Power prediction method, device, equipment, medium and product for wind power icing state

Abstract

Embodiments of the present application disclose a wind power icing state power prediction method, device, equipment, medium and product, comprising: adding icing period labels by using historical measured power generation data, historical measured meteorological data and icing label network; obtaining an icing prediction network by training the historical measured power generation data containing the icing period labels, historical measured meteorological data of a target area, historical meteorological prediction data and historical regional actual output summation data; and predicting future prediction power generation data containing the icing period by using the icing prediction network. By automatically identifying the icing period, the manual labeling error is reduced, the accuracy and consistency of the data labels are improved, the problem of ignoring features caused by the scarcity of icing event data is solved, the icing label and power prediction are decoupled into two stages, different KAN neural networks are used to improve the model training efficiency, manual participation and human error are reduced, and more accurate icing power prediction is provided.

Description

Technical Field

[0001] This invention relates to the field of wind power prediction technology, and in particular to a method, apparatus, equipment, medium and product for predicting power under wind power icing conditions. Background Technology

[0002] With the vigorous development of new energy power, the grid-connected capacity of wind farms is rapidly increasing. Accurate and effective prediction of wind farm output within a target area can assist power system dispatchers in making precise and effective decisions and maintaining grid balance. However, when predicting wind power generation, weather conditions can cause icing on the surfaces of wind turbine blades and other structures. Icing can lead to increased blade weight, shape changes, and regulatory failure, creating unbalanced loads. The presence of ice also reduces the blades' utilization of wind energy and decreases their rotational torque, resulting in reduced power generation and output. Regional wind turbine icing can significantly reduce the overall wind power output within the region, creating power supply gaps and disrupting grid stability and balance.

[0003] Traditionally, forecasting wind turbine power generation under icing conditions, and even under icing conditions, often involves manually labeling historical power generation data from icing periods to create a data sample. This sample is then combined with weather forecast data for modeling. The process involves first labeling the data for the icing period, then building a predictive model based on numerical weather forecast data for that period and the actual wind turbine output. Future numerical weather forecasts are then used to predict wind power output for future icing periods. However, wind turbine icing is an extreme weather event, lasting only a short time within the wind turbine's generating cycle. It represents a "small sample" event relative to the overall power generation data, accounting for only 2% to 3% compared to normal power generation data. Directly using all power generation data, including icing conditions, for model training can easily lead to the model prioritizing predictions under normal conditions, reducing or even ignoring the weight of icing conditions. Furthermore, the traditional reliance on manually labeled icing period data further diminishes the temporal accuracy of the icing data, ultimately resulting in significant deviations in wind power generation forecasts for icing periods. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, medium, and product for predicting power output under icing conditions, in order to address the technical problem of low attention paid to small sample power generation data under icing conditions.

[0005] In a first aspect, embodiments of the present invention provide a power prediction method for wind power under icing conditions, including: S101, collect historical measured power generation data and historical measured meteorological data of each wind farm in the target area, and input them into the icing labeling network to obtain the icing period label data of each wind farm in the target area; S102, Based on the icing period label data, filter out the actual power generation data outside the icing period from the historical measured power generation data to form historical measured normal power generation data, and use the historical measured normal power generation data and the icing period label data to form icing prediction training data. S103, input the training data for icing prediction, historical measured meteorological data, historical meteorological forecast data, and historical total power output data of the target area into the power prediction regression network, train the power prediction regression network, and obtain the icing prediction network. S104. Using the trained icing prediction network, input the future weather forecast data of each wind farm in the target area to obtain the future predicted power generation data or the predicted power generation data of the future icing period.

[0006] Furthermore, S101 includes: Collect historical measured power generation data and historical measured meteorological data of each wind farm in the target area, and preprocess the collected data; Using an icing tagging network, historical measured power generation data is classified based on preprocessed historical measured meteorological data. Actual power generation data under icing that falls within the icing period is then filtered and icing tags are added to form icing period tagging data for each wind farm in the target area.

[0007] Furthermore, the icing label network is a KAN neural network, which has a network structure with a single input layer, two hidden layers, and a single output layer.

[0008] Furthermore, S103 also includes: Using icing prediction training data, historical measured meteorological data, and historical weather forecast data as inputs to the power prediction regression network, and historical regional actual total power output data as regression labels for the power prediction regression network, the network is trained. This allows the power prediction regression network to perform regression modeling based on the icing prediction training data labeled with icing time periods, combined with the differences between measured meteorological data and weather forecast data, and oriented towards the historical regional actual total power output data, thus forming a well-trained icing prediction network.

[0009] Furthermore, S104 includes: Input the future weather forecast data of each wind farm in the target area into the trained icing prediction network to obtain future predicted power generation data including the icing period. or, Using an icing labeling network, based on future weather forecast data for each wind farm in the target area, possible future icing periods are extracted. The weather forecast data for these periods is then input into a trained icing prediction network to obtain predicted power generation data for future icing periods.

[0010] Furthermore, the trained icing prediction network is a KAN neural network, which has a network structure with a single input layer, two hidden layers, and a single output layer. Secondly, embodiments of the present invention provide a power prediction device for wind power icing conditions, comprising: The icing period labeling module is used to obtain icing period label data for each wind farm in the target area based on the historical measured power generation data and historical measured meteorological data of each wind farm in the target area, using the icing labeling network. The training data integration module is used to filter out actual power generation data outside of the icing period from historical measured power generation data, and combine it with the icing period label data to form icing prediction training data; The prediction network training module uses the icing prediction training data and the historical measured meteorological data, historical meteorological forecast data and historical total power output data of the target area as inputs to the power prediction regression network to train the power prediction regression network and form the icing prediction network. The icing power output prediction module uses a trained icing prediction network and inputs future weather forecast data for each wind farm in the target area to obtain future predicted power generation data or predicted power generation data for future icing periods.

[0011] Thirdly, embodiments of the present invention provide an electronic device, including: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the above-described power prediction method for wind power icing conditions.

[0012] Fourthly, embodiments of the present invention provide a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the power prediction method for wind power icing conditions described above.

[0013] Fifthly, embodiments of the present invention also provide a computer product, including a computer program, which, when executed by a processor, implements the power prediction method for wind power icing conditions as described above.

[0014] This invention provides a method, apparatus, device, medium, and product for predicting wind power output under icing conditions. The method introduces a KAN neural network as an icing labeling network. First, it identifies the actual power generation data within the icing period and adds icing labels. Then, it trains the icing prediction network using the labeled data and normal data. The first KAN neural network is used as a part of a second KAN neural network (icing prediction network) to accurately classify and label small sample data during the icing period, increasing the attention given to small sample data in the neural network. Combined with the characteristics of the side activation function, it enhances the learning ability for nonlinear and small sample data, making it particularly suitable for accurate prediction of wind power output under small sample weather conditions such as icing. It not only leverages the KAN network's ability to fit measured meteorological and power data to automatically identify icing periods, reducing manual labeling errors and improving the accuracy and consistency of data labels, but also solves the problem of scarce icing event data (only accounting for 2%-3%), where traditional models easily overlook its features. By decoupling "icing label generation" and "power prediction" into two-stage processes, and using two different KAN neural networks in each stage, the efficiency and interpretability of model training are improved, which facilitates engineering deployment and maintenance. The automated label generation and model training process reduces human intervention, lowers labor costs and human error, and provides more accurate icing power prediction. This helps power grid dispatchers to formulate response strategies in advance, ensuring the stable operation of the power grid and the security of power supply. Attached Figure Description

[0015] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a power prediction method for wind power under icing conditions according to Embodiment 1 of the present invention; Figure 2 This is a flowchart of a power prediction method for wind power icing conditions as described in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the power prediction device for wind power icing conditions as described in Embodiment 3 of the present invention; Figure 4 This is a structural diagram of the electronic device described in Embodiment 4 of the present invention. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0017] Example 1 Figure 1 This is a flowchart of a power prediction method for wind power icing conditions according to Embodiment 1 of the present invention. In this embodiment, the power generation data during the icing period is tagged using measured power generation data and meteorological data within the target area. The measured power generation data containing the tagged icing period data is combined with measured meteorological data and meteorological forecast data for the corresponding time period, and the total power generation data including the icing period is used as the target for training a regression model. The trained regression model can accurately predict the power output during possible future icing periods based on future meteorological forecast data. Specifically, it includes the following steps: S101. Collect historical measured power generation data and historical measured meteorological data of each wind farm in the target area, and input them into the icing labeling network to obtain the icing period label data of each wind farm in the target area.

[0018] Because icing is an extreme weather phenomenon with complex causes, not all wind farms within the target area will experience icing on their turbine surfaces simultaneously. Icing often occurs at individual wind farms due to a combination of localized weather factors. The target area typically refers to a power generation unit or monitoring unit within the power grid, i.e., an area where power generation equipment (e.g., multiple wind turbines) is controlled and dispatched, and icing monitoring is conducted on all wind turbines within that area. The conditions for wind turbine surface icing typically require temperatures between -5°C and 0°C, humidity below 85%, and wind speeds between 1 m / s and 10 m / s. Meeting all three meteorological conditions significantly increases the likelihood of icing. Based on these conditions, if a wind farm's output decreases by more than 10% compared to normal weather and wind speed conditions, icing on the turbine surface can be considered to have occurred with high confidence. For example, if a wind farm normally outputs 30 MW at a wind speed of 5 m / s, but its output drops to 27 MW or lower under the aforementioned icing meteorological conditions, icing is considered to have occurred.

[0019] Specifically, S101 includes: collecting historical measured power generation data and historical measured meteorological data of each wind farm in the target area, and preprocessing the collected data.

[0020] The collected historical measured power generation data and historical measured meteorological data from each wind farm within the target area undergo preprocessing such as data organization and cleaning. Historical measured power generation data consists of accumulated historical data (time-series data) from each wind farm over a specific historical period. This data is divided into time units, with power generation data within each unit. For example, using 30-minute intervals, there are 48 time units within a 24-hour day, and 144 time units of time-series power generation data within a 3-day historical period. Similarly, historical measured meteorological data needs to be aligned with the historical measured power generation data in terms of time, specifically by aligning the timestamps (meteorological data timestamps with power generation data timestamps) for subsequent calculations. For instance, data organization follows the data format specified by the power generation company, organizing the data into the format required for neural network training. For example, using Python's pandas methods to organize text-formatted data into dataframe format. Data cleaning involves removing missing values, dead values, and values ​​exceeding limits from the data. A small number of missing values ​​are filled using non-linear interpolation, while a large number of missing values ​​are directly deleted. Dead values ​​are first deleted to convert them into missing values, and then filled using non-linear interpolation. Values ​​exceeding limits need to be verified for accuracy; if the limits remain unchanged, they are treated as dead values. The preprocessed data is presented as a fixed-time-resolution table, which can be in Excel format for easy retrieval by neural networks. The table includes historical measured power generation data and historical measured meteorological data for each wind farm. It can also include historical meteorological forecast data and historical total power generation data for all wind farms, depending on subsequent forecasting needs.

[0021] Using an icing tagging network, historical measured power generation data is classified based on preprocessed historical measured meteorological data. Actual power generation data under icing that falls within the icing period is then filtered and icing tags are added to form icing period tagging data for each wind farm in the target area.

[0022] The icing labeling network employs deep learning algorithms to identify potential icing weather in preprocessed historical meteorological data. It then combines this with historical power generation data showing changes in generator power output due to icing to categorize power generation data for icing periods, distinguishing it from normal periods. Actual power generation data within the icing periods is selected and tagged with icing labels to indicate the characteristics of power output changes during those periods, forming icing period label data for each wind farm in the target area. By utilizing deep learning algorithms to identify icing scenarios, the accuracy of icing scenario identification is improved through fitting, rather than relying on human experience for icing period segmentation. This reduces the impact of human experience on dataset accuracy and avoids human labeling errors. Furthermore, adding labels to small sample data within icing periods enhances the importance of these small samples relative to the overall imbalanced dataset, allowing the model to focus on the characteristics of small sample data and improve prediction accuracy.

[0023] Optionally, the icing labeling network is a KAN neural network (Kolmogorov-Arnold neural network), which has a single input layer, two hidden layers, and a single output layer. Using the KAN neural network, based on the preprocessed historical meteorological data and historical power generation data of each wind farm in the target area, the icing time period of each wind farm is fitted. The power generation data of each wind farm within the icing time period is then selected and icing labels are added, yielding the actual power generation data within the icing time period.

[0024] S102. Based on the icing period label data, filter out the actual power generation data outside the icing period from the historical measured power generation data to form historical measured normal power generation data. Use the historical measured normal power generation data and the icing period label data to form icing prediction training data.

[0025] In historical measured power generation data, the measured power generation data other than those labeled with icing periods are considered as normal power generation data, forming historical normal measured power generation data. This historical normal measured power generation data is then integrated with the icing period labeled data, which represents power generation data within icing periods, to form icing prediction training data containing power generation data from the labeled icing periods. Because icing periods are labeled in this data, when training a neural network using this data, the neural network can identify the characteristics of power generation changes within icing periods based on the icing labels, improving the accuracy of power generation prediction during icing periods. After training, it can accurately predict the power generation of each wind farm within icing periods.

[0026] S103. Input the icing prediction training data, historical measured meteorological data, historical meteorological forecast data, and historical total power output data of the target area into the power prediction regression network, train the power prediction regression network, and obtain the icing prediction network.

[0027] After adding icing labels to the power generation data during icing periods and integrating it with power generation data under normal conditions to form complete historical power generation data, the integrated icing prediction training data is used as the complete measured power generation data. Combined with historical measured meteorological data, power generation can be predicted through a power prediction regression network. The power prediction regression network is trained using the icing prediction training data, historical measured meteorological data, and historical weather forecast data as inputs, and the total historical regional actual power output as the regression term. The icing prediction training data and historical measured meteorological data serve as the data reference basis (for the regression network to learn), historical weather forecast data as the prediction input, and the total historical regional actual power output as the regression term. In other words, the model learns the correlation between wind power output characteristics during icing periods and measured meteorological characteristics in the icing prediction training data and historical measured meteorological data, and combines the correlation between weather forecast data and the total actual power output. During training, the model continuously optimizes its parameters to achieve the required prediction accuracy (within the allowable deviation range or meeting the confidence requirements), thus completing the training and forming a well-trained icing prediction network.

[0028] S104. Using the trained icing prediction network, input the future weather forecast data of each wind farm in the target area to obtain the future predicted power generation data or the predicted power generation data of the future icing period.

[0029] A well-trained icing prediction network can predict wind power output under corresponding meteorological conditions based on meteorological data. After inputting future weather forecast data into the network, it can derive wind power output prediction data corresponding to the forecast duration. Different forecast durations can be input based on prediction needs. Similar to the accuracy characteristics of meteorological data forecasts, the accuracy of future wind power output prediction decreases to some extent with increasing forecast duration; shorter forecast durations result in higher accuracy. Furthermore, prediction accuracy also depends on the accuracy of the meteorological forecast data. It can also be specifically designed to predict wind power output under icing conditions, facilitating power dispatching by grid dispatchers.

[0030] Optionally, the trained icing prediction network is a KAN neural network, which has a single input layer, two hidden layers, and a single output layer. The main function of the KAN neural network is to regress multi-dimensional data to one dimension, i.e., to perform regression-type tasks such as predicting wind power output using meteorological data within a target area. However, since the icing state of wind turbines occurs for a relatively short period in the overall power generation process, it is a "small sample" event compared to traditional neural networks. Relying on manual labeling would lead to problems such as decreased accuracy and data labeling errors. Therefore, we first use the KAN neural network to add icing labels, and then integrate the labeled power generation data under icing conditions with the power generation data under other normal conditions to form a data sample to train the KAN neural network, enabling it to recognize icing weather. At the same time, based on the impact of icing weather on wind power output, it predicts the wind power output and its degree of change during the icing period, thereby improving the accuracy of wind power output prediction.

[0031] This embodiment introduces a KAN neural network as an icing labeling network. First, it identifies and labels the actual power generation data within the icing period. Then, it trains an icing prediction network using the labeled data and normal data. The first KAN neural network is used as a part of a second KAN neural network (icing prediction network), accurately classifying and labeling small sample data during icing periods. This increases the attention given to small sample data in the neural network. Combined with the characteristics of the edge activation function, it enhances the learning ability for nonlinear and small sample data, making it particularly suitable for accurate prediction of wind power output under small sample weather conditions such as icing. This not only leverages the KAN network's ability to fit measured meteorological and power data to automatically identify icing periods, reducing manual labeling errors and improving the accuracy and consistency of data labels, but also solves the problem of scarce icing event data (only accounting for 2%-3%), where traditional models tend to ignore its features. By decoupling "icing label generation" and "power prediction" into two-stage processes, and using two different KAN neural networks in each stage, the efficiency and interpretability of model training are improved, which facilitates engineering deployment and maintenance. The automated label generation and model training process reduces human intervention, lowers labor costs and human error, and provides more accurate icing power prediction. This helps power grid dispatchers to formulate response strategies in advance, ensuring the stable operation of the power grid and the security of power supply.

[0032] Example 2 Figure 2 This is a flowchart of a power prediction method for wind power icing conditions according to Embodiment 2 of the present invention. This embodiment is based on the above embodiment and optimized. In this embodiment, S104 is specifically optimized as follows: Input the future weather forecast data of each wind farm in the target area into the trained icing prediction network to obtain future predicted power generation data including the icing period. or, By inputting the meteorological data of possible icing periods from the future weather forecast data of each wind farm in the target area into the trained icing prediction network, the predicted power generation data for the future icing period is obtained.

[0033] Accordingly, the power prediction method for wind power icing conditions provided in this embodiment specifically includes: S201. Collect historical measured power generation data and historical measured meteorological data of each wind farm in the target area, and input them into the icing labeling network to obtain the icing period label data of each wind farm in the target area.

[0034] S202. Based on the icing period label data, filter out the actual power generation data outside the icing period from the historical measured power generation data to form historical measured normal power generation data. Use the historical measured normal power generation data and the icing period label data to form icing prediction training data.

[0035] S203. The icing prediction training data, along with the historical measured meteorological data, historical meteorological forecast data, and historical total power output data of the target area, are input into the power prediction regression network to train the power prediction regression network and obtain the icing prediction network.

[0036] An optional implementation of this embodiment is to use icing prediction training data, historical measured meteorological data, and historical weather forecast data as inputs to the power prediction regression network, and historical regional actual power output data as regression labels for the power prediction regression network. The power prediction regression network is trained so that it can perform regression modeling based on the icing prediction training data marked with icing time periods, combined with the differences between measured meteorological data and weather forecast data, and oriented towards the historical regional actual power output data, thus forming a trained icing prediction network.

[0037] Using icing prediction training data, historical measured meteorological data, and historical weather forecast data as input data, the icing prediction training data and historical measured meteorological data are used to train the power prediction regression network. By analyzing the correlation between the icing prediction training data labeled with icing and the historical measured meteorological data, the network learns the linkage between wind power output changes during icing periods and measured meteorological conditions, thus enabling the power prediction regression network to predict corresponding wind power output based on meteorological data. Simultaneously, historical weather forecast data is used as input to the regression network, and the historical total actual power output data for the region is used as the regression term for regression fitting modeling. This optimizes the regression network parameters and improves the prediction accuracy of the power prediction regression network. Specifically, based on the deviation between historical weather forecast data and historical measured meteorological data, the model aims to minimize the deviation between the predicted value and the measured data of the historical total power output of each wind farm in the target area including icing periods, until the required requirements or confidence level are met. This completes the training, forming a well-trained icing prediction network.

[0038] S204-1 inputs the future weather forecast data of each wind farm in the target area into the trained icing prediction network to obtain future predicted power generation data including the icing period.

[0039] The trained icing prediction network takes as input future weather forecast data for the target area. Based on the future weather forecast data, the trained icing prediction network can automatically identify whether there will be icing weather in the future. Based on the possible icing weather and the different impacts of different degrees of icing weather on the power output of wind turbines, it can predict the wind power output of each wind turbine under different weather conditions, forming future predicted power generation data for the target area that includes icing periods. or, S204-2 utilizes an icing labeling network to extract potential future icing periods based on future weather forecast data for each wind farm within the target area. The weather forecast data for these periods is then input into a trained icing prediction network to obtain predicted power generation data for future icing periods.

[0040] It can also predict wind power output only during the icing period, based on actual needs, thus shortening the prediction time and improving prediction accuracy. By inputting meteorological data on potential icing periods from future weather forecasts into a trained icing prediction network, the data can be filtered using a neural network. Specifically, an icing label network is extracted from the trained icing prediction network. Using this network, and based on the input future weather forecast data, the output of the icing label network determines the potential icing periods, allowing for wind power prediction during these periods. Alternatively, potential icing periods can be filtered based on human experience. To minimize errors in manual filtering, a longer time series can be created by combining N time resolutions before and after the selected potential icing periods. This data is then input into the icing prediction network, which automatically predicts wind power output. The predicted data also includes wind power output data during the initial and subsequent stages of icing impact, resulting in a more comprehensive and complete picture of wind power output changes during the icing period. This provides predicted power generation data for future icing periods, which power grid dispatchers can use for reference.

[0041] This embodiment utilizes a KAN neural network to automatically identify icing periods and add icing period labels to the power generation data within those periods. Then, historical measured power generation data and historical measured meteorological data, both labeled with icing periods, are used to train a power prediction regression network. Regression modeling is performed using historical measured meteorological forecast data and historical regional actual total power output data to train the icing prediction network. The trained icing prediction network can accurately predict wind power power under icing conditions based on future meteorological forecast data for the target area. This approach not only leverages the stronger function fitting and data pattern discovery capabilities of the KAN neural network but also more accurately captures the complex nonlinear mapping relationship between icing meteorological conditions and sudden drops in wind power, automatically adding icing labels. Icing event identification is entirely data-driven. Furthermore, the automatic identification of icing labels is seamlessly integrated with the power prediction model training, simplifying the model maintenance process, reducing application costs, and obtaining stable regional wind power prediction results, significantly improving the efficiency and convenience of scheduling decisions.

[0042] Example 3 Figure 3 This is a schematic diagram of the structure of a power prediction device for wind power icing conditions according to Embodiment 3 of the present invention. In this embodiment, the power prediction device for wind power icing conditions includes: The icing period labeling module 810 is used to obtain icing period labeling data for each wind farm in the target area based on the historical measured power generation data and historical measured meteorological data of each wind farm in the target area using the icing labeling network. The training data integration module 820 is used to filter out actual power generation data outside of the icing period from historical measured power generation data, and form icing prediction training data with the icing period label data. The prediction network training module 830 uses the icing prediction training data and the historical measured meteorological data, historical meteorological forecast data and historical actual power output data of the target area as inputs to the power prediction regression network to train the power prediction regression network and form the icing prediction network. The icing power output prediction module 840 uses a trained icing prediction network and inputs future weather forecast data for each wind farm in the target area to obtain future predicted power generation data or predicted power generation data for future icing periods.

[0043] This embodiment uses an icing period labeling module to add icing tags to power generation data within icing periods based on historical measured power generation data and historical measured meteorological data from each wind farm in the target area. A training data integration module integrates the labeled icing period power generation data with power generation data under normal conditions to form icing prediction training data. A prediction network training module uses the icing prediction training data, historical measured meteorological data of the target area, historical weather forecast data, and historical total actual power output data of the area to perform regression modeling training on the power prediction regression network, forming an icing prediction network. An icing power output prediction module uses the trained icing prediction network and inputs future weather forecast data from each wind farm in the target area to predict future power generation data including icing periods. This not only leverages the KAN network's ability to fit measured meteorological and power data to automatically identify icing periods, reducing manual labeling errors and improving the accuracy and consistency of data labeling; it also solves the problem of scarce icing event data (only accounting for 2%-3%), where traditional models easily overlook its characteristics. By decoupling "icing label generation" and "power prediction" into two-stage processes, and using two different KAN neural networks in each stage, the efficiency and interpretability of model training are improved, which facilitates engineering deployment and maintenance. The automated label generation and model training process reduces human intervention, lowers labor costs and human error, and provides more accurate icing power prediction. This helps power grid dispatchers to formulate response strategies in advance, ensuring the stable operation of the power grid and the security of power supply.

[0044] The power prediction device for wind power icing conditions provided in the embodiments of the present invention can execute the power prediction method for wind power icing conditions provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0045] Example 4 Figure 4 This is a structural diagram of an electronic device according to Embodiment 4 of the present invention. Figure 4A block diagram is shown of an exemplary electronic device 12 suitable for implementing embodiments of the present invention. Figure 4 The electronic device 12 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0046] like Figure 4 As shown, the electronic device 12 is represented in the form of a general-purpose computing device. The components of the electronic device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0047] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0048] Electronic device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 12, including volatile and non-volatile media, removable and non-removable media.

[0049] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (… Figure 4 Not shown; usually referred to as a "hard drive"). Although Figure 4 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0050] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0051] Electronic device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with the electronic device 12 / server / computer, and / or with any device that enables the electronic device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output (I / O) interface 22. Furthermore, electronic device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. Figure 4 As shown, network adapter 20 communicates with other modules of electronic device 12 via bus 18. It should be understood that, although... Figure 4 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0052] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the power prediction method for wind power icing state provided in the embodiments of the present invention.

[0053] Example 5 Embodiment 5 of the present invention also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the power prediction method for wind power icing conditions as provided in the above embodiments.

[0054] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0055] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0056] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0057] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0058] Example 6 Embodiment 6 of the present invention also provides a computer product, including a computer program, which, when executed by a processor, implements the power prediction method for wind power icing conditions as described in the above embodiments.

[0059] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for predicting power under icing conditions in wind power plants, characterized in that, include: S101, collect historical measured power generation data and historical measured meteorological data of each wind farm in the target area, and input them into the icing labeling network to obtain the icing period label data of each wind farm in the target area; S102, Based on the icing period label data, filter out the actual power generation data outside the icing period from the historical measured power generation data to form historical measured normal power generation data, and use the historical measured normal power generation data and the icing period label data to form icing prediction training data. S103, input the training data for icing prediction, historical measured meteorological data, historical meteorological forecast data, and historical total power output data of the target area into the power prediction regression network, train the power prediction regression network, and obtain the icing prediction network. S104. Using the trained icing prediction network, input the future weather forecast data of each wind farm in the target area to obtain the future predicted power generation data or the predicted power generation data of the future icing period.

2. The method according to claim 1, characterized in that, S101 includes: Collect historical measured power generation data and historical measured meteorological data of each wind farm in the target area, and preprocess the collected data; Using an icing tagging network, historical measured power generation data is classified based on preprocessed historical measured meteorological data. Actual power generation data under icing that falls within the icing period is then filtered and icing tags are added to form icing period tagging data for each wind farm in the target area.

3. The method according to claim 2, characterized in that: The icing label network is a KAN neural network, which has a network structure with a single input layer, two hidden layers, and a single output layer.

4. The method according to claim 1, characterized in that, S103 further includes: Using icing prediction training data, historical measured meteorological data, and historical weather forecast data as inputs to the power prediction regression network, and historical regional actual total power output data as regression labels for the power prediction regression network, the network is trained. This allows the power prediction regression network to perform regression modeling based on the icing prediction training data labeled with icing time periods, combined with the differences between measured meteorological data and weather forecast data, and oriented towards the historical regional actual total power output data, thus forming a well-trained icing prediction network.

5. The method according to claim 1, characterized in that, S104 includes: Input the future weather forecast data of each wind farm in the target area into the trained icing prediction network to obtain future predicted power generation data including the icing period. or, Using an icing labeling network, based on future weather forecast data for each wind farm in the target area, possible future icing periods are extracted. The weather forecast data for these periods is then input into a trained icing prediction network to obtain predicted power generation data for future icing periods.

6. The method according to claim 5, characterized in that: The trained icing prediction network is a KAN neural network, which has a network structure of a single input layer, two hidden layers, and a single output layer.

7. A power prediction device for wind power icing conditions, characterized in that, include: The icing period labeling module is used to obtain icing period label data for each wind farm in the target area based on the historical measured power generation data and historical measured meteorological data of each wind farm in the target area, using the icing labeling network. The training data integration module is used to filter out actual power generation data outside of the icing period from historical measured power generation data, and combine it with the icing period label data to form icing prediction training data; The prediction network training module uses the icing prediction training data and the historical measured meteorological data, historical meteorological forecast data and historical total power output data of the target area as inputs to the power prediction regression network to train the power prediction regression network and form the icing prediction network. The icing power output prediction module uses a trained icing prediction network and inputs future weather forecast data for each wind farm in the target area to obtain future predicted power generation data or predicted power generation data for future icing periods.

8. An electronic device, characterized in that, The electronic device includes: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the power prediction method for wind power icing conditions as described in any one of claims 1-6.

9. A storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the power prediction method for wind power icing conditions as described in any one of claims 1-6.

10. A computer product comprising a computer program, characterized in that: When executed by a processor, the computer program implements the power prediction method for wind power icing conditions as described in any one of claims 1-6.

Images