Fan blade icing prediction method and device and medium

Abstract

The invention discloses a fan blade icing prediction method and device and a medium, and relates to the technical field of equipment monitoring, and the method comprises the steps: obtaining historical meteorological data, blade tip speed, pitch angle and historical icing data, and taking the data as real samples; performing data expansion on the real sample by using the generative adversarial network model to obtain a generated sample; the real samples and the generated samples form a training sample set, the training sample set is used for training the long-short-term memory neural network, and a fan blade icing prediction model is obtained to predict the icing condition of the fan blade. According to the method, the blade tip speed and the pitch angle which reflect relative motion of the blade and the supercooled liquid drop are added in training data, a prediction model obtained through training comprises icing phase change characteristics and relative motion characteristics of icing of a rotating object, description of the icing process of the fan blade is more accurate, and prediction is more accurate; the adversarial generative network model is adopted to expand the training sample set of the prediction model, the problem of insufficient real samples is solved, and the prediction accuracy is further improved.

Description

Technical Field

[0001] The present application relates to the technical field of equipment monitoring, and in particular to a method, device and medium for predicting icing on wind turbine blades. Background Art

[0002] When wind turbines operate under specific meteorological conditions such as low temperature, rain and snow, blade icing may occur. Icing will deteriorate the aerodynamic performance of the blades, resulting in a decrease in wind energy capture capability, and in severe cases, causing grid disconnection and shutdown. In addition, icing will also cause an imbalance in the aerodynamic load on the blades, causing fatigue damage to unit components, and in severe cases, causing tower deformation or even tower collapse accidents. Therefore, research on wind turbine blade icing prediction technology is very important for the safe and efficient operation of wind turbines.

[0003] The freezing of fan blades is a typical phenomenon of freezing on moving surfaces. The freezing principle is that supercooled water droplets hit the blade surface and freeze. At present, there are two technical routes for predicting the icing of fan blades. One is to identify the data anomalies of the fan output power, speed and other data acquisition and monitoring control (SCADA) system when the blades are frozen, and combine the meteorological conditions at the time to determine whether the fan blades are frozen. However, when the blades are slightly frozen, the operating state of the fan will not be greatly affected. When the blades can be determined to be frozen through the analysis of the fan SCADA data, the blades have already frozen more seriously. This method cannot make timely predictions and judgments at the early stage of blade freezing. Another method is to predict the probability and degree of blade freezing by using a neural network model trained with historical meteorological data and blade freezing records. This method does not consider the solid-liquid collision behavior that affects the freezing of fan blades, and the freezing of fan blades is an occasional event. There is a lack of sufficient samples to train the prediction model. Therefore, the accuracy of using the model to predict blade freezing is not high.

[0004] Therefore, how to improve the accuracy of wind turbine blade icing prediction is a problem that technical personnel in this field need to solve. Summary of the invention

[0005] The purpose of the present application is to provide a method, device and medium for predicting icing on wind turbine blades, so as to solve the problem of low accuracy in predicting blade icing using a model.

[0006] In order to solve the above technical problems, the present application provides a method for predicting wind turbine blade icing, comprising:

[0007] Acquire historical meteorological data, historical wind turbine operation data and historical freezing data during the freezing period and before the freezing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle;

[0008] Performing data expansion on the real sample using a generative adversarial network model to obtain a generated sample;

[0009] The real samples and the generated samples form a training sample set, the meteorological data and the wind turbine operation data in the training sample set are used as input features, and the icing data in the training sample set is used as output features;

[0010] The input features and the output features are used to train a long short-term memory neural network to obtain a fan blade icing prediction model, and the fan blade icing prediction model is used to predict the icing condition of the fan blades in a future time interval.

[0011] Optionally, before using the fan blade icing prediction model to predict the icing condition of the fan blades in a future time interval, the method further includes:

[0012] Obtaining mesoscale numerical weather forecast results to determine the probability of precipitation in the target area within the future time interval;

[0013] Obtain weather elements for the formation of supercooled water droplets monitored by ground meteorological observation equipment at the site;

[0014] Based on the precipitation probability and the weather elements, using a priori knowledge base to determine whether supercooled precipitation will occur in the future time interval;

[0015] If supercooling precipitation occurs, the step of using the fan blade icing prediction model to predict the icing condition of the fan blades in a future time interval is entered.

[0016] Optionally, the ground meteorological observation equipment at the site includes: a microwave radiometer radar for monitoring the temperature and humidity of the atmosphere, a comprehensive cloud meter for monitoring cloud indicators, and a wind profiler radar for detecting wind speed and turbulence conditions at different altitudes; wherein the cloud indicators include the location, thickness and cloud height of cumulonimbus clouds.

[0017] Optionally, using a generative adversarial network model to perform data augmentation on the real sample to obtain a generated sample includes:

[0018] A conditional adversarial generative network model is used as a data generator to generate simulated data that conforms to the characteristics of the original data and prior knowledge to expand the real samples and obtain the generated samples.

[0019] Optionally, the training process of the conditional adversarial generation network model includes:

[0020] Conditional information is constructed based on the prior knowledge base of wind turbine blade icing, and the generator and discriminator are composed of convolutional layers, instance normalization and activation functions;

[0021] randomly extracting noise from the real sample, inputting the noise and the condition information into the generator, and generating a sample using the output of the generator;

[0022] Using the discriminator to judge the real sample and the generated sample respectively, and calculating the first loss of the discriminator;

[0023] Updating a first parameter of the discriminator to minimize the first loss;

[0024] Fixing the first parameter, updating the second parameter of the generator to minimize the second loss of the generator;

[0025] The above steps are repeated until convergence to obtain a trained conditional adversarial generation network model.

[0026] Optionally, the historical meteorological data include near-ground temperature, humidity, wind speed, wind direction, and rainfall intensity; the historical freezing data include freezing rate and freezing thickness.

[0027] Optionally, the process of determining the freezing rate includes:

[0028] Use the ice sensor on the windward side of the leading edge of the wind turbine blade to obtain ice thickness data in real time;

[0029] Determine the time data corresponding to the ice thickness data and construct an ice thickness-time curve;

[0030] The freezing rate is determined based on the ice thickness-time curve.

[0031] The present application also provides a wind turbine blade icing prediction device, comprising:

[0032] A real sample acquisition module is used to acquire historical meteorological data, historical wind turbine operation data and historical freezing data during the freezing period and the pre-freezing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle;

[0033] A data expansion module, used to perform data expansion on the real sample using a generative adversarial network model to obtain a generated sample;

[0034] A training sample set determination module, used for forming a training sample set with the real sample and the generated sample, taking the meteorological data and the fan operation data in the training sample set as input features, and taking the icing data in the training sample set as output features;

[0035] The prediction module is used to train the long short-term memory neural network using the input features and the output features to obtain a fan blade icing prediction model, and use the fan blade icing prediction model to predict the icing condition of the fan blades in a future time interval.

[0036] The present application also provides a wind turbine blade icing prediction device, comprising a memory for storing a computer program;

[0037] A processor is used to implement the steps of the wind turbine blade icing prediction method when executing the computer program.

[0038] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the wind turbine blade icing prediction method are implemented.

[0039] The present application provides a method for predicting icing on wind turbine blades, comprising: obtaining historical meteorological data, historical wind turbine operation data and historical icing data during the icing period and before the icing period as real samples; wherein the historical wind turbine operation data include blade tip speed and blade pitch angle; performing data expansion on the real samples using a generative adversarial network model to obtain generated samples; the real samples and the generated samples form a training sample set, the meteorological data and wind turbine operation data in the training sample set are used as input features, and the icing data in the training sample set is used as output features; the input features and the output features are used to train a long short-term memory neural network to obtain a fan blade icing prediction model, and the fan blade icing prediction model is used to predict the icing condition of the fan blades in a future time period. The present application adds the tip speed and pitch angle that reflect the relative motion between the blades and the supercooled droplets to the training data. The fan blade icing prediction model obtained by training includes both the icing phase change characteristics and the relative motion characteristics of the icing of rotating objects. The description of the physical process of fan blade icing is more precise, making the prediction more accurate. The adversarial generative network model is used to expand the training sample set of the fan blade icing prediction model, which solves the problem of insufficient real samples caused by accidental events, improves the training effect of the prediction model, and thus further improves the prediction accuracy.

[0040] The beneficial effects of a wind turbine blade icing prediction device and medium provided in the present application correspond to the method, and the effects are as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] FIG1 is a flow chart of a method for predicting icing on wind turbine blades provided in an embodiment of the present application;

[0043] FIG2 is a diagram of a wind turbine blade icing simulation data generation architecture provided by an embodiment of the present application;

[0044] FIG3 is a structural diagram of a wind turbine blade icing prediction device provided in an embodiment of the present application;

[0045] FIG. 4 is a structural diagram of another wind turbine blade icing prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0047] The core of the present application is to provide a method, device and medium for predicting icing on wind turbine blades, so as to improve the accuracy of predicting icing on wind turbine blades.

[0048] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0049] FIG1 is a flow chart of a method for predicting icing on wind turbine blades provided in an embodiment of the present application. As shown in FIG1 , the method for predicting icing on wind turbine blades includes:

[0050] S10: Acquire historical meteorological data, historical wind turbine operation data and historical freezing data during the freezing period and before the freezing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle.

[0051] S11: Use the generative adversarial network model to perform data expansion on real samples to obtain generated samples.

[0052] S12: The real samples and the generated samples form a training sample set, the meteorological data and the wind turbine operation data in the training sample set are used as input features, and the icing data in the training sample set is used as output features.

[0053] S13: Using the input features and the output features to train the long short-term memory neural network to obtain a fan blade icing prediction model, and using the fan blade icing prediction model to predict the icing condition of the fan blades in a future time interval.

[0054] In step S10, the embodiment of the present application does not limit the meteorological data, fan operation data and icing data; icing of fan blades is a typical icing phenomenon on a moving surface. The relative motion between the fan blades and the supercooled droplets affects the speed and direction of the droplets spreading after hitting the surface, which in turn significantly affects the speed and degree of icing on the blade surface. Historical fan operation data includes blade tip speed and blade pitch angle; historical meteorological data includes near-ground temperature, humidity, wind speed, wind direction, and rainfall intensity; historical icing data includes icing rate and icing thickness.

[0055] Furthermore, the process of determining the icing rate includes: using the icing sensor on the windward surface of the leading edge of the wind turbine blade to obtain ice thickness data in real time; determining the time data corresponding to the ice thickness data, and constructing an ice thickness-time curve; and determining the icing rate based on the ice thickness-time curve. Among them, the icing sensor can be a MEMS strain-type icing sensor that can detect icing with a thickness of 0.1mm-6mm. Micro-Electro-Mechanical System (MEMS), also known as micro-electro-mechanical system, microsystem, micromachine, etc., refers to high-tech devices with a size of a few millimeters or even smaller.

[0056] In step S11, the icing of wind turbine blades is an extreme event, and the number of real samples is limited. Directly using real samples to train the prediction model will cause problems such as model overfitting or non-convergence, which will affect the prediction effect of the prediction model. Therefore, it is necessary to expand and enhance the small sample data set before training the prediction model. Using the Generative Adversarial Network (GAN) model to generate data to supplement the training sample set can help increase data diversity and improve the accuracy and generalization ability of the neural network model.

[0057] In step S12, the wind turbine operation data includes blade tip speed and blade pitch angle; the meteorological data includes near-ground temperature, humidity, wind speed, wind direction, and rainfall intensity; and the icing data includes icing rate and icing thickness.

[0058] In step S13, the factors affecting the freezing of the fan blades should not only consider the meteorological information at the current moment, but also the weather change trend in the past period of time. Therefore, a long short-term memory neural network (LSTM) that can capture the dependency of time series data is used as a fan blade freezing prediction model. The prediction model that can reflect both the freezing phase change characteristics and the relative motion characteristics of the blade surface and the supercooled droplets is trained using meteorological data, fan operation data and freezing data. Using the fan blade freezing prediction model to predict the icing condition of the fan blades in the future time interval specifically includes: inputting the meteorological data and fan operation data at the current moment and a period of time before the current moment (which can be 24 hours before the current moment) into the fan blade freezing prediction model, and using the fan blade freezing prediction model to predict the icing condition of the fan blades in the future time interval, which can be the freezing rate and ice thickness of the fan blades in the next 4 hours.

[0059] A method for predicting icing on wind turbine blades provided in an embodiment of the present application includes: obtaining historical meteorological data, historical wind turbine operation data and historical icing data during the icing period and the pre-icing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle; using an adversarial generative network model to perform data expansion on the real samples to obtain generated samples; the real samples and the generated samples form a training sample set, the meteorological data and wind turbine operation data in the training sample set are used as input features, and the icing data in the training sample set is used as output features; the input features and the output features are used to train a long short-term memory neural network to obtain a fan blade icing prediction model, and the fan blade icing prediction model is used to predict the icing condition of the fan blades in a future time period. The present application adds the tip speed and pitch angle that reflect the relative motion between the blades and the supercooled droplets to the training data. The fan blade icing prediction model obtained by training includes both the icing phase change characteristics and the relative motion characteristics of the icing of rotating objects. The description of the physical process of fan blade icing is more precise, making the prediction more accurate. The adversarial generative network model is used to expand the training sample set of the fan blade icing prediction model, which solves the problem of insufficient real samples caused by accidental events, improves the training effect of the prediction model, and thus further improves the prediction accuracy.

[0060] Based on the above embodiments, before the embodiment of the present application uses the fan blade icing prediction model to predict the icing condition of the fan blades in the future time interval, it also includes: obtaining the results of a mesoscale numerical weather forecast to determine the precipitation probability in the target area within the future time interval; obtaining weather elements formed by supercooled water droplets monitored by ground meteorological observation equipment at the site; based on the precipitation probability and weather elements, using a priori knowledge base to determine whether supercooled precipitation will occur in the future time interval; if supercooled precipitation occurs, entering the step of using the fan blade icing prediction model to predict the icing condition of the fan blades in the future time interval.

[0061] The rime formed by the collision of large-diameter supercooled raindrops with wind turbine blades has the characteristics of hardness and high adhesion pressure, and is the most serious type of icing that affects the safe and efficient operation of wind turbines. In large-scale weather systems, the generation of supercooled raindrops can be explained as the formation of a vertical atmospheric structure of "ice crystal layer-warm layer-cold layer" from top to bottom when cold air meets warm and humid air currents; the cold layer refers to the air layer with a temperature below 0°C about 2km from the ground, and the warm layer refers to the air layer with a temperature above 0°C about 2-4km from the ground, and above that is the ice crystal layer with a temperature below 0°C. Numerical weather forecasting is to predict the state of atmospheric movement and weather phenomena in the future by solving the fluid mechanics and thermodynamics equations that describe the weather evolution process through numerical calculations under certain initial values ​​and boundary conditions. This application receives the results of mesoscale numerical weather forecasts to obtain the probability of precipitation in the region in the future, and then uses the ground meteorological observation equipment at the site to monitor the temperature and humidity of the atmosphere, the position, thickness, cloud height and wind profile of cumulonimbus clouds and other weather factors that affect the formation of supercooled water droplets, and uses the prior knowledge base to determine whether supercooled precipitation will occur, and issues a site-level supercooled precipitation warning. It can be determined by using the prior knowledge base to determine whether the precipitation probability and weather factors are greater than their corresponding thresholds, and then make a judgment on supercooled precipitation.

[0062] The station's ground meteorological observation equipment includes: microwave radiometer radar for monitoring atmospheric temperature and humidity, integrated cloud meter for monitoring cloud indicators, and wind profiler radar for detecting wind speed and turbulence conditions at different altitudes; among them, cloud indicators include cumulonimbus cloud location, thickness and cloud height.

[0063] The embodiment of the present application analyzes the results of numerical weather forecasts + station meteorological observation data, and adopts a method of determining the icing prior knowledge base of fan blades to predict the time when supercooled droplets are generated and the period when fan blades may be covered with ice. If supercooled precipitation occurs in the future, the fan blade icing prediction model is used to predict the icing rate and thickness of the fan blades in the future time period.

[0064] Based on the above embodiments, the embodiments of the present application use a generative adversarial network model to perform data expansion on real samples to obtain generated samples, including: using a conditional generative adversarial network model as a data generator to generate simulated data that conforms to the characteristics of the original data and prior knowledge to expand the real samples to obtain generated samples.

[0065] The training process of the conditional adversarial generative network model includes: constructing conditional information based on the prior knowledge base of wind turbine blade icing, using convolutional layers, instance normalization and activation functions to form a generator and a discriminator; randomly extracting noise from real samples, inputting the noise and conditional information into the generator, and using the generator output to generate samples; using the discriminator to judge the real samples and the generated samples respectively, and calculating the first loss of the discriminator; updating the first parameter of the discriminator to minimize the first loss; fixing the first parameter, updating the second parameter of the generator to minimize the second loss of the generator; repeating the above steps until convergence to obtain a trained conditional adversarial generative network model.

[0066] FIG2 is a schematic diagram of a wind turbine blade icing simulation data generation architecture provided by an embodiment of the present application. As shown in FIG2, condition information c is constructed according to a priori knowledge base of wind turbine blade icing. The near-ground temperature range of wind turbine blade icing is -10°C to 0°C, the wind speed is between 2m / s and 10m / s, and the rainfall intensity does not exceed 0.5g / m 3 , the ice thickness is between 0mm and 6mm. The generator G and the discriminator D are composed of convolutional layers, instance normalization and activation functions. Training process: 1) Randomly extract noise z and conditional information c and input them into the generator G; 2) The generator G generates samples G(z|c); 3) The discriminator D judges the real sample (X,c) and the generated sample (G(z|c),c) respectively, and calculates the loss E_{(X,c)P(X,c)}[log(D(X,c))]- E_{zP(z)}[log(1-D(G(z|c), c)))]; 4) Update the parameters of the discriminator D to minimize the loss of D; 5) Fix the parameters of the discriminator D and update the parameters of the generator G to minimize the loss of G log(D(G(z|c), c)); 6) Repeat the above steps until convergence. Use the trained conditional generative adversarial network (CGAN) model to generate wind turbine blade icing simulation data as generated samples. Due to the conditional constraints formed by prior knowledge, it has higher quality and effectiveness. Combining real samples and generated samples into a training sample set can provide richer and more reliable samples for subsequent prediction model training, which is conducive to improving the accuracy and generalization ability of the wind turbine blade icing prediction model.

[0067] This application starts from the essence of the icing phenomenon of wind turbine blades, targets typical meteorological conditions for supercooled precipitation, uses numerical weather forecasts, and combines meteorological observations at the site to predict the period when supercooled precipitation may occur, so as to provide advance warning. The Generative Adversarial Network (GAN) model is trained using historical meteorological data, blade tip speed, blade pitch angle, and historical icing data during the icing period and before the icing period. The GAN model is used to generate more generated samples, which are then combined with real samples to form a training sample set with a sufficient sample size, and then the long short-term memory neural network (LSTM) model is trained to predict the icing rate and thickness of wind turbine blades. Training and using the Generative Adversarial Network (GAN) model as a data generator can solve the problem of insufficient real samples for wind turbine blade icing as an occasional event. The introduction of input features such as blade tip speed and blade pitch angle can enable the wind turbine blade icing prediction model to reflect the relative motion characteristics of the blade surface and droplets during the icing process of the wind turbine blades.

[0068] In the above embodiment, a method for predicting ice coating on wind turbine blades is described in detail, and the present application also provides an embodiment corresponding to an apparatus for predicting ice coating on wind turbine blades. It should be noted that the present application describes the embodiments of the apparatus part from two perspectives, one is based on the perspective of functional modules, and the other is based on the perspective of hardware.

[0069] FIG3 is a structural diagram of a wind turbine blade icing prediction device provided in an embodiment of the present application. As shown in FIG3 , the wind turbine blade icing prediction device includes:

[0070] The real sample acquisition module 10 is used to acquire historical meteorological data, historical wind turbine operation data and historical freezing data during the freezing period and the pre-freezing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle.

[0071] The data expansion module 11 is used to perform data expansion on real samples using the adversarial generative network model to obtain generated samples.

[0072] The training sample set determination module 12 is used to form a training sample set with real samples and generated samples, and uses the meteorological data and wind turbine operation data in the training sample set as input features and the icing data in the training sample set as output features.

[0073] The prediction module 13 is used to train the long short-term memory neural network using the input features and the output features to obtain a fan blade icing prediction model, and use the fan blade icing prediction model to predict the icing condition of the fan blades in a future time interval.

[0074] Based on the above embodiment, as an optional embodiment, it also includes:

[0075] The precipitation probability determination module is used to obtain the mesoscale numerical weather forecast results to determine the precipitation probability in the target area within the future time interval;

[0076] A weather element acquisition module is used to obtain weather elements formed by supercooled water droplets monitored by ground meteorological observation equipment at the site;

[0077] The judgment module is used to judge whether supercooled precipitation will occur in the future time interval based on the precipitation probability and weather factors using the prior knowledge base; if supercooled precipitation occurs, the step of using the fan blade icing prediction model to predict the icing condition of the fan blades in the future time interval is entered.

[0078] Based on the above embodiment, as an optional embodiment, the data expansion module includes:

[0079] The data expansion unit is used to use the conditional adversarial generative network model as a data generator to generate simulated data that conforms to the characteristics of the original data and prior knowledge to expand the real samples and obtain generated samples.

[0080] Based on the above embodiment, as an optional embodiment, it includes:

[0081] The first construction unit is used to construct conditional information according to a priori knowledge base of wind turbine blade icing, and uses a convolutional layer, instance normalization and activation function to form a generator and a discriminator;

[0082] An input unit, used to randomly extract noise from real samples, input the noise and condition information into the generator, and generate samples using the generator output;

[0083] A judgment unit, used to use a discriminator to judge the real sample and the generated sample respectively, and calculate the first loss of the discriminator;

[0084] A first updating unit, configured to update a first parameter of the discriminator to minimize a first loss;

[0085] The second updating unit is used to fix the first parameter and update the second parameter of the generator to minimize the second loss of the generator; the above steps are repeated until convergence to obtain a trained conditional adversarial generation network model.

[0086] Based on the above embodiment, as an optional embodiment, it includes:

[0087] An ice thickness data acquisition unit is used to acquire ice thickness data in real time using an ice sensor on the windward surface of the leading edge of a wind turbine blade;

[0088] The second construction unit is used to determine the time data corresponding to the ice thickness data and construct an ice thickness-time curve;

[0089] The freezing rate determination unit is used to determine the freezing rate according to the ice thickness-time curve.

[0090] Since the embodiments of the apparatus part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the apparatus part, which will not be repeated here.

[0091] FIG4 is a structural diagram of another wind blade icing prediction device provided in an embodiment of the present application. As shown in FIG4 , the wind blade icing prediction device includes: a memory 20 for storing a computer program;

[0092] The processor 21 is used to implement the steps of the wind turbine blade icing prediction method of the above embodiment when executing the computer program.

[0093] The wind turbine blade icing prediction device provided in this embodiment may include but is not limited to a smart phone, a tablet computer, a laptop computer or a desktop computer.

[0094] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 can be implemented in at least one hardware form of a digital signal processor (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a central processing unit (CPU); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 21 may be integrated with a graphics processing unit (GPU), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an artificial intelligence (AI) processor, which is used to process computing operations related to machine learning.

[0095] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In this embodiment, the memory 20 is at least used to store the following computer program 201, wherein, after the computer program is loaded and executed by the processor 21, it can implement the relevant steps of the wind turbine blade icing prediction method disclosed in any of the aforementioned embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, etc., and the storage method may be temporary storage or permanent storage. Among them, the operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include, but is not limited to, historical meteorological data, historical wind turbine operation data, and historical icing data.

[0096] In some embodiments, the wind turbine blade icing prediction device may further include a display screen 22 , an input / output interface 23 , a communication interface 24 , a power supply 25 , and a communication bus 26 .

[0097] Those skilled in the art will appreciate that the structure shown in FIG. 4 does not constitute a limitation on the wind turbine blade icing prediction device, and may include more or fewer components than those shown in the figure.

[0098] The wind turbine blade icing prediction device provided in the embodiment of the present application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: obtain historical meteorological data, historical wind turbine operation data and historical icing data during the icing period and the pre-icing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle; use the adversarial generative network model to perform data expansion on the real samples to obtain generated samples; the real samples and the generated samples form a training sample set, the meteorological data and wind turbine operation data in the training sample set are used as input features, and the icing data in the training sample set is used as output features; the input features and the output features are used to train the long short-term memory neural network to obtain a fan blade icing prediction model, and the fan blade icing prediction model is used to predict the icing condition of the fan blades in a future time period.

[0099] Finally, the present application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps recorded in the wind turbine blade icing prediction method in the above method embodiment are implemented.

[0100] It is understandable that if the method in the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0101] The above is a detailed introduction to a wind turbine blade icing prediction method, device and medium provided by the present application. The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the embodiments can be referenced to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

[0102] It should also be noted that, in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

Claims

1. A method for predicting wind turbine blade icing, characterized in that: include: Acquire historical meteorological data, historical wind turbine operation data and historical freezing data during the freezing period and before the freezing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle; Performing data expansion on the real sample using a generative adversarial network model to obtain a generated sample; The real samples and the generated samples form a training sample set, the meteorological data and the wind turbine operation data in the training sample set are used as input features, and the icing data in the training sample set is used as output features; The input features and the output features are used to train a long short-term memory neural network to obtain a fan blade icing prediction model, and the fan blade icing prediction model is used to predict the icing condition of the fan blades in a future time interval.

2. The wind turbine blade icing prediction method according to claim 1, characterized in that: Before using the wind turbine blade icing prediction model to predict the icing condition of the wind turbine blades in a future time interval, the method further includes: Obtaining mesoscale numerical weather forecast results to determine the probability of precipitation in the target area within the future time interval; Obtain weather elements for the formation of supercooled water droplets monitored by ground meteorological observation equipment at the site; Based on the precipitation probability and the weather elements, using a priori knowledge base to determine whether supercooled precipitation will occur in the future time interval; If supercooling precipitation occurs, the step of using the fan blade icing prediction model to predict the icing condition of the fan blades in a future time interval is entered.

3. The wind turbine blade icing prediction method according to claim 2, characterized in that: The ground meteorological observation equipment of the station includes: a microwave radiometer radar for monitoring the temperature and humidity of the atmosphere, a comprehensive cloud meter for monitoring cloud indicators, and a wind profiler radar for detecting wind speed and turbulence conditions at different altitudes; wherein the cloud indicators include the location, thickness and cloud height of cumulonimbus clouds.

4. The wind turbine blade icing prediction method according to claim 1, characterized in that: The real sample is augmented with data using a generative adversarial network model to obtain a generated sample, including: A conditional adversarial generative network model is used as a data generator to generate simulated data that conforms to the characteristics of the original data and prior knowledge to expand the real samples and obtain the generated samples.

5. The wind turbine blade icing prediction method according to claim 4, characterized in that: The training process of the conditional adversarial generation network model includes: Conditional information is constructed based on the prior knowledge base of wind turbine blade icing, and the generator and discriminator are composed of convolutional layers, instance normalization and activation functions; randomly extracting noise from the real sample, inputting the noise and the condition information into the generator, and generating a sample using the output of the generator; Using the discriminator to judge the real sample and the generated sample respectively, and calculating the first loss of the discriminator; Updating a first parameter of the discriminator to minimize the first loss; Fixing the first parameter, updating the second parameter of the generator to minimize the second loss of the generator; The above steps are repeated until convergence to obtain a trained conditional adversarial generation network model.

6. The wind turbine blade icing prediction method according to claim 1, characterized in that: The historical meteorological data include near-ground temperature, humidity, wind speed, wind direction, and rainfall intensity; the historical freezing data include freezing rate and freezing thickness.

7. The wind turbine blade icing prediction method according to claim 6, characterized in that: The process of determining the freezing rate includes: Use the ice sensor on the windward side of the leading edge of the wind turbine blade to obtain ice thickness data in real time; Determine the time data corresponding to the ice thickness data and construct an ice thickness-time curve; The freezing rate is determined based on the ice thickness-time curve.

8. A wind turbine blade icing prediction device, characterized in that: include: A real sample acquisition module is used to acquire historical meteorological data, historical wind turbine operation data and historical freezing data during the freezing period and the pre-freezing period as real samples; wherein the historical wind turbine operation data includes blade tip speed and blade pitch angle; A data expansion module, used to perform data expansion on the real sample using a generative adversarial network model to obtain a generated sample; A training sample set determination module, used for forming a training sample set with the real sample and the generated sample, taking the meteorological data and the fan operation data in the training sample set as input features, and taking the icing data in the training sample set as output features; The prediction module is used to train the long short-term memory neural network using the input features and the output features to obtain a fan blade icing prediction model, and use the fan blade icing prediction model to predict the icing condition of the fan blades in a future time interval.

9. A wind turbine blade icing prediction device, characterized in that: comprising a memory for storing a computer program; A processor is used to implement the steps of the wind turbine blade icing prediction method as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the wind turbine blade icing prediction method according to any one of claims 1 to 7 are implemented.

Images