Incoming wind condition combined prediction method, device and equipment based on multi-task learning

Abstract

The invention relates to an incoming flow wind condition combined prediction method, device and equipment based on multi-task learning. The method comprises the following steps: constructing an input data set and an output data set based on incoming flow wind condition measured data, constructing an incoming flow wind condition combined prediction model, performing training and hyper-parameter optimization on the incoming flow wind condition combined prediction model to obtain a target incoming flow wind condition combined prediction model, and predicting the incoming flow wind condition through the target incoming flow wind condition combined prediction model. And obtaining second-level incoming flow wind condition prediction data, wherein the incoming flow wind condition prediction data comprises incoming flow wind speed prediction data and incoming flow wind direction prediction data. Compared with the prior art, the embodiment of the invention has the advantages that the overall dynamic state of the future wind condition can be effectively captured by building and training the incoming wind condition combined prediction model, so that high-precision prediction of the second-level wind speed and the wind direction is realized. The control of the wind turbine generator and the wind power plant is facilitated, and key support is provided for improving the generating capacity, prolonging the service life of equipment, enhancing the stability of a power grid and the like.

Description

Technical Field

[0001] The present disclosure relates to the field of renewable energy power generation technology, and in particular to a method, device and equipment for joint prediction of incoming wind conditions based on multi-task learning. Background Art

[0002] In the field of wind power development, high temporal resolution future wind condition data is crucial. It can provide key support for wind turbine control and wind farm control, thereby improving power generation and grid stability.

[0003] Among the related technologies, current wind forecasting technologies are mainly focused on the prediction of wind speed, while research on the prediction of wind direction is relatively scarce. Most of the research only predicts wind speed or wind direction separately, and fails to capture the overall dynamics of future wind conditions. In addition, the time resolution of most current forecasting technologies is usually 15 minutes or longer, while the wind conditions of wind turbines change frequently and with large amplitudes, and this time resolution cannot meet the needs of wind farm control. Therefore, the above-mentioned wind forecasting technologies have problems such as being unable to capture the overall dynamics of future wind conditions and having too low a time resolution. Summary of the invention

[0004] In order to solve the above technical problems or at least partially solve the above technical problems, the present disclosure provides a method, device and equipment for joint prediction of incoming wind conditions based on multi-task learning.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for joint prediction of incoming wind conditions based on multi-task learning, the method comprising:

[0006] Construct input data set and output data set based on the measured data of incoming wind conditions;

[0007] Building a joint prediction model for incoming wind conditions, wherein the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task;

[0008] Training the incoming wind condition joint prediction model and optimizing hyperparameters to obtain a target incoming wind condition joint prediction model;

[0009] The incoming wind condition is predicted by the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level, wherein the incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data.

[0010] In some embodiments, the step of constructing an input data set and an output data set based on the measured data of the incoming wind conditions includes:

[0011] Preprocessing the measured data of incoming wind conditions to obtain preprocessed measured data of incoming wind conditions;

[0012] Based on the pre-processed measured data of incoming wind conditions, the input data set and the output data set are constructed through the sliding time window algorithm;

[0013] The input data set is a historical wind condition data set at the second level, and the output data set is a future wind condition data set at the second level.

[0014] In some embodiments, the building of the incoming wind condition joint prediction model includes:

[0015] Designing a model architecture of the incoming wind condition joint prediction model;

[0016] A hard parameter sharing mechanism is adopted for the incoming wind speed prediction task and the incoming wind direction prediction task in the incoming wind condition joint prediction model;

[0017] Based on the weights corresponding to the incoming wind speed prediction task and the incoming wind direction prediction task, the loss function of the incoming wind condition joint prediction model is customized.

[0018] In some embodiments, the model architecture of the design of the incoming wind condition joint prediction model includes:

[0019] The long short-term memory network and the fully connected network are used as the core structure of the joint prediction model of incoming wind conditions;

[0020] The long short-term memory network is used to capture the correlation characteristics of wind speed and wind direction in the time series, and the fully connected network is used to predict the incoming wind conditions based on the correlation characteristics.

[0021] In some embodiments, the training of the incoming wind condition joint prediction model and the optimization of hyperparameters to obtain the target incoming wind condition joint prediction model includes:

[0022] Create training sets, validation sets, and test sets based on the input data set and the output data set;

[0023] Initializing the model parameters of the incoming wind condition joint prediction model and setting a hyperparameter search space;

[0024] For each hyperparameter combination in the hyperparameter search space, the inflow wind condition joint prediction model corresponding to each hyperparameter combination is trained using the training set until the inflow wind condition joint prediction model corresponding to each hyperparameter combination converges;

[0025] Using the test set to test the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations, respectively, to obtain the prediction results of the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations;

[0026] The prediction results of the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations are evaluated by using the model evaluation index to obtain the evaluation value of the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations;

[0027] Based on the model evaluation index, the best evaluation value is determined from the evaluation values ​​of the incoming wind condition joint prediction models corresponding to each hyperparameter combination, and the incoming wind condition joint prediction model corresponding to the hyperparameter combination matching the best evaluation value is determined as the target incoming wind condition joint prediction model.

[0028] In some embodiments, the step of preparing a training set, a validation set, and a test set based on the input data set and the output data set includes:

[0029] Normalizing the input data set and the output data set to obtain a normalized input data set and a normalized output data set;

[0030] The normalized input data set and the normalized output data set are divided into a training set, a validation set, and a test set according to a preset ratio.

[0031] In some embodiments, the forecasting of the incoming wind condition by the target incoming wind condition joint forecasting model to obtain the incoming wind condition forecast data at the second level includes:

[0032] Obtain real-time measurement data of incoming wind conditions;

[0033] The real-time measurement data of the incoming wind condition is input into the target incoming wind condition joint prediction model, and the incoming wind condition prediction data at the second level is output through the target incoming wind condition joint prediction model.

[0034] In a second aspect, an embodiment of the present disclosure provides a device for joint prediction of incoming wind conditions based on multi-task learning, the device comprising:

[0035] A data set construction module is used to construct an input data set and an output data set based on the measured data of incoming wind conditions;

[0036] A model building module is used to build a joint prediction model for incoming wind conditions, wherein the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task;

[0037] A model training module is used to train the incoming wind condition joint prediction model and optimize the hyper parameters to obtain a target incoming wind condition joint prediction model;

[0038] The model application module is used to predict the incoming wind conditions through the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level.

[0039] In a third aspect, an embodiment of the present disclosure provides an electronic device, including:

[0040] Memory;

[0041] Processor; and

[0042] Computer programs;

[0043] The computer program is stored in the memory and is configured to be executed by the processor to implement the method as described in the first aspect.

[0044] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method as described in the first aspect.

[0045] In a fifth aspect, an embodiment of the present disclosure further provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, the method described in the first aspect is implemented.

[0046] The embodiment of the present disclosure provides a method, device and equipment for joint prediction of incoming wind conditions based on multi-task learning. An input data set and an output data set are constructed based on measured data of incoming wind conditions, and a joint prediction model of incoming wind conditions is built. The joint prediction model of incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task. The joint prediction model of incoming wind conditions is trained and hyperparameters are optimized to obtain a target joint prediction model of incoming wind conditions. The incoming wind conditions are predicted by the target joint prediction model of incoming wind conditions to obtain second-level incoming wind condition prediction data, and the incoming wind condition prediction data includes incoming wind speed prediction data and incoming wind direction prediction data. Compared with the prior art, the embodiment of the present disclosure builds and trains a joint prediction model of incoming wind conditions. The target joint prediction model of incoming wind conditions can effectively capture the overall dynamics of future wind conditions, thereby achieving high-precision prediction of wind speed and wind direction at the second level. This not only helps the control of wind turbines and wind farms, but also provides key support for increasing power generation, extending equipment life, and enhancing power grid stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0049] Figure 1 A flow chart of a method for joint prediction of incoming wind conditions based on multi-task learning provided in an embodiment of the present disclosure;

[0050] Figure 2 A schematic diagram of the architecture of the joint prediction model for incoming wind conditions provided in an embodiment of the present disclosure;

[0051] Figure 3 A flow chart of a method for joint prediction of incoming wind conditions based on multi-task learning provided by another embodiment of the present disclosure;

[0052] Figure 4 A flow chart of a method for joint prediction of incoming wind conditions based on multi-task learning provided by another embodiment of the present disclosure;

[0053] Figure 5 A schematic diagram of the structure of a device for joint prediction of incoming wind conditions based on multi-task learning provided in an embodiment of the present disclosure;

[0054] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0055] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0057] In the field of wind power development, high temporal resolution future wind data is crucial. It can provide key support for wind turbine control and wind farm control, thereby improving power generation and grid stability. Among the related technologies, current wind condition prediction technology mainly focuses on the prediction of wind speed, while research on wind direction prediction is relatively scarce. Most of the studies only predict wind speed or wind direction separately, and fail to capture the overall dynamics of future wind conditions. In addition, the temporal resolution of most current prediction technologies is usually 15 minutes or longer, while the frequency and amplitude of wind turbine inflow changes are fast, and this temporal resolution cannot meet the needs of wind farm control. Therefore, the above-mentioned wind condition prediction technology has problems such as being unable to capture the overall dynamics of future wind conditions and too low temporal resolution.

[0058] To address this problem, an embodiment of the present disclosure provides a method for joint prediction of incoming wind conditions based on multi-task learning. The method is introduced below in conjunction with a specific embodiment.

[0059] Figure 1 A flow chart of the method for joint prediction of incoming wind conditions based on multi-task learning provided in the embodiment of the present disclosure. The executor of the method is an electronic device, which may be a portable mobile device such as a smart phone, a tablet computer, a laptop computer, or a fixed device such as a personal computer or a server, wherein the server may be a single server or a server cluster, and the server cluster may be a distributed cluster or a centralized cluster. The method can be applied to the scenario of predicting incoming wind conditions, and can accurately predict incoming wind conditions at the second level, meeting the actual engineering requirements of wake deflection control. It can be understood that the method for joint prediction of incoming wind conditions based on multi-task learning provided in the embodiment of the present disclosure can also be applied in other scenarios.

[0060] Below Figure 1 The following is an introduction to the combined forecasting method of incoming wind conditions based on multi-task learning. The specific steps of the method are as follows:

[0061] S101. Construct an input data set and an output data set based on the measured data of incoming wind conditions.

[0062] In this step, the electronic device obtains the measured data of the incoming wind condition, and further constructs an input data set and an output data set based on the measured data of the incoming wind condition. Optionally, the measured data of the incoming wind condition can be the measurement data of the anemometer and wind vane of the wind turbine, or can be the measurement data of a laser radar, without limitation. In some optional implementations, the input data set includes the wind speed data and wind direction data from time t-n+1 to time t, with a total of measured data for n time points. In some optional implementations, the output data set includes the wind speed data and wind direction data from time t+1 to time t+m, with a total of measured data for m time points. Optionally, as Figure 2 As shown, the input data set is a historical wind condition data set at the second level, i.e., wind speed data and wind direction data before a certain moment, and the output data set is a future wind condition data set at the second level, i.e., wind speed data and wind direction data after a certain moment.

[0063] S102, building a joint prediction model for incoming wind conditions, wherein the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task.

[0064] In this step, the electronic device will build a joint prediction model for incoming wind conditions, which includes an incoming wind speed prediction task and an incoming wind direction prediction task. Specifically, the joint prediction model for incoming wind conditions based on multi-task learning can be built based on the correlation between wind speed data and wind direction data.

[0065] S103, training the incoming wind condition joint prediction model and optimizing hyper parameters to obtain a target incoming wind condition joint prediction model.

[0066] In this step, after building the joint prediction model for incoming wind conditions, the electronic device can train the joint prediction model for incoming wind conditions and optimize hyperparameters to obtain a target joint prediction model for incoming wind conditions. The target joint prediction model for incoming wind conditions refers to a trained joint prediction model for incoming wind conditions with optimized hyperparameters.

[0067] S104, predicting the incoming wind condition through the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level, wherein the incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data.

[0068] In this step, after obtaining the target incoming wind condition joint prediction model, the electronic device predicts the incoming wind condition through the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level. The incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data, which can predict the overall dynamics of the wind speed and wind direction of the future wind conditions at the second level.

[0069] The disclosed embodiment constructs an input data set and an output data set based on the measured data of the incoming wind conditions, builds a joint prediction model for the incoming wind conditions, the joint prediction model for the incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task, trains the joint prediction model for the incoming wind conditions and optimizes the hyperparameters to obtain a target joint prediction model for the incoming wind conditions, predicts the incoming wind conditions through the target joint prediction model for the incoming wind conditions, and obtains second-level incoming wind condition prediction data, the incoming wind condition prediction data includes incoming wind speed prediction data and incoming wind direction prediction data. Compared with the prior art, the disclosed embodiment builds and trains a joint prediction model for the incoming wind conditions, and the target joint prediction model for the incoming wind conditions can effectively capture the overall dynamics of future wind conditions, thereby achieving high-precision prediction of wind speed and wind direction at the second level. This not only helps the control of wind turbines and wind farms, but also provides key support for increasing power generation, extending equipment life, and enhancing grid stability.

[0070] Figure 3 A flow chart of a method for joint prediction of incoming wind conditions based on multi-task learning provided by another embodiment of the present disclosure is shown in FIG. Figure 3 As shown, the method includes the following steps:

[0071] S201, preprocessing the measured data of the incoming wind condition to obtain the preprocessed measured data of the incoming wind condition.

[0072] Preprocessing may include but is not limited to reducing time resolution, filtering and eliminating invalid values, etc.

[0073] S202: construct an input data set and an output data set based on the preprocessed measured data of incoming wind conditions through a sliding time window algorithm.

[0074] Furthermore, based on the sliding time window algorithm, the preprocessed incoming wind condition measured data is used to construct an input data set and an output data set. Optionally, the input data set is a historical wind condition data set at the second level, that is, the wind speed data and wind direction data before a certain moment, and the output data set is a future wind condition data set at the second level, that is, the wind speed data and wind direction data after a certain moment. In some optional implementations, the input data set includes wind speed data and wind direction data from time t-n+1 to time t, a total of measured data for n time points. Optionally, the input data set can be expressed as X{[Ws t-n+1 ,…,Ws t ],[Wd t-n+1 ,…,Wd t ]}, where Ws t-n+1 Represents the wind speed data at time t-n+1, ​​Ws t Represents the wind speed data at time t, Wd t-n+1 Indicates the wind direction data at time t-n+1, ​​Wd t In some optional implementations, the output data set includes wind speed data and wind direction data from time t+1 to time t+m, a total of m measured data at time points. Optionally, the output data set can be expressed as Y{[Ws t+1 ,…,Ws t+m ],[Wd t+1 ,…,Wd t+m ]}, where Ws t+1 Represents the wind speed data at time t+1, Ws t+m Represents the wind speed data at time t+m, Wd t+1 Represents the wind direction data at time t+1, Wd t+m Indicates the wind direction data at time t+m.

[0075] S203: Design a model architecture of the incoming wind condition joint prediction model.

[0076] In some embodiments, the model architecture of the design of the joint prediction model for incoming wind conditions includes: using a long short-term memory network and a fully connected network as the core structure of the joint prediction model for incoming wind conditions; wherein the long short-term memory network is used to capture the correlation characteristics of wind speed and wind direction in a time series, and the fully connected network is used to predict the incoming wind conditions based on the correlation characteristics.

[0077] In this step, the main model architecture of the incoming wind condition joint prediction model is selected, such as Figure 2As shown, neural network structure A and neural network structure B are the main model architectures of the joint forecasting model for incoming wind conditions. In some optional implementations, a long short-term memory network (LSTM) and a fully connected network can be selected as the core structure of the model. LSTM is used to capture the long-term dependency, i.e., correlation features, in time series data, while the fully connected network is used to use the correlation features output by LSTM to make the final forecast of incoming wind conditions.

[0078] S204: For the incoming wind speed prediction task and the incoming wind direction prediction task in the incoming wind condition joint prediction model, a hard parameter sharing mechanism is adopted.

[0079] In this step, the electronic device uses a hard parameter sharing mechanism to implement the multi-task model, such as Figure 2 As shown, the input layer of the incoming wind condition joint prediction model is set as a task sharing layer, and multiple tasks, namely the incoming wind speed prediction task and the incoming wind direction prediction task, can share several hidden layers of the model. By utilizing the correlation between the incoming wind speed prediction task and the incoming wind direction prediction task, the model training can cooperate with each other to improve the prediction effect.

[0080] S205. Customizing the loss function of the incoming wind condition joint prediction model based on the weights corresponding to the incoming wind speed prediction task and the incoming wind direction prediction task.

[0081] In this step, the electronic device will assign different weights to different prediction tasks. Customizing a loss function that combines wind speed and wind direction prediction can effectively balance the performance of the model on each task. Different prediction tasks include wind speed and wind direction prediction tasks.

[0082] Optionally, you can set the loss function as follows:

[0083] L=α·L speed +β·L direction

[0084] Where L is the loss function of the combined forecasting model of incoming wind conditions; L speed is the loss function of the wind speed prediction task; L direction is the loss function of the wind direction prediction task, α is the weight coefficient of the wind speed prediction task, and β is the weight coefficient of the wind direction prediction task. The weight coefficient is used to balance the importance of the two tasks. Optionally, the loss function can be mean absolute error MAE (Mean Absolute Error), root mean square error RMSE (Root Mean Square Error), etc., without specific limitation.

[0085] S206: training the incoming wind condition joint prediction model and optimizing hyper parameters to obtain a target incoming wind condition joint prediction model.

[0086] Specifically, the implementation process and principle of S206 and S103 are the same, and will not be repeated here.

[0087] S207: Obtain real-time measurement data of incoming wind conditions.

[0088] In this step, the electronic device obtains real-time measurement data of incoming wind conditions, including real-time measurement data of wind speed and real-time measurement data of wind direction.

[0089] S208, inputting the real-time measurement data of the incoming wind condition into the target incoming wind condition joint prediction model, and outputting the incoming wind condition prediction data at the second level through the target incoming wind condition joint prediction model.

[0090] After acquiring the real-time measurement data of the incoming wind condition, the electronic device inputs the real-time measurement data of the incoming wind condition into the target incoming wind condition joint prediction model, and outputs the second-level incoming wind condition prediction data through the target incoming wind condition joint prediction model.

[0091] The disclosed embodiment preprocesses the measured data of the incoming wind condition to obtain the preprocessed measured data of the incoming wind condition, and constructs the input data set and the output data set based on the preprocessed measured data of the incoming wind condition through the sliding time window algorithm. Further, the model architecture of the incoming wind condition joint prediction model is designed, and the incoming wind speed prediction task and the incoming wind direction prediction task in the incoming wind condition joint prediction model are used. The loss function of the incoming wind condition joint prediction model is customized based on the weights corresponding to the incoming wind speed prediction task and the incoming wind direction prediction task. Then, the incoming wind condition joint prediction model is trained and hyperparameters are optimized to obtain the target incoming wind condition joint prediction model, and then the real-time measurement data of the incoming wind condition is obtained, and the real-time measurement data of the incoming wind condition is input into the target incoming wind condition joint prediction model, and the incoming wind condition prediction data of the second level is output through the target incoming wind condition joint prediction model. Through this method, the overall dynamics of future wind conditions can be effectively captured, thereby achieving high-precision prediction of wind speed and wind direction at the second level. This not only helps control wind turbines and wind farms, but also provides key support for increasing power generation, extending equipment life, and enhancing grid stability. By adopting a hard parameter sharing mechanism, the mutual collaboration between different tasks is effectively achieved, the prediction accuracy is improved, and the risk of overfitting is reduced.

[0092] Figure 4 A flow chart of a method for joint prediction of incoming wind conditions based on multi-task learning provided by another embodiment of the present disclosure is shown in FIG. Figure 4 As shown, the method includes the following steps:

[0093] S301 . Construct an input data set and an output data set based on the measured data of incoming wind conditions.

[0094] Specifically, the implementation process and principle of S301 and S101 are the same, and will not be repeated here.

[0095] S302: Building a joint prediction model for incoming wind conditions, wherein the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task.

[0096] Specifically, the implementation process and principle of S302 and S102 are the same, and will not be repeated here.

[0097] S303: Create a training set, a validation set, and a test set based on the input data set and the output data set.

[0098] In this step, the electronic device generates a training set, a validation set, and a test set based on the input data set and the output data set.

[0099] In some embodiments, S303 may include but is not limited to S3031 and S3032:

[0100] S3031. Normalize the input data set and the output data set to obtain a normalized input data set and a normalized output data set.

[0101] In this step, the electronic device may perform normalization processing on the input data set and the output data set to obtain a normalized input data set and a normalized output data set.

[0102] S3032. Divide the normalized input data set and the normalized output data set into a training set, a validation set, and a test set according to a preset ratio.

[0103] Furthermore, the electronic device divides the normalized input data set and the normalized output data set into a training set, a validation set, and a test set according to a preset ratio. The preset ratio may be 7:1:2, or other ratios, without specific limitation. It is necessary to ensure that the division is random. This is to prevent the model from overfitting under a specific wind condition sequence, thereby affecting the prediction effect.

[0104] S304: Initialize the model parameters of the incoming wind condition joint prediction model and set a hyperparameter search space.

[0105] In this step, the model parameters are initialized and the hyperparameter search space is set. The hyperparameter search space can be used to achieve hyperparameter optimization.

[0106] S305. For each hyperparameter combination in the hyperparameter search space, the joint prediction model of incoming wind conditions corresponding to each hyperparameter combination is trained using the training set until the joint prediction models of incoming wind conditions corresponding to each hyperparameter combination converge.

[0107] In this step, for each hyperparameter combination in the hyperparameter search space, the training set is used to train the joint prediction model of the incoming wind condition corresponding to each hyperparameter combination. Through continuous training, the loss function loss value of the joint prediction model of the incoming wind condition corresponding to each hyperparameter combination tends to be stable. At this time, the joint prediction model of the incoming wind condition corresponding to each hyperparameter combination converges. The hyperparameter combination, taking the learning rate η and the regularization strength λ as an example, refers to a specific learning rate and regularization strength such as (0.01, 0.1), which means that the learning rate is set to 0.01, and the regularization strength is set to 0.1. The relationship between hyperparameters is often mutually coupled and mutually influential. For example, the learning rate and the regularization strength jointly determine the training process and the final generalization performance of the model.

[0108] S306. Using the test set, the joint prediction model for incoming wind conditions corresponding to each of the hyperparameter combinations is tested to obtain prediction results of the joint prediction model for incoming wind conditions corresponding to each of the hyperparameter combinations.

[0109] In this step, the test set is used to test the joint prediction model of the incoming wind condition corresponding to each hyperparameter combination to obtain the prediction results of the joint prediction model of the incoming wind condition corresponding to each hyperparameter combination.

[0110] S307. Use the model evaluation index to evaluate the prediction results of the incoming wind condition joint prediction model corresponding to each hyperparameter combination, and obtain the evaluation value of the incoming wind condition joint prediction model corresponding to each hyperparameter combination.

[0111] After obtaining the prediction results of the joint prediction model for incoming wind conditions corresponding to each hyperparameter combination, the prediction results of the joint prediction model for incoming wind conditions corresponding to each hyperparameter combination are evaluated respectively using the model evaluation index. The evaluation value of the joint prediction model for incoming wind conditions corresponding to each hyperparameter combination can be obtained through the model evaluation index.

[0112] S308. Determine the best evaluation value from the evaluation values ​​of the incoming wind condition joint prediction models corresponding to each hyperparameter combination based on the model evaluation index, and determine the incoming wind condition joint prediction model corresponding to the hyperparameter combination matching the best evaluation value as the target incoming wind condition joint prediction model.

[0113] In this step, the electronic device will determine the best evaluation value from the evaluation values ​​of the joint prediction models of incoming wind conditions corresponding to each hyperparameter combination based on the model evaluation index, and determine the joint prediction model of incoming wind conditions corresponding to the hyperparameter combination that matches the best evaluation value in the joint prediction models of incoming wind conditions corresponding to each hyperparameter combination as the target joint prediction model of incoming wind conditions. The best evaluation value is the evaluation value that can reflect the best model performance among the various evaluation values. Specifically, whether to select the maximum evaluation value or the minimum evaluation value depends on the type of model evaluation index selected. Exemplarily, if the model evaluation index is accuracy, etc., the larger the evaluation value, the better the model, and the maximum evaluation value is selected as the best evaluation value; if the model evaluation index is RMSE, etc., the smaller the evaluation value, the better the model, and the minimum evaluation value is selected as the best evaluation value.

[0114] S309, predicting the incoming wind condition through the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level, wherein the incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data.

[0115] Specifically, the implementation process and principle of S309 and S104 are the same, and will not be repeated here.

[0116] The embodiment of the present disclosure constructs an input data set and an output data set based on the measured data of the incoming wind conditions, and builds a joint prediction model for the incoming wind conditions, wherein the joint prediction model for the incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task. Further, based on the input data set and the output data set, a training set, a validation set, and a test set are produced, the model parameters of the joint prediction model for the incoming wind conditions are initialized, and a hyperparameter search space is set. For each hyperparameter combination in the hyperparameter search space, the joint prediction model for the incoming wind conditions corresponding to each hyperparameter combination is trained using the training set until the joint prediction model for the incoming wind conditions corresponding to each hyperparameter combination converges. Next, the joint prediction model for the incoming wind conditions corresponding to each hyperparameter combination is tested using the test set to obtain the prediction results of the joint prediction model for the incoming wind conditions corresponding to each hyperparameter combination, and the prediction results of the joint prediction model for the incoming wind conditions corresponding to each hyperparameter combination are evaluated using the model evaluation index to obtain the evaluation value of the joint prediction model for the incoming wind conditions corresponding to each hyperparameter combination. Then, based on the model evaluation index, the best evaluation value is determined from the evaluation values ​​of the joint prediction models of the incoming wind conditions corresponding to each hyperparameter combination, and the joint prediction model of the incoming wind conditions corresponding to the hyperparameter combination matching the best evaluation value is determined as the target joint prediction model of the incoming wind conditions. Then, the incoming wind conditions are predicted by the target joint prediction model of the incoming wind conditions to obtain the second-level incoming wind condition prediction data, and the incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data. Through this method, the overall dynamics of future wind conditions can be effectively captured, thereby achieving high-precision prediction of wind speed and wind direction at the second level. This not only helps the control of wind turbines and wind farms, but also provides key support for increasing power generation, extending equipment life, and enhancing power grid stability.

[0117] Figure 5 The structural diagram of the device for joint prediction of wind conditions based on multi-task learning provided in the embodiment of the present disclosure. The device for joint prediction of wind conditions based on multi-task learning can be the electronic device as described in the above embodiment, or the device for joint prediction of wind conditions based on multi-task learning can be a component or assembly in the electronic device. The device for joint prediction of wind conditions based on multi-task learning provided in the embodiment of the present disclosure can execute the processing flow provided in the embodiment of the method for joint prediction of wind conditions based on multi-task learning, such as Figure 5As shown, the device 40 for joint prediction of incoming wind conditions based on multi-task learning includes: a data set construction module 41, a model construction module 42, a model training module 43, and a model application module 44; wherein, the data set construction module 41 is used to construct an input data set and an output data set based on the measured data of the incoming wind conditions; the model construction module 42 is used to build a joint prediction model for incoming wind conditions, and the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task; the model training module 43 is used to train the joint prediction model for incoming wind conditions and optimize hyperparameters to obtain a target joint prediction model for incoming wind conditions; the model application module 44 is used to predict the incoming wind conditions through the target joint prediction model for incoming wind conditions to obtain second-level incoming wind condition prediction data.

[0118] Optionally, when the data set construction module 41 constructs an input data set and an output data set based on the measured data of the incoming wind conditions, it is specifically used to: preprocess the measured data of the incoming wind conditions to obtain preprocessed measured data of the incoming wind conditions; construct an input data set and an output data set based on the preprocessed measured data of the incoming wind conditions through a sliding time window algorithm; wherein the input data set is a historical wind condition data set, and the output data set is a future wind condition data set.

[0119] Optionally, when the model building module 42 builds the joint prediction model for incoming wind conditions, it is specifically used to: design the model architecture of the joint prediction model for incoming wind conditions; adopt a hard parameter sharing mechanism for the incoming wind speed prediction task and the incoming wind direction prediction task in the joint prediction model for incoming wind conditions; and customize the loss function of the joint prediction model for incoming wind conditions based on the corresponding weights of the incoming wind speed prediction task and the incoming wind direction prediction task.

[0120] Optionally, when the model building module 42 designs the model architecture of the incoming wind condition joint prediction model, it is specifically used to: use the long short-term memory network and the fully connected network as the core structure of the incoming wind condition joint prediction model; wherein the long short-term memory network is used to capture the correlation characteristics of wind speed and wind direction in the time series, and the fully connected network is used to predict the incoming wind condition based on the correlation characteristics.

[0121] Optionally, the model training module 43 trains the joint prediction model for incoming wind conditions and optimizes hyperparameters to obtain a target joint prediction model for incoming wind conditions, and is specifically used to: create a training set, a validation set, and a test set based on an input data set and an output data set; initialize the model parameters of the joint prediction model for incoming wind conditions, and set a hyperparameter search space; for each hyperparameter combination in the hyperparameter search space, use the training set to train the joint prediction model for incoming wind conditions corresponding to each hyperparameter combination until the joint prediction models for incoming wind conditions corresponding to each hyperparameter combination converge; use the test set to respectively train the joint prediction model for incoming wind conditions corresponding to each hyperparameter combination. The joint prediction models of incoming wind conditions corresponding to the various hyperparameter combinations are tested to obtain prediction results of the joint prediction models of incoming wind conditions corresponding to the various hyperparameter combinations; the prediction results of the joint prediction models of incoming wind conditions corresponding to the various hyperparameter combinations are evaluated using model evaluation indicators to obtain evaluation values ​​of the joint prediction models of incoming wind conditions corresponding to the various hyperparameter combinations; based on the model evaluation indicators, the best evaluation value is determined from the evaluation values ​​of the joint prediction models of incoming wind conditions corresponding to the various hyperparameter combinations, and the joint prediction model of incoming wind conditions corresponding to the hyperparameter combination that matches the best evaluation value is determined as the target joint prediction model of incoming wind conditions.

[0122] Optionally, when the model training module 43 produces a training set, a validation set and a test set based on the input data set and the output data set, it is specifically used to: normalize the input data set and the output data set to obtain a normalized input data set and a normalized output data set; divide the normalized input data set and the normalized output data set into a training set, a validation set and a test set according to a preset ratio.

[0123] Optionally, when the model application module 44 predicts the incoming wind condition through the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level, it is specifically used to: obtain real-time measurement data of the incoming wind condition; input the real-time measurement data of the incoming wind condition into the target incoming wind condition joint prediction model, and output the incoming wind condition prediction data at the second level through the target incoming wind condition joint prediction model.

[0124] Figure 5 The device for joint prediction of incoming wind conditions based on multi-task learning in the illustrated embodiment can be used to execute the technical solution of the above-mentioned method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0125] Figure 6 Schematic diagram of the structure of an electronic device in the embodiment of the present disclosure. Figure 6 , which shows a structural schematic diagram of an electronic device 600 suitable for implementing the embodiments of the present disclosure. Figure 6The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0126] like Figure 6 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603 to implement the method for joint prediction of incoming wind conditions based on multi-task learning as described in the present disclosure. In RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0127] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0128] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart, thereby realizing the combined prediction method of incoming wind conditions based on multi-task learning as described above. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0129] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0130] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0131] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0132] The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device:

[0133] Construct input data set and output data set based on the measured data of incoming wind conditions;

[0134] Building a joint prediction model for incoming wind conditions, wherein the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task;

[0135] Training the incoming wind condition joint prediction model and optimizing hyperparameters to obtain a target incoming wind condition joint prediction model;

[0136] The incoming wind condition is predicted by the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level, wherein the incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data.

[0137] Optionally, when the above one or more programs are executed by the electronic device, the electronic device may also execute other steps described in the above embodiments.

[0138] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0139] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0140] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware, wherein the name of a unit does not, in some cases, limit the unit itself.

[0141] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0142] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0143] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0144] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0145] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A method for joint prediction of incoming wind conditions based on multi-task learning, characterized in that: The method comprises: Construct input data set and output data set based on the measured data of incoming wind conditions; Building a joint prediction model for incoming wind conditions, wherein the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task; Training the incoming wind condition joint prediction model and optimizing hyperparameters to obtain a target incoming wind condition joint prediction model; The incoming wind condition is predicted by the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level, wherein the incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data.

2. The method according to claim 1, characterized in that The step of constructing an input data set and an output data set based on the measured data of the incoming wind conditions includes: Preprocessing the measured data of incoming wind conditions to obtain preprocessed measured data of incoming wind conditions; Based on the pre-processed measured data of incoming wind conditions, the input data set and the output data set are constructed through the sliding time window algorithm; The input data set is a historical wind condition data set at the second level, and the output data set is a future wind condition data set at the second level.

3. The method according to claim 1, characterized in that The construction of the incoming wind condition joint prediction model includes: Designing a model architecture of the incoming wind condition joint prediction model; A hard parameter sharing mechanism is adopted for the incoming wind speed prediction task and the incoming wind direction prediction task in the incoming wind condition joint prediction model; Based on the weights corresponding to the incoming wind speed prediction task and the incoming wind direction prediction task, the loss function of the incoming wind condition joint prediction model is customized.

4. The method according to claim 3, characterized in that The model architecture of the design of the incoming wind condition joint prediction model includes: The long short-term memory network and the fully connected network are used as the core structure of the joint prediction model of incoming wind conditions; The long short-term memory network is used to capture the correlation characteristics of wind speed and wind direction in the time series, and the fully connected network is used to predict the incoming wind conditions based on the correlation characteristics.

5. The method according to claim 1, characterized in that The training of the incoming wind condition joint prediction model and the optimization of hyper parameters to obtain the target incoming wind condition joint prediction model includes: Create training sets, validation sets, and test sets based on the input data set and the output data set; Initializing the model parameters of the incoming wind condition joint prediction model and setting a hyperparameter search space; For each hyperparameter combination in the hyperparameter search space, the inflow wind condition joint prediction model corresponding to each hyperparameter combination is trained using the training set until the inflow wind condition joint prediction model corresponding to each hyperparameter combination converges; Using the test set to test the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations, respectively, to obtain the prediction results of the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations; The prediction results of the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations are evaluated by using the model evaluation index to obtain the evaluation value of the combined prediction model of incoming wind conditions corresponding to each of the hyperparameter combinations; Based on the model evaluation index, the best evaluation value is determined from the evaluation values ​​of the incoming wind condition joint prediction models corresponding to each hyperparameter combination, and the incoming wind condition joint prediction model corresponding to the hyperparameter combination matching the best evaluation value is determined as the target incoming wind condition joint prediction model.

6. The method according to claim 5, characterized in that The step of preparing a training set, a validation set, and a test set based on the input data set and the output data set includes: Normalizing the input data set and the output data set to obtain a normalized input data set and a normalized output data set; The normalized input data set and the normalized output data set are divided into a training set, a validation set, and a test set according to a preset ratio.

7. The method according to claim 1, characterized in that The method of predicting the incoming wind condition by using the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level includes: Obtain real-time measurement data of incoming wind conditions; The real-time measurement data of the incoming wind condition is input into the target incoming wind condition joint prediction model, and the incoming wind condition prediction data at the second level is output through the target incoming wind condition joint prediction model.

8. A device for joint prediction of incoming wind conditions based on multi-task learning, characterized in that: include: A data set construction module is used to construct an input data set and an output data set based on the measured data of incoming wind conditions; A model building module is used to build a joint prediction model for incoming wind conditions, wherein the joint prediction model for incoming wind conditions includes an incoming wind speed prediction task and an incoming wind direction prediction task; A model training module is used to train the incoming wind condition joint prediction model and optimize the hyper parameters to obtain a target incoming wind condition joint prediction model; The model application module is used to predict the incoming wind condition through the target incoming wind condition joint prediction model to obtain the incoming wind condition prediction data at the second level, wherein the incoming wind condition prediction data includes the incoming wind speed prediction data and the incoming wind direction prediction data.

9. An electronic device, characterized in that: include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and is configured to be executed by the processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images