Yaw optimization control method and device based on multi-source data fusion and medium

Abstract

The invention relates to the technical field of wind turbine generators, particularly provides a yaw optimization control method and equipment based on multi-source data fusion and a medium, and aims to solve the technical problem that a traditional yaw optimization control method cannot realize accurate optimization control. In order to achieve the purpose, the yaw optimization control method based on multi-source data fusion comprises the steps that a label data set of a wind turbine generator set is constructed, and the label data set comprises first yaw angles of the wind turbine generator set under different wind conditions; acquiring a wind speed and direction prediction sequence in a future preset time period; determining a second yaw angle based on the first yaw angle and the wind speed and direction prediction sequence; and performing yaw optimization control on the wind turbine generator based on the second yaw angle. Optimization control based on the second yaw angle is achieved, the windward angle of the draught fan can be dynamically adjusted, wind energy can be captured to the maximum degree to improve the power generation efficiency, safe operation of the unit can be guaranteed through load constraint, and yaw optimization control over the wind turbine unit is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of wind turbine technology, specifically providing a yaw optimization control method, device, and medium based on multi-source data fusion. Background Technology

[0002] With the continuous development of the wind power industry, the yaw control system plays an increasingly crucial role in improving the energy efficiency of wind turbines. Precise yaw control ensures that the wind turbine is always optimally aligned with the wind flow, thereby maximizing wind energy capture efficiency.

[0003] However, due to the influence of variable meteorological factors such as wind speed and direction, traditional yaw optimization control methods cannot achieve precise optimization control, specifically manifested in response lag leading to a decline in power generation performance. Therefore, there is an urgent need for a yaw optimization control scheme that integrates multi-source data and balances simulation accuracy with real-time prediction capabilities. Summary of the Invention

[0004] To overcome the aforementioned shortcomings, this application is proposed to provide a solution, or at least a partial solution, to the technical problem that traditional yaw optimization control methods cannot achieve precise optimization control. This application provides a yaw optimization control method, device, and medium based on multi-source data fusion.

[0005] In a first aspect, this application provides a yaw optimization control method based on multi-source data fusion, the method comprising:

[0006] Construct a labeled dataset for wind turbines, wherein the labeled dataset includes the first yaw angle of the wind turbines under different wind conditions;

[0007] Obtain the wind speed and direction prediction sequence for a future preset time period;

[0008] The second yaw angle is determined based on the first yaw angle and the wind speed and direction prediction sequence;

[0009] The wind turbine is subjected to yaw optimization control based on the second yaw angle.

[0010] In one embodiment of the yaw optimization control method based on multi-source data fusion in this application, the construction of the wind turbine tag dataset includes:

[0011] Obtain the attribute parameters of the wind turbine;

[0012] A wind turbine model is built based on the attribute parameters of the wind turbine.

[0013] A turbulent dynamic wind file is generated based on the wind turbine model, wherein the turbulent dynamic wind file includes wind speed, wind direction and turbulence intensity;

[0014] The turbulent dynamic wind files are tagged to obtain standard wind condition combinations;

[0015] Determine the first yaw angle corresponding to each standard wind condition;

[0016] By integrating all standard wind conditions and their corresponding first yaw angles, the labeled dataset is obtained.

[0017] In one embodiment of the yaw optimization control method based on multi-source data fusion in this application, the tagging of the turbulent dynamic wind file includes:

[0018] The wind speed intervals from the cut-in wind speed to the cut-out wind speed in the turbulent dynamic wind file are divided into wind speed intervals; the wind direction intervals in the turbulent dynamic wind file are divided into wind direction intervals; and the turbulence intensity in the turbulent dynamic wind file is divided into low, medium and high levels.

[0019] In one embodiment of the yaw optimization control method based on multi-source data fusion in this application, determining the first yaw angle corresponding to each of the standard wind conditions includes:

[0020] Using the yaw angle as the iterative variable, the utility function value under different yaw angles is calculated iteratively, wherein the utility function is determined based on the output power of the wind turbine under the standard wind conditions and the load over-threshold penalty term;

[0021] The calculation stops when the number of iterations reaches the iteration threshold, and the optimal utility function under the iteration threshold is obtained.

[0022] The yaw angle corresponding to the optimal utility function value is taken as the first yaw angle.

[0023] In one embodiment of the yaw optimization control method based on multi-source data fusion in this application, obtaining the wind speed and direction prediction sequence for a future preset time period includes:

[0024] Collect the first wind speed and direction sequence;

[0025] The first wind speed and direction sequence is standardized.

[0026] The standardized first wind speed and direction sequence is input into a pre-trained echo state network to obtain a wind speed and direction prediction sequence for a future preset time period.

[0027] In one embodiment of the yaw optimization control method based on multi-source data fusion in this application, the pre-trained echo state network is obtained through the following steps:

[0028] Obtain the second wind speed and direction sequence as the wind speed and direction training set;

[0029] The data in the wind speed and direction training set are preprocessed to obtain a preprocessed dataset;

[0030] Construct an echo state network model;

[0031] The echo state network model is trained using the preprocessed dataset to obtain the pretrained echo state network.

[0032] In one embodiment of the yaw optimization control method based on multi-source data fusion in this application, obtaining the second yaw angle based on the first yaw angle and the wind speed and direction prediction sequence includes:

[0033] Construct a backpropagation neural network model;

[0034] The backpropagation neural network model is trained using the first yaw angle to obtain a pre-trained backpropagation neural network;

[0035] The wind speed and direction prediction sequence is input into the pre-trained backpropagation neural network to obtain the second yaw angle.

[0036] In one embodiment of the yaw optimization control method based on multi-source data fusion in this application, training the backpropagation neural network using the first yaw angle and the predicted wind speed and direction includes:

[0037] Obtain constraint data for wind turbine units;

[0038] The historical wind speed and direction prediction values ​​and the constraint data are combined as a sample input feature vector, which is then input into the backpropagation neural network to output the third yaw angle.

[0039] Calculate the prediction error between the first yaw angle and the third yaw angle;

[0040] The network parameters of the backpropagation neural network are adjusted based on the prediction error until the prediction error between the first yaw angle and the third yaw angle is less than a preset threshold or the number of iterations is greater than a preset number, thereby obtaining a pre-trained backpropagation neural network.

[0041] In a second aspect, an electronic device is provided, comprising:

[0042] At least one processor;

[0043] And, a memory communicatively connected to the at least one processor;

[0044] The memory stores a computer program, which, when executed by the at least one processor, is the aforementioned yaw optimization control method based on multi-source data fusion.

[0045] In a third aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and run by a processor to perform the yaw optimization control method based on multi-source data fusion as described in any of the preceding claims.

[0046] The above-described technical solutions of this application have at least one or more of the following beneficial effects:

[0047] The yaw optimization control method based on multi-source data fusion in this application includes: constructing a labeled dataset of wind turbine units, wherein the labeled dataset includes the first yaw angle of the wind turbine units under different wind conditions; obtaining a wind speed and direction prediction sequence for a future preset time period; determining a second yaw angle based on the first yaw angle and the wind speed and direction prediction sequence; and performing yaw optimization control on the wind turbine units based on the second yaw angle. This achieves optimized control based on the second yaw angle, which can dynamically adjust the wind turbine's angle of attack, maximizing wind energy capture to improve power generation efficiency while ensuring safe operation of the unit through load constraints, effectively improving the yaw optimization control of wind turbine units. Attached Figure Description

[0048] The disclosure of this application will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this application. Furthermore, similar numbers in the drawings are used to denote similar components, wherein:

[0049] Figure 1 This is a schematic diagram of the main process of a yaw optimization control method based on multi-source data fusion in one embodiment of this application;

[0050] Figure 2 This is a schematic diagram of the structure of an echo state network in one embodiment of this application;

[0051] Figure 3 This is a schematic diagram of the backpropagation neural network structure in one embodiment of this application;

[0052] Figure 4 This is a schematic diagram of the main structure of a yaw optimization control device based on multi-source data fusion in one embodiment of this application;

[0053] Figure 5 This is a schematic diagram of the structure of an electronic device in one embodiment of this application. Detailed Implementation

[0054] Some embodiments of this application are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of this application and are not intended to limit the scope of protection of this application.

[0055] In the description of this application, "module" and "processor" can include hardware, software, or a combination of both. A module can include hardware circuitry, various suitable sensors, communication ports, memory, and can also include software components, such as program code, or a combination of software and hardware. A processor can be a central processing unit, microprocessor, image processor, digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Non-transitory computer-readable storage media includes any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B, or A and B. The terms "at least one A or B" or "at least one of A and B" have a similar meaning to "A and / or B" and can include only A, only B, or A and B. The singular terms "a" or "this" can also include plural forms.

[0056] Currently, due to the influence of variable meteorological factors such as wind speed and direction, traditional yaw optimization control methods cannot achieve precise optimization control. Therefore, this application proposes a yaw optimization control method, device, and medium based on multi-source data fusion.

[0057] See appendix Figure 1 , Figure 1 This is a schematic flowchart of the main steps of a yaw optimization control method based on multi-source data fusion according to one embodiment of this application.

[0058] like Figure 1 As shown, the yaw optimization control method based on multi-source data fusion in this application embodiment mainly includes the following steps S10-S40.

[0059] Step S10: Construct a label dataset for the wind turbine, wherein the label dataset includes the first yaw angle of the wind turbine under different wind conditions.

[0060] The labeled dataset of wind turbines refers to the data set showing the correspondence between different wind conditions and the first yaw angle under different standardized wind conditions.

[0061] The first yaw angle is the theoretically optimal yaw angle in the labeled dataset corresponding to a specific wind condition.

[0062] Step S20: Obtain the wind speed and direction prediction sequence for a future preset time period.

[0063] The preset time period in the future can be a specific time period in the future, such as 30 seconds, 1 minute, or half an hour in the future.

[0064] Wind speed and direction prediction sequences are continuous data sequences generated by predicting the trends of wind speed and direction changes within a preset time period in the future. They can reflect the dynamic changes in wind conditions in the short term.

[0065] Step S30: Determine the second yaw angle based on the first yaw angle and the wind speed and direction prediction sequence.

[0066] Step S40: Perform yaw optimization control on the wind turbine based on the second yaw angle.

[0067] Specifically, based on the second yaw angle command, the yaw actuator of the wind turbine is driven to adjust the direction of the nacelle, so that the wind turbine blades are always at the optimal windward angle to adapt to future wind conditions, thereby realizing a closed-loop control process that improves power generation efficiency and reduces load risk.

[0068] Based on steps S10-S40 above, a labeled dataset for the wind turbine is first constructed, including the first yaw angle of the wind turbine under different wind conditions; a wind speed and direction prediction sequence for a future preset time period is obtained; a second yaw angle is determined based on the first yaw angle and the wind speed and direction prediction sequence; and yaw optimization control of the wind turbine is performed based on the second yaw angle. This achieves the goal of dynamically adjusting the wind turbine's windward angle based on the second yaw angle, which can maximize wind energy capture to improve power generation efficiency, and ensure the safe operation of the unit through load constraints, effectively improving the yaw optimization control of the wind turbine.

[0069] The following provides further explanation of steps S10 to S40.

[0070] Specifically, step S10 can be implemented through steps S101 to S106.

[0071] Step S101: Obtain the attribute parameters of the wind turbine.

[0072] The attribute parameters of a wind turbine refer to the key physical and aerodynamic parameters of the core components of the target wind turbine, including the mass, stiffness, and aerodynamic shape of the blades, tower, and nacelle. These parameters directly determine the structural characteristics, stress patterns, and wind energy capture capabilities of the wind turbine, and are the core basis for building an accurate simulation model to ensure that the model can reproduce the actual operating state of the wind turbine.

[0073] Step S102: Build a wind turbine model based on the attribute parameters of the wind turbine.

[0074] Specifically, a wind turbine model can be configured in simulation software based on the turbine's attribute parameters, and a dynamic yaw angle adjustment module can be built using Simulink. The constructed wind turbine model is essentially a digital twin, capable of simulating power output and load variations under different yaw angles and wind conditions. For example, OpenFAST can be used as a simulation software example.

[0075] Step S103: Generate a turbulent dynamic wind file based on the wind turbine model, wherein the turbulent dynamic wind file includes wind speed, wind direction and turbulence intensity.

[0076] Turbulent dynamic wind files can be generated by calling the built-in wind turbine model in simulation software, containing wind speed, wind direction, and turbulence intensity. This file simulates the dynamic characteristics of a real wind field. Wind speed and wind direction are the core environmental variables affecting the wind turbine's upwind efficiency, while turbulence intensity reflects the stability of the wind field. Together, these three factors constitute the complete external environment input for wind turbine operation, ensuring that the simulation scenario closely matches actual wind field conditions.

[0077] Step S104: Tag the turbulent dynamic wind file to obtain standard wind condition combinations.

[0078] In a specific embodiment of the yaw optimization control method based on multi-source data fusion in this application, the step of tagging the turbulent dynamic wind file includes: dividing the wind speed interval from the cut-in wind speed to the cut-out wind speed in the turbulent dynamic wind file into wind speed intervals; dividing the wind direction interval in the turbulent dynamic wind file into wind direction intervals; and classifying the turbulence intensity in the turbulent dynamic wind file into low, medium, and high levels.

[0079] The preset wind speed interval and preset wind direction interval can be preset values, which can be adapted to the actual scenario without specific limitations.

[0080] The cut-in wind speed and cut-out wind speed can be given wind speed values ​​in the simulated scenario.

[0081] The wind direction can be a given wind direction value in the simulated scenario. For example, [-20°, 20°] can be used as an example of the wind direction.

[0082] For example, the wind speed from the wind speed entering the fan to the wind speed exiting the fan can be divided into 2 m / s, the wind direction [-20°, 20°] can be divided into 2°, and the turbulence intensity can be divided into three levels: low, medium, and high according to the actual scenario.

[0083] By labeling complex wind field data, a systematic and standardized combination of wind conditions is transformed, providing a unified scenario benchmark for subsequent batch solutions of the optimal yaw angle.

[0084] Step S105: Determine the first yaw angle corresponding to each standard wind condition.

[0085] Specifically, step S105 can be implemented through the following steps S1051 to S1053.

[0086] Step S1051: Using the yaw angle as the iteration variable, iteratively calculate the utility function value under different yaw angles, wherein the utility function is determined based on the output power of the wind turbine under the standard wind conditions and the load over-threshold penalty term.

[0087] The utility function is the core indicator for evaluating the quality of yaw angles, and its expression is:

[0088] in, For utility function, This represents the average power output of the wind turbine under the current standard wind conditions and current yaw angle. This is a weighting coefficient, which can be preset according to the characteristics of the wind turbine and the application scenario, and is used to balance the priority of power and load. This is a penalty term for loads exceeding a certain threshold. Specifically, when critical loads such as blade root bending moment and tower base bending moment exceed a safety threshold, a penalty term is calculated based on the magnitude of the exceedance. The greater the load exceedance, the larger the penalty value; if the threshold is not exceeded, the penalty term is 0. For example, when critical loads such as blade root bending moment and tower base bending moment exceed the safety threshold, a penalty term can be calculated based on... Calculate the penalty term, where For preset coefficients, These are the actual monitoring values ​​of the critical loads. This refers to the safety threshold corresponding to the critical load.

[0089] Specifically, under standard wind conditions, a multi-objective optimization algorithm can be used to drive iterative changes in the yaw angle, with the yaw angle as the sole object of iterative adjustment. That is, by continuously changing the value of the yaw angle, the corresponding values ​​can be calculated using a wind turbine simulation model. and This allows us to obtain utility function values ​​at different yaw angles, balancing the relationship between power and load while maximizing power output and minimizing load risk. Constraints can also be set during the iterative calculation process, such as allowing the blade load to be lower than the maximum allowable load value.

[0090] Step S1052: Stop the calculation when the number of iterations reaches the iteration number threshold, and obtain the optimal utility function under the iteration number threshold.

[0091] The threshold for the number of iterations is a value obtained in advance through experiments. It can be adjusted adaptively according to the actual scenario, and no specific limit is set for it.

[0092] Specifically, the particle swarm optimization algorithm is used in the Simulink-MATLAB environment for automatic iteration to quickly find the optimal solution that satisfies the constraints. During the iteration process, the utility function values ​​corresponding to all yaw angles are recorded in real time, and the current maximum value is tracked. When the number of iterations reaches the iteration threshold, yaw angle adjustment and utility function calculation are stopped. At this point, the largest value is selected from the utility function values ​​recorded in all iteration steps, which is the optimal utility function value under the iteration threshold. This value represents the optimal performance index obtainable within the preset computational cost under the current standard wind conditions. It should be noted that using the particle swarm optimization algorithm to find the optimal solution that satisfies the constraints is a conventional optimization algorithm, and will not be elaborated upon here.

[0093] Step S1053: Take the yaw angle corresponding to the optimal utility function value as the first yaw angle.

[0094] Specifically, once the optimal utility function value is found, the yaw angle corresponding to the optimal utility function value can be obtained, and this yaw angle is defined as the first yaw angle under the current standard wind conditions.

[0095] Step S106: Integrate all standard wind conditions and their corresponding first yaw angles to obtain the label dataset.

[0096] Specifically, all standard wind conditions are associated with their corresponding first yaw angles to form a complete labeled dataset.

[0097] Through high-fidelity simulation and dataset construction, yaw angles under different wind conditions are simulated using digital wind farm and wind turbine models. A multi-objective optimization algorithm is employed to maximize power output and minimize load risk, while ensuring blade load remains within safe limits. The optimal yaw angle is iteratively solved using a particle swarm optimization algorithm, and a theoretically optimal yaw angle dataset is constructed to provide a reference for subsequent decision-making training.

[0098] The above is a further explanation of step S10. Step S20 will be further explained below.

[0099] Specifically, step S20 can be implemented through steps S201 to S203.

[0100] Step S201: Collect the first wind speed and direction sequence.

[0101] The first wind speed and direction sequence refers to the time series data of wind speed and direction of the wind turbines located in the wind farm, which is a preset time period away from the current time. Specifically, it can be obtained from the wind farm's SCADA (Supervisory and Data Acquisition) system or wind speed and direction sensors deployed on-site. The first wind speed and direction sequence records the wind speed value and wind direction angle within a specific period (such as the most recent 30 seconds or 1 minute) using continuous timestamps as the dimension, forming an ordered time series data set.

[0102] Step S202: Standardize the first wind speed and direction sequence.

[0103] Standardization is a preprocessing step that converts the first wind speed and direction sequence into a uniform scale and eliminates the influence of dimensions. The commonly used method is standardization, which involves calculating the mean or standard deviation of the wind speed and direction data in the first wind speed and direction sequence to obtain standardized data.

[0104] Step S203: Input the standardized first wind speed and direction sequence into the pre-trained echo state network to obtain the wind speed and direction prediction sequence for the future preset time period.

[0105] Specifically, the standardized first wind speed and direction sequence is input into a pre-trained echo state network, which outputs a wind speed and direction prediction sequence for a preset time period in the future. For example, it outputs wind speed and direction prediction sequences for the next 30 seconds or 1 minute.

[0106] Echo-state networks (ERNs) are a type of artificial recurrent neural network that combines the characteristics of RNNs. Internal neurons contain feedback loops, allowing them to remain active even without external input and exhibiting a memory effect. This characteristic enables ERNs to handle many nonlinear prediction problems and achieve better results than traditional recurrent neural networks.

[0107] Echo state network structure diagram as follows Figure 2 As shown, specifically, the basic structure of the echo state network consists of three main components: an input layer, a reservoir layer, and an output layer. The input layer contains... Each neuron receives input information and transmits it to the reservoir layer; the reservoir layer contains... The internal neurons process and map the input information; the output layer contains... Each neuron decodes and outputs the information processed by the reservoir layer. The connection between the input layer and the reservoir layer is as follows: The weight matrix is ​​of order, and the interconnections within the reserve pool layer are... The weight matrix is ​​of order [-1, 1]. Before training, the weight matrix is ​​fixed and is a symmetric matrix randomly selected in the interval [-1, 1]. The connection between the reservoir and the output layer is... The weight matrix is ​​of order 1, and its weight coefficients are usually adjusted by the least squares method.

[0108] For the input layer and the reservoir layer The weight matrix is ​​of order W, where W represents the interconnection of neurons within the reservoir layer. The weight matrix before training of the echo-state network. W is usually fixed and is a symmetric matrix that is randomly selected and uniformly distributed in the interval [-1,1]. For the space between the reservoir layer and the output layer The weights are typically adjusted using the least squares method. For the output layer to the reservoir layer The feedback matrix. To ensure the echo-state characteristics of the echo-state network, the weight matrix... The spectral radius must be less than 1. Generally, the matrix... Scaling to ,in spectral radius, Scaling factor (0 < <1).

[0109] Specifically, the pre-trained echo state network can be obtained through the following steps S2031 to S2034.

[0110] Step S2031: Obtain the second wind speed and direction sequence as the wind speed and direction training set.

[0111] The second wind speed and direction sequence can be selected from the rich historical wind speed and direction sequences in the wind field SCADA system.

[0112] Step S2032: Preprocess the data in the wind speed and direction training set to obtain a preprocessed dataset.

[0113] Specifically, preprocessing can involve removing noisy data and filling in missing data in the wind speed and direction training set, as well as standardizing the data in the wind speed and direction training set, where standardization can involve calculating the mean or standard deviation.

[0114] Step S2033: Construct the echo state network model.

[0115] Specifically, it can be constructed as follows: Figure 2 The diagram shows three main components: an input layer, a reservoir layer, and an output layer. The reservoir weights are randomly initialized and fixed.

[0116] Step S2034: Train the echo state network model using the preprocessed dataset to obtain the pre-trained echo state network.

[0117] Specifically, only the weight matrix between the reservoir layer and the output layer. These are the core parameters that need to be trained, while other parameters are fixed after random initialization. Further, during training, the standardized historical wind speed and direction training sequences are first input into the echo state network model. A global reservoir state matrix capturing time-series dependencies is generated through a fixed-structure reservoir, while simultaneously constructing the corresponding future wind speed and direction training target matrix. Then, using the ridge regression algorithm as the core, the output layer weights are efficiently solved by minimizing the loss function of the prediction error and the regularization term, with the regularization parameter determined through cross-validation to avoid overfitting. Finally, the weight training effect is verified, and the input layer and reservoir weights of the echo state network model are fixed to form a pre-trained model with wind speed and direction prediction capabilities.

[0118] The above is a further explanation of step S20. Step S30 will be further explained below.

[0119] Specifically, step S30 can be implemented through steps S301 to S304.

[0120] Step S301: Construct a backpropagation neural network model.

[0121] Specifically, such as Figure 3 As shown, a backpropagation neural network model typically includes an input layer, one or more hidden layers, and an output layer, with each layer being interconnected by neurons through connection weights.

[0122] Step S302: Train the backpropagation neural network model using the first yaw angle and the wind speed and direction prediction sequence to obtain a pre-trained backpropagation neural network.

[0123] Specifically, during training, input data is passed from the input layer through the hidden layers to the output layer and compared with the expected output to calculate the error. If the error exceeds the allowable range, the connection weights and neuron thresholds of each layer are adjusted layer by layer through the error backpropagation mechanism. The error is continuously reduced through multiple iterations of training until the preset accuracy requirement is reached, thus obtaining the pre-trained backpropagation neural network.

[0124] Specifically, step S302 can be implemented through the following steps S3021 to S3024.

[0125] Step S3021: Obtain the constraint data of the wind turbine.

[0126] The constraint data for wind turbines are key boundary conditions ensuring their safe operation. Specifically, these refer to the safety thresholds and real-time operational status data of critical components, including structural load constraints and operational status constraints. Structural load constraints may include blade root bending moment thresholds, tower base bending moment thresholds, and nacelle vibration thresholds, determined by turbine design parameters and industry safety standards. Operational status constraints include the maximum rotational speed of the yaw mechanism, yaw angle adjustment range limits, and current turbine fault status indicators, reflecting the turbine's actual operational capabilities. Constraint data can be obtained through the wind turbine's SCADA (Supervisory Control and Data Acquisition) system.

[0127] Step S3022: Combine the historical wind speed and direction prediction values ​​with the constraint data as a sample input feature vector, input it into the backpropagation neural network, and output the third yaw angle.

[0128] Historical wind speed and direction predictions can be obtained from historical data and used to train a backpropagation neural network.

[0129] Specifically, historical wind speed and direction predictions can be input into the constructed backpropagation neural network for training, with constraint data serving as the constraints during the training process. The input layer receives historical wind speed and direction predictions and passes them to the hidden layer. The hidden layer performs a non-linear transformation on the input features using an activation function (such as the ReLU function) to capture the complex relationship between the wind speed and direction predictions, constraint data, and the third yaw angle. Finally, the output layer outputs the specific yaw angle value, which is the third yaw angle.

[0130] Step S3023: Calculate the prediction error between the first yaw angle and the third yaw angle.

[0131] Specifically, the first yaw angle is used as the true label, and the third yaw angle is used as the model prediction value. The degree of deviation between the two is calculated through a preset error evaluation index, where the mean square error (MSE) can be used as an example of an error evaluation index.

[0132] Step S3024: Adjust the network parameters of the backpropagation neural network based on the prediction error until the prediction error between the first yaw angle and the third yaw angle is less than a preset threshold or the number of iterations is greater than a preset number, thereby obtaining a pre-trained backpropagation neural network.

[0133] The preset threshold and preset number of times are both pre-set values, which can be adjusted adaptively according to the actual scenario. No specific limits are set here.

[0134] Specifically, the calculated prediction error is propagated back from the output layer to the hidden and input layers. The gradient of the error with respect to the connection weights and neuron biases (i.e., the rate at which the error changes with the parameters) is calculated layer by layer to determine the contribution of each parameter to the error. Based on the calculated gradients, the weights and biases of each layer are adjusted according to a preset learning rate, so that the error in the next prediction tends to decrease. In other words, model parameters are adjusted using error backpropagation and gradient descent, ultimately obtaining the optimal model parameters when the prediction error is less than a preset threshold or the number of iterations exceeds a preset number.

[0135] Step S303: Input the wind speed and direction prediction sequence into the pre-trained backpropagation neural network to obtain the second yaw angle.

[0136] Specifically, the standardized wind speed and direction prediction sequence is input into a pre-trained backpropagation neural network. The output layer of the backpropagation neural network outputs a second yaw angle, which is the optimal actual yaw control angle output by the backpropagation neural network for the current real-time wind conditions. Finally, the yaw actuator of the wind turbine can adjust the nacelle direction according to the second yaw angle command, so that the wind turbine blades are always at the optimal windward angle to adapt to future wind conditions, ultimately achieving the dual goals of improving power generation efficiency and ensuring operational safety.

[0137] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effect of this application, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the scope of protection of this application.

[0138] Furthermore, this application also provides a yaw optimization control device based on multi-source data fusion.

[0139] See appendix Figure 4 , Figure 4 This is a main structural block diagram of a yaw optimization control device based on multi-source data fusion according to an embodiment of this application.

[0140] like Figure 4 As shown, the yaw optimization control device based on multi-source data fusion in this embodiment mainly includes a construction module 11, an acquisition module 12, a determination module 13, and a control module 14. In some embodiments, one or more of the construction module 11, acquisition module 12, determination module 13, and control module 14 can be combined into a single module.

[0141] In some embodiments, the construction module 11 can be configured to construct a label dataset for a wind turbine, wherein the label dataset includes the first yaw angle of the wind turbine under different wind conditions.

[0142] The acquisition module 12 can be configured to acquire a wind speed and wind direction prediction sequence for a future preset time period.

[0143] The determining module 13 can be configured to determine the second yaw angle based on the first yaw angle and the wind speed and direction prediction sequence.

[0144] Control module 14 can be configured to perform yaw optimization control on the wind turbine based on the second yaw angle.

[0145] In one implementation, a description of the specific function can be found in steps S10-S40.

[0146] The aforementioned yaw optimization control device based on multi-source data fusion is used to execute... Figure 1 The embodiments of the yaw optimization control method based on multi-source data fusion shown are similar in technical principle, technical problem solved and technical effect. Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the yaw optimization control device based on multi-source data fusion can be found in the embodiments of the yaw optimization control method based on multi-source data fusion, and will not be repeated here.

[0147] Furthermore, it should be understood that since the various modules are only provided to illustrate the functional units of the device described in this application, the physical devices corresponding to these modules may be the processor itself, or a part of the processor's software, hardware, or a combination of both. Therefore, the number of modules shown in the figures is merely illustrative.

[0148] Those skilled in the art will understand that the various modules in the device can be adaptively split or combined. Such splitting or combining of specific modules will not cause the technical solution to deviate from the principles of this application; therefore, the technical solutions after splitting or combining will fall within the protection scope of this application.

[0149] Those skilled in the art will understand that all or part of the processes in the method of the above-described embodiment can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0150] Furthermore, this application also provides an electronic device, which may include at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program, which, when executed by the at least one processor, implements the yaw optimization control method based on multi-source data fusion as described in any of the above embodiments. See also Figure 5 As shown, Figure 5 The structure of an electronic device, including a processor 100 and a memory 200, is illustrated by way of example.

[0151] Furthermore, this application also provides a computer-readable storage medium. In one embodiment of the computer-readable storage medium according to this application, the computer-readable storage medium can be configured to store a program that executes the yaw optimization control method based on multi-source data fusion of the above-described method embodiments. This program can be loaded and run by a processor to implement the above-described yaw optimization control method based on multi-source data fusion. For ease of explanation, only the parts related to the embodiments of this application are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of this application. The computer-readable storage medium can be a memory device formed by various electronic devices. Optionally, in the embodiments of this application, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0152] The technical solution of this application has been described in conjunction with the specific embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of this application is obviously not limited to these specific embodiments. Without departing from the principles of this application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of this application.

Claims

1. A yaw optimization control method based on multi-source data fusion, characterized in that, The method includes: Construct a labeled dataset for wind turbines, wherein the labeled dataset includes the first yaw angle of the wind turbines under different wind conditions; Obtain the wind speed and direction prediction sequence for a future preset time period; The second yaw angle is determined based on the first yaw angle and the wind speed and direction prediction sequence; The wind turbine is subjected to yaw optimization control based on the second yaw angle.

2. The yaw optimization control method based on multi-source data fusion according to claim 1, characterized in that, The labeled dataset for constructing wind turbine generators includes: Obtain the attribute parameters of the wind turbine; A wind turbine model is built based on the attribute parameters of the wind turbine. A turbulent dynamic wind file is generated based on the wind turbine model; The turbulent dynamic wind files are tagged to obtain standard wind condition combinations; Determine the first yaw angle corresponding to each standard wind condition; By integrating all standard wind conditions and their corresponding first yaw angles, the labeled dataset is obtained.

3. The yaw optimization control method based on multi-source data fusion according to claim 2, characterized in that, The turbulent dynamic wind file includes cut-in wind speed, cut-out wind speed, wind direction, and turbulence intensity; the tagging of the turbulent dynamic wind file includes: The wind speed intervals from the cut-in wind speed to the cut-out wind speed in the turbulent dynamic wind file are divided into wind speed intervals; the wind direction intervals in the turbulent dynamic wind file are divided into wind direction intervals; and the turbulence intensity in the turbulent dynamic wind file is divided into low, medium and high levels.

4. The yaw optimization control method based on multi-source data fusion according to claim 2, characterized in that, Determining the first yaw angle corresponding to each of the standard wind conditions includes: Using the yaw angle as the iteration variable, the utility function value under different yaw angles is calculated iteratively, wherein the utility function is determined based on the output power of the wind turbine under the standard wind conditions and the load over-threshold penalty term; The calculation stops when the number of iterations reaches the iteration threshold, and the optimal utility function under the iteration threshold is obtained. The yaw angle corresponding to the optimal utility function value is taken as the first yaw angle.

5. The yaw optimization control method based on multi-source data fusion according to claim 1, characterized in that, The step of obtaining the wind speed and direction prediction sequence for a future preset time period includes: Collect the first wind speed and direction sequence; The first wind speed and direction sequence is standardized. The standardized first wind speed and direction sequence is input into a pre-trained echo state network to obtain a wind speed and direction prediction sequence for a future preset time period.

6. The yaw optimization control method based on multi-source data fusion according to claim 5, characterized in that, The pre-trained echo state network is obtained through the following steps: Obtain the second wind speed and direction sequence as the wind speed and direction training set; The data in the wind speed and direction training set are preprocessed to obtain a preprocessed dataset; Construct an echo state network model; The echo state network model is trained using the preprocessed dataset to obtain the pretrained echo state network.

7. The yaw optimization control method based on multi-source data fusion according to claim 1, characterized in that, The step of obtaining the second yaw angle based on the first yaw angle and the wind speed and direction prediction sequence includes: Construct a backpropagation neural network model; The backpropagation neural network model is trained using the first yaw angle to obtain a pre-trained backpropagation neural network; The wind speed and direction prediction sequence is input into the pre-trained backpropagation neural network to obtain the second yaw angle.

8. The yaw optimization control method based on multi-source data fusion according to claim 7, characterized in that, The step of training the backpropagation neural network using the first yaw angle includes: Obtain constraint data for wind turbine units; The historical wind speed and direction prediction values ​​and the constraint data are combined as a sample input feature vector, which is then input into the backpropagation neural network to output the third yaw angle. Calculate the prediction error between the first yaw angle and the third yaw angle; The network parameters of the backpropagation neural network are adjusted based on the prediction error until the prediction error between the first yaw angle and the third yaw angle is less than a preset threshold or the number of iterations is greater than a preset number, thereby obtaining a pre-trained backpropagation neural network.

9. An electronic device, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores a computer program, which, when executed by the at least one processor, implements the yaw optimization control method based on multi-source data fusion as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by a processor to perform the yaw optimization control method based on multi-source data fusion as described in any one of claims 1 to 8.

Images