters D1 and D2 of the inner and outer surfaces of the hub and the inner surface diameter D3 of the shroud using the following formula:

[0121]

[0122] The air flow velocity u on the inner and outer surfaces of the hub can be obtained c1 and u s1 , air velocity u on the inner surface of the shroud s2 .

[0123] Convection heat transfer coefficient h inside the hub c1Calculation method:

[0124] Since the air velocity inside the wind wheel is much lower than that on the outer surface of the shroud, the convection heat transfer coefficient h inside the hub is calculated according to the following formula: c1 :

[0125]

[0126] Where: ρ c and ρ0 are the air densities inside the hub and on the sea, respectively, u c1 is the air flow velocity inside the hub.

[0127] Convection heat transfer coefficient h on the outer surface of the hub s1 and the convection heat transfer coefficient h on the inner surface of the shroud s2 The calculation method refers to the convection heat transfer coefficient h inside the hub c1 .

[0128] Calculation method of thermal resistance between hub and fairing:

[0129]

[0130] Where δ1 and δ2 are the thickness of the hub and the fairing, respectively, and λ1 and λ2 are the thermal conductivities of the corresponding materials.

[0131] The inverse of the total heat transfer resistance from the inside of the hub to the outer surface of the fairing, that is, the calculation formula of the total heat transfer coefficient K1 is:

[0132]

[0133] 2.1.3 Establishing the heat balance equation

[0134] When the hub is in thermal equilibrium, the outer surfaces of the shroud and blades are also in equilibrium, and the heat load absorbed by the two should be equal to the heat load dissipated. Since convection heat transfer and thermal radiation act together on the outer surfaces of the shroud and blades, the outside world and the outer surfaces of the two are convection-solar thermal radiation coupled heat transfer. In addition, because the generator set is installed at sea and is located in a relatively open area, the overall geometric difference between the cabin and the external environment is large. Therefore, the thermal radiation process between the outer surface of the shroud and the external environment can be regarded as a "small object to large space radiation model." Therefore, the thermal balance equation of the outer surface of the shroud is:

[0135]

[0136] Φ sun1 =α s1 ·Φ sun ·f p1 ·A2(2)

[0137] Where: T sur1, T air , T in are the outer surface temperature of the shroud, the ambient air temperature, and the internal temperature of the hub, respectively. ε1 is the emissivity of the outer surface of the shroud, and σ is the Stefan-Boltzmann constant, which is approximately: σ = 5.67 × 10 -8 W / (m 2 ·K 4 ). Φ sun1 is the solar radiation heat load on the deflector surface, Φ in is the heat load transmitted to the surface of the air guide shield through the hub. s1 is the absorption rate of the outer surface of the shroud, f p1 is the correction coefficient, and A2 is the outer surface area of ​​the shroud.

[0138] According to Fourier's law of heat conduction, when in thermal equilibrium, the heat load Φ transferred from the hub to the outer surface of the shroud is in for:

[0139] Φ in =A1K1(T in -T sur1 ) (3)

[0140] Where: A1 is the heat transfer area between the air duct and the hub, and K1 is the total heat transfer coefficient from the inside of the hub to the outer surface of the air duct.

[0141] In order to allow the cooling system to directly and quickly calculate the control command, it is necessary to obtain the internal temperature of the wheel hub T in and heat load Φ in The intuitive relationship between the two is that the thermal radiation term of the outer surface of the shroud can be linearized at a specific temperature. Because the wind speed at sea is high and the surface area of ​​the shroud is large, the temperature of the outer surface of the shroud will not change significantly before and after the operation of the wind turbine. Therefore, the outer surface temperature of the shroud when only solar radiation exists can be used as the linearization reference temperature T sur1* In engineering calculations, due to the low thermal conductivity of the air shroud, the heat conduction load from the air shroud surface to the hub can be ignored when calculating the reference temperature. The reference temperature T can be obtained from this. sur1* The calculation formula (4):

[0142]

[0143] After solving for the reference temperature, perform Taylor expansion on (1) at the reference temperature to simplify the heat balance equation:

[0144]

[0145] Comparing formulas (4) and (5), we can get the heat load Φ in and the wheel hub outer surface temperature T sur1, reference temperature T sur1* The relationship between them is formula (6):

[0146]

[0147] Then, combining formulas (3) and (6), we get formula (7):

[0148]

[0149] make:

[0150] ΔT=T in -T sur1* (8)

[0151]

[0152] Where: ΔT is the temperature difference between the internal temperature of the hub and the reference temperature of the outer surface of the fairing, and R is the equivalent heat transfer resistance.

[0153] Therefore, (7) can be finally expressed as:

[0154]

[0155] That is, by introducing the reference temperature T sur1* , which can visually represent the heat load Φ transferred from the hub to the fairing in and the hub internal temperature T in The relationship between them.

[0156] 2.1.4 Thermal load variation Δμ and hub target temperature T set The correspondence between

[0157] From formula (10), we can get:

[0158] ΔT=Φ in R(11)

[0159] From formula (11), we can see that at the reference temperature T sur1* When the thermal resistance R is determined, the heat load Φ transmitted to the fairing by the control hub is in The temperature difference ΔT between the internal temperature of the hub and the reference temperature can be controlled, and finally the internal temperature of the hub T can be controlled. in effect.

[0160] When the cooling system is not enabled, the initial heat load transferred from the hub to the fairing is recorded as Φ0. Due to differences in wind turbine models and operating environments, the specific value should be reasonably selected and determined by the wind turbine user through experiments or tests. From this, the temperature difference ΔT0 between the internal temperature of the hub and the reference temperature in the thermal equilibrium state when the cooling system is not enabled can be obtained, and its calculation formula is as follows:

[0161] ΔT0=Φ0R(12)

[0162] To achieve a lower temperature inside the hub, the heat load transferred from the hub to the shroud must be reduced. Under the action of the centrifugal fan, the heat load transferred from the hub to the outer surface of the shroud is ultimately reduced to Φ1. The ratio of Φ1 to Φ0 is recorded as the heat load change Δμ, which is calculated as follows:

[0163]

[0164] After the heat load transferred from the hub to the outer surface of the fairing is reduced to Φ1, the temperature difference ΔT between the hub interior and the reference temperature is also reduced accordingly. The final temperature difference ΔT is:

[0165] ΔT=ΔμΔT0(14)

[0166] Where ΔT is:

[0167] ΔT=T set -T sur1* (15)

[0168] At the hub target temperature T set After determination, the corresponding thermal load variation Δμ can be calculated by formulas (14) and (15). That is, the thermal load variation Δμ is used as the adjustment amount of the thermal load transferred from the hub to the outer surface of the fairing.

[0169] 2.1.5 Ventilation volume of centrifugal fans Corresponding relationship with heat load change Δμ

[0170] When the wind turbine is running, the total heat load inside the hub is recorded as Φ. Due to the differences in wind turbine models and operating environments, the specific value should be reasonably selected and determined by the wind turbine user through experiments or tests. When the wind turbine reaches thermal equilibrium as a whole, the heat load Φ1 transferred from the hub to the shroud and the heat load Φ transferred from the hub to the three blades are equal. blade The sum should be equal to the total heat load Φ inside the hub, that is:

[0171] Φ=Φ1+3Φ blade (16)

[0172] According to the method described in 2.1.4, the control target T set When the corresponding heat load change Δμ and heat load Φ1 are known, the corresponding heat load change Δμ and heat load Φ1 can be obtained. Through formula (16), further calculation can be obtained that the heat load Φ that the hub should transfer to the inside of the blade at this target temperature is blade .

[0173] Since the heat transfer on the outer surface of the blade and the guide cover is similar coupled heat transfer, the blade parameters are substituted into formula (1) to obtain formula (17).

[0174]

[0175] Where: Φ sun2 is the solar radiation heat load on the blades, and its calculation method is the same as Φ sun1 Φ blade is the heat load from the hub to the blade. ε2 is the emissivity of the blade outer surface, A3 is the outer surface area of ​​the blade, T sur2 is the outer surface temperature of the blade. s4 is the convective heat transfer coefficient of the outer surface of the blade, and its calculation method is the same as h s3 same.

[0176] The heat load Φ introduced into the blade can be obtained by calculating formula (17): blade When the blade reaches thermal equilibrium, the outer surface temperature T sur2 .

[0177] Since the average cross-sectional area of ​​the blade cavity of the offshore wind turbine is large and the air flow velocity entering the blade is low, the internal heat balance equation of each blade after the heat load is passed through it is:

[0178]

[0179] Φ blade =A4K2(T blade -T sur2 )(19)

[0180] Where: A4 is the heat transfer area of ​​the blade, δ3 is the blade thickness, λ3 is the blade thermal conductivity, T blade is the air temperature inside the blade. c2 The specific value should be selected and determined by the user through experiments or tests. K2 is the total heat transfer coefficient from the inner to the outer surface of the blade.

[0181] According to the heat exchanger formula, the heat balance equation of the air in the hub and the air in the blade is:

[0182]

[0183] Where: is the mass flow rate of air entering the blade, i.e., the ventilation volume, C p is the specific heat capacity of air at constant pressure, T set The target temperature set for the inside of the hub.

[0184] Combining formulas (19) and (20) yields formula (21):

[0185]

[0186] According to formulas (17) and (21), according to the temperature target T set and the heat load Φ transferred from the hub to the blades blade The centrifugal fan ventilation rate that matches the target temperature can be determined Ultimately achieve centrifugal fan ventilation volume Corresponding to the thermal load change Δμ.

[0187] 2.2 Mathematical model of the controlled object

[0188] In this heat dissipation control system, the speed of the motor in the cooling fan directly affects the fan's airflow. Changing the motor's rotation speed via the frequency converter in the centrifugal fan can alter the temperature inside the hub. In actual operation, the cooling system is subject to numerous influences, making it difficult to define a specific mathematical model for the control object. Therefore, to facilitate subsequent implementation, demonstration, and understanding, this paper, following common engineering methods, treats the transfer function of the cooling control system, consisting of components such as the centrifugal fan, as a second-order inertia plus a pure lag link.

[0189] The transfer function G(s) of the heat dissipation control system is defined as:

[0190]

[0191] The transfer function G1(s) is:

[0192]

[0193] The transfer function describes the response process of the centrifugal fan from the initial state air volume to the target air volume that matches the heat load change Δμ.

[0194] K1 is the static gain, which indicates the amplitude of the centrifugal fan's ventilation volume response to the target air volume.

[0195] T1 is a time constant, which indicates the speed at which the ventilation volume of the centrifugal fan responds to the target air volume, and is related to the characteristics of the centrifugal fan itself.

[0196] The values ​​of these parameters need to be determined by the specific operation test of the centrifugal fan.

[0197] The transfer function G2(s) is:

[0198]

[0199] K2=ΔT0(25)

[0200] This transfer function describes the response process of the temperature difference ΔT between the internal temperature of the hub and the reference temperature transitioning from an initial stable state to another stable state when the ventilation volume changes after the cooling system is turned on.

[0201] K2 is the static gain, which represents the initial temperature difference ΔT0 between the inside of the wheel hub and the reference temperature when the cooling system is not turned on;

[0202] T2 is the time constant, which indicates the response speed of the temperature difference ΔT to the change of ventilation volume, reflecting the dynamic characteristics of the system, which is related to the overall heat capacity of the system;

[0203] τ1 is the lag time, which represents the delay in the system's response during temperature control. This is the time delay between when the ventilation rate changes and when the temperature begins to change. The values ​​of these parameters must be determined through simulation or actual machine operation tests.

[0204] Since external environmental parameters will change over time, it is necessary to update the relevant heat transfer parameters and transfer function G2(s) according to the changing environmental parameters. The specific steps are as follows:

[0205] a. Set the sampling period

[0206] According to the environmental conditions of the wind turbine, a reasonable sampling period is set to update the system parameters and adjust the control strategy in time.

[0207] b. Get the latest environmental data

[0208] Whenever the scheduled sampling time is reached, the latest ambient temperature T is obtained air , ambient wind speed u s3 and solar radiation Φ sun .

[0209] c. Calculate update parameters

[0210] The system's heat transfer parameters are calculated and updated using the latest environmental data. Due to the thermal inertia of the rotor material itself, the change in the hub's internal temperature has a certain lag compared to the change in the shroud's external surface temperature. In order to more accurately reflect the impact of external conditions on the hub's internal temperature, the reference temperature T in the heat transfer parameters is updated every time a predetermined sampling period is reached. sur1* The update follows the following formula:

[0211] T sur1* (t) = T sur1* (t-1)+ΔT sur1* (t)(26)

[0212] ΔT sur1* (t) is the increment of the reference temperature, which combines the change in the reference temperature and the change in the ambient temperature. The specific formula is as follows:

[0213] ΔT sur1* (t) = k s (T sur1*new (t)-Tsur1* (t-1))+k a (T air (t)-T air (t-1))(27)

[0214] in:

[0215] T sur1* (t-1) is the reference temperature at the last sampling moment;

[0216] T sur1*new (t) is the new reference temperature calculated based on the sampling data at the current moment;

[0217] T air (t-1) is the ambient temperature at the last sampling moment;

[0218] T air (t) is the ambient temperature at the current sampling moment.

[0219] k s and k a It is the weight coefficient, which reflects the influence of the reference temperature and the change of ambient temperature on the increment of the reference temperature. Its specific value should be reasonably selected and determined by the user through experiments or tests.

[0220] d. Update transfer function

[0221] According to the new heat transfer parameters, the transfer function G2(s) is updated.

[0222] e. Adjust control strategy

[0223] Use the updated parameters and integrate them into the system's control loop to ensure that the system can adjust the control strategy according to changes in the external environment.

[0224] The construction process of transfer function G2(s) is as follows Figure 3 shown

[0225] 3. Control strategy

[0226] 3.1 Control Objectives

[0227] For wind turbines, their internal components have a maximum survival temperature T max The internal temperature of the wheel hub must be lower than the maximum survival temperature T max , in order to ensure the safety of the generator components. And in order to extend the service life of the components as much as possible, the heat dissipation requirements are usually further increased when the external environment allows, so that the target temperature T set Set higher than the maximum survival environment temperature T maxThe specific target temperature should be determined according to the specific wind turbine model and the external environment in which the wind turbine is located. In order to ensure the stable operation of the wind turbine and the demonstration of the subsequent embodiments, the following targets are set:

[0228] (1) Hub internal temperature T in Should be controlled at the target temperature T set within a certain range;

[0229] (2) Centrifugal fan ventilation volume should be kept within its operating range.

[0230] 3.2 Closed-loop control system composed of neural network PID controller

[0231] According to the method described in 2.1.5, the temperature target T set The corresponding heat load change Δμ can be determined and the ventilation volume of the centrifugal fan can be realized Quantitative control that matches the change in heat load Δμ.

[0232] However, since some processes in the theoretical calculation are simplified, there are usually uncertainties in the system, which may affect the performance of the heat dissipation system, resulting in differences between the actual heat dissipation effect and the theoretical calculation. In order to solve the above problems, on the basis of open-loop quantitative control, the present invention introduces feedback-based neural network PID control to deal with the effects of nonlinear factors in the heat transfer process and inaccurate heat transfer mathematical models. The neural network PID controller can not only adapt to the nonlinear characteristics of the system, but also continuously optimize the control process through online learning, thereby further improving the control effect of the system. At the same time, in order to avoid the problem that the neural network easily falls into the local optimum, the present invention will use an improved particle swarm algorithm to optimize the BP neural network PID. The specific implementation method is shown in 3.3 and 3.4.

[0233] 3.3BP neural network PID control method

[0234] BP-PID control combines the learning ability of BP neural network and the adjustment technology of PID controller to achieve the adjustment of PID parameter K. p , K i , K d Real-time adaptive adjustment. The variables of the input layer neurons include: actual temperature value y(k) (i.e. the current wheel hub temperature T in ), the expected output temperature value r(k) (ie the target temperature T set), system error e(k) (i.e., temperature deviation), and control variable U(k) (i.e., heat load change Δμ), a total of four variables. By processing these input data through a neural network, the parameters of the PID controller can be dynamically adjusted. In a BP neural network, the nodes in the input layer, intermediate hidden layer, and output layer are represented by j, i, and l, respectively.

[0235] According to the input variables of the BP neural network, the input neurons of the neural network input layer are input and output for:

[0236]

[0237] According to the activation function f(x), the input net of the intermediate hidden layer node can be obtained i (2) (k) and output Specifically, the formula is as follows:

[0238]

[0239] Where: is the connection weight between the input layer and the hidden layer, the activation function H represents the number of neurons in the middle hidden layer.

[0240] Further get the input of the output layer and output

[0241]

[0242]

[0243] Where: is the connection weight between the intermediate hidden layer neurons and the output layer neurons, the activation function

[0244]

[0245] The incremental PID control algorithm used in this control system is specifically expressed as follows:

[0246] Error calculation:

[0247] e(k)=r(k)-y(k)=T set -T in

[0248] Control increment calculation:

[0249] ΔU(k)=K p (Δe(k))+K i e(k)+K d (Δ2 e(k))

[0250] Among them: Δe(k)=e(k)-e(k-1),Δ 2 e(k)=e(k)-2e(k-1)+e(k-2)

[0251] Control quantity update:

[0252] U(k)=U(k-1)+ΔU(k)

[0253] Using the gradient descent method, the parameter update formula can be obtained:

[0254] Weight update of hidden layer and output layer:

[0255]

[0256] Where:

[0257]

[0258] Weight update of input layer and hidden layer:

[0259]

[0260] Where:

[0261]

[0262] Where: g′(x) is the derivative of the output layer activation function, f′(x) is the derivative of the intermediate hidden layer activation function, α is the momentum factor, and η is the learning rate.

[0263] 3.4 Particle Swarm Optimization (PSO) Method for Optimizing BP Neural Network

[0264] The particle swarm optimization (PSO) simulates the search process of particles in the solution space, using individual and global optimal positions to guide particle update speed and position, thereby continuously approaching the optimal solution. In this paper, the particle swarm optimization is used to optimize the parameter settings of the BP neural network PID controller to improve control performance.

[0265] The control system sets the solution space of the particle swarm algorithm to the following range according to the characteristics of the system:

[0266] [Kp max ,Ki max ,Kd max ]

[0267] [Kp min ,Ki min ,Kd min ]

[0268] The particle velocity and position update formula in the particle swarm algorithm is as follows:

[0269]

[0270] Where: is the current velocity of the particle, is the current position of the particle. ω is the inertia factor, c1 and c2 are learning factors used to control the speed at which the particle moves toward its individual optimal position pBest and global optimal position gBest. r1 and r2 are random numbers in the interval [0, 1] that introduce randomness into the search.

[0271] The inertia factor ω adjustment formula is:

[0272] ω=μ+δ*N(0,1)

[0273] μ=μ min +(μ max -μ min )*rand(0,1)

[0274] Where: δ is a value used to represent the degree of deviation between the inertia factor ω and its mathematical expectation. This term is used to control the weight error in the value, so that the inertia factor evolves in a direction that is favorable to the expected weight. In addition, the maximum value μ max With the minimum value μ min Adjust the base weight μ.

[0275] The fitness function is the integral of the absolute error:

[0276]

[0277] Where: K is the running time of the simulation system.

[0278] The optimization steps of the particle swarm algorithm for the BP neural network PID controller parameters are as follows:

[0279] 1. Initialization

[0280] Set the particle swarm size and dimension: the dimension is 3, corresponding to the three parameters K of the PID controller p ,K i ,K d .

[0281] Initialize position and velocity: Randomly initialize the position and velocity of each particle, where the particle position represents the parameter value of the PID controller.

[0282] 2. Fitness calculation

[0283] Simulation environment operation: Run the temperature control model in the simulation environment to generate response data.

[0284] Fitness evaluation: Calculate the absolute error integral of the system response to measure the performance of each particle.

[0285] 3. Set the initial optimal position

[0286] Individual and Global Optimal Positions: Determine the individual and global optimal positions for each particle.

[0287] 4. Iterative Optimization Process

[0288] 1). Dynamically adjust the inertia factor:

[0289] ω=μ+δ·N(0,1)

[0290] μ=μ min +(μ max -μ min )·rand(0,1)

[0291] Among them: δ controls the degree of deviation of the inertia factor, μ min ,μ max are the upper and lower bounds of the change of the basic weight μ respectively.

[0292] 2). Update particle speed and position:

[0293] For each particle:

[0294] ①. Update speed:

[0295]

[0296] ②. Update location:

[0297]

[0298] ③. Calculate the fitness value:

[0299] Use the fitness function F(K) to calculate the current fitness value of each particle.

[0300] ④. Update individual best position and global best position:

[0301] For each particle: If the current fitness value is better than the historical best value, update the individual best position to the current position. If the current fitness value is better than the global best value, update the global best position to the current particle's position.

[0302] 5. Check termination conditions:

[0303] Check whether the maximum number of iterations is reached or the fitness value is converged. If so, terminate the iteration and use the PID parameters obtained in the final iteration as the initial PID parameters of the BP neural network.

[0304] 6. BP neural network controller initialization and parameter adjustment

[0305] ①. Initialize BP neural network controller:

[0306] The optimal PID parameters obtained by the particle swarm optimization algorithm are used as the initial parameters of the BP neural network PID controller and the input layer and hidden layer weights of the BP neural network are initialized.

[0307] ②. Calculate and adjust control parameters:

[0308] According to the input variables of the neural network, the PID control parameters are calculated through the BP neural network, and the PID controller output is dynamically adjusted to optimize the control effect.

[0309] In the optimization step of the present invention:

[0310] Each set of PID parameters K p ,K i ,K d Each parameter is used as the position of a particle in the particle swarm algorithm, and the impact of each set of parameters on the controller performance is evaluated through the fitness function. The fitness function quantifies the quality of the controller response based on key performance indicators of the control system (such as overshoot, rise time, settling time, etc.).

[0311] In each iteration, the particle's position and velocity are updated based on the particle's own historical best position and the global best position. The velocity update formula takes into account the particle's inertia, the influence of individual best performance, and the influence of global group information to optimize the selection of PID parameter combinations. The particle's new position Corresponding to the updated PID parameters, these parameters will be applied and verified in the next simulation or control process.

[0312] To improve system performance, the present invention introduces a dynamically adjusted inertia factor ω. Adjustment of the inertia factor helps to find a balance between global exploration and local fine search, thereby optimizing the performance of the algorithm at different search stages.

[0313] After multiple iterations, the algorithm finally converges to a set of optimal PID parameters, which provide the best control effect under the selected performance indicators.

[0314] Figure 4 Schematic diagram of the process of optimizing BP neural network PID using particle swarm algorithm.

[0315] 4. Control system operation

[0316] When the cooling centrifugal fan is started, the neural network PID controller will be based on the current hub temperature T in With the target temperature T setset The PID parameters are dynamically adjusted based on the temperature deviation and other parameters, and the value of the heat load change Δμ is calculated. The centrifugal fan adjusts the ventilation volume of the centrifugal fan according to the corresponding relationship between the ventilation volume and the heat load change until the internal temperature of the hub reaches the set target value.

[0317] In order to verify the effect of the wind turbine hub heat dissipation control method based on neural network PID in the present invention, a simulation model was built in Simulink based on the stable environmental parameters and wind turbine operating parameters at a certain moment. Figure 5 The control framework of the wind turbine cooling system is shown.

[0318] In the embodiment of the present invention, the initial conditions are set as follows: the range of the thermal load variation Δμ is 0.4 to 0.75, the reference temperature of the outer surface of the fairing is about 18.5°C, and the internal temperature of the hub is stable at 34°C. Under these conditions, the target temperature is changed to 32°C, and in the subsequent process, the dynamic performance of the heat dissipation system based on neural network PID control is tested with a temperature change gradient of 2°C, and compared with the fixed parameter PID control. The initial PID parameters of the fixed parameter PID controller and the neural network PID controller are obtained by the particle swarm optimization algorithm, as shown in Figure 6 As shown, where K p =0.234928,K i =0.00694847,K d =0.751389.

[0319] The operation process of the control system is as follows Figure 7 ,Depend on Figure 8 It can be seen that the neural network PID controller has better stability and smaller overshoot than the fixed parameter PID controller. The method proposed by the present invention can ultimately achieve dynamic adjustment of the heat dissipation centrifugal fan air volume according to different temperature targets.

[0320] Some parts of the present invention are well known to those skilled in the art and are not described in detail.

[0321] The method according to the present invention described above can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored in a remote recording medium or a non-transitory machine-readable medium downloaded via a network and then stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It can be understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, a wind turbine hub cooling system control method based on a neural network PID described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the processing shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for performing the processing shown herein.

[0322] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A wind turbine hub heat dissipation system control method based on neural network PID, characterized in that: The following steps are involved: Step 1: Establish a mathematical model of the control system, including obtaining important parameters, establishing the thermal balance equation and the transfer function of the temperature control process Obtain important parameters, including: obtaining external environment parameters and heat transfer parameters; Obtain external environmental parameters, including: ambient temperature Tair, ambient solar radiation Φ sun and ambient wind speed u s3 ; Calculate heat transfer parameters, including: Calculate the air velocity u on the inner surface of the hub inside the wind wheel c1 , air velocity u on the outer surface of the hub s1 and the air velocity u on the inner surface of the shroud s2 ; and further calculate the convection heat transfer coefficient hc1 inside the hub and the convection heat transfer coefficient hc1 on the outer surface of the hub s1 and the convection heat transfer coefficient h on the inner surface of the shroud s2 , the convection heat transfer coefficient hs3 of the outer surface of the air shroud; calculate the thermal resistance r between the hub and the air shroud, and finally use the above-obtained parameters to calculate the total heat transfer coefficient K1 from the inside of the hub to the outer surface of the air shroud; Establish the thermal balance equation, including: calculating the reference temperature T of the outer surface of the shroud sur1* , calculate the equivalent thermal resistance R; and based on the heat load Φ0 transferred from the hub to the outer surface of the air shroud when the cooling system is not turned on, calculate the temperature difference ΔT0 between the internal temperature of the hub and the reference temperature in this state; Establish the transfer function of the temperature control process: Based on the system parameters and heat transfer parameters of the wind turbine itself, the transfer function G(s) of the hub internal temperature control process is constructed, and a certain sampling period is set to update the environmental parameters, heat transfer parameters and transfer function G(s); Step 2: Develop a control strategy, including: Determine the control target, including setting the target temperature range and centrifugal fan air volume range to ensure the hub is at the appropriate temperature; Step 3: Simulation operation of the control system A closed-loop control system is constructed based on the mathematical model of the control system, and the particle swarm algorithm is used in the simulation environment to provide the optimal initial PID parameters for the neural network PID controller. Monitor the parameter values ​​of the neural network input variables, including the internal temperature of the wheel hub. The neural network PID controller will further adjust the PID parameters based on the parameter values ​​of the input variables. The neural network PID controller will output control instructions based on the current PID parameters and dynamically adjust the ventilation volume of the centrifugal fan to maintain the temperature within the target range; Step 4: Conduct actual machine verification of the control system to ensure its stability and reliability under different working conditions; when the wind turbine is running, synchronously start the air cooling system control device, monitor the temperature data in real time and judge the preset cooling demand. If the unit meets the cooling demand, further start the cooling centrifugal fan to test the cooling system.

2. The wind turbine hub heat dissipation system control method according to claim 1, characterized in that: The air flow rate of each part inside the wind wheel is calculated, and the calculation formula of the air flow rate inside the hub is: Where D1 is the diameter of the inner surface of the hub, and n is the rated rotation speed of the wind wheel; Air velocity u on the outer surface of the hub s1 and the air velocity u on the inner surface of the shroud s2 The calculation method refers to the air flow velocity u on the inner surface of the hub c1 ; The heat transfer coefficient h of the outer surface of the guide cover is calculated in the heat transfer parameter calculation. s3 , the formula is: Where x s3 is the length of the air flow section on the outer surface of the air guide cover, λ s3 is the thermal conductivity of the air on the outer surface of the shroud, Nu s3 is the Nusselt number of the outer surface of the shroud; The calculation of the convection heat transfer coefficient h inside the hub c1 , the formula is: Where, ρ c and ρ0 are the air densities inside the hub and on the sea, respectively, u c1 is the air velocity inside the hub; Convection heat transfer coefficient h on the outer surface of the hub s1 and the convection heat transfer coefficient h on the inner surface of the shroud s2 The calculation method refers to the convection heat transfer coefficient h inside the hub c1 ; The thermal resistance r between the hub and the fairing is calculated using the following formula: Where δ1 and δ2 are the thickness of the hub and the fairing, respectively, and λ1 and λ2 are the thermal conductivities of the corresponding materials; Use the parameters calculated above to calculate the inverse of the total heat transfer resistance, that is, the total heat transfer coefficient K1. The formula is as follows:

3. The wind turbine hub heat dissipation system control method according to claim 1, characterized in that: The heat balance equation is established as follows: Where, Φ in is the heat load transferred from the hub to the fairing in thermal equilibrium, and ΔT is the internal temperature of the hub, T in The reference temperature T of the outer surface of the shroud sur1* The temperature difference between the two; R is the equivalent thermal resistance, which represents the heat transfer capacity from the inside of the hub to the external environment and is defined by the following formula: Where a1 is the heat transfer area between the shroud and the hub, K1 is the total heat transfer coefficient from the hub interior to the shroud exterior; A2 is the shroud exterior area, h s3 is the convection heat transfer coefficient of the outer surface of the guide cover, ε1 is the emissivity of the outer surface of the guide cover, σ is the Stefan-Boltzmann constant, T sur1* is the reference temperature of the outer surface of the shroud.

4. The wind turbine hub heat dissipation system control method according to claim 1, characterized in that: The transfer function of the temperature control process is as follows: Where Δμ represents the change in heat load transferred from the hub to the air shroud, and ΔT represents the temperature difference between the hub internal temperature and the reference temperature of the air shroud outer surface. The transfer function G1(s) represents the response of the cooling centrifugal fan from the initial air volume to the new target air volume that matches the heat load change Δμ. The transfer function G2(s) represents the response of the temperature difference ΔT between the hub internal temperature and the reference temperature from one initial stable state to another stable state when the ventilation volume changes. s is the Laplace transform variable, which represents the complex frequency.

5. The wind turbine hub heat dissipation system control method according to claim 4, characterized in that: The updating steps of the environmental parameters, heat transfer parameters and transfer function G(s) are as follows: a. Sampling period setting: Set a suitable sampling update period according to the environment in which the wind turbine is located; b. Get the latest environmental data: Whenever the predetermined sampling period is reached, get the latest ambient temperature T air , ambient wind speed u s3 and solar radiation Φ sun ; c. Calculate and update heat transfer parameters: recalculate heat transfer parameters using new environmental parameters; d. Update transfer function: Update the transfer function G2(s) according to the new parameters; e. Adjust control strategy: Use the updated parameters and integrate them into the system’s control loop to ensure that the system can adjust the control strategy according to changes in the external environment.

6. The wind turbine hub heat dissipation system control method according to claim 1, characterized in that: The specific steps of constructing the neural network PID controller method are as follows: 1) Initialize the neural network parameters, including: Initialize the connection weights between the input layer and the hidden layer and the connection weights between the hidden layer and the output layer Initialize momentum factor α and learning rate η; 2) Obtain input data, including: current hub temperature T in , target temperature T set , system error e(k) and control quantity U(k); 3) According to the input data, use the formula to calculate the input of the hidden layer node and output And the input of the output layer node and output 4) Calculate the control quantity increment ΔU(k) ​​based on the current system error e(k) and historical errors; 5) Update the control variable U(k) according to the control variable increment ΔU(k); 6) According to the current output layer error Use gradient descent method to update the connection weights between the output layer and the hidden layer According to the hidden layer error Use gradient descent method to update the connection weights between the input layer and the hidden layer 7) Repeat steps 2) to 6) and continuously iteratively update the neural network parameters until the set stopping condition is reached; 8) Finally, the ventilation volume of the centrifugal fan is adjusted according to the control quantity U(k) output by the neural network PID to control the hub temperature.

7. The wind turbine hub heat dissipation system control method according to claim 1, characterized in that: The particle swarm algorithm is used to optimize the parameters of the neural network PID controller, and the method comprises the following steps: 1) Set the particle swarm size and dimension to 3, corresponding to the three parameters Kp, Ki, and Kd of the PID controller; randomly initialize the position and velocity of each particle, and the particle position represents the parameter value of the PID controller; 2) Based on the current particle position, run the temperature control model in the simulation environment to generate response data; 3) In each iteration, the absolute error integral value is used to calculate the fitness value of each particle, and the initial individual optimal position and the global optimal position are set; 4) Iterative optimization process: The inertia factor is calculated using a dynamic adjustment formula to balance the exploration between global search and local search, as follows: ω=μ+δ·N(0,1) Where ω is the inertia factor, μ is the basic weight, δ is the coefficient that controls the degree of deviation between the inertia factor and its expected value; N(0,1) is a value randomly drawn from a normal distribution with mean 0 and standard deviation 1; For each particle, the next speed and position are calculated based on the current speed and position using the particle swarm algorithm update formula. The specific formula is as follows: Where, is the velocity of particle i in the next iteration, is the velocity of particle i in the current iteration, is the individual best position of particle i in the current iteration; is the position of particle i in the next iteration, is the position of particle i in the current iteration, gBest (t) is the global best position of the entire particle swarm in the current iteration, c1 and c2 are learning factors used to control the influence of individual historical best position and global best position on particle velocity, r1 and r2 are random numbers in the interval [0,1]; Update the individual best position and global best position of each particle according to the current fitness value of each particle; 5) Check whether the maximum number of iterations has been reached or whether the fitness value has converged; if the termination condition is met, the iteration ends; 6) Neural network controller initialization and parameter adjustment: The optimal PID parameters obtained by the particle swarm optimization algorithm are used as the initial parameters of the neural network PID controller, and the input layer and hidden layer weights of the neural network are initialized; According to the input variables of the neural network, the PID control parameters are calculated through the neural network, and the PID controller output is dynamically adjusted to optimize the control effect.

Images