Wind power prediction method, device, electronic equipment and storage medium

Abstract

The present application discloses a wind power prediction method, device, electronic device and storage medium, which relates to the field of new energy wind power generation technology. The method includes obtaining numerical weather forecast data of a wind farm; processing the numerical weather forecast data based on a wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, wherein the offset scenario includes a positive offset or a negative offset; if the offset scenario of the numerical weather forecast data is a positive offset, processing the numerical weather forecast data based on a first wind power prediction model to obtain a first wind power prediction result; if the offset scenario of the numerical weather forecast data is a negative offset, processing the numerical weather forecast data based on a second wind power prediction model to obtain a second wind power prediction result. In this way, the purpose of correcting the wind power can be achieved, so that the predicted wind power is accurate; and the modeling cost and cost are not high.

Description

Technical Field

[0001] The present application relates to the technical field of new energy wind power generation, and in particular to a wind power prediction method, device, electronic equipment and storage medium. Background Art

[0002] In the renewable energy wind power generation technology, the short-term wind power forecast can predict the wind power for the next 1 to 3 days at a time, which provides a basis for power dispatching plans in short time scales. Common short-term wind power forecasting methods are usually based on numerical weather prediction (NWP) data for power mapping modeling. Improving the accuracy of wind power forecasting is of great significance for the safe and stable operation of the power system, formulating reasonable dispatching plans, reducing wind power curtailment and power rationing, and maintaining the economic operation of the power system. However, during the peak and valley periods of wind power, the wind power forecast value usually has significant deviations, which are mainly manifested as higher valley periods or lower peak periods. This is strongly related to the NWP wind speed and the characteristics of the prediction model itself.

[0003] At present, in order to improve the accuracy of wind power prediction, most studies consider data-driven optimization and improvement of prediction models or adopt refined modeling schemes of wind turbine clustering, but these schemes have high modeling cost and cumbersome model updating process. Therefore, how to control modeling cost and cost while improving the accuracy of wind power prediction is an urgent problem to be solved. Summary of the invention

[0004] In view of this, embodiments of the present application provide a wind power prediction method, device, electronic device, and storage medium.

[0005] According to the first aspect of the present application, an embodiment of the present application provides a wind power prediction method, including:

[0006] Obtain numerical weather forecast data for wind farms;

[0007] The numerical weather forecast data is processed based on the wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, wherein the offset scenario includes a positive offset or a negative offset;

[0008] If the offset scenario of the numerical weather forecast data is a positive offset, the numerical weather forecast data is processed based on the first wind power prediction model to obtain a first wind power prediction result;

[0009] If the offset scenario of the numerical weather forecast data is a negative offset, the numerical weather forecast data is processed based on the second wind power prediction model to obtain a second wind power prediction result;

[0010] Among them, the first loss function during the training of the first wind power prediction model can enable the first wind power prediction model to update the first wind power prediction model parameters in the direction in which the output result deviates negatively from the expected value; the second loss function during the training of the second wind power prediction model can enable the second wind power prediction model to update the second wind power prediction model parameters in the direction in which the output result deviates positively from the expected value.

[0011] Optionally, the wind speed deviation identification model training process includes:

[0012] Acquire training samples and corresponding labels; the training samples include historical numerical weather forecast data and weather generation data of the wind farm, the historical numerical weather forecast data include historical numerical weather forecast data with positive offsets during wind speed peak and valley periods, and historical numerical weather forecast data with negative offsets during wind speed peak and valley periods; the weather generation data is generated based on historical numerical weather forecast data with positive offsets or historical numerical weather forecast data with negative offsets in which the proportion of data volume during wind speed peak and valley periods is less than a first threshold; the labels include positive offset labels or negative offset labels;

[0013] The training samples and corresponding labels are used to train the neural network with a directed acyclic graph structure to obtain a wind speed deviation recognition model.

[0014] Optionally, the step of generating weather generation data comprises:

[0015] Get historical numerical weather forecast data and corresponding labels;

[0016] Based on the labels corresponding to the historical numerical weather forecast data, determine the proportion of positive offset data and the proportion of negative offset data in the peak and valley wind speed periods;

[0017] Based on the proportion of positive offset data volume and the proportion of negative offset data volume, determine the target historical numerical weather forecast data whose data volume proportion is less than the first threshold; the target historical numerical weather forecast data is the historical numerical weather forecast data with positive offset or the historical numerical weather forecast data with negative offset;

[0018] The target historical numerical weather forecast data are processed based on the pre-trained WGAN-GP network to generate weather generation data; the weather generation data and the target historical numerical weather forecast data meet similar conditions.

[0019] Optionally, the number of directed acyclic graph structures in the neural network with a directed acyclic graph structure is 1.

[0020] Optionally, the first loss function calculation formula is as follows:

[0021] ;

[0022] in, is the first loss function, Indicates the number of output samples, represents the expected value, represents the model output value, represents the weight coefficient, represents the weight modification operator, ( )>0.5; ; is the relative error between the expected value and the output value, is the threshold value.

[0023] Optionally, the second loss function calculation formula is as follows:

[0024] ;

[0025] in, is the second loss function, Indicates the number of output samples, represents the expected value, represents the model output value, represents the weight coefficient, represents the weight modification operator, ( )>0.5; ; is the relative error between the expected value and the output value, is the threshold value.

[0026] Optionally, the first wind power prediction model and the second wind power prediction model are obtained by training a temporal convolutional network based on a multi-head attention mechanism.

[0027] According to the second aspect of the present application, an embodiment of the present application provides a wind power prediction device, including:

[0028] An acquisition module, used to acquire numerical weather forecast data of the wind farm;

[0029] A first processing module is used to process the numerical weather forecast data based on the wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, where the offset scenario includes a positive offset or a negative offset;

[0030] A second processing module is used for processing the numerical weather forecast data based on the first wind power prediction model to obtain a first wind power prediction result if the offset scenario of the numerical weather forecast data is a positive offset;

[0031] A third processing module is used for processing the numerical weather forecast data based on the second wind power prediction model to obtain a second wind power prediction result if the offset scenario of the numerical weather forecast data is a negative offset;

[0032] Among them, the first loss function during the training of the first wind power prediction model can enable the first wind power prediction model to update the first wind power prediction model parameters in the direction in which the output result deviates negatively from the expected value; the second loss function during the training of the second wind power prediction model can enable the second wind power prediction model to update the second wind power prediction model parameters in the direction in which the output result deviates positively from the expected value.

[0033] According to a third aspect of the present application, an embodiment of the present application provides an electronic device, including:

[0034] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the wind power prediction method as in the first aspect or any embodiment of the first aspect.

[0035] According to the fourth aspect of the present application, an embodiment of the present application provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute a wind power prediction method as in the first aspect or any embodiment of the first aspect.

[0036] The wind power prediction method, device, electronic device and storage medium provided in the embodiments of the present application obtain numerical weather forecast data of a wind farm; process the numerical weather forecast data based on a wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, wherein the offset scenario includes a positive offset or a negative offset; if the offset scenario of the numerical weather forecast data is a positive offset, process the numerical weather forecast data based on a first wind power prediction model to obtain a first wind power prediction result; if the offset scenario of the numerical weather forecast data is a negative offset, process the numerical weather forecast data based on a second wind power prediction model to obtain a second wind power prediction result; wherein the first loss function during the training of the first wind power prediction model enables the first wind power prediction model to update the first wind power prediction model parameters in the direction in which the output result deviates negatively from the expected value. ; The second loss function during the training of the second wind power prediction model can enable the second wind power prediction model to update the second wind power prediction model parameters in the direction in which the output result deviates positively from the expected value; in this way, when the numerical weather forecast data is positively offset, the first wind power prediction model can be used to predict wind power, so that the first wind power prediction result is offset in the opposite direction, thereby achieving the purpose of correcting the wind power and making the predicted wind power accurate; when the numerical weather forecast data is negatively offset, the second wind power prediction model can be used to predict wind power, so that the second wind power prediction result is offset in the opposite direction, thereby achieving the purpose of correcting the wind power and making the predicted wind power accurate; and, since there is no need to adopt data-driven optimization improvements or a refined modeling scheme for wind turbine clustering, the modeling cost and cost are not high.

[0037] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of a flow chart of a wind power prediction method in an embodiment of the present application;

[0039] Figure 2 A schematic diagram of the three-dimensional structure of the first loss function and the second loss function in an embodiment of the present application;

[0040] Figure 3 This is a schematic diagram of the statistical indicator distribution results after the negative offset samples are generated by using the pre-trained WGAN-GP network in an embodiment of the present application;

[0041] Figure 4This is a schematic diagram of the structure of a wind power prediction device in an embodiment of the present application;

[0042] Figure 5 This is a schematic diagram of the hardware structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0044] The present application embodiment provides a method for predicting wind power. Figure 1 As shown, including:

[0045] S101, obtaining numerical weather forecast data of the wind farm.

[0046] In this embodiment, the numerical weather forecast data includes but is not limited to meteorological data such as wind speed, wind direction, temperature, and humidity.

[0047] S102 processes the numerical weather forecast data based on the wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, where the offset scenario includes a positive offset or a negative offset.

[0048] In this embodiment, a wind speed deviation recognition model can be pre-trained to classify the deviation scenario of the numerical weather forecast data and determine whether it is a positive deviation scenario or a negative deviation scenario. A positive deviation can indicate that the numerical weather forecast data will be higher than the actual weather data, and a negative deviation can indicate that the numerical weather forecast data will be lower than the actual weather data.

[0049] In some embodiments, all data in the numerical weather forecast data may be processed by a wind speed deviation identification model to determine the deviation scenario of the numerical weather forecast data.

[0050] In some embodiments, part of the numerical weather forecast data may also be processed using a wind speed offset recognition model, for example, the wind speed in the numerical weather forecast data may be processed to determine an offset scenario of the numerical weather forecast data.

[0051] S103: If the offset scenario of the numerical weather forecast data is a positive offset, the numerical weather forecast data is processed based on the first wind power prediction model to obtain a first wind power prediction result. The first loss function during the training of the first wind power prediction model enables the first wind power prediction model to update the first wind power prediction model parameters in a direction in which the output result deviates negatively from the expected value.

[0052] In this embodiment, the first wind power prediction model may be pre-trained to predict the wind power of the wind farm in a short period of time when the numerical weather forecast data is in a positive offset scenario.

[0053] In this embodiment, due to the first loss function during the training of the first wind power prediction model, the first wind power prediction model can update the first wind power prediction model parameters in the direction in which the output result deviates negatively from the expected value. Therefore, when the wind power prediction is performed by the trained first wind power prediction model, when the numerical weather forecast data is positively offset, the first wind power prediction result can be offset in the opposite direction, thereby achieving the purpose of correcting the wind power, thereby predicting accurate wind power.

[0054] In some embodiments, the first wind power prediction model may be obtained by training a temporal convolutional network based on a multi-head attention mechanism.

[0055] In some embodiments, the first loss function calculation formula is as follows:

[0056] ;

[0057] in, is the first loss function, Indicates the number of output samples, represents the expected value, represents the model output value, represents the weight coefficient, represents the weight modification operator, ( )>0.5; ; is the relative error between the expected value and the output value, is the threshold value.

[0058] In some embodiments, the training process of the first wind power prediction model may be:

[0059] Obtain historical numerical weather forecast data and the corresponding historical wind power; use the historical numerical weather forecast data as training data, and the corresponding historical wind power as the expected value, i.e., the label, train the temporal convolutional network based on the multi-head attention mechanism, and calculate the first loss function using the first loss function calculation formula to update the model parameters and obtain the first wind power prediction model.

[0060] S104: If the offset scenario of the numerical weather forecast data is a negative offset, the numerical weather forecast data is processed based on the second wind power prediction model to obtain a second wind power prediction result. The second loss function during the training of the second wind power prediction model enables the second wind power prediction model to update the second wind power prediction model parameters in a direction in which the output result deviates positively from the expected value.

[0061] In this embodiment, the second wind power prediction model may be pre-trained to predict the wind power of the wind farm in a short period of time when the numerical weather forecast data is in a negative offset scenario.

[0062] In this embodiment, due to the second loss function during the training of the second wind power prediction model, the second wind power prediction model can update the second wind power prediction model parameters in the direction in which the output result deviates positively from the expected value. Therefore, when the wind power prediction is performed by the trained second wind power prediction model, when the numerical weather forecast data is negatively offset, the first wind power prediction result can be offset in the opposite direction, thereby achieving the purpose of correcting the wind power and thus predicting accurate wind power.

[0063] In some embodiments, the second wind power prediction model may be obtained by training a temporal convolutional network based on a multi-head attention mechanism.

[0064] In some embodiments, the second loss function is calculated as follows:

[0065] ;

[0066] in, is the second loss function, Indicates the number of output samples, represents the expected value, represents the model output value, represents the weight coefficient, represents the weight modification operator, ( )>0.5; ; is the relative error between the expected value and the output value, is the threshold value.

[0067] Specifically, Figure 2The three-dimensional structure of the first loss function and the second loss function is shown, where the error ( ) is the relative error between the expected value and the output value. It can be seen that when When is negative (actual value < predicted value), as the weight coefficient and weight correction operator increase, the model loss decreases, and the output value shifts in the direction higher than the expected value. When it is positive (actual value > predicted value), the loss value also decreases with the increase of weight coefficient and weight correction operator. At this time, the output value shifts in the direction below the expected value.

[0068] In some embodiments, the training process of the second wind power prediction model may be:

[0069] Obtain historical numerical weather forecast data and the corresponding historical wind power; use the historical numerical weather forecast data as training data, and the corresponding historical wind power as the expected value, i.e., the label, train the temporal convolutional network based on the multi-head attention mechanism, and calculate the second loss function using the second loss function calculation formula to update the model parameters and obtain the second wind power prediction model.

[0070] The wind power prediction method provided in the embodiment of the present application can perform wind power prediction through the first wind power prediction model when the numerical weather forecast data is positively offset, so that the first wind power prediction result is offset in the opposite direction, thereby achieving the purpose of correcting the wind power and making the predicted wind power accurate; when the numerical weather forecast data is negatively offset, the wind power prediction can be performed through the second wind power prediction model, so that the second wind power prediction result is offset in the opposite direction, thereby achieving the purpose of correcting the wind power and making the predicted wind power accurate; and since there is no need to adopt data-driven optimization improvement or a refined modeling scheme for wind turbine clustering, the modeling cost and cost are not high.

[0071] In an optional embodiment, the wind speed deviation identification model training process includes:

[0072] Acquire training samples and corresponding labels; the training samples include historical numerical weather forecast data and weather generation data of the wind farm, the historical numerical weather forecast data include historical numerical weather forecast data of positive offsets during wind speed peak and valley periods, and historical numerical weather forecast data of negative offsets during wind speed peak and valley periods; the weather generation data is generated based on historical numerical weather forecast data of positive offsets or historical numerical weather forecast data of negative offsets in wind speed peak and valley periods, the data volume ratio of which is less than a first threshold value; the labels include positive offset labels or negative offset labels; use the training samples and the corresponding labels to train a neural network with a directed acyclic graph structure to obtain a wind speed offset recognition model.

[0073] In this embodiment, there is an imbalance in the amount of positive offset data and the amount of negative offset data in the wind farm during the peak and valley periods of wind speed. For example, Table 1 counts the proportion of positive offset and negative offset samples in the three representative wind farms during the peak and valley periods of wind speed. Among the three wind farms, the number of positive offset samples during the peak wind speed period is significantly more than the number of negative offset samples, and the average proportion exceeds 90.46%, although the proportion of samples of the two types of offsets during the valley period in wind farm I is close. However, there is still an uneven distribution of wind speed offset samples in wind farms II and III. The imbalance between data classes is an important factor leading to the low generalization of the model. Therefore, weather generation data can be generated based on the historical numerical weather forecast data of positive offset or negative offset whose data volume proportion in the peak and valley periods of wind speed is less than the first threshold, so as to balance the imbalance between data classes. For example, if the amount of data corresponding to the negative offset data during the peak and valley periods of the wind farm accounts for less than 50%, the weather generation data can be generated based on the negative offset data during the peak and valley periods of the wind farm. The structure of the weather generation data is the same as that of the numerical weather forecast data, and can include meteorological data such as wind speed, wind direction, temperature, and humidity.

[0074] Table 1

[0075]

[0076] In some embodiments, the step of generating weather generation data may include:

[0077] Acquire historical numerical weather forecast data and corresponding labels; determine the proportion of positive offset data and the proportion of negative offset data in wind speed peak and valley periods based on the labels corresponding to the historical numerical weather forecast data; determine target historical numerical weather forecast data whose data proportion is less than a first threshold based on the proportion of positive offset data and the proportion of negative offset data; the target historical numerical weather forecast data is historical numerical weather forecast data with positive offset or historical numerical weather forecast data with negative offset; process the target historical numerical weather forecast data based on a pre-trained WGAN-GP network to generate weather generation data; the weather generation data and the target historical numerical weather forecast data meet similarity conditions.

[0078] Specifically, since the labels of the historical numerical weather forecast data can be used to determine whether the historical numerical weather forecast data is positively offset or negatively offset, the amount of positively offset data and the amount of negatively offset data in the peak and valley periods of wind speed can be counted, thereby calculating the proportion of positively offset data and the proportion of negatively offset data. Then, the proportion of positively offset data and the proportion of negatively offset data can be compared with the first threshold to determine whether they are less than the first threshold. If they are less than the first threshold, the corresponding historical numerical weather forecast data with positive offset or the historical numerical weather forecast data with negative offset is the target historical numerical weather forecast data. Then, some extreme data in the target historical numerical weather forecast data can be eliminated, and then the target historical numerical weather forecast data can be processed based on the pre-trained WGAN-GP network to generate weather generation data.

[0079] Figure 3 The distribution results of statistical indicators after the negative offset samples are generated by the pre-trained WGAN-GP network are shown. The statistical indicators include the maximum value (Max), the minimum value (Min), the mean (Mean) and the variance (Var). Under the same meteorological element (wind speed, humidity, temperature or wind direction), the more concentrated distribution represents the statistical indicators of the generated samples, and the more dispersed distribution represents the statistical indicators of the original negative offset samples. It can be seen that when the weather generation data is generated based on the pre-trained WGAN-GP network, the generated samples are closer to the original negative offset samples.

[0080] In this embodiment, after obtaining the training samples and the corresponding labels, the neural network with a directed acyclic graph structure may be trained to obtain a wind speed deviation recognition model.

[0081] In some embodiments, the number of directed acyclic graph structures in a neural network with a directed acyclic graph structure can be set to 1.

[0082] In specific implementation, when determining the number of directed acyclic graph structures in a neural network with a directed acyclic graph structure, it can be determined by comparison. For example, Table 2 shows the statistical results of offset scene recognition of neural networks with different numbers of directed acyclic graph structures after training samples and corresponding label training. It can be seen that the recognition accuracy of the neural network using a single directed acyclic graph structure is 0.77% and 1.58% higher than that of the neural network containing 2 and three directed acyclic graph structures on average, the recall rate is 2.86% and 10.11% higher on average, the F1 score is 2.38% and 7.49% higher respectively, and the model training time is 3.6s and 31.5s lower on average. The neural network with a single directed acyclic graph structure is significantly better than the neural network with multiple directed acyclic graph structures in terms of classification accuracy and running time. Therefore, it can be determined that the number of directed acyclic graph structures in the neural network with a directed acyclic graph structure is 1.

[0083] Table 2

[0084]

[0085] In the present application, weather generation data is generated by generating historical numerical weather forecast data of positive offset or historical numerical weather forecast data of negative offset based on the data volume ratio in wind speed peak and valley periods that is less than a first threshold, so that different types of data in the training data can be balanced to improve the accuracy of wind speed offset recognition model identification.

[0086] The present application also provides a wind power prediction device, such as Figure 4 As shown, including:

[0087] An acquisition module 41 is used to acquire numerical weather forecast data of the wind farm;

[0088] A first processing module 42 is used to process the numerical weather forecast data based on the wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, where the offset scenario includes a positive offset or a negative offset;

[0089] A second processing module 43 is configured to process the numerical weather forecast data based on the first wind power prediction model to obtain a first wind power prediction result if the offset scenario of the numerical weather forecast data is a positive offset;

[0090] A third processing module 44 is configured to process the numerical weather forecast data based on a second wind power prediction model to obtain a second wind power prediction result if the offset scenario of the numerical weather forecast data is a negative offset;

[0091] Among them, the first loss function during the training of the first wind power prediction model can enable the first wind power prediction model to update the first wind power prediction model parameters in the direction in which the output result deviates negatively from the expected value; the second loss function during the training of the second wind power prediction model can enable the second wind power prediction model to update the second wind power prediction model parameters in the direction in which the output result deviates positively from the expected value.

[0092] The wind power prediction device provided in the embodiment of the present application can perform wind power prediction through the first wind power prediction model when the numerical weather forecast data is positively offset, so that the first wind power prediction result is offset in the opposite direction, thereby achieving the purpose of correcting the wind power and making the predicted wind power accurate; when the numerical weather forecast data is negatively offset, the wind power prediction can be performed through the second wind power prediction model, so that the second wind power prediction result is offset in the opposite direction, thereby achieving the purpose of correcting the wind power and making the predicted wind power accurate; and since there is no need to adopt data-driven optimization improvement or a refined modeling scheme of wind turbine clustering, the modeling cost and cost are not high.

[0093] According to an embodiment of the present application, the present application also provides an electronic device and a readable storage medium.

[0094] Figure 5 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0095] like Figure 5 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the electronic device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0096] Multiple components in the electronic device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the electronic device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0097] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as a wind power prediction method. For example, in some embodiments, the wind power prediction method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the wind power prediction method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the wind power prediction method in any other appropriate manner (e.g., by means of firmware).

[0098] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0099] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0100] In the context of the present application, 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 device, 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.

[0101] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0102] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by 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), and the Internet.

[0103] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0104] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution disclosed in this application can be achieved, and this document is not limited here.

[0105] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0106] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A wind power prediction method, characterized in that: include: Obtain numerical weather forecast data for wind farms; Processing the numerical weather forecast data based on a wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, wherein the offset scenario includes a positive offset or a negative offset; If the offset scenario of the numerical weather forecast data is a positive offset, the numerical weather forecast data is processed based on a first wind power prediction model to obtain a first wind power prediction result; If the offset scenario of the numerical weather forecast data is a negative offset, processing the numerical weather forecast data based on a second wind power prediction model to obtain a second wind power prediction result; Among them, the first loss function during the training of the first wind power prediction model can enable the first wind power prediction model to update the first wind power prediction model parameters in the direction in which the output result deviates negatively from the expected value; the second loss function during the training of the second wind power prediction model can enable the second wind power prediction model to update the second wind power prediction model parameters in the direction in which the output result deviates positively from the expected value.

2. The wind power prediction method according to claim 1, characterized in that: The wind speed deviation identification model training process includes: Acquire training samples and corresponding labels; the training samples include historical numerical weather forecast data and weather generation data of the wind farm, the historical numerical weather forecast data include historical numerical weather forecast data with positive offsets during wind speed peak and valley periods, and historical numerical weather forecast data with negative offsets during wind speed peak and valley periods; the weather generation data is generated based on historical numerical weather forecast data with positive offsets or historical numerical weather forecast data with negative offsets in which the proportion of data volume during wind speed peak and valley periods is less than a first threshold; the labels include positive offset labels or negative offset labels; The training samples and corresponding labels are used to train a neural network with a directed acyclic graph structure to obtain a wind speed deviation recognition model.

3. The wind power prediction method according to claim 2, characterized in that: The steps of generating the weather generation data include: Obtain the historical numerical weather forecast data and corresponding labels; Based on the labels corresponding to the historical numerical weather forecast data, determining the proportion of positive offset data and the proportion of negative offset data in the wind speed peak and valley periods; Based on the positive offset data volume ratio and the negative offset data volume ratio, determine the target historical numerical weather forecast data whose data volume ratio is less than the first threshold; the target historical numerical weather forecast data is the historical numerical weather forecast data with positive offset or the historical numerical weather forecast data with negative offset; The target historical numerical weather forecast data is processed based on a pre-trained WGAN-GP network to generate the weather generation data; the weather generation data and the target historical numerical weather forecast data meet a similarity condition.

4. The wind power prediction method according to claim 2, characterized in that: The number of directed acyclic graph structures in the neural network with the directed acyclic graph structure is 1.

5. The wind power prediction method according to claim 1, characterized in that: The first loss function calculation formula is as follows: ； in, is the first loss function, Indicates the number of output samples, represents the expected value, represents the model output value, represents the weight coefficient, represents the weight modification operator, ( )>0.5; ; is the relative error between the expected value and the output value, is the threshold value.

6. The wind power prediction method according to claim 1, characterized in that: The second loss function calculation formula is as follows: ； in, is the second loss function, Indicates the number of output samples, represents the expected value, represents the model output value, represents the weight coefficient, represents the weight modification operator, ( )>0.5; ; is the relative error between the expected value and the output value, is the threshold value.

7. The wind power prediction method according to claim 1, characterized in that: The first wind power prediction model and the second wind power prediction model are obtained by training a temporal convolutional network based on a multi-head attention mechanism.

8. A wind power prediction device, characterized in that: include: An acquisition module, used to acquire numerical weather forecast data of the wind farm; A first processing module, configured to process the numerical weather forecast data based on a wind speed offset recognition model to obtain an offset scenario of the numerical weather forecast data, wherein the offset scenario includes a positive offset or a negative offset; A second processing module is configured to process the numerical weather forecast data based on a first wind power prediction model to obtain a first wind power prediction result if the offset scenario of the numerical weather forecast data is a positive offset; A third processing module is used for processing the numerical weather forecast data based on a second wind power prediction model to obtain a second wind power prediction result if the offset scenario of the numerical weather forecast data is a negative offset; Among them, the first loss function during the training of the first wind power prediction model can enable the first wind power prediction model to update the first wind power prediction model parameters in the direction in which the output result deviates negatively from the expected value; the second loss function during the training of the second wind power prediction model can enable the second wind power prediction model to update the second wind power prediction model parameters in the direction in which the output result deviates positively from the expected value.

9. An electronic device, characterized in that: include: at least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the wind power prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the wind power prediction method according to any one of claims 1 to 7.

Images