Wind farm blade icing prediction method, device, equipment and medium

Abstract

The application discloses a wind farm blade icing prediction method, device, equipment and medium, and relates to the field of wind power generation. The method comprises the following steps: acquiring predicted meteorological data of a wind farm in a future preset time period, wherein the predicted meteorological data comprises temperature and humidity; inputting the predicted meteorological data into a blade icing prediction model trained in advance, processing the predicted meteorological data by the blade icing prediction model, and outputting a blade icing prediction result of the wind farm, wherein the blade icing prediction result represents a blade icing condition in the future preset time period, the blade icing prediction model is trained according to historical sample data of the wind farm, the historical sample data comprises meteorological data in a historical icing time and meteorological data in a historical non-icing time, the meteorological data comprises temperature and humidity, and the historical sample data has a label for representing a blade icing condition corresponding to the meteorological data. According to the embodiment of the application, the safety risk of the wind farm can be reduced.

Description

Technical Field

[0001] This application belongs to the field of wind power generation, and in particular relates to a method, device, equipment and medium for predicting icing of wind farm blades. Background Technology

[0002] During winter and spring, when wind turbines operate in low-temperature environments, water droplets from the natural environment may condense on the turbine blades, forming a thin layer of ice – a condition known as blade icing. Once icy, the ice layer gradually grows thicker, altering the aerodynamic characteristics of the blades, increasing the dynamic load on the wind turbine, and posing safety risks.

[0003] Blade icing is a natural phenomenon and cannot be controlled. To ensure the safety of wind turbine generators, it is necessary to monitor the icing status of the blades. Currently, this relies heavily on manual observation of the blades to determine if icing has occurred. However, real-time observation of blade icing has a time lag. By the time manual observation detects icing, the wind turbine generators in the wind farm have already been operating with iced blades for a considerable period, making it difficult to detect the icing in time and posing a significant safety risk to the wind farm. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, and medium for predicting blade icing in wind farms, which can detect blade icing in a timely and accurate manner, thereby reducing the safety risks of wind farms.

[0005] In a first aspect, embodiments of this application provide a method for predicting blade icing in a wind farm, comprising: acquiring predicted meteorological data of the wind farm within a predetermined future time period, the predicted meteorological data including temperature and humidity; inputting the predicted meteorological data into a pre-trained blade icing prediction model, the blade icing prediction model processing the predicted meteorological data, and outputting a blade icing prediction result for the wind farm, the blade icing prediction result being used to characterize the blade icing situation within the predetermined future time period, wherein the blade icing prediction model is trained based on historical sample data of the wind farm, the historical sample data including meteorological data during historical icing periods and meteorological data during historical non-icing periods, the meteorological data including temperature and humidity, the historical sample data having labels, the labels being used to characterize the blade icing situation corresponding to the meteorological data.

[0006] In some possible embodiments, before inputting the predicted meteorological data into the pre-trained blade icing prediction model, the method further includes: acquiring historical sample data and the icing shutdown ratio corresponding to the meteorological data in the historical sample data, wherein the icing shutdown ratio includes the ratio of the number of wind turbines shut down due to blade icing in the wind farm to the total number of wind turbines in the wind farm; generating labels based on the icing shutdown ratio; establishing a correspondence between the labels and the historical sample data; and using at least a portion of the labeled historical sample data as a training dataset to train a machine learning model to obtain the blade icing prediction model.

[0007] In some possible embodiments, the historical sample data also includes meteorological data within a first preset duration prior to at least one data collection time in the historical icing time.

[0008] In some possible embodiments, the meteorological data used to train the leaf icing prediction model during historical icing periods include first-type icing meteorological data, second-type icing meteorological data, and third-type icing meteorological data.

[0009] The first type of icing meteorological data includes meteorological data with corresponding labels indicating leaf icing and temperatures below the preset icing temperature and humidity above the preset icing humidity. The second type of icing meteorological data includes meteorological data with corresponding labels indicating leaf icing, labels indicating leaf icing appearing continuously within the previous second preset time period, and temperatures below the preset icing temperature and humidity below or equal to the preset icing humidity. The third type of icing meteorological data includes meteorological data with corresponding labels indicating leaf icing, labels indicating leaf icing appearing continuously within the previous third preset time period, and temperatures above or equal to the preset icing temperature.

[0010] In some possible embodiments, the blade icing prediction model includes a blade icing classification prediction model and / or a blade icing proportion prediction model; when the blade icing prediction model includes a blade icing classification prediction model, the blade icing prediction result includes whether the blade is iced or not, and the label represents whether the blade is iced or not; when the blade icing prediction model includes a blade icing proportion prediction model, the blade icing prediction result includes an icing shutdown proportion, and the label represents the icing shutdown proportion, which includes the ratio of the number of wind turbine generators in the wind farm that are shut down due to blade icing to the total number of wind turbine generators in the wind farm.

[0011] In some possible embodiments, the blade icing prediction results include the icing shutdown ratio, which is the ratio of the number of wind turbines in the wind farm that are shut down due to blade icing to the total number of wind turbines in the wind farm.

[0012] After inputting the predicted meteorological data into the pre-trained blade icing prediction model, and having the model process the data to output the blade icing prediction results for the wind farm, the process further includes: calculating the predicted number of wind turbines in the wind farm within the preset area that will be shut down due to icing, based on the predicted blade icing results and the number of wind turbines in the preset area; determining the total number of wind turbines in the preset area that will be shut down due to icing, based on the predicted number; and executing the turbine protection strategy corresponding to the number of wind turbines shut down due to icing in the preset area.

[0013] In some possible embodiments, before using at least a portion of the labeled historical sample data as a training dataset for model training, the method further includes: selecting one wind farm as a standard wind farm from two or more wind farms; normalizing the historical sample data of the standard wind farm and obtaining a normalization scaling factor; and using the normalization scaling factor to normalize the historical sample data of wind farms other than the standard wind farm.

[0014] Using at least a portion of labeled historical sample data as a training dataset for model training includes: using at least a portion of labeled, normalized historical sample data as a training dataset for model training.

[0015] In some possible embodiments, the predicted meteorological data is consistent with the types of meteorological data, which also include one or more of the following: wind speed, precipitation, cloud cover, air pressure, and dew point temperature.

[0016] Secondly, embodiments of this application provide a wind farm blade icing prediction device, comprising: a data acquisition module for acquiring predicted meteorological data of the wind farm within a preset future time period, the predicted meteorological data including temperature and humidity; and a model processing module for inputting the predicted meteorological data into a pre-trained blade icing prediction model, the blade icing prediction model processing the predicted meteorological data, and outputting a wind farm blade icing prediction result, the blade icing prediction result being used to characterize the blade icing situation within the preset future time period. The blade icing prediction model is trained based on historical sample data of the wind farm, the historical sample data including meteorological data during historical icing periods and meteorological data during historical non-icing periods, the meteorological data including temperature and humidity, and the historical sample data having labels, the labels being used to characterize the blade icing situation corresponding to the meteorological data.

[0017] Thirdly, embodiments of this application provide a wind farm blade icing prediction device, comprising: a processor and a memory storing computer program instructions; the processor executes the computer program instructions to implement the wind farm blade icing prediction method of the first aspect.

[0018] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the wind farm blade icing prediction method of the first aspect.

[0019] This application provides a method, apparatus, device, and medium for predicting blade icing in wind farms. It utilizes a pre-trained blade icing prediction model to process input meteorological data from a wind farm over a predetermined time period, thereby obtaining the blade icing prediction result output by the model. The blade icing prediction model is trained on labeled historical sample data from the wind farm. This historical sample data includes meteorological data from historical icing periods and meteorological data from historical non-icing periods. The labels on the historical sample data characterize the blade icing situation corresponding to the meteorological data. The blade icing prediction model trained on labeled historical sample data from the wind farm can accurately describe the nonlinear dynamic process of blade icing evolution, thus accurately predicting blade icing conditions in the predetermined time period based on forecast meteorological data, enabling timely and accurate detection of blade icing, and reducing safety risks to the wind farm. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart illustrating a method for predicting wind farm blade icing according to an embodiment of this application;

[0022] Figure 2 A schematic diagram illustrating an example of wind farm blade icing prediction hit rate provided in an embodiment of this application;

[0023] Figure 3 A flowchart of a method for predicting wind farm blade icing according to another embodiment of this application;

[0024] Figure 4 A flowchart of a method for predicting wind farm blade icing, provided in another embodiment of this application;

[0025] Figure 5A schematic diagram showing a comparison between the actual icing shutdown capacity and the predicted icing shutdown capacity in a certain region, provided as an example of an embodiment of this application;

[0026] Figure 6 A schematic diagram of the structure of a wind farm blade icing prediction device provided in an embodiment of this application;

[0027] Figure 7 A schematic diagram of the structure of a wind farm blade icing prediction device provided in an embodiment of this application. Detailed Implementation

[0028] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0029] During winter and spring, when wind turbines operate in low-temperature environments, water droplets from the natural environment may condense on the turbine blades, forming a thin layer of ice – a condition known as blade icing. Once icing occurs, the ice layer gradually thickens, altering the aerodynamic characteristics of the blades and increasing the dynamic load on the wind turbine, thus posing a safety risk. Blade icing is a natural phenomenon and cannot be controlled. To ensure the safety of wind turbines, it is necessary to monitor the icing status of the blades. Currently, this relies heavily on manual observation of the blades to determine if icing has occurred. However, real-time observation of blade icing has a time lag; by the time manual observation detects icing, the wind turbines in the wind farm have already been operating with iced blades for a considerable period, making timely detection difficult and posing a significant safety risk to the wind farm. In some cases, physical mathematical models are used to simulate natural phenomena and model the icing formation mechanism in order to monitor blade icing. However, the icing formation mechanism is complex and variable, making it difficult to comprehensively simulate the physical formation of blade icing. Neglected factors can become sources of error, thus reducing the accuracy of blade icing predictions. In other cases, mathematical statistics can be used to monitor blade icing and make predictions based on data collected over various time periods. However, due to the instability of weather, the weather conditions in the same month can vary from year to year. This single time-series prediction method of mathematical statistics cannot accurately describe the nonlinear dynamic process of blade icing evolution, also reducing the accuracy of blade icing predictions.

[0030] This application provides a method, apparatus, equipment, and medium for predicting blade icing in wind farms. It utilizes meteorological data from historical blade icing periods and historical non-icing periods to train a blade icing prediction model capable of outputting predictions of blade icing conditions based on meteorological data. The blade icing prediction model, trained using historical meteorological data and the occurrence of blade icing within those periods, accurately describes the nonlinear dynamic process of blade icing evolution. By inputting predicted meteorological data for a predetermined future time period into the blade icing prediction model, a blade icing prediction result characterizing the blade icing situation within that period can be obtained, enabling the prediction of blade icing in wind farms, timely and accurate detection of blade icing, and reducing safety risks to wind farms.

[0031] The first aspect of this application provides a method for predicting icing on wind farm blades, which can be applied to the prediction of icing conditions on wind farm blades. This method for predicting icing on wind farm blades can be executed by a wind farm blade icing prediction device, equipment, etc. Figure 1 A flowchart of a method for predicting wind farm blade icing according to an embodiment of this application is shown below. Figure 1 As shown, the method for predicting icing on wind farm blades may include steps S101 and S102.

[0032] In step S101, the predicted meteorological data of the wind farm for a future preset time period is obtained.

[0033] The number of wind farms to be monitored is not limited here. When it is necessary to monitor a wind farm, the predicted meteorological data of that wind farm for a future preset time period is obtained. When it is necessary to monitor multiple wind farms in a preset area, such as a region or province, the predicted meteorological data of each wind farm for a future preset time period is obtained.

[0034] Forecasted meteorological data includes meteorological data for a predetermined future time period, which may include, but is not limited to, temperature and humidity. In some examples, the forecasted meteorological data may also include one or more of the following: wind speed, precipitation, cloud cover, air pressure, and dew point temperature. The predetermined future time period can be set according to the scenario, needs, experience, etc., and is not limited here. For example, the predetermined future time period may be 3 days in the future. Forecasted meteorological data can be obtained from official meteorological agencies, or it can be predicted by a pre-trained meteorological prediction model. Other channels that can obtain forecasted meteorological data are also within the protection scope of the embodiments of this application, and will not be elaborated here.

[0035] In step S102, the predicted meteorological data is input into the pre-trained blade icing prediction model, which processes the predicted meteorological data and outputs the blade icing prediction results for the wind farm.

[0036] The blade icing prediction model is trained based on historical sample data from the wind farm. This model is a machine learning model and may include, but is not limited to, one or more of Long Short-Term Memory (LSTM) models, XGBoost models, etc. Any blade icing prediction model capable of being trained to output blade icing information based on input meteorological data is within the scope of protection of this application. Historical sample data includes meteorological data from historical icing periods and historical non-icing periods. Historical icing periods include the time intervals during which blade icing occurred at the wind farm, and the meteorological data during these periods reflects the characteristics of the meteorological conditions leading to blade icing. Historical non-icing periods include the time intervals during which blade icing did not occur at the wind farm, and the meteorological data during these periods reflects the characteristics of the meteorological conditions that did not lead to blade icing. The meteorological data may include temperature and humidity. The predicted meteorological data is consistent with the types of meteorological data mentioned above. In some examples, meteorological data may also include one or more of the following: wind speed, precipitation, cloud cover, air pressure, and dew point temperature. Abundant meteorological data can further improve the accuracy of predicting leaf icing. In some examples, historical sample data may also include meteorological data from a first preset duration prior to at least one data collection time in historical icing events. The first preset duration can be set according to the scenario, requirements, experience, etc., for example, the first preset duration may be 8 hours. The labels for the meteorological data from the first preset duration prior to at least one data collection time in historical icing events can be obtained statistically from the leaf icing conditions during that first preset duration. The meteorological data from the first preset duration prior to at least one data collection time in historical icing events can serve as additional historical sample data, increasing the number of historical sample data used to train the leaf icing prediction model, thereby further improving the accuracy and reliability of the trained leaf icing prediction model.

[0037] Historical sample data is labeled, and the labels are used to characterize the blade icing situation corresponding to the meteorological data. Blade icing situation may include, but is not limited to, whether blade icing has occurred at the wind farm and / or the proportion of wind turbine generators in the wind farm that have been shut down due to blade icing. The labels for historical sample data can be obtained based on the blade icing situation of the wind farm collected within the corresponding time period. The blade icing prediction model trained using labeled historical sample data can output a predicted blade icing situation for the wind farm based on the input meteorological data. In this embodiment, by inputting the predicted meteorological data for a future preset time period into the blade icing prediction model, the blade icing prediction result output by the model can be obtained. This blade icing prediction result can be used to characterize the blade icing situation within the future preset time period.

[0038] In this embodiment, a pre-trained blade icing prediction model can be used to process the predicted meteorological data of a wind farm within a preset future time period, thereby obtaining the blade icing prediction result output by the model. The blade icing prediction model is trained based on labeled historical sample data from the wind farm. This historical sample data includes meteorological data during historical icing periods and meteorological data during historical non-icing periods. The labels on the historical sample data characterize the blade icing situation corresponding to the meteorological data. The blade icing prediction model trained on labeled historical sample data from the wind farm can accurately describe the nonlinear dynamic process of blade icing evolution, thus accurately predicting the blade icing situation within a preset future time period based on the predicted meteorological data, enabling timely and accurate detection of blade icing and reducing the safety risks of the wind farm.

[0039] For example, Figure 2 This is a schematic diagram illustrating an example of the wind farm blade icing prediction hit rate provided in an embodiment of this application, as shown below. Figure 2 As shown, the horizontal axis represents the wind farm number, and the vertical axis represents the hit rate. The blade icing prediction method of this application was applied to 51 wind farms, resulting in blade icing predictions for each of the 51 wind farms. The predicted blade icing results were compared with the actual blade icing conditions at the wind farms to obtain the hit rate. The hit rate characterizes the consistency between the predicted blade icing results and the actual blade icing conditions at the wind farms. Specifically, it is the ratio of the accurately predicted blade icing duration to the total duration, where the total duration is the sum of the accurately predicted duration and the inaccurately predicted duration. The hit rate can be calculated as TP / (TP+FP), where TP is the accurately predicted duration of blade icing, and FP is the inaccurately predicted duration. Figure 2It can be seen that, among the 51 wind farms, the blade icing prediction accuracy rate reached over 60% in 98% of the wind farms and over 80% in 94% of the wind farms. The blade icing situation predicted by the wind farm blade icing prediction method in this application embodiment is basically consistent with the actual blade icing situation, and the accuracy of the blade icing prediction can be guaranteed.

[0040] In some embodiments, the blade icing prediction model needs to be trained before it can be used to predict the future blade icing of a wind farm. Figure 3 A flowchart illustrating a method for predicting wind farm blade icing according to another embodiment of this application. Figure 3 and Figure 1 The difference is that, Figure 3 The wind farm blade icing prediction method shown may also include steps S103 to S106.

[0041] In step S103, historical sample data and the proportion of shutdowns due to icing corresponding to the meteorological data in the historical sample data are obtained.

[0042] The icing shutdown ratio is the ratio of the number of wind turbines shut down due to blade icing to the total number of wind turbines in the wind farm. Continuing to operate a wind turbine when blade icing occurs poses a significant safety risk; therefore, the turbine is shut down upon detection of blade icing. The icing shutdown ratio ranges from [0, 1].

[0043] In step S104, a tag is generated based on the icing shutdown ratio.

[0044] In some examples, the value of the label can be equal to the proportion of shutdowns due to icing; that is, the label can represent the proportion of shutdowns due to icing.

[0045] In other examples, the label value can be either a first value or a second value. The first and second values ​​differ: the first value indicates that the wind turbine blades in the wind farm are not iced, while the second value indicates that the blades are iced. If the icing-induced shutdown rate is less than or equal to a preset threshold, the label value is the first value; if the icing-induced shutdown rate is greater than the preset threshold, the label value is the second value. The preset threshold can be set based on the scenario, requirements, experience, etc. For example, if the preset threshold is 0, meaning that more than one wind turbine in the wind farm corresponding to the historical sample data is shut down due to blade icing, the label for the historical sample data is the second value; if no wind turbine in the wind farm corresponding to the historical sample data is shut down due to blade icing, the label for the historical sample data is the first value.

[0046] In step S105, the correspondence between the labels and historical sample data is established.

[0047] Establish a correspondence between labels and historical sample data, so that the historical sample data has labels.

[0048] In step S106, at least a portion of the labeled historical sample data is used as the training dataset to train a machine learning model, thereby obtaining a leaf icing prediction model.

[0049] A leaf icing prediction model can be trained using all labeled historical sample data as the training dataset. Alternatively, a portion of the labeled historical sample data can be selected as the training dataset, and another portion as the test dataset. The leaf icing model can be trained using the training dataset and tested using the test dataset to determine if its accuracy meets the requirements. For example, 80% of the labeled historical sample data can be used as the training dataset, and the remaining 20% ​​as the test dataset.

[0050] In some examples, the meteorological data used to train the leaf icing prediction model during historical icing periods, i.e., the training dataset, may include Category I icing meteorological data, Category II icing meteorological data, and Category III icing meteorological data.

[0051] The first category of icing meteorological data includes meteorological data with corresponding labels indicating blade icing, temperatures below the preset icing temperature, and humidity above the preset icing humidity. The label indicating blade icing indicates that blade icing occurred at the wind farm when the first category of icing meteorological data was collected. The preset icing temperature and preset icing humidity can be used to determine whether icing has begun, whether icing is ongoing, and whether the icing has begun to melt. The preset icing temperature and preset icing humidity can be set according to the scenario, requirements, experience, etc., and are not limited here. For example, the preset icing temperature could be 5℃, and the preset icing humidity could be 70%. Temperatures below the preset icing temperature and humidity above the preset icing humidity indicate that the wind farm was under low-temperature and high-humidity meteorological conditions when the first category of icing meteorological data was collected. The first category of icing meteorological data can be meteorological data during the icing stage.

[0052] The second type of icing meteorological data includes corresponding tags indicating blade icing, tags continuously appearing within a second preset time period indicating blade icing, and meteorological data where the temperature is lower than the preset icing temperature and the humidity is lower than or equal to the preset icing humidity. A tag indicating blade icing means that blade icing occurred at the wind farm when the second type of icing meteorological data was collected. The continuous appearance of tags indicating blade icing within the second preset time period before the collection of the second type of icing meteorological data indicates that blade icing had already occurred at the wind farm before the collection of the second type of icing meteorological data. The second preset time period can be set according to the scenario, requirements, experience, etc., and is not limited here; for example, the second preset time period can be 3 hours. The specific content of the preset icing temperature and preset icing humidity can be found in the relevant descriptions in the above embodiments, and will not be repeated here. A temperature lower than the preset icing temperature and humidity lower than or equal to the preset icing humidity indicates that the wind farm was under low temperature and low humidity meteorological conditions when the second type of icing meteorological data was collected. The second type of icing meteorological data can be meteorological data from a continuous icing phase.

[0053] The third type of icing meteorological data includes meteorological data with corresponding tags indicating blade icing, and data where tags indicating blade icing appear continuously within a third preset time period and the temperature is higher than or equal to a preset icing temperature. A tag indicating blade icing means that blade icing occurred at the time the third type of icing meteorological data was collected. The continuous appearance of tags indicating blade icing within a third preset time period before the collection of the third type of icing meteorological data indicates that blade icing had already occurred at the wind farm before the collection of the third type of icing meteorological data. The third preset time period can be set according to the scenario, requirements, experience, etc., and is not limited here; for example, the third preset time period can be 3 hours. Specific details regarding the preset icing temperature and preset icing humidity can be found in the relevant descriptions in the above embodiments and will not be repeated here. A temperature higher than or equal to the preset icing temperature indicates that the wind farm was under non-low-temperature meteorological conditions when the third type of icing meteorological data was collected. The third type of icing meteorological data can be meteorological data during the icing melting stage.

[0054] The historical meteorological data used to train the leaf icing prediction model includes meteorological data during the freezing stage (Category I icing meteorological data), meteorological data during the continuous icing stage (Category II icing meteorological data), and meteorological data during the icing melting stage (Category III icing meteorological data). This enriches the types of historical sample data used to train the leaf icing prediction model, enabling the model to learn more comprehensive and complete information about the leaf icing stages during training, thereby improving the accuracy and completeness of the trained model.

[0055] In some examples, the blade icing prediction model may include a blade icing classification prediction model and / or a blade icing proportion prediction model. The blade icing classification prediction model is used to predict whether blade icing will occur in a wind farm. When the blade icing prediction model includes a blade icing classification prediction model, the blade icing prediction result may include whether the blades are iced or not. Correspondingly, the labels on the historical sample data used to train the blade icing prediction model can characterize whether the blades are iced or not. The blade icing proportion prediction model is used to predict the proportion of wind turbine generators shut down due to icing in a wind farm. When the blade icing prediction model includes a blade icing proportion prediction model, the blade icing prediction result includes the proportion of wind turbine generators shut down due to icing. Correspondingly, the labels on the historical sample data used to train the blade icing prediction model can characterize the proportion of wind turbine generators shut down due to icing. The proportion of wind turbine generators shut down due to icing is the ratio of the number of wind turbine generators shut down due to blade icing to the total number of wind turbine generators in the wind farm. In some cases, the blade icing classification prediction model can be obtained from the blade icing ratio prediction model. The icing shutdown ratio output by the blade icing ratio prediction model can be converted into a prediction result of whether the blades are iced or not, based on a preset ratio threshold, and used as the output of the blade icing classification prediction model. For example, if the icing shutdown ratio is less than or equal to the preset ratio threshold, the blade icing prediction result output by the blade icing classification prediction model indicates that the blades are not iced; if the icing shutdown ratio is greater than the preset ratio threshold, the blade icing prediction result output by the blade icing classification prediction model indicates that the blades are iced.

[0056] By implementing a blade icing classification prediction model and / or a blade icing ratio prediction model, the blade icing prediction results can represent richer and more flexible information, making it easier to meet the needs of wind farm safety management.

[0057] Because meteorological conditions vary among different wind farms, and these variations can be significant, to avoid the adverse effects of large differences in meteorological conditions on model training, the historical sample data of each wind farm can be normalized before step S106. Specifically, one wind farm can be selected as a standard wind farm from two or more wind farms; the historical sample data of the standard wind farm is normalized to obtain a normalization ratio coefficient; the normalization ratio coefficient is then used to normalize the historical sample data of wind farms other than the standard wind farm. Correspondingly, step S106 can be further refined as follows: at least a portion of the labeled, normalized historical sample data is used as the training dataset for model training to obtain the blade icing prediction model.

[0058] A standard wind farm serves as a reference point. Historical sample data from the standard wind farm can be normalized to ensure that all values ​​fall within a defined range. The scaling factor that keeps the historical sample data within this range is the normalization factor. For example, a scaler, MinMaxScalaer, can be created. This scaler multiplies the historical sample data values ​​from the standard wind farm by the normalization factor, resulting in a maximum value of 1 and a minimum value of 0 for the multiplied historical sample data. Similarly, the historical sample data from other wind farms (excluding the standard wind farm) can be multiplied by the normalization factor to achieve normalization. The resulting historical sample data is the normalized historical sample data.

[0059] By normalizing standard wind farm data, a normalized scaling factor is obtained. This normalized scaling factor is then used to scale historical sample data from other wind farms (excluding the standard wind farm) proportionally. This reduces the adverse effects of significant differences in meteorological conditions between different wind farms on model training, while also preserving the differences between wind farms. Using the normalized historical sample data for model training improves the efficiency of model training and further enhances the accuracy of the resulting blade icing prediction model.

[0060] In some embodiments, the number of wind turbine generators that will be shut down due to icing in the preset area can be obtained based on the blade icing prediction results of multiple wind farms in the preset area, thereby facilitating the adoption of protection strategies for wind farms in the preset area. Figure 4 A flowchart illustrating a method for predicting wind farm blade icing, provided in yet another embodiment of this application. Figure 4 and Figure 1 The difference is that, Figure 4 The wind farm blade icing prediction method shown also includes steps S107 to S109.

[0061] In step S107, based on the blade icing prediction results of the wind farm in the preset area and the number of wind turbine generators in the wind farm in the preset area, the predicted number of wind turbine generators that will be shut down due to icing in the wind farm in the preset area is calculated.

[0062] The blade icing prediction result includes the icing shutdown ratio. For details on the icing shutdown ratio, please refer to the relevant descriptions in the above embodiments, which will not be repeated here. The preset area can be set according to the scenario, requirements, experience, etc. The preset area can be a certain region that needs to be monitored, such as a province or a city. The product of the number of wind turbine generators in the wind farm and the icing shutdown ratio in the blade icing prediction result corresponding to the wind farm is the number of wind turbine generators in the wind farm that are shut down due to icing.

[0063] In step S108, the number of wind turbine generators that are shut down due to icing in the preset area is determined based on the predicted number of wind turbine generators that are shut down due to icing in the preset area.

[0064] The sum of the number of wind turbine generators that stopped due to icing in each wind farm within the preset area is the total number of wind turbine generators that stopped due to icing within the preset area.

[0065] In some examples, the predicted icing shutdown capacity within a preset area can be calculated based on the number of wind turbines shut down due to icing and the capacity of the wind turbines in the wind farm. The predicted icing shutdown capacity is the sum of the capacities of the wind turbines shut down due to icing in the preset area. The accuracy of the icing shutdown capacity predicted by the wind farm blade icing prediction method in this embodiment is also high. For example, Figure 5 This is a schematic diagram comparing the actual icing-induced shutdown capacity and the predicted icing-induced shutdown capacity in a certain region, provided as an example of an embodiment of this application. The horizontal axis represents time, and the vertical axis represents the shutdown capacity, in megawatts (MW). The actual icing-induced shutdown capacity is the icing-induced shutdown capacity determined by actual observation, while the predicted icing-induced shutdown capacity is the icing-induced shutdown capacity predicted using the wind farm blade icing prediction method provided in this embodiment of the application. Figure 5 As shown, during the initial icing phase, specifically January 14th and 15th, the blades began to ice over, causing a rapid increase in the region's outage capacity, which reached its peak. During the middle phase of continuous icing, specifically January 16th and 17th, due to meteorological conditions such as temperature and humidity consistently meeting icing requirements, a high proportion of wind turbines in the region were out of service. During the later phase of icing melting, specifically from January 18th to 21st, as the icing melted, the region's outage capacity rapidly decreased to zero. The wind farm blade icing prediction method provided in this application embodiment performs well, with the actual icing outage capacity and the predicted icing outage capacity in the region basically matching. Especially during the initial icing phase and the middle phase of continuous icing, the prediction accuracy of the wind farm blade icing prediction method in this application embodiment is high, providing effective decision-making guidance for wind farms and the power grid, ensuring the safe and stable operation of wind farms and the power grid.

[0066] In step S109, a unit protection strategy corresponding to the number of wind turbine generators shut down due to icing within a preset area is executed.

[0067] The number of wind turbine generators shut down due to icing within a preset area can guide the control strategy of the wind farm within that area. A pre-defined correspondence between the number of wind turbine generators shut down due to icing and the corresponding turbine protection strategy can be established. Once the predicted number of wind turbine generators shut down due to icing within the preset area is obtained, the corresponding turbine protection strategy can be executed. The turbine protection strategy can be used to maintain the safe and stable operation of the wind farm and the power grid connected to it. For example, the turbine protection strategy may include, but is not limited to, the handling and adjustment of wind turbine generators in the wind farm. Other strategies that can maintain the safe and stable operation of the wind farm and the power grid connected to it in the event of icing shutdowns are also within the scope of protection of this application's embodiments.

[0068] By predicting the number of wind turbines that will shut down due to icing within a preset area, the icing shutdown situation in the preset area can be known in advance. This allows for the implementation of appropriate turbine protection strategies, effectively reducing economic losses caused by blade icing, extending the service life of wind turbines, and ensuring the safe operation of electronic systems, including those of wind farms and power grids.

[0069] The second aspect of this application provides a device for predicting icing on wind farm blades. Figure 6 This is a schematic diagram of the structure of a wind farm blade icing prediction device provided in an embodiment of this application, as shown below. Figure 6 As shown, the wind farm blade icing prediction device 200 may include a data acquisition module 201 and a model processing module 202.

[0070] The data acquisition module 201 can be used to acquire the predicted meteorological data of the wind farm in the future within a preset time period. The predicted meteorological data includes temperature and humidity.

[0071] The model processing module 202 can be used to input the predicted meteorological data into the pre-trained blade icing prediction model, and the blade icing prediction model processes the predicted meteorological data to output the blade icing prediction results of the wind farm.

[0072] The blade icing prediction results are used to characterize the blade icing situation within a predetermined future time period. The blade icing prediction model is trained based on historical sample data from the wind farm. The historical sample data includes meteorological data from historical icing periods and historical non-icing periods. The meteorological data includes temperature and humidity. The historical sample data is labeled, and the labels are used to characterize the blade icing situation corresponding to the meteorological data.

[0073] In some examples, the predicted meteorological data is consistent with the types of meteorological data, which also include one or more of the following: wind speed, precipitation, cloud cover, air pressure, and dew point temperature.

[0074] In some examples, historical sample data also includes meteorological data for a first preset duration prior to at least one data collection time in the historical icing period.

[0075] In this embodiment, a pre-trained blade icing prediction model can be used to process the predicted meteorological data of a wind farm within a preset future time period, thereby obtaining the blade icing prediction result output by the model. The blade icing prediction model is trained based on labeled historical sample data from the wind farm. This historical sample data includes meteorological data during historical icing periods and meteorological data during historical non-icing periods. The labels on the historical sample data characterize the blade icing situation corresponding to the meteorological data. The blade icing prediction model trained on labeled historical sample data from the wind farm can accurately describe the nonlinear dynamic process of blade icing evolution, thus accurately predicting the blade icing situation within a preset future time period based on the predicted meteorological data, enabling timely and accurate detection of blade icing and reducing the safety risks of the wind farm.

[0076] In some embodiments, the wind farm blade icing prediction device 200 may further include a model training module.

[0077] The model training module can be used to: acquire historical sample data and the icing shutdown ratio corresponding to the meteorological data in the historical sample data, the icing shutdown ratio including the ratio of the number of wind turbines shut down due to blade icing to the total number of wind turbines in the wind farm; generate labels based on the icing shutdown ratio; establish the correspondence between the labels and the historical sample data; and use at least a portion of the labeled historical sample data as the training dataset to train the machine learning model to obtain the blade icing prediction model.

[0078] In some examples, the meteorological data used to train the leaf icing prediction model during historical icing periods includes three types of icing meteorological data: Type 1, Type 2, and Type 3. Type 1 icing meteorological data includes meteorological data with corresponding labels representing leaf icing, temperatures below a preset icing temperature, and humidity above a preset icing humidity. Type 2 icing meteorological data includes meteorological data with corresponding labels representing leaf icing, labels representing leaf icing that have appeared continuously within a previous second preset time period, and temperatures below the preset icing temperature with humidity below or equal to the preset icing humidity. Type 3 icing meteorological data includes meteorological data with corresponding labels representing leaf icing, labels representing leaf icing that have appeared continuously within a previous third preset time period, and temperatures above or equal to the preset icing temperature.

[0079] In some examples, the leaf icing prediction model includes a leaf icing classification prediction model and / or a leaf icing proportion prediction model.

[0080] When the leaf icing prediction model includes a leaf icing classification prediction model, the leaf icing prediction result includes whether the leaf is iced or not, and the label represents whether the leaf is iced or not.

[0081] When the blade icing prediction model includes a blade icing ratio prediction model, the blade icing prediction results include the icing shutdown ratio. The label represents the icing shutdown ratio, which is the ratio of the number of wind turbines shut down due to blade icing to the total number of wind turbines in the wind farm.

[0082] In some embodiments, the blade icing prediction result includes an icing shutdown ratio, which is the ratio of the number of wind turbines shut down due to blade icing in the wind farm to the total number of wind turbines in the wind farm. The wind farm blade icing prediction device 200 may further include a region protection module. The region protection module can be used to: calculate the predicted number of wind turbines shut down due to icing in the preset region based on the blade icing prediction result and the number of wind turbines in the preset region; determine the number of wind turbines shut down due to icing in the preset region based on the predicted number of wind turbines shut down due to icing in the preset region; and execute a turbine protection strategy corresponding to the number of wind turbines shut down due to icing in the preset region.

[0083] In some embodiments, the wind farm blade icing prediction device 200 may further include a normalization processing module. The normalization processing module may be used to: select one wind farm as a standard wind farm from two or more wind farms; normalize the historical sample data of the standard wind farm and obtain a normalization ratio coefficient; and use the normalization ratio coefficient to normalize the historical sample data of wind farms other than the standard wind farm.

[0084] Correspondingly, the above-mentioned model training module can be specifically used to: use at least a portion of the labeled, normalized historical sample data as the training dataset for model training.

[0085] The third aspect of this application also provides a device for predicting icing on wind farm blades. Figure 7 This is a schematic diagram of a wind farm blade icing prediction device provided in one embodiment of this application. Figure 7 As shown, the wind farm blade icing prediction device 300 includes a memory 301, a processor 302, and a computer program stored in the memory 301 and capable of running on the processor 302.

[0086] In some examples, the processor 302 described above may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits that may be configured to implement the embodiments of this application.

[0087] Memory 301 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the wind farm blade icing prediction method according to embodiments of this application.

[0088] The processor 302 reads the executable program code stored in the memory 301 to run the computer program corresponding to the executable program code, so as to implement the wind farm blade icing prediction method in the above embodiment.

[0089] In some examples, the wind farm blade icing prediction device 300 may also include a communication interface 303 and a bus 304. For example, Figure 7 As shown, the memory 301, processor 302, and communication interface 303 are connected through bus 304 and complete communication with each other.

[0090] The communication interface 303 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application. Input devices and / or output devices can also be connected through the communication interface 303.

[0091] Bus 304 includes hardware, software, or both, that couples the components of the wind farm blade icing prediction device 300 together. For example, and not limitingly, bus 304 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-E) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0092] A fourth aspect of this application also provides a computer-readable storage medium storing computer program instructions. When executed by a processor, these computer program instructions can implement the wind farm blade icing prediction method described in the above embodiments and achieve the same technical effect. To avoid repetition, further details are omitted here. The aforementioned computer-readable storage medium may include non-transitory computer-readable storage media, such as read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, etc., and is not limited thereto.

[0093] This application provides a computer program product. When the instructions in the computer program product are executed by the processor of an electronic device, the electronic device performs the wind farm blade icing prediction method in the above embodiment and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0094] It should be clarified that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. For the device embodiments, equipment embodiments, computer-readable storage medium embodiments, and computer program product embodiments, the relevant parts can be referred to the description section of the method embodiments. This application is not limited to the specific steps and structures described above and shown in the figures. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application. Furthermore, for the sake of brevity, detailed descriptions of known methods and techniques are omitted here.

[0095] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0096] Those skilled in the art will understand that the above embodiments are exemplary and not restrictive. Different technical features appearing in different embodiments can be combined to achieve beneficial effects. Based on a study of the drawings, specification, and claims, those skilled in the art should be able to understand and implement other variations of the disclosed embodiments. In the claims, the term "comprising" does not exclude other means or steps; the quantifier "a" does not exclude a plurality; the terms "first" and "second" are used to identify names and not to indicate any particular order. No reference numerals in the claims should be construed as limiting the scope of protection. The functionality of multiple parts appearing in the claims can be implemented by a single hardware or software module. The appearance of certain technical features in different dependent claims does not mean that these technical features cannot be combined to achieve beneficial effects.

Claims

1. A method of predicting blade icing for a wind farm, characterized in that, The method comprises: obtaining predicted meteorological data of a wind farm in a future preset time period, the predicted meteorological data comprising temperature and humidity; inputting the predicted meteorological data into a pre-trained blade icing prediction model, processing the predicted meteorological data by the blade icing prediction model, and outputting a blade icing prediction result of the wind farm, the blade icing prediction result being used to represent a blade icing condition in the future preset time period, the blade icing prediction result comprising blade icing or no blade icing, and an icing shutdown ratio, wherein the blade icing prediction model is a machine learning model, the blade icing prediction model being trained according to historical sample data of the wind farm, the historical sample data comprising meteorological data in a historical icing time and meteorological data in a historical non-icing time, the meteorological data comprising temperature and humidity, the historical sample data having a label, the label being used to represent a blade icing condition corresponding to the meteorological data, the historical meteorological data in the historical icing time used to train the blade icing prediction model comprising first-type icing meteorological data in an icing stage, second-type icing meteorological data in a continuous icing stage, and third-type icing meteorological data in an icing melting stage; The method further comprises: selecting one wind farm as a standard wind farm among two or more wind farms; normalizing the historical sample data of the standard wind farm and obtaining a normalization proportion coefficient; normalizing the historical sample data of the wind farms other than the standard wind farm by using the normalization proportion coefficient.

2. The method of claim 1, wherein, Before the predicted meteorological data is input into the pre-trained blade icing prediction model, the method further comprises: obtaining the historical sample data and an icing shutdown ratio corresponding to the meteorological data in the historical sample data, the icing shutdown ratio comprising a ratio of the number of wind turbine generators stopped due to blade icing to the number of wind turbine generators in the wind farm; generating the label according to the icing shutdown ratio; establishing a correspondence between the label and the historical sample data; performing machine learning model training on at least part of the historical sample data with the label as a training data set to obtain the blade icing prediction model.

3. The method of claim 1, wherein, The historical sample data further comprises meteorological data in a first preset time period before at least one data acquisition time in the historical icing time.

4. The method of claim 1, wherein, The historical meteorological data used to train the blade icing prediction model comprises first-type icing meteorological data, second-type icing meteorological data, and third-type icing meteorological data, The first icing weather data includes weather data corresponding to the label indicating blade icing, temperature lower than a preset icing temperature, and humidity higher than a preset icing humidity. The second icing weather data includes weather data corresponding to the label indicating blade icing, the label indicating blade icing appearing continuously within a second preset time period, temperature lower than the preset icing temperature, and humidity lower than or equal to the preset icing humidity. The third icing weather data includes weather data corresponding to the label indicating blade icing, the label indicating blade icing appearing continuously within a third preset time period, and temperature higher than or equal to the preset icing temperature.

5. The method of claim 1, wherein, The blade icing prediction model includes a blade icing classification prediction model and / or a blade icing proportion prediction model. In a case where the blade icing prediction model includes the blade icing classification prediction model, the blade icing prediction result includes blade icing or no blade icing, and the label indicates blade icing or no blade icing. In a case where the blade icing prediction model includes the blade icing proportion prediction model, the blade icing prediction result includes the icing shutdown proportion, and the label indicates the icing shutdown proportion, the icing shutdown proportion including a ratio of a number of wind generating sets in a wind farm that are shutdown due to blade icing to a number of wind generating sets in the wind farm.

6. The method of claim 1, wherein, The blade icing prediction result includes an icing shutdown proportion, the icing shutdown proportion including a ratio of a number of wind generating sets in a wind farm that are shutdown due to blade icing to a number of wind generating sets in the wind farm, After the predicted weather data is input into a blade icing prediction model trained in advance, the predicted weather data is processed by the blade icing prediction model, and a blade icing prediction result of a wind farm is output, the method further includes: According to the blade icing prediction result of the wind farm in the preset area and the number of wind generating sets in the wind farm in the preset area, a number of wind generating sets in the wind farm in the preset area that are predicted to be shutdown due to icing is calculated; According to the number of wind generating sets in the wind farm in the preset area that are predicted to be shutdown due to icing, a number of wind generating sets in the preset area that are shutdown due to icing is determined; A unit protection strategy corresponding to the number of wind generating sets in the preset area that are shutdown due to icing is executed.

7. The method of claim 2, wherein, The model training includes: The model training includes:

8. The method according to any one of claims 1 to 7, characterized in that, The predicted weather data is consistent with a type of weather data, and the weather data further includes one or more of the following: Wind speed, precipitation, cloud cover, air pressure, and dew point temperature.

9. A device for predicting icing of blades of a wind farm, characterized in that The method includes: A data acquisition module is configured to acquire predicted weather data of a wind farm within a future preset time period, the predicted weather data including temperature and humidity. The model processing module is configured to input the predicted meteorological data into a blade icing prediction model that is pre-trained, process the predicted meteorological data by the blade icing prediction model, and output a blade icing prediction result of the wind farm, the blade icing prediction result being used to represent a blade icing condition in a future preset time period, the blade icing prediction result including blade icing or no blade icing, and an icing shutdown ratio. The blade icing prediction model is a machine learning model, and is trained according to historical sample data of the wind farm, the historical sample data including meteorological data in a historical icing time and meteorological data in a historical non-icing time, the meteorological data including temperature and humidity, the historical sample data having a label used to represent a blade icing condition corresponding to the meteorological data, the historical sample data in the historical icing time used to train the blade icing prediction model including first icing meteorological data in an icing stage, second icing meteorological data in a continuous icing stage, and third icing meteorological data in an icing melting stage. The normalization processing module is configured to select one wind farm as a standard wind farm in two or more wind farms, normalize the historical sample data of the standard wind farm, and obtain a normalization proportion coefficient, and normalize the historical sample data of wind farms other than the standard wind farm by using the normalization proportion coefficient.

10. A device for predicting icing of blades of a wind farm, characterized in that, The computer program instructions are executed by the processor to implement the method for predicting blade icing of a wind farm according to any one of claims 1 to 8. The computer program instructions are executed by the processor to implement the method for predicting blade icing of a wind farm according to any one of claims 1 to 8. The computer program instructions are executed by the processor to implement the method for predicting blade icing of a wind farm according to any one of claims 1 to 8.

11. A computer readable storage medium, characterized in that, ​

Images