Wind turbine generator blade icing degree prediction method, device, equipment and medium

Abstract

The invention discloses a wind turbine generator blade icing degree prediction method and device, equipment and a medium, relates to the technical field of wind power generation, and aims to obtain unit environment monitoring information and weather forecast information of a wind power plant under different time granularities, and perform standardization processing and serialization processing to obtain a wind turbine generator blade icing degree prediction result. Performing feature extraction on the processed unit environment monitoring information and weather forecast information by using a neural network model to obtain environment monitoring time sequence features and weather forecast time sequence features; calculating the environment monitoring time sequence characteristics and the weather forecast time sequence characteristics based on an attention mechanism to obtain an environment monitoring weight coefficient and a weather forecast weight coefficient; and inputting the environment monitoring weight coefficient, the environment monitoring time sequence characteristics, the weather forecast weight coefficient and the weather forecast time sequence characteristics into a neural network model, outputting a blade icing degree probability, and predicting the wind turbine blade icing degree according to the blade icing degree probability, thereby improving the overall operation efficiency of a wind power plant, and ensuring the safety of the unit, personnel and equipment assets.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, and in particular to a method, device, equipment and medium for predicting the icing degree of blades of a wind turbine generator set. Background Art

[0002] As the demand for renewable energy continues to increase, wind energy has been widely developed as an important clean energy source. The operating efficiency of wind turbines is closely related to their design and working environment, and the performance of the blades plays a vital role in this. However, wind turbines operate in an environment with low temperature and high humidity. Under such conditions, the blades are prone to icing, which affects their aerodynamic performance and leads to a decrease in power generation efficiency. Icing not only increases the weight of the blades, but also increases the risk of blade damage. At the same time, the rotation of the blades may cause ice to fall off during ice throwing or melting, posing a threat to the safety of personnel and equipment assets.

[0003] As can be seen from the above, how to predict the degree of icing on wind turbine blades, improve the overall operating efficiency of wind farms, and ensure the safety of turbines, personnel, and equipment assets is a problem to be solved in this field. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for predicting the icing degree of wind turbine blades, which can realize the prediction of the icing degree of wind turbine blades, improve the overall operating efficiency of wind farms, and ensure the safety of wind turbines, personnel and equipment assets. The specific scheme is as follows:

[0005] In a first aspect, the present application discloses a method for predicting the degree of icing of wind turbine blades, comprising:

[0006] Obtain unit environmental monitoring information and weather forecast information at different time granularities of wind farms;

[0007] Standardizing and serializing the unit environmental monitoring information and weather forecast information, and extracting features from the processed unit environmental monitoring information and weather forecast information using a pre-built neural network model to obtain environmental monitoring time series features and weather forecast time series features;

[0008] Based on the attention mechanism, weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated respectively to obtain an environmental monitoring weight coefficient and a weather forecast weight coefficient;

[0009] The environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics are input into the neural network model for calculation to output the probability of blade icing degree, and the blade icing degree of the wind turbine in the wind farm is predicted according to the probability of blade icing degree.

[0010] Optionally, the unit environmental monitoring information includes historical unit temperature information, historical unit humidity information and historical unit wind speed information; the data types in the weather forecast information include temperature, humidity and wind speed.

[0011] Optionally, before extracting features from the processed unit environment monitoring information and the weather forecast information using the pre-built neural network model, the method further includes:

[0012] Construct a neural network model based on gated recurrent units.

[0013] Optionally, the weight coefficient calculation formula is:

[0014] ;

[0015] ;

[0016] ;

[0017] Among them, H is the time series characteristics of environmental monitoring or weather forecast, , are the query vector weight matrix and the key vector weight matrix respectively, Q and K are the query vector matrix and the key vector matrix respectively, It is the environmental monitoring weight coefficient or the weather forecast weight coefficient.

[0018] Optionally, the step of inputting the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient, and the weather forecast time series characteristics into the neural network model for calculation to output the probability of blade icing degree includes:

[0019] The environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics are input into the neural network model and calculated using a calculation formula to obtain a blade icing probability matrix; the calculation formula is:

[0020] ;

[0021] in, is the blade icing probability matrix, , is the environmental monitoring weight coefficient and the weather forecast weight coefficient, , It is the time series characteristics of environmental monitoring and weather forecast;

[0022] The blade icing probability with the largest value in the blade icing probability matrix is ​​used as the blade icing degree probability, and the blade icing degree probability is output.

[0023] Optionally, after predicting the icing degree of blades of wind turbines in the wind farm according to the blade icing degree probability, the method further includes:

[0024] Calculate the loss function of the neural network model based on the probability of blade icing degree;

[0025] The loss function is minimized, and the optimal parameters of the model are determined using the gradient descent method, and the optimal parameters of the model are used to perform real-time optimization and adjustment on the neural network model.

[0026] Optionally, the loss function of the neural network model based on the blade icing degree probability calculation includes:

[0027] A loss function calculation formula is constructed, and the blade icing degree probability is input into the loss function calculation formula to calculate the loss function of the neural network model; the loss function calculation formula is:

[0028] ;

[0029] Among them, Y is the actual classification label of the sample, is the blade icing probability matrix, and L is the loss function.

[0030] In a second aspect, the present application discloses a device for predicting the degree of icing of blades of a wind turbine generator set, comprising:

[0031] An information acquisition module is used to obtain unit environmental monitoring information and weather forecast information at different time granularities of the wind farm;

[0032] A feature extraction module, used for standardizing and serializing the unit environmental monitoring information and the weather forecast information, and extracting features from the processed unit environmental monitoring information and the weather forecast information using a pre-built neural network model to obtain environmental monitoring time series features and weather forecast time series features;

[0033] A weight coefficient calculation module, used to calculate the weight coefficients of the environmental monitoring time series characteristics and the weather forecast time series characteristics based on the attention mechanism, so as to obtain the environmental monitoring weight coefficient and the weather forecast weight coefficient;

[0034] A prediction module is used to input the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics into the neural network model for calculation to output the probability of blade icing degree, and predict the blade icing degree of the wind turbine in the wind farm according to the blade icing degree probability.

[0035] In a third aspect, the present application discloses an electronic device, comprising:

[0036] Memory, used to store computer programs;

[0037] The processor is used to execute the computer program to implement the above-mentioned method for predicting the degree of icing of wind turbine blades.

[0038] In a fourth aspect, the present application discloses a computer storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned method for predicting the degree of icing of wind turbine blades disclosed above are implemented.

[0039] It can be seen that the present application provides a method for predicting the degree of icing on blades of a wind turbine generator set, including obtaining environmental monitoring information and weather forecast information of the unit at different time granularities of a wind farm; standardizing and serializing the environmental monitoring information and weather forecast information of the unit, and using a pre-built neural network model to extract features of the processed environmental monitoring information and weather forecast information of the unit to obtain environmental monitoring time series features and weather forecast time series features; calculating weight coefficients of the environmental monitoring time series features and the weather forecast time series features based on an attention mechanism to obtain environmental monitoring weight coefficients and weather forecast weight coefficients; inputting the environmental monitoring weight coefficient, the environmental monitoring time series features, the weather forecast weight coefficient and the weather forecast time series features into the neural network model for calculation to output a probability of the degree of icing on the blades, and predicting the degree of icing on the blades of the wind turbine generator set in the wind farm according to the probability of the degree of icing on the blades. The present application obtains unit environmental monitoring information and weather forecast information at different time granularities of a wind farm. Different time granularities can achieve more detailed prediction of blade icing conditions. A neural network model is used to extract features of the processed unit environmental monitoring information and weather forecast information to obtain environmental monitoring time series features and weather forecast time series features. Weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated based on the attention mechanism, respectively. From a data-driven perspective, rapid calculation of independent variable features and dependent variable weight coefficients is achieved, and complex correlation analysis is avoided to obtain environmental monitoring weight coefficients and weather forecast weight coefficients. Finally, a neural network model is used for calculation to output the probability of blade icing degree and predict the icing degree of wind turbine blades in the wind farm. The neural network model can capture the complex factors of icing more comprehensively, thereby improving the accuracy and reliability of the prediction, and ultimately providing a more scientific and reasonable operation and maintenance strategy for the operation of wind farm units, so that the units can continue to operate efficiently in a safe state and ensure the safety of units, personnel and equipment assets. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0041] Figure 1 A flow chart of a method for predicting the degree of icing on blades of a wind turbine generator system disclosed in the present application;

[0042] Figure 2 A flow chart of a method for predicting the degree of icing on blades of a wind turbine generator system disclosed in the present application;

[0043] Figure 3 A framework diagram of a neural network model based on a gated recurrent unit disclosed in this application;

[0044] Figure 4 A specific wind turbine blade icing degree prediction flow chart disclosed in the present application;

[0045] Figure 5 This is a schematic diagram of the structure of a wind turbine blade icing degree prediction device disclosed in the present application;

[0046] Figure 6 A structural diagram of an electronic device provided for this application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] As the demand for renewable energy continues to increase, wind energy has been widely developed as an important clean energy source. The operating efficiency of wind turbines is closely related to their design and working environment, and the performance of the blades plays a vital role in this. However, wind turbines operate in an environment with low temperature and high humidity. Under such conditions, the blades are prone to icing, which affects their aerodynamic performance and leads to a decrease in power generation efficiency. Icing not only increases the weight of the blades, but also increases the risk of blade damage. At the same time, the rotation of the blades may cause ice to fall off during ice throwing or melting, posing a threat to the safety of personnel and equipment assets. As can be seen from the above, how to predict the degree of icing on wind turbine blades, improve the overall operating efficiency of wind farms, and ensure the safety of units, personnel and equipment assets is a problem to be solved in this field.

[0049] See also Figure 1 As shown, the embodiment of the present invention discloses a method for predicting the icing degree of blades of a wind turbine generator set, which may specifically include:

[0050] Step S11: Obtaining unit environmental monitoring information and weather forecast information at different time granularities of the wind farm.

[0051] In this embodiment, the unit environment monitoring information includes historical unit temperature information, historical unit humidity information and historical unit wind speed information; the data types in the weather forecast information include temperature, humidity and wind speed.

[0052] Taking three different time granularities as examples, namely short-term (hours), medium-term (weeks), and long-term (months), a collection of unit environmental monitoring information and a collection of weather forecast information are constructed:

[0053] ;

[0054] in, is the unit environmental monitoring information at time t, is the short-term temperature information of the unit at time t, is the short-term humidity information of the unit at time t, is the short-term wind speed information of the unit at time t, is the medium-term temperature information of the unit at time t, is the medium-term humidity information of the unit at time t, is the medium-term wind speed information of the unit at time t, is the long-term wind speed information of the unit at time t, is the long-term temperature information of the unit at time t, is the long-term wind speed information of the unit at time t;

[0055] ;

[0056] Similarly, The meaning of the parameters in The parameters in have the same meaning, except that is the unit environmental monitoring information at time t, is the weather forecast information at time t.

[0057] Step S12: Standardize and serialize the unit environmental monitoring information and weather forecast information, and use a pre-built neural network model to extract features from the processed unit environmental monitoring information and weather forecast information to obtain environmental monitoring time series features and weather forecast time series features.

[0058] In this embodiment, the standardization process is: standardize each column of data. Taking the standardization of the short-term temperature information of the unit as an example, the standardization calculation formula is as follows:

[0059] ;

[0060] ;

[0061] in, is the short-term temperature data of the unit at time t, is the short-term temperature data set of the unit, is the standardized short-term temperature data of the unit at time t, min represents the minimum value calculation function, and max represents the maximum value calculation function;

[0062] The short-term / medium-term / long-term temperature, humidity and wind speed data of the unit and weather forecast are standardized according to the above standardization formula, and the standardized unit environmental monitoring data is recorded as , the standardized weather forecast data is recorded as ;

[0063] The serialization process is as follows: the standardized data is serialized to predict the blade icing state of the unit at time t+m. , the serialization results are as follows:

[0064] ;

[0065] ;

[0066] Among them, j is the historical time window size, k is the prediction window interval, is the historical environmental monitoring data set of the unit, It is a set of weather forecast data from t+1 to t+m.

[0067] Step S13: Based on the attention mechanism, weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated respectively to obtain the environmental monitoring weight coefficient and the weather forecast weight coefficient.

[0068] In this embodiment, the time series characteristics of the computing environment monitoring Environmental monitoring weight coefficient , Weather Forecast Time Series Characteristics The weather forecast weight coefficient is .

[0069] Step S14: input the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics into the neural network model for calculation to output the probability of blade icing degree, and predict the blade icing degree of the wind turbine in the wind farm according to the probability of blade icing degree.

[0070] In this embodiment, the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics are input into the neural network model and calculated using a calculation formula to obtain a blade icing probability matrix; the calculation formula is:

[0071] ;

[0072] in, is the blade icing probability matrix, , is the environmental monitoring weight coefficient and the weather forecast weight coefficient, , It is the time series characteristics of environmental monitoring and weather forecast;

[0073] The blade icing probability with the largest value in the blade icing probability matrix is ​​used as the blade icing degree probability, and the blade icing degree probability is output, and the blade icing degree of the wind turbine in the wind farm is predicted according to the blade icing degree probability.

[0074] The specific blade icing probability calculation and icing degree prediction process is as follows: Calculate the feature output based on the time series features and their weight coefficients:

[0075] ;

[0076] ;

[0077] Among them, H is the data feature input, V is the value vector matrix, is the weight matrix representing the value vector, is the data feature weight coefficient matrix, Att is the feature output calculated based on the data feature weight matrix;

[0078] Then, the blade icing probability matrix is ​​calculated using the characteristic output and the calculation formula of the blade icing probability matrix.

[0079] In this embodiment, the unit environmental monitoring information and weather forecast information at different time granularities of the wind farm are obtained; the unit environmental monitoring information and weather forecast information are standardized and serialized, and the processed unit environmental monitoring information and weather forecast information are subjected to feature extraction using a pre-constructed neural network model to obtain environmental monitoring time series features and weather forecast time series features; weight coefficients of the environmental monitoring time series features and the weather forecast time series features are respectively calculated based on the attention mechanism to obtain environmental monitoring weight coefficients and weather forecast weight coefficients; the environmental monitoring weight coefficients, the environmental monitoring time series features, the weather forecast weight coefficients and the weather forecast time series features are input into the neural network model for calculation to output the probability of blade icing degree, and the blade icing degree of the wind turbine unit in the wind farm is predicted according to the blade icing degree probability. The present application obtains unit environmental monitoring information and weather forecast information at different time granularities of a wind farm. Different time granularities can achieve more detailed prediction of blade icing conditions. A neural network model is used to extract features of the processed unit environmental monitoring information and weather forecast information to obtain environmental monitoring time series features and weather forecast time series features. Weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated based on the attention mechanism, respectively. From a data-driven perspective, rapid calculation of independent variable features and dependent variable weight coefficients is achieved, and complex correlation analysis is avoided to obtain environmental monitoring weight coefficients and weather forecast weight coefficients. Finally, a neural network model is used for calculation to output the probability of blade icing degree and predict the icing degree of wind turbine blades in the wind farm. The neural network model can capture the complex factors of icing more comprehensively, thereby improving the accuracy and reliability of the prediction, and ultimately providing a more scientific and reasonable operation and maintenance strategy for the operation of wind farm units, so that the units can continue to operate efficiently in a safe state and ensure the safety of units, personnel and equipment assets.

[0080] See also Figure 2 As shown, the embodiment of the present invention discloses a method for predicting the icing degree of blades of a wind turbine generator set, which may specifically include:

[0081] Step S21: Obtaining unit environmental monitoring information and weather forecast information at different time granularities of the wind farm.

[0082] Step S22: Standardize and serialize the unit environmental monitoring information and weather forecast information, build a neural network model based on the gated recurrent unit, and use the pre-built neural network model to extract features of the processed unit environmental monitoring information and the weather forecast information to obtain environmental monitoring time series features and weather forecast time series features.

[0083] In this embodiment, the unit environmental monitoring information and weather forecast information are standardized and serialized, a neural network model is constructed based on GRU (Gated Recurrent Unit), and the environmental monitoring time series features and weather forecast time series features corresponding to the processed unit environmental monitoring information and weather forecast information are extracted.

[0084] In this embodiment, the neural network model framework constructed based on the gated recurrent unit is as follows: Figure 3 As shown. Assume that the input of the neural network model built based on the gated recurrent unit is , the model calculation process is as follows:

[0085] ;

[0086] ;

[0087] ;

[0088] ;

[0089] in, To update the gate output, To reset the gate output, is the intermediate calculation variable, is the output result of the hidden layer, f is the sigmoid activation function, , They are and Update the gate weight matrix, , They are and Reset the gate weight matrix, , They are and Cell weight matrix.

[0090] In this embodiment, the neural network model constructed based on GRU extracts the environmental monitoring time series features and the weather forecast time series features, and the calculation expression is as follows:

[0091] ;

[0092] ;

[0093] Among them, GRU is a neural network model built based on GRU. The time series features extracted from the historical environmental monitoring data of the unit, Time series features extracted for the weather forecast data set.

[0094] Step S23: Based on the attention mechanism, weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated respectively to obtain the environmental monitoring weight coefficient and the weather forecast weight coefficient.

[0095] In this embodiment, the weight coefficient calculation formula is:

[0096] ;

[0097] ;

[0098] ;

[0099] Among them, H is the time series characteristics of environmental monitoring or weather forecast, , are the query vector weight matrix and the key vector weight matrix respectively, Q and K are the query vector matrix and the key vector matrix respectively, It is the environmental monitoring weight coefficient or the weather forecast weight coefficient.

[0100] Step S24: input the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics into the neural network model for calculation to output the blade icing degree probability, and predict the blade icing degree of the wind turbine in the wind farm according to the blade icing degree probability.

[0101] For example, according to the blade icing condition, the types of wind turbine blade icing degree are divided into four categories: no icing, light icing, moderate icing, and severe icing. No icing is marked as 0, light icing is marked as 1, moderate icing is marked as 2, and severe icing is marked as 3. The probability of blade icing degree is determined to match which type of blade icing degree, and then the final wind turbine blade icing degree is predicted.

[0102] Step S25: Calculate the loss function of the neural network model based on the probability of blade icing degree, minimize the loss function, and use the gradient descent method to determine the optimal parameters of the model, and use the optimal parameters of the model to perform real-time optimization and adjustment on the neural network model.

[0103] In this embodiment, a loss function calculation formula is constructed, and the blade icing degree probability is input into the loss function calculation formula to calculate the loss function of the neural network model; the loss function calculation formula is:

[0104] ;

[0105] Among them, Y is the actual classification label of the sample, is the blade icing probability matrix, L is the loss function;

[0106] Then, the loss function L is minimized, and the optimal parameters of the model are determined using the gradient descent method, and the neural network model is optimized and adjusted in real time using the optimal parameters of the model.

[0107] The specific process of predicting the icing degree of wind turbine blades proposed in this application is as follows: Figure 4 As shown, first, the environmental monitoring information and weather forecast information of the wind farm unit at different time granularities are collected, and the collected information is standardized and serialized; then, a neural network model is constructed based on GRU, and the neural network model is used to extract features of the processed environmental monitoring information and weather forecast information of the unit to obtain environmental monitoring time series features and weather forecast time series features; then, the environmental monitoring weight coefficient and the weather forecast weight coefficient are calculated based on the attention mechanism; then, the environmental monitoring weight coefficient, the environmental monitoring time series features, the weather forecast weight coefficient and the weather forecast time series features are input into the neural network model for calculation to output the probability of blade icing degree; finally, the icing degree of the wind turbine blades in the wind farm is predicted according to the probability of blade icing degree, a loss function calculation formula is constructed, and the loss function of the neural network model is calculated, the loss function is minimized, and the optimal parameters of the model are determined by the gradient descent method to realize real-time optimization and adjustment of the neural network model.

[0108] The present application comprehensively constructs a neural network model based on historical data and real-time monitoring of weather forecasts, enriches the input of information, especially weather forecast information, so that the model can more comprehensively capture the complex factors of icing, thereby improving the accuracy and reliability of the prediction; the present invention can classify the icing types of wind turbine blades into four types, realizes a more detailed prediction of blade icing conditions, and can provide a more scientific and reasonable operation and maintenance strategy for the operation of wind farm units, so that the units can continue to operate efficiently in a safe state and obtain maximum power generation capacity and efficiency; the present invention adopts an attention mechanism to realize the rapid calculation of independent variable characteristics and dependent variable weight coefficients from a data-driven perspective, avoiding complex correlation analysis; in addition, the present application can also identify icing risks in advance, which can effectively reduce the accident rate caused by blade icing and ensure the safety of equipment and operators.

[0109] In this embodiment, the unit environmental monitoring information and weather forecast information at different time granularities of the wind farm are obtained; the unit environmental monitoring information and weather forecast information are standardized and serialized, and the processed unit environmental monitoring information and weather forecast information are subjected to feature extraction using a pre-constructed neural network model to obtain environmental monitoring time series features and weather forecast time series features; weight coefficients of the environmental monitoring time series features and the weather forecast time series features are respectively calculated based on the attention mechanism to obtain environmental monitoring weight coefficients and weather forecast weight coefficients; the environmental monitoring weight coefficients, the environmental monitoring time series features, the weather forecast weight coefficients and the weather forecast time series features are input into the neural network model for calculation to output the probability of blade icing degree, and the blade icing degree of the wind turbine unit in the wind farm is predicted according to the blade icing degree probability. The present application obtains unit environmental monitoring information and weather forecast information at different time granularities of a wind farm. Different time granularities can achieve more detailed prediction of blade icing conditions. A neural network model is used to extract features of the processed unit environmental monitoring information and weather forecast information to obtain environmental monitoring time series features and weather forecast time series features. Weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated based on the attention mechanism, respectively. From a data-driven perspective, rapid calculation of independent variable features and dependent variable weight coefficients is achieved, and complex correlation analysis is avoided to obtain environmental monitoring weight coefficients and weather forecast weight coefficients. Finally, a neural network model is used for calculation to output the probability of blade icing degree and predict the icing degree of wind turbine blades in the wind farm. The neural network model can capture the complex factors of icing more comprehensively, thereby improving the accuracy and reliability of the prediction, and ultimately providing a more scientific and reasonable operation and maintenance strategy for the operation of wind farm units, so that the units can continue to operate efficiently in a safe state and ensure the safety of units, personnel and equipment assets.

[0110] See also Figure 5 As shown, the embodiment of the present invention discloses a device for predicting the degree of icing of blades of a wind turbine generator set, which may specifically include:

[0111] The information acquisition module 11 is used to obtain the unit environmental monitoring information and weather forecast information of the wind farm at different time granularities;

[0112] The feature extraction module 12 is used to perform standardization and serialization processing on the unit environmental monitoring information and the weather forecast information, and use a pre-built neural network model to perform feature extraction on the processed unit environmental monitoring information and the weather forecast information to obtain environmental monitoring time series features and weather forecast time series features;

[0113] A weight coefficient calculation module 13 is used to calculate the weight coefficients of the environmental monitoring time series characteristics and the weather forecast time series characteristics based on the attention mechanism to obtain the environmental monitoring weight coefficient and the weather forecast weight coefficient;

[0114] The prediction module 14 is used to input the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics into the neural network model for calculation to output the probability of blade icing degree, and predict the blade icing degree of the wind turbine in the wind farm according to the blade icing degree probability.

[0115] In this embodiment, the unit environmental monitoring information and weather forecast information at different time granularities of the wind farm are obtained; the unit environmental monitoring information and weather forecast information are standardized and serialized, and the processed unit environmental monitoring information and weather forecast information are subjected to feature extraction using a pre-constructed neural network model to obtain environmental monitoring time series features and weather forecast time series features; weight coefficients of the environmental monitoring time series features and the weather forecast time series features are respectively calculated based on the attention mechanism to obtain environmental monitoring weight coefficients and weather forecast weight coefficients; the environmental monitoring weight coefficients, the environmental monitoring time series features, the weather forecast weight coefficients and the weather forecast time series features are input into the neural network model for calculation to output the probability of blade icing degree, and the blade icing degree of the wind turbine unit in the wind farm is predicted according to the blade icing degree probability. The present application obtains unit environmental monitoring information and weather forecast information at different time granularities of a wind farm. Different time granularities can achieve more detailed prediction of blade icing conditions. A neural network model is used to extract features of the processed unit environmental monitoring information and weather forecast information to obtain environmental monitoring time series features and weather forecast time series features. Weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated based on the attention mechanism, respectively. From a data-driven perspective, rapid calculation of independent variable features and dependent variable weight coefficients is achieved, and complex correlation analysis is avoided to obtain environmental monitoring weight coefficients and weather forecast weight coefficients. Finally, a neural network model is used for calculation to output the probability of blade icing degree and predict the icing degree of wind turbine blades in the wind farm. The neural network model can capture the complex factors of icing more comprehensively, thereby improving the accuracy and reliability of the prediction, and ultimately providing a more scientific and reasonable operation and maintenance strategy for the operation of wind farm units, so that the units can continue to operate efficiently in a safe state and ensure the safety of units, personnel and equipment assets.

[0116] In some specific embodiments, the unit environment monitoring information includes historical unit temperature information, historical unit humidity information and historical unit wind speed information; the data types in the weather forecast information include temperature, humidity and wind speed.

[0117] In some specific embodiments, the feature extraction module 12 may specifically include:

[0118] A model building module for building neural network models based on gated recurrent units.

[0119] In some specific embodiments, the weight coefficient calculation formula is:

[0120] ;

[0121] ;

[0122] ;

[0123] Among them, H is the time series characteristics of environmental monitoring or weather forecast, , are the query vector weight matrix and the key vector weight matrix respectively, Q and K are the query vector matrix and the key vector matrix respectively, It is the environmental monitoring weight coefficient or the weather forecast weight coefficient.

[0124] In some specific embodiments, the prediction module 14 may specifically include:

[0125] An input module is used to input the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics into the neural network model and calculate using a calculation formula to obtain a blade icing probability matrix; the calculation formula is:

[0126] ;

[0127] in, is the blade icing probability matrix, , is the environmental monitoring weight coefficient and the weather forecast weight coefficient, , It is the time series characteristics of environmental monitoring and weather forecast;

[0128] The output module is used to take the blade icing probability with the largest value in the blade icing probability matrix as the blade icing degree probability, and output the blade icing degree probability.

[0129] In some specific embodiments, the prediction module 14 may specifically include:

[0130] A loss function calculation module, used for calculating the loss function of the neural network model based on the probability of blade icing degree;

[0131] The model real-time optimization and adjustment module is used to minimize the loss function, determine the optimal parameters of the model using the gradient descent method, and use the optimal parameters of the model to perform real-time optimization and adjustment on the neural network model.

[0132] In some specific embodiments, the prediction module 14 may specifically include:

[0133] A calculation formula building module is used to build a loss function calculation formula, and input the blade icing degree probability into the loss function calculation formula to calculate the loss function of the neural network model; the loss function calculation formula is:

[0134] ;

[0135] Among them, Y is the actual classification label of the sample, is the blade icing probability matrix, and L is the loss function.

[0136] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the wind turbine blade icing degree prediction method performed by the electronic device disclosed in any of the aforementioned embodiments.

[0137] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0138] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon include an operating system 221, a computer program 222 and data 223, etc. The storage method can be temporary storage or permanent storage.

[0139] Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20, so as to realize the operation and processing of the data 223 in the memory 22 by the processor 21, which can be Windows, Unix, Linux, etc. In addition to including a computer program that can be used to complete the wind turbine blade icing degree prediction method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. In addition to including data transmitted from an external device received by the wind turbine blade icing degree prediction device, the data 223 can also include data collected by its own input and output interface 25, etc.

[0140] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0141] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium, in which a computer program is stored. When the computer program is loaded and executed by a processor, the steps of the method for predicting the degree of icing of wind turbine blades disclosed in any of the aforementioned embodiments are implemented.

[0142] Finally, it should be noted that, in this article, 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 "include", "comprise" 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, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0143] The above is a detailed introduction to a method, device, equipment and storage medium for predicting the degree of icing on wind turbine blades provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for predicting the degree of icing of wind turbine blades, characterized in that: include: Obtain unit environmental monitoring information and weather forecast information at different time granularities of wind farms; Standardizing and serializing the unit environmental monitoring information and weather forecast information, and extracting features from the processed unit environmental monitoring information and weather forecast information using a pre-built neural network model to obtain environmental monitoring time series features and weather forecast time series features; Based on the attention mechanism, weight coefficients of the environmental monitoring time series features and the weather forecast time series features are calculated respectively to obtain an environmental monitoring weight coefficient and a weather forecast weight coefficient; The environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics are input into the neural network model for calculation to output the probability of blade icing degree, and the blade icing degree of the wind turbine in the wind farm is predicted according to the probability of blade icing degree.

2. The method for predicting the icing degree of wind turbine blades according to claim 1, characterized in that: The unit environmental monitoring information includes historical unit temperature information, historical unit humidity information and historical unit wind speed information; the data types in the weather forecast information include temperature, humidity and wind speed.

3. The method for predicting the icing degree of wind turbine blades according to claim 1, characterized in that: Before the pre-built neural network model is used to extract features from the processed unit environment monitoring information and the weather forecast information, the method further includes: Construct a neural network model based on gated recurrent units.

4. The method for predicting the icing degree of wind turbine blades according to claim 1, characterized in that: The weight coefficient calculation formula is: ； ； ； Among them, H is the time series characteristics of environmental monitoring or weather forecast, , are the query vector weight matrix and the key vector weight matrix respectively, Q and K are the query vector matrix and the key vector matrix respectively, It is the environmental monitoring weight coefficient or the weather forecast weight coefficient.

5. The method for predicting the icing degree of wind turbine blades according to claim 1, characterized in that: The step of inputting the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient, and the weather forecast time series characteristics into the neural network model for calculation to output the probability of blade icing degree includes: The environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics are input into the neural network model and calculated using a calculation formula to obtain a blade icing probability matrix; the calculation formula is: ； in, is the blade icing probability matrix, , is the environmental monitoring weight coefficient and the weather forecast weight coefficient, , It is the time series characteristics of environmental monitoring and weather forecast; The blade icing probability with the largest value in the blade icing probability matrix is ​​used as the blade icing degree probability, and the blade icing degree probability is output.

6. The method for predicting the icing degree of wind turbine blades according to any one of claims 1 to 5, characterized in that: After predicting the icing degree of the blades of the wind turbines in the wind farm according to the blade icing degree probability, the method further includes: Calculate the loss function of the neural network model based on the probability of blade icing degree; The loss function is minimized, and the optimal parameters of the model are determined using the gradient descent method, and the optimal parameters of the model are used to perform real-time optimization and adjustment on the neural network model.

7. The method for predicting the icing degree of wind turbine blades according to claim 6, characterized in that: The loss function of the neural network model based on the blade icing degree probability calculation includes: A loss function calculation formula is constructed, and the blade icing degree probability is input into the loss function calculation formula to calculate the loss function of the neural network model; the loss function calculation formula is: ； Among them, Y is the actual classification label of the sample, is the blade icing probability matrix, and L is the loss function.

8. A device for predicting the degree of icing of wind turbine blades, characterized in that: include: An information acquisition module is used to obtain unit environmental monitoring information and weather forecast information at different time granularities of the wind farm; A feature extraction module, used for standardizing and serializing the unit environmental monitoring information and the weather forecast information, and extracting features from the processed unit environmental monitoring information and the weather forecast information using a pre-built neural network model to obtain environmental monitoring time series features and weather forecast time series features; A weight coefficient calculation module, used to calculate the weight coefficients of the environmental monitoring time series characteristics and the weather forecast time series characteristics based on the attention mechanism, so as to obtain the environmental monitoring weight coefficient and the weather forecast weight coefficient; A prediction module is used to input the environmental monitoring weight coefficient, the environmental monitoring time series characteristics, the weather forecast weight coefficient and the weather forecast time series characteristics into the neural network model for calculation to output the probability of blade icing degree, and predict the blade icing degree of the wind turbine in the wind farm according to the blade icing degree probability.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the method for predicting the degree of icing on blades of a wind turbine generator set according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program; wherein, when the computer program is executed by a processor, the method for predicting the degree of icing of a wind turbine blade according to any one of claims 1 to 7 is implemented.

Images