Training method and device of wind speed prediction model, equipment and medium

Abstract

The invention relates to a training method and device of a wind speed prediction model, equipment and a medium. Acquiring a first historical actually measured wind speed of the wind power plant in a first historical time period before the historical prediction time period, a second historical actually measured wind speed of the wind power plant in a second historical time period before the historical prediction time period, and a historical prediction wind speed in the historical prediction time period, the difference between the first historical time period and the second historical time period is greater than a preset duration; training a preset network by using the first historical measured wind speed and the historical predicted wind speed to obtain an initial model; if the similarity between the second historical actually-measured wind speed of any target time window in the second historical time period and the second historical actually-measured wind speed of the previous time window of the target time window is smaller than a preset similarity threshold value, the second historical actually-measured wind speed of the target time window and the historical predicted wind speed are utilized; and finely adjusting the initial model to form a wind speed prediction model of the wind power plant. Therefore, wind speed prediction is accurately carried out based on the wind speed prediction model.

Description

Technical Field

[0001] The present disclosure relates to the technical field of wind speed prediction, and in particular to a training method, device, equipment and medium for a wind speed prediction model. Background Art

[0002] In the wind power industry, real-time and accurate wind speed forecast results are key inputs for wind farm wake optimization and operation control. They can also optimize the power output of wind turbines to reduce equipment wear and tear, thereby achieving long-term sustainability of wind farms.

[0003] In the related art, the predicted wind speed data is mainly determined by relying on historical wind speed data with low time resolution over a long period of time. However, the prediction timeliness of this wind speed data prediction method is low, resulting in low accuracy of wind speed prediction, which cannot meet the actual operation control requirements of wind farms. Summary of the invention

[0004] In order to solve the above technical problems, the present disclosure provides a training method, device, equipment and medium for a wind speed prediction model.

[0005] In a first aspect, the present disclosure provides a method for training a wind speed prediction model, comprising:

[0006] Acquire a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before a historical predicted time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical predicted time period, the historical predicted wind speed is the predicted wind speed of the wind farm in the historical predicted time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset duration;

[0007] Using the first historical measured wind speed and the historical predicted wind speed, a preset network is trained to obtain an initial model;

[0008] Determine the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window;

[0009] If the similarity is less than a preset similarity threshold, the initial model is fine-tuned using the second historical measured wind speed in the target time window and the historical predicted wind speed to obtain a wind speed prediction model for the wind farm.

[0010] In a second aspect, the present disclosure provides a training device for a wind speed prediction model, comprising:

[0011] an acquisition module, configured to acquire a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before a historical predicted time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical predicted time period, the historical predicted wind speed is the predicted wind speed of the wind farm in the historical predicted time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset duration;

[0012] A training module, used to train a preset network using the first historical measured wind speed and the historical predicted wind speed to obtain an initial model;

[0013] A similarity determination module, used to determine the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window;

[0014] The fine-tuning module is used to fine-tune the initial model using the second historical measured wind speed and the historical predicted wind speed in the target time window if the similarity is less than a preset similarity threshold, so as to obtain a wind speed prediction model for the wind farm.

[0015] In a third aspect, an embodiment of the present disclosure further provides an electronic device, the device comprising:

[0016] one or more processors;

[0017] a storage device for storing one or more programs,

[0018] When one or more programs are executed by one or more processors, the one or more processors implement the method provided in the first aspect.

[0019] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method provided in the first aspect is implemented.

[0020] Compared with the prior art, the technical solution provided by the embodiments of the present disclosure has the following advantages:

[0021] A training method, device, equipment and medium for a wind speed prediction model of an embodiment of the present disclosure include: obtaining a first historical measured wind speed, a second historical measured wind speed and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before the historical prediction time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical prediction time period, the historical predicted wind speed is the predicted wind speed of the wind farm in the historical prediction time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset time length; using the first historical measured wind speed and the historical predicted wind speed, training a preset network to obtain an initial model; determining the similarity between the second historical measured wind speed of any target time window in the second historical time period and the second historical measured wind speed of a previous time window of the target time window; if the similarity is less than a preset similarity threshold, using the second historical measured wind speed and the historical predicted wind speed of the target time window to fine-tune the initial model to obtain a wind speed prediction model for the wind farm. Therefore, we first use the historical long-term wind speed dataset to train the model, and then use the historical short-term wind speed dataset to fine-tune the model, so that the model can be updated when the similarity of the historical short-term wind speed dataset changes greatly, thereby enhancing the generalization ability and prediction accuracy of the model, and using the trained wind speed prediction model to achieve high-time resolution wind speed prediction, which improves the real-time and accuracy of the wind speed prediction results, and provides strong support for the operation and scheduling of wind farms. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] 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.

[0023] 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.

[0024] Figure 1 A schematic diagram of a flow chart of a training method for a wind speed prediction model provided in an embodiment of the present disclosure;

[0025] Figure 2 A schematic diagram of the structure of a wind speed prediction model training device provided in an embodiment of the present disclosure;

[0026] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0027] 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.

[0028] 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.

[0029] Most wind speed prediction methods in related technologies rely on data at a time scale of 15 minutes or even longer, and the prediction timeliness cannot meet the actual operation and control requirements of wind farms.

[0030] In order to solve the above problems, the following Figure 1 The training method of the wind speed prediction model provided in the embodiment of the present disclosure is described. In the embodiment of the present disclosure, the training method of the wind speed prediction model can be executed by an electronic device or a server. Among them, the electronic device may include a device with a communication function such as a tablet computer, a desktop computer, a laptop computer, and may also include a device simulated by a virtual machine or a simulator. The server may include a single server or a server cluster. This embodiment is specifically explained with an electronic device as the execution subject.

[0031] Figure 1 A flow chart of a training method for a wind speed prediction model provided by an embodiment of the present disclosure is shown.

[0032] like Figure 1 As shown, the training method of the wind speed prediction model may include the following steps.

[0033] S110. Obtain a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of the wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before the historical prediction time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical prediction time period, the historical predicted wind speed is the predicted wind speed of the wind farm within the historical prediction time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset duration.

[0034] In this embodiment, when performing model training, the electronic device obtains a training data set in the wind farm, specifically obtains the first historical measured wind speed in the first historical time period before the historical prediction time period, the second historical predicted wind speed in the second historical time period before the historical prediction time period, and the historical predicted wind speed within the historical prediction time period, so as to train the wind speed prediction model using the acquired training data set.

[0035] Since the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than the preset duration, the first historical time period refers to the long historical time period after the historical moment and before the current moment, and the second historical time period refers to the short historical time period after the historical moment and before the current moment. The first historical measured wind speed and the historical predicted wind speed can be called a long historical wind speed data set, and the second historical measured wind speed and the historical predicted wind speed can be called a short historical wind speed data set.

[0036] For example, if the historical prediction time period is 10:00-10:09 yesterday, the historical prediction wind speed is the predicted wind speed from 10:00 to 10:09 yesterday; if the first historical time period is the 24 hours before 10:00-10:09 yesterday, the first historical measured wind speed is the historical measured wind speed for the 24 hours before 10:00-10:09 yesterday; if the second historical time period is the 2 hours before 10:00-10:09 yesterday, the second historical measured wind speed is the historical measured wind speed for the 2 hours before 10:00-10:09 yesterday.

[0037] Optionally, the preset duration is a duration set in advance based on experience to distinguish between a historical long time period and a historical short time period.

[0038] The first historical measured wind speed and the second historical measured wind speed both include at least one of the historical measured wind speed of the wind turbine, the historical measured wind speed of the laser radar, and the historical measured wind speed of the wind tower. Optionally, the resolution of the first historical measured wind speed and the second historical measured wind speed can be set to 1 minute.

[0039] In order to improve the accuracy of model training, the first historical measured wind speed, the second historical measured wind speed and the historical predicted wind speed are preprocessed by data cleaning and normalization to improve the quality and consistency of the training data set.

[0040] Optionally, data cleaning processing includes abnormal data removal and data interpolation.

[0041] S120: Using the first historical measured wind speed and the historical predicted wind speed, a preset network is trained to obtain an initial model.

[0042] In this embodiment, the electronic device first uses the first historical measured wind speed and the historical predicted wind speed as a historical long-term wind speed data set. When using the historical long-term wind speed data set to train the preset network, it is necessary to ensure that the first historical measured wind speed and the historical predicted wind speed in the historical long-term wind speed data set are used to train the network according to the timestamp correspondence, until the loss value determined based on the first historical measured wind speed and the historical estimated wind speed is less than or equal to a preset threshold value, it is determined that the preset network under the current training times meets the preset cutoff condition, and an initial model is obtained.

[0043] Among them, the specific implementation method of S120 includes but is not limited to the following methods: based on the input layer in the preset network, processing the first historical measured wind speed to obtain the shallow features of the first historical measured wind speed; based on the feature processing layer in the preset network, processing the shallow features of the first historical measured wind speed to obtain the deep features of the first historical measured wind speed; based on the output layer in the preset network, performing full connection processing and feature activation processing on the deep features of the first historical measured wind speed to obtain the first historical estimated wind speed of the first historical measured wind speed; based on the historical predicted wind speed and the first historical estimated wind speed, iteratively training the preset network until the preset network under the current training times meets the preset cutoff condition to obtain an initial model.

[0044] Among them, based on the feature processing layer in the preset network, the shallow features of the first historical measured wind speed are processed to obtain the deep features of the first historical measured wind speed. The specific implementation method includes but is not limited to the following methods: based on the convolution layer in the feature processing layer, the shallow features are convolved to obtain convolution features; based on the activation function layer in the feature processing layer, the convolution features are nonlinearly activated to obtain nonlinear features; based on the pooling layer in the feature processing layer, the nonlinear features are pooled to obtain pooling features; based on the regularization layer in the feature processing layer, the pooling features are regularized to obtain deep features.

[0045] Optionally, the preset network includes but is not limited to a long short-term memory network (LSTM), a convolutional neural network (CNN), and a Seq2Seq model built based on a gated recurrent unit (GRU).

[0046] Therefore, the first historical measured wind speed in the training data set is layered by using different network layers of the preset network to determine the first historical estimated wind speed under each training number, and the preset network parameters are iteratively adjusted through the first historical estimated wind speed and the historical predicted wind speed until the preset cutoff conditions are met to obtain the initial model.

[0047] S130: Determine the similarity between the second historical measured wind speed of any target time window in the second historical time period and the second historical measured wind speed of the previous time window of the target time window.

[0048] Considering that the accuracy of the initial model trained only with a wind speed data set of a long historical time period with low time resolution is low, the present embodiment may also consider a wind speed data set of a short historical time period with high time resolution to fine-tune the initial model. Specifically, for two adjacent time windows in the second historical time period, the electronic device needs to determine the similarity between the second historical measured wind speed of any target time window in the second historical time period and the second historical measured wind speed of the previous time window of the target time window, so as to determine whether to fine-tune the initial model based on the similarity.

[0049] In some embodiments, the similarity between the second historical measured wind speed of any target time window in the second historical time period and the second historical measured wind speed of the previous time window of the target time window is determined by the Euclidean distance. Specifically, the larger the Euclidean distance between the second historical measured wind speed of any target time window in the second historical time period and the second historical measured wind speed of the previous time window of the target time window, the smaller the similarity between the two. Conversely, the smaller the Euclidean distance between the second historical measured wind speed of any target time window in the second historical time period and the second historical measured wind speed of the previous time window of the target time window, the greater the similarity between the two.

[0050] In other embodiments, other similarity calculation methods may be used to determine the similarity between the first historical measured wind speed and the second historical measured wind speed.

[0051] S140: If the similarity is less than a preset similarity threshold, fine-tune the initial model using the second historical measured wind speed and the historical predicted wind speed in the target time window to obtain a wind speed prediction model for the wind farm.

[0052] In this embodiment, the electronic device obtains a preset similarity threshold and compares the similarity with the preset similarity threshold. If it is determined that the similarity is less than the preset similarity threshold, it means that the wind speed data sets within the historical short time period with high time resolution are quite different, and the initial model cannot be used directly to accurately predict the wind speed. Therefore, it is necessary to use the wind speed data sets within the historical short time period with high time resolution to fine-tune the initial model before making wind speed predictions.

[0053] Among them, the specific implementation method of S140 includes but is not limited to the following methods: using the initial model to process the second historical measured wind speed in the target time window to obtain the second historical estimated wind speed in the historical prediction time period; based on the second historical estimated wind speed and the historical predicted wind speed, calculating the loss value of the initial model; based on the loss value, iteratively adjusting the model parameters of the output layer of the initial model, and keeping the model parameters of the input layer and the model parameters of the hidden layer of the initial model unchanged, until the loss value of the initial model under the current training times is less than the preset loss value threshold, thereby obtaining the wind speed prediction model of the wind farm.

[0054] Specifically, when the wind speed dataset within a historical short period of time with high temporal resolution changes significantly, the transfer learning method is adopted to use the second historical measured wind speed and historical predicted wind speed of the target time window to freeze the layers of the model except the output layer, and only the model parameters of the output layer of the initial model are iteratively adjusted to achieve model fine-tuning using the wind speed dataset within a historical short period of time with high temporal resolution, so that the obtained wind speed prediction model of the wind farm can better adapt to the data within the historical short period of time, thereby achieving the purpose of improving the prediction accuracy.

[0055] The preset loss value threshold is a loss value predetermined based on experience at which the adjustment of the model parameters of the output layer is stopped.

[0056] In this way, when the wind speed dataset within a short historical period of high temporal resolution changes significantly, the wind speed dataset within a short historical period of high temporal resolution is used to adjust the model parameters of the output layer of the initial model, and the model parameters of other network layers are not adjusted. This can not only achieve fine-tuning of the basic wind speed prediction model by using transfer learning, so that the model can adapt to the complex and changeable environment of the wind farm in real time, but also reduce the adjustment amount of the model parameters.

[0057] In some embodiments, after executing S130, the method further includes: if the similarity is greater than or equal to a preset similarity threshold, using the initial model as a wind speed prediction model for the wind farm.

[0058] It is understandable that if the similarity is determined to be greater than the preset similarity threshold, it means that the wind speed dataset in the historical short period of time with high temporal resolution has not changed significantly. In this case, the initial model can be used directly to accurately predict the wind speed without the need to use transfer learning technology to fine-tune the model, thereby improving the training efficiency of the wind speed prediction model.

[0059] A training method for a wind speed prediction model of an embodiment of the present disclosure includes: obtaining a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before the historical prediction time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical prediction time period, the historical predicted wind speed is the predicted wind speed of the wind farm within the historical prediction time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset time length; using the first historical measured wind speed and the historical predicted wind speed, training a preset network to obtain an initial model; determining the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window; if the similarity is less than a preset similarity threshold, using the second historical measured wind speed and the historical predicted wind speed of the target time window to fine-tune the initial model to obtain a wind speed prediction model for the wind farm. Therefore, we first use the historical long-term wind speed dataset to train the model, and then use the historical short-term wind speed dataset to fine-tune the model, so that the model can be updated when the similarity of the historical short-term wind speed dataset changes greatly, thereby enhancing the generalization ability and prediction accuracy of the model, and using the trained wind speed prediction model to achieve high-time resolution wind speed prediction, which improves the real-time and accuracy of the wind speed prediction results, and provides strong support for the operation and scheduling of wind farms.

[0060] In some embodiments of the present disclosure, after the electronic device has trained the wind speed prediction model of the wind farm, the method also includes the following steps: obtaining the third historical measured wind speed of the wind farm, wherein the third historical measured wind speed is the historical measured wind speed of the wind farm in the third historical time period before the current prediction time period; using the wind speed prediction model of the wind farm, processing the third historical measured wind speed to obtain the predicted wind speed data of the wind farm in the current prediction time period.

[0061] Specifically, after the wind speed prediction model of the wind farm is trained, the model can be put into use. First, the third historical measured wind speed of the wind farm in the third historical time period before the current prediction time period collected at the wind farm is input into the wind speed prediction model of the wind farm. Then, the input layer, feature processing layer and output layer of the wind speed prediction model of the wind farm are sequentially passed through to predict the predicted wind speed data of the wind farm at future times.

[0062] For example, the current prediction time period is 10:00-10:09 today, and the third historical time period before the current prediction time period is 2 hours before 10:00-10:09 today. Then the third historical measured wind speed is the historical measured wind speed 2 hours before 10:00-10:09 today, and the predicted wind speed data for the current prediction time period is the predicted wind speed from 10:00-10:09 today.

[0063] It can be understood that, due to the use of transfer learning methods to train and fine-tune the wind speed prediction model of the wind farm, the model can be updated to adapt to the meteorological changes of the wind farm by considering the real-time meteorological conditions of the wind farm, thereby improving the accuracy of real-time wind speed prediction with high temporal resolution.

[0064] The disclosed embodiment also provides a training device for a wind speed prediction model for implementing the above-mentioned training method for a wind speed prediction model, and the device is configured in an electronic device or a server. The electronic device may include a tablet computer, a desktop computer, a laptop computer, and other devices with communication functions, and may also include a virtual machine or a device simulated by a simulator. The server may include a single server or a server cluster. This embodiment is specifically explained using an electronic device as a training device for a wind speed prediction model. Figure 2 Provide explanation.

[0065] Figure 2 A schematic structural diagram of a wind speed prediction model training device provided by an embodiment of the present disclosure is shown.

[0066] like Figure 2 As shown, the training device 200 of the wind speed prediction model may include:

[0067] An acquisition module 210 is used to acquire a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is a historical measured wind speed of the wind farm in a first historical time period before a historical predicted time period, the second historical measured wind speed is a historical measured wind speed of the wind farm in a second historical time period before the historical predicted time period, the historical predicted wind speed is a predicted wind speed of the wind farm in the historical predicted time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset duration;

[0068] A training module 220, configured to train a preset network using the first historical measured wind speed and the historical predicted wind speed to obtain an initial model;

[0069] A similarity determination module 230, configured to determine the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window;

[0070] The fine-tuning module 240 is used to fine-tune the initial model using the second historical measured wind speed and the historical predicted wind speed in the target time window to obtain a wind speed prediction model for the wind farm if the similarity is less than a preset similarity threshold.

[0071] A training device for a wind speed prediction model of an embodiment of the present disclosure includes: obtaining a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before the historical prediction time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical prediction time period, the historical predicted wind speed is the predicted wind speed of the wind farm within the historical prediction time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset time length; using the first historical measured wind speed and the historical predicted wind speed, training a preset network to obtain an initial model; determining the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window; if the similarity is less than a preset similarity threshold, using the second historical measured wind speed and the historical predicted wind speed of the target time window to fine-tune the initial model to obtain a wind speed prediction model for the wind farm. Therefore, we first use the historical long-term wind speed dataset to train the model, and then use the historical short-term wind speed dataset to fine-tune the model, so that the model can be updated when the similarity of the historical short-term wind speed dataset changes greatly, thereby enhancing the generalization ability and prediction accuracy of the model, and using the trained wind speed prediction model to achieve high-time resolution wind speed prediction, which improves the real-time and accuracy of the wind speed prediction results, and provides strong support for the operation and scheduling of wind farms.

[0072] In some embodiments of the present disclosure, the training module 220 includes:

[0073] A first processing unit is used to process the first historical measured wind speed based on the input layer in the preset network to obtain the shallow characteristics of the first historical measured wind speed;

[0074] A second processing unit is used to process the shallow features of the first historical measured wind speed based on the feature processing layer in the preset network to obtain the deep features of the first historical measured wind speed;

[0075] A third processing unit is used to perform full connection processing and feature activation processing on the deep features of the first historical measured wind speed based on the output layer in the preset network to obtain a first historical estimated wind speed of the first historical measured wind speed;

[0076] An iterative training unit is used to iteratively train the preset network based on the historical measured wind speed and the first historical estimated wind speed until the preset network under the current training times meets the preset cutoff condition to obtain the initial model.

[0077] In some embodiments of the present disclosure, the second processing unit is specifically configured to:

[0078] Based on the convolution layer in the feature processing layer, convolution processing is performed on the shallow features to obtain convolution features;

[0079] Based on the activation function layer in the feature processing layer, performing nonlinear activation processing on the convolution feature to obtain a nonlinear feature;

[0080] Based on the pooling layer in the feature processing layer, the nonlinear feature is pooled to obtain a pooling feature;

[0081] Based on the regularization layer in the feature processing layer, the pooled features are regularized to obtain the deep features.

[0082] In some embodiments of the present disclosure, the fine-tuning module 240 is specifically used to:

[0083] Using the initial model, processing the second historical measured wind speed of the target time window to obtain a second historical estimated wind speed of the historical prediction time period;

[0084] Calculating a loss value of the initial model based on the second historical estimated wind speed and the historical predicted wind speed;

[0085] Based on the loss value, the model parameters of the output layer of the initial model are iteratively adjusted, and the model parameters of the input layer and the model parameters of the hidden layer of the initial model are kept unchanged, until the loss value of the initial model under the current training times is less than the preset loss value threshold, so as to obtain the wind speed prediction model of the wind farm.

[0086] In some embodiments of the present disclosure, the device further includes:

[0087] A determination module is configured to use the initial model as a wind speed prediction model for the wind farm if the similarity is greater than or equal to the preset similarity threshold.

[0088] In some embodiments of the present disclosure, the first historical measured wind speed and the second historical measured wind speed both include at least one of a historical measured wind speed of a wind turbine generator set, a historical measured wind speed of a lidar, and a historical measured wind speed of a wind tower.

[0089] In some embodiments of the present disclosure, the device further includes:

[0090] A third historical measured wind speed acquisition module, used to acquire a third historical measured wind speed of the wind farm, wherein the third historical measured wind speed is a historical measured wind speed of the wind farm in a third historical time period before a current prediction time period;

[0091] The wind speed prediction module is used to process the third historical measured wind speed using the wind speed prediction model of the wind farm to obtain predicted wind speed data of the wind farm in the current prediction time period.

[0092] It should be noted that Figure 2 The wind speed prediction model training device 200 shown can be executed Figure 1 The various steps in the method embodiment shown in the figure are implemented Figure 1 The various processes and effects in the method embodiment shown are not described in detail here.

[0093] Figure 3 A schematic structural diagram of an electronic device provided by an embodiment of the present disclosure is shown.

[0094] like Figure 3 As shown, the electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0095] Specifically, the processor 301 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0096] The memory 302 may include a large capacity memory for information or instructions. By way of example and not limitation, the memory 302 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a tape, or a universal serial bus (USB) drive or a combination of two or more of these. Where appropriate, the memory 302 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 302 may be inside or outside the integrated gateway device. In a particular embodiment, the memory 302 is a non-volatile solid-state memory. In a particular embodiment, the memory 302 includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (Electrically Erasable Programmable ROM, EEPROM), an electrically rewritable ROM (EAROM) or a flash memory, or a combination of two or more of these.

[0097] The processor 301 reads and executes the computer program instructions stored in the memory 302 to perform the steps of the training method of the wind speed prediction model provided in the embodiment of the present disclosure.

[0098] In one example, the electronic device may further include a transceiver 303 and a bus 304. Figure 3 As shown, the processor 301, the memory 302 and the transceiver 303 are connected via a bus 304 and communicate with each other.

[0099] The bus 304 includes hardware, software, or both. For example, but not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a Memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, the bus 304 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0100] The following is an embodiment of a computer-readable storage medium provided in an embodiment of the present disclosure. The computer-readable storage medium and the training method of the wind speed prediction model in the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the computer-readable storage medium, reference can be made to the embodiment of the training method of the wind speed prediction model in the above-mentioned embodiments.

[0101] This embodiment provides a storage medium containing computer executable instructions. When the computer executable instructions are executed by a computer processor, they are used to perform a training method for a wind speed prediction model. The method includes:

[0102] Acquire a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before a historical predicted time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical predicted time period, the historical predicted wind speed is the predicted wind speed of the wind farm in the historical predicted time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset duration;

[0103] Using the first historical measured wind speed and the historical predicted wind speed, a preset network is trained to obtain an initial model;

[0104] Determine the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window;

[0105] If the similarity is less than a preset similarity threshold, the initial model is fine-tuned using the second historical measured wind speed in the target time window and the historical predicted wind speed to obtain a wind speed prediction model for the wind farm.

[0106] Of course, the storage medium containing computer executable instructions provided in the embodiment of the present disclosure is not limited to the above method operations, and the computer executable instructions can also execute related operations in the training method of the wind speed prediction model provided in any embodiment of the present disclosure.

[0107] Through the above description of the implementation methods, technicians in the relevant field can clearly understand that the present disclosure can be implemented with the help of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present disclosure can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer cloud platform (which can be a personal computer, a server, or a network cloud platform, etc.) to execute the training method of the wind speed prediction model provided in each embodiment of the present disclosure.

[0108] Note that the above are only preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art will understand that the present disclosure is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present disclosure. Therefore, although the present disclosure is described in more detail through the above embodiments, the present disclosure is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present disclosure, and the scope of the present disclosure is determined by the scope of the appended claims.

Claims

1. A training method for a wind speed prediction model, characterized in that: include: Acquire a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before a historical predicted time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical predicted time period, the historical predicted wind speed is the predicted wind speed of the wind farm in the historical predicted time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset duration; Using the first historical measured wind speed and the historical predicted wind speed, a preset network is trained to obtain an initial model; Determine the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window; If the similarity is less than a preset similarity threshold, the initial model is fine-tuned using the second historical measured wind speed in the target time window and the historical predicted wind speed to obtain a wind speed prediction model for the wind farm.

2. The method according to claim 1, characterized in that The method of using the first historical measured wind speed and the historical predicted wind speed to train a preset network to obtain an initial model includes: Based on the input layer in the preset network, the first historical measured wind speed is processed to obtain the shallow characteristics of the first historical measured wind speed; Based on the feature processing layer in the preset network, shallow features of the first historical measured wind speed are processed to obtain deep features of the first historical measured wind speed; Based on the output layer in the preset network, performing full connection processing and feature activation processing on the deep features of the first historical measured wind speed to obtain a first historical estimated wind speed of the first historical measured wind speed; Based on the historical predicted wind speed and the first historical estimated wind speed, the preset network is iteratively trained until the preset network under the current training times meets a preset cutoff condition, thereby obtaining the initial model.

3. The method according to claim 2, characterized in that The processing of the shallow features of the first historical measured wind speed based on the feature processing layer in the preset network to obtain the deep features of the first historical measured wind speed includes: Based on the convolution layer in the feature processing layer, convolution processing is performed on the shallow features to obtain convolution features; Based on the activation function layer in the feature processing layer, performing nonlinear activation processing on the convolution feature to obtain a nonlinear feature; Based on the pooling layer in the feature processing layer, the nonlinear feature is pooled to obtain a pooling feature; Based on the regularization layer in the feature processing layer, the pooled features are regularized to obtain the deep features.

4. The method according to claim 1, characterized in that: The method of fine-tuning the initial model by using the second historical measured wind speed and the historical predicted wind speed in the target time window to obtain a wind speed prediction model for the wind farm includes: Using the initial model, processing the second historical measured wind speed of the target time window to obtain a second historical estimated wind speed of the historical prediction time period; Calculating a loss value of the initial model based on the second historical estimated wind speed and the historical predicted wind speed; Based on the loss value, the model parameters of the output layer of the initial model are iteratively adjusted, and the model parameters of the input layer and the model parameters of the hidden layer of the initial model are kept unchanged, until the loss value of the initial model under the current training times is less than the preset loss value threshold, so as to obtain the wind speed prediction model of the wind farm.

5. The method according to claim 1, characterized in that Also includes: If the similarity is greater than or equal to the preset similarity threshold, the initial model is used as the wind speed prediction model of the wind farm.

6. The method according to any one of claims 1 to 5, characterized in that: The first historical measured wind speed and the second historical measured wind speed both include at least one of the historical measured wind speed of the wind turbine generator set, the historical measured wind speed of the laser radar, and the historical measured wind speed of the wind tower.

7. The method according to any one of claims 1 to 5, characterized in that: Also includes: Acquire a third historical measured wind speed of the wind farm, wherein the third historical measured wind speed is a historical measured wind speed of the wind farm in a third historical time period before a current prediction time period; The third historical measured wind speed is processed using the wind speed prediction model of the wind farm to obtain predicted wind speed data of the wind farm in the current prediction time period.

8. A training device for a wind speed prediction model, characterized in that: include: an acquisition module, configured to acquire a first historical measured wind speed, a second historical measured wind speed, and a historical predicted wind speed of a wind farm, wherein the first historical measured wind speed is the historical measured wind speed of the wind farm in a first historical time period before a historical predicted time period, the second historical measured wind speed is the historical measured wind speed of the wind farm in a second historical time period before the historical predicted time period, the historical predicted wind speed is the predicted wind speed of the wind farm in the historical predicted time period, the first historical time period is greater than the second historical time period, and the time difference between the first historical time period and the second historical time period is greater than a preset duration; A training module, used to train a preset network using the first historical measured wind speed and the historical predicted wind speed to obtain an initial model; A similarity determination module, used to determine the similarity between the second historical measured wind speed of any target time window within the second historical time period and the second historical measured wind speed of a previous time window of the target time window; The fine-tuning module is used to fine-tune the initial model using the second historical measured wind speed and the historical predicted wind speed in the target time window if the similarity is less than a preset similarity threshold, so as to obtain a wind speed prediction model for the wind farm.

9. An electronic device, characterized in that: include: processor; A memory for storing executable instructions; The processor is used to read the executable instructions from the memory and execute the executable instructions to implement the method described in any one of claims 1 to 7.

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

Images