A method and system for predicting coupled forces on offshore wind turbine platforms

Abstract

The present invention belongs to the technical field of intelligent prediction of coupled forces of offshore wind turbine platforms, and discloses a method and system for predicting coupled forces of offshore wind turbine platforms, the method comprising obtaining a force data set of offshore wind turbine platforms required for training by using a computational fluid dynamics method; constructing a physical information neural network architecture with a fully connected structure; using physical equations related to wind loads on the fan blades of the wind turbine, wind loads on the semi-submersible platform, and wave loads on the semi-submersible platform to constrain the loss function of the physical neural network, thereby forming a coupled force prediction model of the offshore wind turbine platform based on the physical information neural network. The coupled force prediction model of the offshore wind turbine platform based on the physical information neural network of the present invention outputs the wind loads on the fan blades of the wind turbine, the wind loads on the semi-submersible platform, and the wave loads on the semi-submersible platform. The output result is in good agreement with the measured data set, with an error of no more than 2%.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent prediction of coupled forces of an offshore wind turbine platform, and in particular relates to a method and system for predicting coupled forces of an offshore wind turbine platform. Background Art

[0002] As a renewable energy source, offshore wind power generation has attracted wide attention. Offshore wind power generation platforms face complex marine environments and variable wind and wave loads, so the stress state of the platform needs to be accurately predicted and analyzed during its design and operation.

[0003] The stress analysis of offshore wind turbine platforms needs to consider the coupling effects of multiple environmental factors such as wind speed, waves, and tides. These factors interact with each other in a complex manner, and traditional methods are often difficult to fully consider them during the calculation process. Secondly, the existing forecasting methods are not accurate enough in predicting dynamically changing stress states, especially under extreme environmental conditions, which may lead to large deviations in the prediction results and affect the safety and economy of the platform. Therefore, the development of a coupled stress prediction method for offshore wind turbine platforms based on physical information neural networks can fully consider the coupling effects of wind and wave loads while ensuring high accuracy, which has important theoretical value and application prospects. Summary of the invention

[0004] In order to overcome the problems existing in the related art, the disclosed embodiments of the present invention provide a method and system for predicting coupled forces of an offshore wind turbine platform.

[0005] The technical solution is as follows: A method for predicting coupled forces on an offshore wind turbine platform comprises the following steps:

[0006] S1, using computational fluid dynamics methods to obtain the offshore wind turbine platform force data set required for training;

[0007] S2, constructing a physical information neural network architecture with a fully connected structure;

[0008] S3, using the wind load on the fan blade surface , wind load on semi-submersible platform , Wave loads on semi-submersible platforms The physical equations are used to constrain the loss function of the physical neural network, and a coupled force prediction model for the offshore wind turbine platform based on the physical information neural network is formed.

[0009] In step S1, a computational fluid dynamics method is used, comprising:

[0010] S101, establishing a simulation model, taking the semi-submersible platform as the object, constructing a fluid domain, the water plane of the platform column is the interface between the wind area and the water area, the direction perpendicular to the fan blade rotation plane is the incoming flow direction, the front of the fan blade plane is the velocity inlet, and the rear of the fan blade plane is the velocity outlet; the force-bearing parts include: the wind load force on the fan blade surface in the rotating state, the wind load force on the semi-submersible platform, and the wave load force on the semi-submersible platform;

[0011] S102, input environmental parameters into the simulation model, including wind speed, blade plane force area, semi-submersible platform wind force area, semi-submersible platform wave force volume, wind density, seawater density and wind resistance coefficient, drag force coefficient; the obtained data set is the wind load on the fan blade surface of the wind turbine, the wind load on the semi-submersible platform and the wave load on the semi-submersible platform.

[0012] In step S2, a fully connected physical information neural network architecture is constructed, including:

[0013] (1) The input layer has 6 parameters, namely the current moment , fan blade swept area ,Semi-submersible platform windward area , wind speed , seawater velocity , the volume of wave action on the semi-submersible platform ;

[0014] (2) There are three hidden layers, each containing 64, 128 and 64 neurons, and the activation function uses the ReLU function;

[0015] (3) The output layer consists of three parameters, namely, the wind load on the fan blade surface , Wind loads on semi-submersible platforms Wave loads on semi-submersible platforms .

[0016] Furthermore, in the constructed physical information neural network, from the input layer to the first layer, the expression is:

[0017] ;

[0018] In the formula, is the output of the first layer, is the activation function, is the weight matrix of the first layer, is the input vector, ; is the bias vector of the first layer;

[0019] The expression from the first layer to the second layer is:

[0020] ;

[0021] In the formula, is the weight matrix of the second layer, is the bias vector of the second layer, is the output of the second layer;

[0022] The expression from the second layer to the third layer is:

[0023] ;

[0024] In the formula, is the weight matrix of the third layer, is the bias vector of the third layer, is the output of the third layer;

[0025] The expression from the third layer to the output layer is:

[0026] ;

[0027] In the formula, is the output of the network, , , ]; is the weight matrix of the output layer, is the bias vector of the output layer.

[0028] In step S3, the loss function of the physical neural network is Depend on and The composition is expressed as:

[0029] ;

[0030] In the formula, is the physical loss constrained by the physical equations, is the network mean square error data loss;

[0031] in,

[0032] ;

[0033] ;

[0034] In the formula, For quantity, To calculate the total number of steps, is a function of x and t, is a differential expression symbol for the frontal area, is a differential expression symbol for the upwind volume, is the differential expression symbol of the wave action volume, To calculate the average, is the calculated value for this time, is the air density, is the diameter of the structure, is the absolute value of the flow velocity, is the drag coefficient, is the thrust coefficient, is the water density, is the inertia coefficient, is the action volume, is the acceleration;

[0035] The above loss function is added to the physical neural network architecture constructed in step S2 to form an offshore wind turbine platform coupled force prediction model based on a physical information neural network.

[0036] After step S3, it is necessary to train the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network. Use the data in the training set constructed in step S1, input six-dimensional parameters, and the network outputs three-dimensional prediction values. Then calculate the physical equation loss, and then calculate the mean square error loss of the prediction data of the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network. If the calculated value meets the result accuracy requirements, the final result is output.

[0037] Another object of the present invention is to provide a coupled force prediction system for an offshore wind turbine platform, which implements the coupled force prediction method for an offshore wind turbine platform, and the system comprises:

[0038] A force data set acquisition module is used to obtain the offshore wind turbine platform force data set required for training using a computational fluid dynamics method;

[0039] Physical information neural network building module, used to build a physical information neural network architecture with a fully connected structure;

[0040] Physical neural network loss function constraint module, used to utilize the wind load on the fan blade surface , wind load on semi-submersible platform , Wave loads on semi-submersible platforms The physical equations are used to constrain the loss function of the physical neural network, and a coupled force prediction model for the offshore wind turbine platform based on the physical information neural network is formed.

[0041] Furthermore, the offshore wind turbine platform coupled force prediction system is mounted on a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, and the processor implements the functions of the offshore wind turbine platform coupled force prediction system described above when executing the computer program.

[0042] Furthermore, the offshore wind turbine platform coupled force prediction system is mounted on a computer-readable storage medium, and the computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the functions of the offshore wind turbine platform coupled force prediction system described above can be realized.

[0043] Furthermore, the offshore wind turbine platform coupling force prediction system is mounted on an information data processing terminal, and the information data processing terminal is used to provide a user input interface to implement the functions of the offshore wind turbine platform coupling force prediction system as described above when executed on an electronic device.

[0044] Combining all the above technical solutions, the beneficial effects of the present invention are as follows: The present invention uses computational fluid dynamics methods to establish a force data set for an offshore wind turbine platform for training and verification of a physical information neural network. The wind load formula and Morrison formula are added to the network loss function as physical constraint equations, so that the forecast results conform to the real laws of physics, forming a method for intelligent forecasting of the "wind-wave" coupled forces of an offshore wind turbine platform based on a physical information neural network. The speed of calculating forces is increased by more than 10 times compared to traditional mechanical methods. The accuracy rate is increased by more than 3%, and the calculation results are very stable without any error points. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings herein are incorporated in and constitute a 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;

[0046] Figure 1 It is a flow chart of a method for predicting coupled forces of an offshore wind turbine platform provided by an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of the physical information neural network architecture provided by an embodiment of the present invention;

[0048] Figure 3 It is a flow chart for training a coupled force prediction model of an offshore wind turbine platform based on a physical information neural network provided by an embodiment of the present invention;

[0049] Figure 4 Schematic diagram of a coupled force prediction system for an offshore wind turbine platform provided by an embodiment of the present invention;

[0050] In the figure: 1. Force data set acquisition module; 2. Physical information neural network construction module; 3. Physical neural network loss function constraint module. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

[0052] The innovation of the present invention is that the present invention uses computational fluid dynamics to obtain the offshore wind turbine platform force data set required for training; the physical equations related to the wind load on the fan blades, the wind load on the semi-submersible platform, and the wave load on the semi-submersible platform are used to constrain the loss function of the physical neural network to form an offshore wind turbine platform coupled force prediction model based on the physical information neural network. This is the first time in the industry that this technology has been used to calculate loads.

[0053] Embodiment 1, as Figure 1 As shown, the coupled force prediction method for an offshore wind turbine platform provided by an embodiment of the present invention includes:

[0054] S1, using computational fluid dynamics methods to obtain the offshore wind turbine platform force data set required for training;

[0055] Exemplarily, the computational fluid dynamics method comprises:

[0056] S101, firstly, a simulation model is established, and a computational domain of a data set is established using a computational fluid dynamics method;

[0057] Taking the semi-submersible platform as the research object, the fluid domain is constructed. The waterline of the platform column is the interface between the wind area and the water area. The direction perpendicular to the fan blade rotation plane is the incoming flow direction. The front of the fan blade plane is the velocity inlet, and the rear of the fan blade plane is the velocity outlet. The parts subject to force include:

[0058] The fan blade, that is, the circular plane of the fan blade in the rotating state, is subject to wind load, that is, the wind load on the fan blade surface ;

[0059] The part above the waterline of the semi-submersible platform is subject to wind load, that is, the wind load on the semi-submersible platform ;

[0060] The part below the waterline of the semi-submersible platform is subject to wave load, that is, the wave load on the semi-submersible platform .

[0061] S102, input environmental parameters into the simulation model, including wind speed, blade surface force area, semi-submersible platform wind force area, semi-submersible platform wave force volume, wind density, seawater density and wind resistance coefficient, drag coefficient. The resulting data set is the wind load on the fan blade surface. ; Wind load on semi-submersible platform Wave loads on semi-submersible platforms .

[0062] S2, constructs a physical information neural network architecture with a fully connected structure; Figure 2 shown.

[0063] The constructed physical information neural network with a fully connected structure satisfies the mapping relationship between the environmental parameters and the forces described in step S1. It includes an input layer, a hidden layer, and an output layer. The constructed physical information neural network includes:

[0064] (1) The input layer has 6 parameters, namely the current moment , fan blade swept area ,Semi-submersible platform windward area , wind speed , seawater velocity , the volume of wave action on the semi-submersible platform .

[0065] (2) Hidden layer. In order to capture complex nonlinear relationships, three hidden layers are used, each containing 64, 128, and 64 neurons respectively. The activation function uses the ReLU function.

[0066] (3) The output layer consists of three parameters, namely, the wind load on the fan blade surface ; Wind load on semi-submersible platform Wave loads on semi-submersible platforms .

[0067] For example, the output of each layer can be expressed by the following mathematical formula:

[0068] The expression from the input layer to the first layer is:

[0069] ;

[0070] In the formula, is the output of the first layer, is the activation function, is the weight matrix of the first layer, is the input vector, ; is the bias vector of the first layer;

[0071] The expression from the first layer to the second layer is:

[0072] ;

[0073] In the formula, is the weight matrix of the second layer, is the bias vector of the second layer, is the output of the second layer;

[0074] The expression from the second layer to the third layer is:

[0075] ;

[0076] In the formula, is the weight matrix of the third layer, is the bias vector of the third layer, is the output of the third layer;

[0077] The expression from the third layer to the output layer is:

[0078] ;

[0079] In the formula, is the output of the network, , , ]; is the weight matrix of the output layer, is the bias vector of the output layer.

[0080] S3, using the wind load on the fan blade surface , wind load on semi-submersible platform , Wave loads on semi-submersible platforms The physical equations are used to constrain the loss function of the physical neural network, and a coupled force prediction model for the offshore wind turbine platform based on the physical information neural network is formed.

[0081] For example, the wind load on the fan blade surface is calculated as follows:

[0082] ;

[0083] In the formula, is the wind load on the fan blade surface, is the thrust coefficient (depending on the fan design and operating conditions), is the air density, is the swept area of ​​the fan blades;

[0084] Exemplarily, the calculation formula for the wind load on the semi-submersible platform is:

[0085] ;

[0086] In the formula, is the wind load on the semi-submersible platform, is the drag coefficient, which depends on the shape of the object, It is the windward area of ​​the semi-submersible platform.

[0087] Since wind speed varies with height or time, or Integrate above to calculate the sum of wind loads as wind speed varies with location and time:

[0088] ;

[0089] ;

[0090] Exemplarily, the wave load calculation formula for the semi-submersible platform is:

[0091] ;

[0092] In the formula, is the water density; is the wave resistance coefficient; is the diameter of the structure; is the absolute value of flow velocity; is the seawater velocity; is the inertia coefficient; is the action volume; is the acceleration.

[0093] In actual calculation, the effective action volume is considered and integrated along the direction of wave velocity to obtain the resultant wave load on the structure:

[0094] ;

[0095] Exemplarily, the present invention innovatively adds the above two physical equations as constraints into the loss function of the network, wherein the loss function consists of two parts: is the physical loss constrained by the physical equations, is the network mean square error data loss;

[0096] ;

[0097] in,

[0098] ;

[0099] ;

[0100] In the formula, For quantity, To calculate the total number of steps, is a function of x and t, is a differential expression symbol for the frontal area, is a differential expression symbol for the upwind volume, is the differential expression symbol of the wave action volume, To calculate the average, is the calculated value for this time, is the air density, is the diameter of the structure, is the absolute value of the flow velocity, is the drag coefficient, is the thrust coefficient, is the water density, is the inertia coefficient, is the action volume, is the acceleration;

[0101] Through the above formula, the present invention effectively accelerates the calculation speed of the neural network, effectively increases the calculation accuracy of the neural network, and makes the calculation results of the neural network conform to the actual physical properties.

[0102] As another example, the above loss function is added to the physical neural network architecture constructed in step S2 to form a coupled force prediction model of an offshore wind turbine platform based on a physical information neural network.

[0103] Another exemplary example is Figure 3 As shown, the present invention trains the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network, uses the data in the training set constructed in step S1, inputs six-dimensional parameters, and the network outputs three-dimensional prediction values, and then calculates the physical equation loss, and then calculates the mean square error loss of the prediction data of the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network. If the calculated value meets the result accuracy requirement, the final result is output. If the accuracy requirement is not met, the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network is trained, and optimized with the AdamW optimizer. When the number of iteration steps exceeds 20,000 steps, the L-BFGS optimizer is used for optimization. Using ADAM first and then LBFGS, and combining the two is an innovation, which makes up for the poor effect when the ADAM calculation step length is too long.

[0104] Embodiment 2, as Figure 4 As shown, the offshore wind turbine platform coupled force prediction system provided by the embodiment of the present invention includes:

[0105] The force data set obtaining module 1 is used to obtain the offshore wind turbine platform force data set required for training by using the computational fluid dynamics method.

[0106] Physical information neural network building module 2, used to build a physical information neural network architecture with a fully connected structure;

[0107] Physical neural network loss function constraint module 3 is used to utilize the wind load on the fan blade surface ; Wind load on semi-submersible platform , Wave loads on semi-submersible platforms The relevant physical equations constrain the loss function of the physical neural network to form a coupled force prediction model for offshore wind turbine platforms based on physical information neural network.

[0108] To further illustrate the effects of the embodiments of the present invention, the following experiment is conducted: an offshore wind turbine platform in a specific ocean area is used as the research object, environmental parameters are collected, and the current time , fan blade swept area ,Semi-submersible platform windward area , wind speed , seawater velocity , the volume of wave action on the semi-submersible platform The input is input into the loss function of the physical neural network constrained by the relevant physical equations, and the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network is obtained. After calculation, the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network outputs the wind load on the fan blade surface. ; Wind load on semi-submersible platform Wave loads on semi-submersible platforms The output results are in good agreement with the measured data set, with an error of no more than 2%.

[0109] Experiments show that the existing technology only uses the traditional computational fluid dynamics method to simulate the platform force, and does not have intelligent solution. At the same time, it takes a long time and a lot of work to measure these data. The intelligent prediction of the offshore wind turbine platform coupled force prediction model based on the physical information neural network of the present invention can minimize the calculation of the offshore wind turbine platform coupled force in terms of human and physical costs.

[0110] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for predicting coupled forces on an offshore wind turbine platform, characterized in that: The method comprises the following steps: S1, using computational fluid dynamics methods to obtain the offshore wind turbine platform force data set required for training; S2, constructing a physical information neural network architecture with a fully connected structure; S3, using the wind load F on the fan blade surface y , wind load F on the semi-submersible platform w , wave load F on semi-submersible platform s The physical equations of the constrained physical neural network loss function are used to form a coupled force prediction model for offshore wind turbine platforms based on physical information neural networks. In step S2, a fully connected physical information neural network architecture is constructed, including: (1) The input layer consists of 6 parameters, namely the current time t, the swept area A of the fan blades y ,Semi-submersible platform windward area A w , wind speed V, seawater velocity U, wave action volume V on the semi-submersible platform t ; (2) There are three hidden layers, each containing 64, 128, and 64 neurons, and the activation function uses the ReLU function; (3) The output layer consists of three parameters, namely, the wind load F on the fan blade surface y 、Wind load F on semi-submersible platform w and the wave load F on the semi-submersible platform s ; In step S3, the loss function Loss of the physical neural network total By Loss function and Loss data The composition is expressed as: Loss total =Loss function +Loss data In the formula, Loss function is the physical loss constrained by the physical equation, Loss data is the network mean square error data loss; in, In the formula, n is the number, N is the total number of calculation steps, (x, t) is the function of x and t, A wn A is a differential expression for the frontal area. yn V is a differential expression symbol for the windward volume, tn is the differential expression symbol of the wave action volume, To calculate the average value, F n is the calculated value, ρ a is the air density, D is the diameter of the structure, |U| is the absolute value of the flow velocity, C d is the drag coefficient, C T is the thrust coefficient, ρ w is the water density, C m is the inertia coefficient, V t is the action volume, is the acceleration; The above loss function is added to the physical neural network architecture constructed in step S2 to form an offshore wind turbine platform coupled force prediction model based on a physical information neural network.

2. The method for predicting coupled forces of an offshore wind turbine platform according to claim 1, characterized in that: In step S1, a computational fluid dynamics method is used, comprising: S101, establishing a simulation model, taking the semi-submersible platform as the object, constructing a fluid domain, the water plane of the platform column is the interface between the wind area and the water area, the direction perpendicular to the fan blade rotation plane is the incoming flow direction, the front of the fan blade plane is the velocity inlet, and the rear of the fan blade plane is the velocity outlet; the force-bearing parts include: the wind load force on the fan blade surface in the rotating state, the wind load force on the semi-submersible platform, and the wave load force on the semi-submersible platform; S102, input environmental parameters into the simulation model, including wind speed, blade plane force area, semi-submersible platform wind force area, semi-submersible platform wave force volume, wind density, seawater density and wind resistance coefficient, drag force coefficient; the obtained data set is the wind load on the fan blade surface of the wind turbine, the wind load on the semi-submersible platform and the wave load on the semi-submersible platform.

3. The method for predicting coupled forces of an offshore wind turbine platform according to claim 1, characterized in that: In the constructed physical information neural network, from the input layer to the first layer, the expression is: h1=ReLU(W1·x+b1) Where h1 is the output of the first layer, ReLU is the activation function, W1 is the weight matrix of the first layer, x is the input vector, x = [t, A y ,A w ,V,U,V t ]; b1 is the bias vector of the first layer; The expression from the first layer to the second layer is: h2=ReLU(W2·h1+b2) Where W2 is the weight matrix of the second layer, b2 is the bias vector of the second layer, and h2 is the output of the second layer; The expression from the second layer to the third layer is: h3=ReLU(W3·h2+b3) Where W3 is the weight matrix of the third layer, b3 is the bias vector of the third layer, and h3 is the output of the third layer; The expression from the third layer to the output layer is: y=W4·h4+b4 In the formula, y is the output of the network, y=[F y ,F w ,F s ]; W4 is the weight matrix of the output layer, and b4 is the bias vector of the output layer.

4. The method for predicting coupled forces of an offshore wind turbine platform according to claim 1, characterized in that: After step S3, it is necessary to train the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network. Use the data in the training set constructed in step S1, input six-dimensional parameters, and the network outputs three-dimensional prediction values. Then calculate the physical equation loss, and then calculate the mean square error loss of the prediction data of the coupled force prediction model of the offshore wind turbine platform based on the physical information neural network. If the calculated value meets the result accuracy requirements, the final result is output.

5. A coupled force prediction system for an offshore wind turbine platform, characterized in that: The system implements the method for predicting coupled forces of an offshore wind turbine platform as claimed in any one of claims 1 to 4, and the system comprises: A force data set obtaining module (1) is used to obtain the offshore wind turbine platform force data set required for training using a computational fluid dynamics method; A physical information neural network building module (2), used to build a physical information neural network architecture with a fully connected structure; The physical neural network loss function constraint module (3) is used to use the wind load F on the fan blade surface y , wind load F on the semi-submersible platform w , wave load F on semi-submersible platform s The physical equations are used to constrain the loss function of the physical neural network, and a coupled force prediction model for the offshore wind turbine platform based on the physical information neural network is formed.

6. The offshore wind turbine platform coupled force prediction system according to claim 5, characterized in that: The offshore wind turbine platform coupled force prediction system is mounted on a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor. When the processor executes the computer program, the functions of the offshore wind turbine platform coupled force prediction system described above are implemented.

7. The offshore wind turbine platform coupled force prediction system according to claim 5, characterized in that: The offshore wind turbine platform coupled force prediction system is mounted on a computer-readable storage medium, and the computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the functions of the offshore wind turbine platform coupled force prediction system described above can be realized.

8. The offshore wind turbine platform coupled force prediction system according to claim 5, characterized in that: The offshore wind turbine platform coupling force prediction system is mounted on an information data processing terminal, and the information data processing terminal is used to provide a user input interface to implement the functions of the offshore wind turbine platform coupling force prediction system as described above when executed on an electronic device.

Images