Combined floating wind power platform and intelligent optimization method for power generation efficiency

Abstract

The present invention proposes a combined floating wind power platform and an intelligent optimization method for its power generation efficiency. The combined floating wind power platform is provided with six groups of buoyancy tanks, which are distributed in a regular hexagon and connected by trusses. A frustum component and a tower mounting base are provided on the upper part of the buoyancy tank, a tower is mounted on the tower mounting base, and two groups of horizontal axis wind turbines and four groups of vertical axis wind turbines are symmetrically mounted on the six groups of towers; support rods and mooring cables are provided on the lower part of the buoyancy tanks, a heave plate is provided under the support rod, a ballast tank is installed under the heave plate, and the mooring cables are connected to gravity anchors; after the combined floating wind power platform is installed, half of the buoyancy tanks are below sea level and half are above sea level, and the combined floating wind power platform is fixed in the seawater by connecting gravity anchors with mooring cables. The combined floating wind power platform is symmetrically distributed in a regular hexagon, has uniform force, and a stable structure. The combined floating wind power platform can efficiently obtain wind energy at different heights and angles through the intelligent matching optimization method.

Description

Technical Field

[0001] The present invention relates to the field of offshore wind power technology, and in particular to a combined floating wind power platform and an intelligent optimization method for power generation efficiency thereof. Background Art

[0002] In the development of wind power technology, wind turbines are mainly divided into two categories: horizontal axis and vertical axis. Horizontal axis wind energy utilization efficiency is relatively high and is the most widely used in the wind power field. However, horizontal axis wind turbines are placed on top of the supporting structure, which has an eccentric effect on the foundation platform, making the entire structural system top-heavy and unstable. Horizontal axis wind turbines need to constantly change direction to maintain perpendicularity to the wind direction to obtain energy. In engineering projects, the blade tips are generally far from the water surface, so the wind energy utilization rate near the water surface is low. For vertical axis wind turbines, the transmission system is installed at the bottom, which will not affect the wind turbine tower. It can accept wind from all directions and does not require a windward adjustment system. However, due to the limitations of the supporting structure, vertical axis wind turbines are difficult to build high in engineering projects. Therefore, the wind energy utilization rate is low at high places above the water surface. To this end, based on the development needs of wind power, the applicant has proposed a combined floating wind turbine platform and an intelligent optimization method for its power generation efficiency. The combined floating wind turbine platform is symmetrically distributed in a regular hexagon, has uniform force, and a stable structure. It can fully capture wind energy at different heights and angles, thereby improving the efficiency and production capacity of a single floating platform. It also provides an intelligent matching optimization method to adjust the interference of the blades of a single device, optimize the deployment of multiple wind turbine platforms, and ensure that the wake effect is minimized. Summary of the Invention

[0003] To solve the above technical problems, the present invention proposes a combined floating wind power platform and an intelligent optimization method for its power generation efficiency, by arranging six groups of buoyancy tanks, the six groups of buoyancy tanks are distributed in a regular hexagon, the buoyancy tanks are connected by trusses, a frustum component and a tower mounting base are arranged on the upper part of the buoyancy tank, the tower is installed on the tower mounting base, and two groups of horizontal axis wind turbines and four groups of vertical axis wind turbines are symmetrically installed on the six groups of towers; support rods and mooring cables are arranged at the lower part of the buoyancy tanks, a heave plate is arranged under the support rod, a ballast tank is installed under the heave plate, and a gravity anchor is connected to the mooring cable; after the combined floating wind power platform is installed at sea, half of the buoyancy tanks are below sea level and half are above sea level, and it is fixed in the sea water by connecting the gravity anchor with the mooring cable. The combined floating wind power platform with symmetrical distribution of regular hexagons is evenly stressed, has a stable structure, can fully obtain wind energy at different heights and angles, and greatly improves the efficiency and production capacity of a single floating platform.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] The combined floating wind power platform includes a horizontal axis wind turbine, a long tower, a short tower, a vertical axis wind turbine, a tower mounting base, a frustum component, a buoyancy tank, a truss, a vertical support rod, an oblique support rod, a heave plate, a ballast tank, a mooring cable and a gravity anchor, and is characterized in that: the combined floating wind power platform is provided with six groups of buoyancy tanks, the six groups of buoyancy tanks are distributed in a regular hexagon, the buoyancy tanks are connected by trusses, a frustum component is provided on the upper part of the buoyancy tank, a tower mounting base is provided on the frustum component, two groups of long towers and four groups of short towers are symmetrically installed on the six groups of tower mounting bases, the horizontal axis wind turbine is installed on the long tower, and the vertical axis wind turbine is installed on the short tower; vertical support rods, oblique support rods and mooring cables are installed at the bottom of the buoyancy tank, a heave plate is provided under the vertical support rods and the oblique support rods, the ballast tank is connected under the heave plate, and the lower end of the mooring cable is connected to the gravity anchor.

[0006] Furthermore, the combined floating wind power platform is provided with horizontal axis wind turbines and vertical axis wind turbines, the horizontal axis wind turbines are in two groups, and the vertical axis wind turbines are in four groups. The two groups of horizontal axis wind turbines and the four groups of vertical axis wind turbines are symmetrically installed on the upper part of the six groups of buoyancy cabins.

[0007] Furthermore, the combined floating wind turbine platform is provided with horizontal axis wind turbines and vertical axis wind turbines, which are arranged alternately in plane and elevation, and the blade tips of the horizontal axis wind turbine blades are 3-5 meters higher than the blade tips of the vertical axis wind turbine blades at the lowest position of rotation.

[0008] Furthermore, the tower mounting base and the frustum component are modular prefabricated parts.

[0009] Furthermore, the cross section of the truss is circular or square.

[0010] Furthermore, the vertical support rods and the oblique support rods have circular cross-sections.

[0011] The present invention provides a method for intelligently optimizing power generation efficiency of a combined floating wind power platform based on a convolutional neural network, comprising the following steps:

[0012] 1) Grid deployment of wind turbine platforms;

[0013] A grid-based deployment method is used for initial deployment to ensure that a sufficiently large area on the sea surface can be covered and locations with strong wind energy can be found;

[0014] 2) Parameter measurement;

[0015] High-precision sensors on each wind turbine platform are used to measure sea surface parameters, including sea surface wind speed, sea surface wind direction, turbulence level, temperature and humidity, which are closely related to wind energy at various locations on the sea surface.

[0016] 3) Convolutional neural network training;

[0017] Sea surface wind speed and direction are significant features, which means that network-specific optimization of wind speed and direction features is required to increase their expressive power. Multi-layer cumulative convolution is used to enhance their expressive power, and the filling formula is used to fill the two missing units.

[0018] 4) Effectiveness evaluation;

[0019] Conduct wind energy evaluation on the predicted power generation efficiency according to the power generation efficiency evaluation formula to find areas with better power generation efficiency;

[0020] 5) Field interpolation processing;

[0021] Newton polynomial interpolation is performed on the entire field to obtain the overall power generation efficiency of the wind turbine platform on the sea surface after initial deployment;

[0022] 6) Individual power generation efficiency matching;

[0023] When the power generation efficiency of the horizontal-axis wind turbine and the power generation efficiency of the vertical-axis wind turbine at the predicted highest point are smaller, the fan blade speed can be operated without restriction. When the power generation efficiency of the horizontal-axis wind turbine and the power generation efficiency of the vertical-axis wind turbine at the predicted highest point are both larger, the horizontal-axis wind blades and the vertical-axis wind blades in the equipment can be operated within the allowable range. However, when the power generation efficiency of the horizontal-axis wind turbine and the power generation efficiency of the vertical-axis wind turbine are extremely unbalanced, adjustments need to be made according to the vertical-axis wind blade adjustment formula or the horizontal-axis wind blade adjustment formula to maximize the energy collection of the entire system.

[0024] 7) SVM binary classification wake effect;

[0025] Using the SVM binary classification algorithm, we extract the wind speed, direction, turbulence level, temperature, wind energy value, and the spacing between each wind turbine at the location with the highest power generation efficiency. This is used to predict whether the wake effect of all wind turbines when arranging them to maximize power generation efficiency will exceed a set threshold.

[0026] 8) Optimized deployment;

[0027] Arrange all devices at the location where power generation efficiency is maximized with minimal spacing. Then use the SVM algorithm to predict the wake effect at that time. When the predicted wake effect value is less than the set threshold, the power generation efficiency can be maximized. If the predicted wake effect value is higher than the threshold, the spacing is increased and the prediction is repeated until the wake effect is still below the threshold when the power generation efficiency is maximized, thus achieving maximum power generation efficiency and low wake effect.

[0028] As a further improvement of the present invention, the filling formula in step 3) of the power generation efficiency intelligent optimization method based on convolutional neural network is expressed as:

[0029] The filling formula is expressed as:

[0030]

[0031]

[0032] k=1, 2, 3, ..., n

[0033] 0≤α≤2

[0034] 0≤β≤2

[0035] Among them S k The new wind speed related feature value is filled in for each layer, α is a hyperparameter, k means the current filling value is in the kth convolution layer, n means there are n convolution layers, W s It represents the first layer initial input value of the sea surface wind speed after normalization, d k Fill in the wind direction related eigenvalues for each layer, β is a hyperparameter, W d It represents the first layer initial input value of the sea surface wind direction after normalization.

[0036] As a further improvement of the present invention, the power generation efficiency evaluation formula in step 4) of the power generation efficiency intelligent optimization method based on convolutional neural network is expressed as:

[0037] In step 3), the power generation efficiency of each platform is predicted, including the power generation efficiency of the horizontal axis wind turbine and the vertical axis wind turbine at each point. The following needs to be evaluated to find the area with better power generation efficiency. The power generation efficiency evaluation formula is:

[0038]

[0039] Where P is the power generation efficiency assessment value, y1 is the power generation efficiency of the horizontal axis wind turbine at the location, and y2 is the power generation efficiency of the vertical axis wind turbine at the location.

[0040] As a further improvement of the present invention, the vertical axis fan blade adjustment formula in step 6) of the power generation efficiency intelligent optimization method based on convolutional neural network is expressed as:

[0041] The vertical axis fan blade adjustment formula is as follows:

[0042]

[0043] y1-y2>y2

[0044] Where V2 is the speed of the vertical-axis fan blades after adjustment when the horizontal-axis fan efficiency is much greater than the vertical-axis fan efficiency, and v2 is the speed of the vertical-axis fan blades at the vertical-axis fan efficiency y2. y1 and y2 are the horizontal-axis fan efficiency and vertical-axis fan efficiency, respectively.

[0045] The horizontal axis fan blade adjustment formula in step 6) is expressed as:

[0046] The horizontal axis fan blade adjustment formula is as follows:

[0047]

[0048] y2-y1>y1

[0049] Among them, V1 is the speed of the horizontal axis fan blade after adjustment when the vertical axis fan power generation efficiency is much greater than the horizontal axis fan power generation efficiency, v1 is the speed of the horizontal axis fan blade under the horizontal axis fan power generation efficiency y1. y1 and y2 are the horizontal axis fan power generation efficiency and the vertical axis fan power generation efficiency respectively.

[0050] The present invention provides a combined floating wind power platform and an intelligent optimization method for its power generation efficiency. Six groups of buoyancy tanks are arranged in a regular hexagonal shape. The buoyancy tanks are connected by trusses. A frustum component and a tower mounting base are arranged on the upper part of the buoyancy tank. The tower is installed on the tower mounting base. Two groups of horizontal axis wind turbines and four groups of vertical axis wind turbines are symmetrically installed on the six groups of towers. Support rods and mooring cables are arranged at the lower part of the buoyancy tanks. A heave plate is arranged under the support rod. A ballast tank is installed under the heave plate. The mooring cables are connected to gravity anchors. After the combined floating wind power platform is installed at sea, half of the buoyancy tanks are below sea level and the other half are above sea level. The combined floating wind power platform is fixed in the seawater by connecting gravity anchors via mooring cables. The combined floating wind power platform with a regular hexagonal symmetrical distribution is evenly stressed and has a stable structure. The advantages brought about are:

[0051] 1. The combined floating wind power platform provided by this patent can fully capture wind energy at different heights and angles, greatly improving the efficiency and production capacity of a single floating platform;

[0052] 2. The combined floating wind power platform provided by this patent adopts six groups of buoyancy cabins, which are generally symmetrically distributed in a regular hexagon and evenly stressed. The vertical axis wind turbine can not only improve the unit wind energy utilization rate, but also improve the stability of the platform structure;

[0053] 3. The modular floating wind turbine platform provided by this patent is easy and simple to install, and has flexible construction and installation. It can be assembled on land and directly transported to a fixed location offshore for installation. Alternatively, it can utilize its self-floating characteristics and be towed to the construction site as a whole, where it can be connected to mooring cables and gravity anchors for installation. This reduces the use of large lifting machinery during offshore construction and reduces construction costs.

[0054] 4. This patent provides an intelligent optimization method for power generation efficiency based on convolutional neural networks, which is applicable to a new type of offshore wind turbine power generation platform. It can coordinately control the efficiency of horizontal-axis and vertical-axis wind turbines, improve the utilization rate of wind energy per unit space, improve power generation efficiency, reduce costs, and ensure a more stable structure.

[0055] 5. This patent provides an intelligent optimization method for power generation efficiency based on convolutional neural networks. It uses convolutional neural networks to predict the power generation efficiency of each individual wind turbine platform, and optimizes the power generation efficiency matching between the horizontal axis wind turbine and the vertical axis wind turbine. It can improve the power generation efficiency of the individual wind turbine platform and ensure the normal operation of the equipment.

[0056] 6. This patent provides a convolutional neural network-based intelligent optimization method for power generation efficiency, which uses the SVM algorithm to evaluate and predict the wake effect of a wind turbine platform group, thereby maximizing the power generation efficiency of the wind turbine platform group and ensuring that the wake effect is within the threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0058] Figure 2 It is a schematic diagram of the top view of the structure of the present invention;

[0059] Figure 3 This is a flow chart of the intelligent optimization method for power generation efficiency based on convolutional neural networks;

[0060] Figure 4 This is a schematic diagram of the gridded deployment of wind turbine platforms for the intelligent optimization method of power generation efficiency based on convolutional neural networks;

[0061] Figure 5 Schematic diagram of the convolutional neural network architecture for power generation efficiency prediction based on the convolutional neural network intelligent optimization method for power generation efficiency.

[0062] The markings in the figure are: 1. Horizontal axis wind turbine; 2. Long tower; 3. Short tower; 4. Vertical axis wind turbine; 5. Tower mounting base; 6. Cone component; 7. Buoyancy tank; 8. Truss; 9. Sea level; 10. Vertical support rod; 11. Diagonal support rod; 12. Heave plate; 13. Ballast tank; 14. Mooring cable; 15. Gravity anchor. DETAILED DESCRIPTION

[0063] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0064] like Figure 1-2 As shown: As shown is a combined floating wind power platform and an intelligent optimization method for power generation efficiency thereof, comprising a horizontal axis wind turbine 1, a long tower 2, a short tower 3, a vertical axis wind turbine 4, a tower mounting base 5, a frustum component 6, a buoyancy tank 7, a truss 8, a vertical support rod 10, an inclined support rod 11, a heave plate 12, a ballast tank 13, a mooring cable 14 and a gravity anchor 15; as shown Figure 1-2As shown, the combined floating wind power platform is provided with six groups of buoyancy cabins 7, which are made of special steel materials welded together and have strong resistance to seawater corrosion. The exterior of the buoyancy cabins 7 is painted with protective paint. The six groups of buoyancy cabins 7 are distributed in a regular hexagon, and the buoyancy cabins 7 are connected by trusses 8. The cross-section of the trusses 8 is circular or rectangular. Taking the circle as an example, the diameter is 3 meters to 5 meters and the wall thickness is 4.5 centimeters to 8 centimeters. The trusses 8 are made of the same special steel materials as the buoyancy cabins 7. The trusses 8 are brought together in the center after connecting the six groups of buoyancy cabins 7. A frustum component 6 is provided on the upper part of the buoyancy cabin 7, and a tower mounting base is provided on the frustum component 6. The base 5 shown in FIG5 is a block diagram of a floating platform, and the base 5 and the frustum component 6 shown in FIG5 are modular prefabricated parts, which can be modularly installed and replaced. Two groups of long towers 2 and four groups of short towers 3 are symmetrically installed on the six groups of tower mounting bases 5. The horizontal axis wind turbine 1 is installed on the long tower 2, and the vertical axis wind turbine 4 is installed on the short tower 3. The horizontal axis wind turbine 1 and the vertical axis wind turbine 4 are staggered with each other in the plane and the elevation. The blade tip of the horizontal axis wind turbine 1 is 3-5 meters higher than the blade tip of the vertical axis wind turbine 4 at the lowest position of rotation. The combined use of the horizontal axis wind turbine 1 and the vertical axis wind turbine 4 can fully obtain wind energy at different heights and angles, greatly improving the efficiency and production capacity of a single floating platform. A vertical support rod 10, an oblique support rod 11 and a mooring cable 14 are installed at the bottom of the buoyancy tank 7. The vertical support rod 10 and the oblique support rod 11 are circular in cross section, with a diameter of 1 meter to 1.5 meters and a wall thickness of 1.5 centimeters to 3 centimeters. They are made of the same special steel as the buoyancy tank 7. A heave plate 12 is set under the vertical support rod 10 and the oblique support rod 11. The ballast tank 13 is connected under the heave plate 12. The ballast tank 13 can lower the center of gravity of the entire floating platform and increase the stability of the platform. The lower end of the mooring cable 14 is connected to the gravity anchor 15. The mooring cable 14 and the gravity anchor 15 can fix the combined floating wind power platform on the seabed to ensure that it is in position under the action of wind and waves. The displacement cannot be too large; after the combined floating wind power platform shown is installed, half of the buoyancy cabin 7 is below the sea level 9 and the other half is above the sea level 9 at sea. It has good wind and wave resistance, and the buoyancy cabin 7 and the truss 8 are symmetrically distributed in a regular hexagon as a whole. The force is evenly distributed and the structure is stable. It can fully obtain wind energy at different heights and angles. It is easy and simple to install and flexible to build and install. It can be directly transported to a fixed position in the open sea for installation after being assembled on land. It can also take advantage of its self-floating characteristics and complete the installation by floating and towing the whole machine to the construction site and connecting it with the mooring cable 14 and the gravity anchor 15, thereby reducing the use of large-scale lifting machinery during offshore construction and reducing construction costs.

[0065] like Figure 3 Shown is a flow chart of the intelligent optimization method for power generation efficiency based on convolutional neural network provided by this application.

[0066] Step S1: Grid deployment of wind turbine platforms.

[0067] like Figure 4 Shown is a schematic diagram of the gridded deployment of a wind turbine platform for the intelligent optimization method for power generation efficiency based on a convolutional neural network provided in this application.

[0068] For this application, the initial deployment of the wind turbine platform is required first. This application adopts a grid-based deployment method to ensure that it can cover a large enough area on the sea surface and find locations with strong wind energy.

[0069] Step S2: Parameter measurement.

[0070] In step S1, the initial grid deployment of wind turbine platforms is completed. In this step, high-precision sensors on each wind turbine platform are used to measure sea surface parameters, including wind speed, direction, turbulence, temperature, and humidity, which are closely related to power generation efficiency at each location on the sea surface.

[0071] Step S3: Convolutional neural network training

[0072] like Figure 5 Shown is a schematic diagram of the convolutional neural network architecture for power generation efficiency prediction based on the convolutional neural network intelligent optimization method for power generation efficiency provided in this application.

[0073] In step S2, characteristic parameters closely related to sea surface wind energy are obtained and need to be processed for subsequent network convolution analysis. First, all features are normalized to maintain uniform dimensions.

[0074] At the network level, in this application, sea surface wind speed and direction are significant features, requiring network-specific optimization of these features to increase their expressive power. This application proposes using a multi-layer cumulative convolutional approach to enhance this expressive power.

[0075] The initial input parameters are the above-mentioned sea breeze speed, sea breeze direction, turbulence level, temperature and humidity.

[0076] Conv1 uses a 3x1 convolution with a stride of 1 and padding with a zero on each side. This convolution ensures that the input and output remain consistent during the first convolution. Subsequent convolutions (Conv2, Conv3, etc.) also use 3x1 convolutions with a stride of 1, but without padding, resulting in the output layer having two fewer units than the input layer. Furthermore, to increase computation speed and accuracy, the ReLU activation function is used throughout this network architecture. The two missing units are filled with the characteristic values of factors related to sea breeze speed and direction.

[0077] The filling formula is expressed as:

[0078]

[0079]

[0080] k=1,2,3,…,n

[0081] 0≤α≤2 (4)

[0082] 0≤β≤2 (5)

[0083] where s k The newly added wind speed related feature value for each layer, α is a hyperparameter, k indicates that the current filling value is in the kth convolution layer. n indicates that there are n convolution layers in total. s It represents the first layer initial input value of the sea surface wind speed after normalization. k Fill in the wind direction related eigenvalues for each layer, β is a hyperparameter, W d It represents the first layer initial input value of the sea surface wind direction after normalization.

[0084] After building the network, train and evaluate the algorithm model. The mean square error (MSE) evaluation standard is used. Training is complete when both the training and test sets have high accuracy.

[0085] Step S4: Performance evaluation.

[0086] like Figure 4 As shown, in step S3, the power generation efficiency of each platform is predicted, including the power generation efficiency of the horizontal axis wind turbine and the vertical axis wind turbine at each point. The following needs to be evaluated for power generation efficiency to find the area with better power generation efficiency. The power generation efficiency evaluation formula is:

[0087]

[0088] Where P is the power generation efficiency assessment value, y1 is the power generation efficiency of the horizontal axis wind turbine at the location, and y2 is the power generation efficiency of the vertical axis wind turbine at the location.

[0089] The power generation efficiency evaluation value of each node after initial deployment can be obtained through the above formula.

[0090] Step S5: Field interpolation processing.

[0091] In S4, the power generation efficiency evaluation values of each location are completed. In this step, Newton polynomial interpolation can be performed on the entire field, such as Figure 4 As shown in Figure 1, the interpolated data object is the data on each straight line. The target value is the power generation efficiency evaluation value of each other point on the data line. This method can be used to obtain the overall power generation efficiency of the sea surface where the wind turbine platform is located after initial deployment.

[0092] Step S6: Individual power generation efficiency matching

[0093] In step S6, the field interpolation process is completed, meaning that the power generation efficiency of the entire sea surface covered by the wind turbine platform is known. The following utilizes the predicted maximum power generation efficiency. This application utilizes a new wind blade platform that includes vertical-axis wind turbine blades and horizontal-axis wind turbine blades. However, the vertical blades, horizontal-axis blades, and vertical-axis blades have different operating characteristics and efficiency curves, yet this application applies to the same platform. Within a certain wind speed range, the horizontal-axis blades may have higher power generation efficiency; while within other wind speed ranges, the vertical-axis blades may be more efficient. When the predicted maximum power generation efficiency of the horizontal-axis wind turbine and the vertical-axis wind turbine are both low, the blade speed can be operated without restriction. When the predicted maximum power generation efficiency of the horizontal-axis wind turbine and the vertical-axis wind turbine are both high, the horizontal-axis wind turbine blades and the vertical-axis wind turbine blades on the device can be operated within the allowable operating range. However, when the power generation efficiency of the horizontal-axis wind turbine and the vertical-axis wind turbine is significantly unbalanced, adjustments are required to maximize energy harvesting for the entire system. For example, when the wind direction is primarily horizontal, vertical-axis blades are generally more efficient at capturing wind energy, while horizontal-axis blades may be less efficient. In this case, the rotation of the horizontal-axis blades can be reduced by controlling their speed or power output, thereby adjusting the power generation efficiency.

[0094] The vertical axis fan blade adjustment formula is as follows:

[0095]

[0096] y1-y2>y2 (8)

[0097] Where V2 is the speed of the vertical-axis fan blades after adjustment when the horizontal-axis fan efficiency is much greater than the vertical-axis fan efficiency, and v2 is the speed of the vertical-axis fan blades at the vertical-axis fan efficiency y2. y1 and y2 are the horizontal-axis fan efficiency and vertical-axis fan efficiency, respectively.

[0098] The horizontal axis fan blade adjustment formula is as follows:

[0099]

[0100] y2-y1>y1 (10)

[0101] Where V1 is the adjusted rotational speed of the horizontal-axis fan blades when the vertical-axis fan efficiency is significantly greater than that of the horizontal-axis fan, and v1 is the rotational speed of the horizontal-axis fan blades at the horizontal-axis fan efficiency y1. y1 and y2 are the horizontal-axis fan efficiency and vertical-axis fan efficiency, respectively. This formula limits the rotational speed of the lower-efficiency unit when there's a significant difference between the horizontal-axis and vertical-axis fan efficiencies, ensuring that more wind energy is effectively utilized by the higher-efficiency blades.

[0102] S7: SVM binary classification wake effect.

[0103] In steps S5 and S6, the prediction of the highest power generation efficiency point on the sea surface and the operating state when the wind blades are at the highest power generation efficiency are completed. The following is an overall optimization of multiple wind turbine fleets. First, the wake effect of multiple wind turbine fleets needs to be considered. This application uses the SVM binary classification algorithm to extract the wind speed value, wind direction, turbulence level, temperature, wind energy value, and the spacing value between each wind turbine at the location with the highest power generation efficiency. This is to predict whether the wake effect when all devices are arranged to the location where the power generation efficiency is maximized is higher than the set threshold.

[0104] The SVM algorithm training process includes data labeling and normalization. The kernel function used for this SVM algorithm is a polynomial kernel function. Ultimately, the SVM classification accuracy should reach 0.95.

[0105] S8: Optimized deployment.

[0106] In S7, the SVM binary classification wake effect algorithm is trained. In this step, the entire fleet of wind turbines is deployed. First, all turbines are arranged with minimal spacing to maximize power generation efficiency. Then, the SVM algorithm is used to predict the wake effect at that location. If the predicted wake effect value is less than a set threshold, power generation efficiency is maximized. If the predicted wake effect value is higher than a threshold, the spacing is increased and the prediction is repeated. This continues until the wake effect remains below the threshold at maximum power generation efficiency, achieving maximum efficiency and low wake effect.

[0107] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent variation based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. An intelligent optimization method for power generation efficiency of a combined floating wind power platform based on a convolutional neural network, characterized in that: The combined floating wind power platform comprises a horizontal axis wind turbine (1), a long tower (2), a short tower (3), a vertical axis wind turbine (4), a tower mounting base (5), a truncated cone component (6), a buoyancy tank (7), a truss (8), a vertical support rod (10), an inclined support rod (11), a heave plate (12), a ballast tank (13), a mooring cable (14) and a gravity anchor (15), and is characterized in that: the combined floating wind power platform is provided with six groups of buoyancy tanks (7), the six groups of buoyancy tanks (7) are distributed in a regular hexagon, the buoyancy tanks (7) are connected by trusses (8), and the upper part of the buoyancy tanks (7) is provided with a truncated cone component. The truncated cone component (6) is provided with a tower mounting base (5), and two groups of long towers (2) and four groups of short towers (3) are symmetrically mounted on the six groups of tower mounting bases (5), the horizontal axis wind turbine (1) is mounted on the long tower (2), and the vertical axis wind turbine (4) is mounted on the short tower (3); a vertical support rod (10), an oblique support rod (11) and a mooring cable (14) are mounted on the bottom of the buoyancy tank (7), a vertical swing plate (12) is provided below the vertical support rod (10) and the oblique support rod (11), the vertical swing plate (12) is connected to the ballast tank (13) below, and the lower end of the mooring cable (14) is connected to the gravity anchor (15); The combined floating wind power platform is provided with horizontal axis wind turbines (1) and vertical axis wind turbines (4), wherein the horizontal axis wind turbines (1) are in two groups and the vertical axis wind turbines (4) are in four groups. The two groups of horizontal axis wind turbines (1) and the four groups of vertical axis wind turbines (4) are symmetrically mounted on the upper part of the six groups of buoyancy cabins (7), and the steps include: 1) Grid deployment of wind turbine platforms; A grid-based deployment method is used for initial deployment to ensure that a sufficiently large area on the sea surface can be covered and locations with strong wind energy can be found; 2) Parameter measurement; High-precision sensors on each wind turbine platform are used to measure sea surface parameters, including sea surface wind speed, sea surface wind direction, turbulence level, temperature and humidity, which are closely related to wind energy at various locations on the sea surface. 3) Convolutional neural network training: Sea surface wind speed and direction are significant features, which requires network-specific optimization to increase their expressive power. Multi-layer cumulative convolution is used to enhance their expressive power, and a filling formula is used to fill in the two missing units. 4) Effectiveness evaluation; Evaluate the predicted power generation efficiency according to the power generation efficiency evaluation formula to find the area with better power generation efficiency; 5) Field interpolation processing; Newton polynomial interpolation is performed on the entire field to obtain the overall power generation efficiency of the wind turbine platform on the sea surface after initial deployment; 6) Individual power generation efficiency matching; When the horizontal-axis wind turbine power generation efficiency and the vertical-axis wind turbine power generation efficiency at the predicted highest point are smaller, the fan blade speed can be operated without restriction. When the horizontal-axis wind turbine power generation efficiency and the vertical-axis wind turbine power generation efficiency at the predicted highest point are both larger, the horizontal-axis wind blades and the vertical-axis wind blades in the equipment can be operated within the allowable range. However, when the horizontal-axis wind turbine power generation efficiency and the vertical-axis wind turbine power generation efficiency are extremely unbalanced, the vertical-axis wind blade adjustment formula or the horizontal-axis wind blade adjustment formula should be adjusted as needed to maximize the energy collection of the entire system. 7) SVM binary classification wake effect; Using the SVM binary classification algorithm, we extract the wind speed, wind direction, turbulence level, temperature, wind energy value, and the spacing between each wind turbine at the location with the highest power generation efficiency. This is used to predict whether the wake effect of all devices arranged to maximize power generation efficiency will exceed the set threshold. 8) Optimized deployment; All devices are arranged at the location where power generation efficiency is maximized with the minimum spacing. Then, the SVM algorithm is used to predict the wake effect at that time. When the predicted wake effect value is less than the set threshold, the power generation efficiency can be maximized. If the predicted wake effect value is higher than the threshold, the spacing is increased and the prediction is repeated until the wake effect is still below the threshold when the power generation efficiency is maximized, thus achieving maximum power generation efficiency and low wake effect.

2. The method for intelligent optimization of power generation efficiency based on a convolutional neural network according to claim 1, characterized in that: The filling formula in step 3) is expressed as: The filling formula is expressed as: ； ； ； ； in Fill in the wind speed related characteristic values for each layer, is a hyperparameter, Indicates that the current filling value is in the kth convolution layer, Indicates that there are n convolutional layers. It represents the first layer initial input value of the sea surface wind speed after normalization. Fill in the wind direction related eigenvalues for each layer, is a hyperparameter, It represents the first layer initial input value of the sea surface wind direction after normalization.

3. The method for intelligent optimization of power generation efficiency based on a convolutional neural network according to claim 1, characterized in that: The power generation efficiency evaluation formula in step 4) is expressed as: In step 3), the power generation efficiency of each platform was predicted, including the power generation efficiency of the horizontal axis wind turbine and the vertical axis wind turbine at each point. The following needs to be evaluated to find the area with better power generation efficiency. The power generation efficiency evaluation formula is: ； in, is the power generation efficiency evaluation value, is the power generation efficiency of the horizontal axis wind turbine at the location, is the power generation efficiency of the vertical axis wind turbine at that location.

4. The method for intelligent optimization of power generation efficiency based on a convolutional neural network according to claim 1, characterized in that: The vertical axis fan blade adjustment formula in step 6) is expressed as: The vertical axis fan blade adjustment formula is as follows: ； ； in, When the power generation efficiency of the horizontal axis fan is much greater than that of the vertical axis fan, the speed of the vertical axis fan blades is adjusted. For vertical axis fan blades The speed of the vertical axis wind turbine at the power generation efficiency, They are the power generation efficiency of horizontal axis wind turbine and vertical axis wind turbine respectively; The horizontal axis fan blade adjustment formula in step 6) is expressed as: The horizontal axis fan blade adjustment formula is as follows: ； ； in, When the power generation efficiency of the vertical axis fan is much greater than that of the horizontal axis fan, the speed of the horizontal axis fan blades is adjusted. For horizontal axis fan blades The speed of the horizontal axis wind turbine at the power generation efficiency, They are the power generation efficiency of horizontal axis wind turbine and vertical axis wind turbine respectively.

5. The method for intelligent optimization of power generation efficiency based on convolutional neural networks according to claim 1, characterized in that: The combined floating wind power platform is provided with a horizontal axis wind turbine (1) and a vertical axis wind turbine (4), wherein the horizontal axis wind turbine (1) and the vertical axis wind turbine (4) are arranged alternately in a plane and in a vertical plane, and the blade tip of the horizontal axis wind turbine (1) is 3-5 meters higher than the blade tip of the vertical axis wind turbine (4) at the lowest position of rotation.

6. The method for intelligent optimization of power generation efficiency based on convolutional neural networks according to claim 1, characterized in that: The tower mounting base (5) and the frustum component (6) of the combined floating wind power platform are modular prefabricated parts.

7. The method for intelligent optimization of power generation efficiency based on convolutional neural networks according to claim 1, characterized in that: The cross section of the truss (8) of the combined floating wind power platform is circular or square.

8. The method for intelligent optimization of power generation efficiency based on convolutional neural networks according to claim 1, characterized in that: The vertical support rods (10) and the oblique support rods (11) of the combined floating wind power platform have circular cross sections.

Images