Long- and short-term memory correction prediction method and device for ice thickness of fan blades

Abstract

The embodiment of the present application provides a long-short-term memory correction forecasting method for the ice thickness of wind turbine blades. The method includes: determining the forecasting area for which numerical forecasting of freezing rain and ice is required; dividing the forecasting area into equidistant grids; obtaining initial field data and boundary condition data for numerical forecasting of freezing rain and ice, the initial field data refers to the data on each grid point at the time when the numerical model starts integration, inputting the initial field data and boundary condition data consistent with the grid of the forecasting area into the numerical model, obtaining the first freezing rain and ice value of the wind turbine blades in the forecasting area at a preset time point in the future; inputting the first freezing rain and ice value into the long-short-term memory neural network; obtaining the second freezing rain and ice value obtained by the long-short-term memory neural network after correcting the first freezing rain and ice value, and determining the second freezing rain and ice value as the freezing rain and ice value of the wind turbine blades in the forecasting area at the preset time point in the future.

Description

Technical Field

[0001] The present application relates to the field of power grid dispatching, and specifically to a method, device, storage medium and processor for long- and short-term memory correction and prediction of ice thickness on wind turbine blades. Background Art

[0002] With the development of wind power generation technology, wind farms are located in mountainous areas with high altitudes, which makes freezing rain and icing of wind turbines frequent. However, in the prior art, the wind and rain weather is usually predicted according to the weather forecast, but this method cannot accurately predict the icing of wind turbine blades. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a method, device, storage medium and processor for long- and short-term memory correction prediction of ice thickness on wind turbine blades.

[0004] In order to achieve the above-mentioned object, the first aspect of the present application provides a long- and short-term memory correction prediction method for ice thickness of wind turbine blades, comprising:

[0005] Determine the forecast areas where numerical forecast of freezing rain and icing is needed;

[0006] Divide the area to be forecast into equidistant grids;

[0007] Obtain the initial field data and boundary condition data for numerical prediction of freezing rain and icing. The initial field data refers to the data at each grid point at the beginning of the integration of the numerical model, and the boundary condition data refers to the data in the boundary area of ​​the numerical model;

[0008] Interpolating the initial field data and the boundary condition data according to the grid of the area to be predicted so that the grid of the initial field data and the boundary condition data is consistent with the grid of the area to be predicted;

[0009] Inputting the initial field data and boundary condition data consistent with the grid of the area to be forecasted into the numerical model to drive the numerical model to integrate forward and obtain the first freezing rain ice coverage value of the wind turbine blades in the area to be forecasted at a preset time point in the future;

[0010] Inputting the first freezing rain ice coverage value into the long short-term memory neural network;

[0011] A second freezing rain coverage value is obtained after the long short-term memory neural network corrects the first freezing rain coverage value, and a third freezing rain coverage value obtained by weighted summation of the second freezing rain coverage values ​​is determined as the freezing rain coverage value of the wind turbine blades in the area to be forecasted at a preset time point in the future.

[0012] In an embodiment of the present application, a long and short-term memory correction and prediction method for ice coverage thickness of wind turbine blades includes multiple long and short-term memory neural networks, each of which predicts data on different time scales; obtaining a second freezing rain coverage value obtained after the long and short-term memory neural network corrects the first freezing rain coverage value, and determining a third freezing rain coverage value obtained by weighted summation of the second freezing rain coverage values ​​as the freezing rain coverage value of the wind turbine blades in the area to be predicted at a preset future time point. The method includes: obtaining a second freezing rain coverage value obtained after each long and short-term memory neural network corrects the first freezing rain coverage value; weighted summing multiple second freezing rain coverage values ​​according to the weight of each long and short-term memory neural network to obtain a third freezing rain coverage value; and determining the third freezing rain coverage value as the freezing rain coverage value of the wind turbine blades in the area to be predicted at a preset future time point.

[0013] In an embodiment of the present application, a long short-term memory correction forecasting method for wind turbine blade ice thickness comprises a time scale including at least one of a seasonal scale, a monthly scale and a daily scale, wherein the long short-term memory neural network predicts seasonal scale data, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of wind turbine blades in the area to be forecasted in the next three months based on the input data; the long short-term memory neural network predicts monthly scale data, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of wind turbine blades in the area to be forecasted in the next month based on the input data; the long short-term memory neural network predicts daily scale data, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of wind turbine blades in the area to be forecasted tomorrow based on the input data.

[0014] In an embodiment of the present application, a long and short-term memory correction prediction method for ice coverage thickness of wind turbine blades is provided, the method also includes a training step for a long and short-term memory neural network, the training step includes: obtaining a plurality of historical freezing rain icing values ​​corresponding to different time scales, the historical freezing rain icing values ​​being historical prediction values ​​obtained by inputting initial field data and boundary condition data into a numerical model within a historical time period; respectively inputting the historical freezing rain icing values ​​of different time scales into different long and short-term memory neural networks to obtain a long and short-term memory neural network that can perform predictions for data of each time scale respectively.

[0015] In an embodiment of the present application, in a long short-term memory correction prediction method for wind turbine blade ice thickness, when training a long short-term memory neural network, the input data and output data of the long short-term memory neural network are as shown in formula (1):

[0016] (1);

[0017] Among them, Y is the predicted value of freezing rain and icing on wind turbine blades output by the long short-term memory neural network, S is the season type of the year, M is the monthly mark, D is the date mark, Y1 is the predicted data for the current moment in the numerical mode, Y- is the predicted data for the previous moment in the numerical mode, and Re is the predicted value for the previous moment output by the long short-term memory neural network.

[0018] In an embodiment of the present application, a long short-term memory correction prediction method for the ice thickness of a wind turbine blade is provided, wherein the long short-term memory neural network includes a forget gate , Memory Gate and output gate ; Among them, the forget gate determines the unit state at the previous moment The unit state preserved to the current moment The calculation formula is shown in formula (2):

[0019] (2);

[0020] in, represents the output of the forget gate, Represents the sigmoid layer. Each data needs to go through the sigmoid function. represents the weight matrix, is the module output at the previous moment, [ , ] represents the concatenation of two matrices. Represents bias.

[0021] The memory gate determines the current input The unit status The unit state preserved to the current moment The calculation formula is shown in formula (3) and formula (4):

[0022] (3);

[0023] (4);

[0024] The updated cell state is shown in formula (5):

[0025] (5);

[0026] in, is the input gate, is the current cell state, is the unit state at the previous moment, The current input cell state determines what information can be added to the cell state.

[0027] The calculation formula of the output gate is shown in formulas (6) and (7):

[0028] (6);

[0029] (7);

[0030] in, is the output gate, It is the module output at the current moment.

[0031] In an embodiment of the present application, a long-short-term memory correction prediction method for ice thickness on wind turbine blades inputs initial field data and boundary condition data consistent with the grid of the area to be predicted into a numerical model, including: determining a future preset time point required for prediction; determining a parameterization scheme for the initial field data and boundary condition data consistent with the grid of the area to be predicted based on the future preset time point; and inputting the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model in a manner set by the parameterization scheme.

[0032] In an embodiment of the present application, a long- and short-term memory correction and prediction device for ice thickness of wind turbine blades includes:

[0033] The area selection module is used to determine the forecast area where the freezing rain and ice numerical forecast needs to be carried out;

[0034] The data processing module is used to divide the area to be forecasted into equidistant grids; obtain the initial field data and boundary condition data for numerical prediction of freezing rain and icing, the initial field data refers to the data on each grid point at the time when the numerical model starts integration, and the boundary condition data refers to the data in the boundary area of ​​the numerical model; interpolate the initial field data and boundary condition data according to the grid of the area to be forecasted, so that the grids of the initial field data and boundary condition data are consistent with the grids of the area to be forecasted;

[0035] The first prediction module is used to input the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model to drive the numerical model to integrate forward and obtain the first freezing rain ice coverage value of the wind turbine blades in the area to be predicted at a preset time point in the future;

[0036] The second prediction module is used to input the first freezing rain and icing value into the long short-term memory neural network; obtain the second freezing rain and icing value obtained after the long short-term memory neural network corrects the first freezing rain and icing value, and determine the third freezing rain and icing value obtained by weighted summation of the second freezing rain and icing values ​​as the freezing rain and icing value of the wind turbine blades in the area to be forecasted at a preset time point in the future.

[0037] A second aspect of the present invention provides a processor configured to execute the above-mentioned long- and short-term memory correction prediction method of ice coating thickness on wind turbine blades.

[0038] A third aspect of the present invention provides a machine-readable storage medium having instructions stored thereon, which when executed by a processor configures the processor to execute the above-mentioned long- and short-term memory correction prediction method for ice thickness on wind turbine blades.

[0039] By means of the long and short-term memory correction forecasting method, device and processor of the ice thickness of the wind turbine blades, the forecasting area to be forecasted where the numerical forecast of freezing rain and icing is required is determined; the forecasting area to be forecasted is divided into equidistant grids; initial field data and boundary condition data for numerical forecasting of freezing rain and icing are obtained, the initial field data refers to the data on each grid point at the time when the numerical model starts integration, and the boundary condition data refers to the data in the boundary area of ​​the numerical model; the initial field data and boundary condition data are interpolated according to the grid of the forecasting area to make the grid of the initial field data and boundary condition data consistent with the grid of the forecasting area; the initial field data and boundary condition data consistent with the grid of the forecasting area are input into the numerical model to drive the numerical model to integrate forward, and a first freezing rain and icing value of the wind turbine blades in the forecasting area at a preset time point in the future is obtained; the first freezing rain and icing value is input into the long and short-term memory neural network; a second freezing rain and icing value obtained after the long and short-term memory neural network corrects the first freezing rain and icing value is obtained, and the second freezing rain and icing value is determined as the freezing rain and icing value of the wind turbine blades in the forecasting area at the preset time point in the future. Through the above technical solution, a scientific and reasonable method can be used to guide power grid companies to optimize power generation scheduling to reduce freezing rain and icing, and improve the accuracy of numerical prediction of freezing rain and icing on wind turbines.

[0040] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:

[0042] Figure 1 A schematic diagram of a flow chart of a method for long- and short-term memory correction and prediction of ice thickness on wind turbine blades according to an embodiment of the present application is shown;

[0043] Figure 2 A schematic diagram of the framework of a long- and short-term memory correction and prediction device for ice thickness on wind turbine blades according to an embodiment of the present application is shown;

[0044] Figure 3The internal structure of a computer device according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0046] Figure 1 The following schematically shows a flow chart of a data collection method for a winding machine according to an embodiment of the present application. Figure 1 As shown, in one embodiment of the present application, a method for long- and short-term memory correction and prediction of ice thickness of wind turbine blades is provided, comprising the following steps:

[0047] Step 202, determining the forecast area for which numerical forecast of freezing rain and icing is required.

[0048] Step 204, dividing the area to be forecasted into equidistant grids.

[0049] Step 206, obtaining initial field data and boundary condition data for numerical prediction of freezing rain and icing, the initial field data refers to the data at each grid point at the time when the numerical model starts integration, and the boundary condition data refers to the data in the boundary area of ​​the numerical model.

[0050] Step 208, interpolating the initial field data and the boundary condition data according to the grid of the area to be forecasted, so that the grid of the initial field data and the boundary condition data is consistent with the grid of the area to be forecasted.

[0051] Step 210, inputting the initial field data and boundary condition data consistent with the grid of the area to be forecasted into the numerical model to drive the numerical model to integrate forward and obtain the first freezing rain icing value of the wind turbine blades in the area to be forecasted at a preset time point in the future.

[0052] Step 212, inputting the first freezing rain ice coverage value into the long short-term memory neural network.

[0053] Step 214, obtain a second freezing rain and icing value obtained after the long short-term memory neural network corrects the first freezing rain and icing value, and determine the second freezing rain and icing value as the freezing rain and icing value of the wind turbine blades in the forecast area at a preset time point in the future.

[0054] Freezing rain and icing refer to supercooled water droplets with larger particle sizes, which are deposited on objects under the action of gravity and frozen into frost or mixed frost. First, the area where the numerical forecast of freezing rain and icing needs to be carried out can be determined, which is called the forecast area. Then, the forecast area can be divided into equidistant grids. For example, the area selected for the forecast of freezing rain and icing is Hunan Province. The specific longitude and latitude range of Hunan Province is 108° to 114° east longitude and 24° to 31° north latitude. The area can be divided into equidistant grids of 3 km × 3 km. Furthermore, the processor can obtain the initial field data and boundary condition data for the numerical prediction of freezing rain and icing. Among them, the initial field data refers to the data on each grid point at the time when the numerical model starts integration (i.e.), and the boundary condition data refers to the data on the grid points around the area calculated by the numerical model. The initial field data and boundary condition data are interpolated according to the grid of the area to be forecasted. Interpolation refers to interpolating a continuous function based on discrete data so that this continuous curve passes through all given discrete data points. Specifically, the initial field data of each grid is written into the corresponding grid so that the grid of the initial field data and boundary condition data is consistent with the grid of the area to be forecasted. Then, the initial field data and boundary condition data consistent with the grid of the area to be forecasted can be input into the numerical model. The numerical model refers to the use of a computer to apply the earth fluid dynamic equation and related physical laws to reproduce or predict the historical weather or future weather of the world or region, so as to drive the numerical model to integrate forward and calculate the meteorological element values ​​at each moment in the future, that is, to obtain the first freezing rain icing value of the wind turbine blades in the area to be forecasted at a preset time point in the future, and input the first freezing rain icing value into the long short-term memory neural network. Further, the second freezing rain icing value obtained by the long short-term memory neural network after correcting the first freezing rain icing value can be obtained, and the second freezing rain icing value is determined as the freezing rain icing value of the wind turbine blades in the area to be forecasted at a preset time point in the future.

[0055] In one embodiment, the long short-term memory neural network includes multiple, each long short-term memory neural network predicts data at different time scales, and the time scale includes at least one of the seasonal scale, the monthly scale, and the daily scale. The long short-term memory neural network predicts the data at the seasonal scale, which means that the long short-term memory neural network predicts the freezing rain and ice covering values ​​of the wind turbine blades in the area to be forecasted in the next three months based on the input data. The long short-term memory neural network predicts the data at the monthly scale, which means that the long short-term memory neural network predicts the freezing rain and ice covering values ​​of the wind turbine blades in the area to be forecasted in the next month based on the input data. The long short-term memory neural network predicts the data at the daily scale, which means that the long short-term memory neural network predicts the freezing rain and ice covering values ​​of the wind turbine blades in the area to be forecasted tomorrow based on the input data. After obtaining the long short-term memory neural network that can predict the data at each time scale, the second freezing rain and ice covering value obtained by correcting the first freezing rain and ice covering value by each long short-term memory neural network can be obtained. A weighted summation of multiple second freezing rain and icing values ​​is performed according to the weight of each long short-term memory neural network, and a third freezing rain and icing value is obtained after the weighted summation. The third freezing rain and icing value is determined as the freezing rain and icing value of the wind turbine blades in the area to be forecasted at a preset time point in the future.

[0056] In one embodiment, before the long short-term memory neural network is put into actual use, the long short-term memory neural network can be trained first. Specifically, the training steps of the long short-term memory neural network include: obtaining multiple historical freezing rain and ice coverage values ​​corresponding to different time scales, and the historical freezing rain and ice coverage values ​​are historical prediction values ​​obtained by inputting the initial field data and boundary condition data of each grid into the numerical model during the historical time period. The historical freezing rain and ice coverage values ​​of different time scales of each grid are respectively input into different long short-term memory neural networks to obtain a long short-term memory neural network that can predict the data of each time scale of each grid. Specifically, when training the long short-term memory neural network, the input data and output data of the long short-term memory neural network are shown in formula (1):

[0057] (1);

[0058] Among them, Y is the predicted value of freezing rain and icing on wind turbine blades output by the long short-term memory neural network, S is the season type of the year, M is the monthly mark, D is the date mark, Y1 is the predicted data for the current moment in the numerical mode, Y- is the predicted data for the previous moment in the numerical mode, and Re is the predicted value for the previous moment output by the long short-term memory neural network.

[0059] In one embodiment, a long short-term memory correction prediction method for wind turbine blade ice thickness, a long short-term memory neural network includes a forget gate , Memory Gate and output gate ; Among them, the forget gate determines the unit state at the previous moment The unit state preserved to the current moment The calculation formula is shown in formula (2):

[0060] (2);

[0061] in, represents the output of the forget gate, Represents the sigmoid layer. Each data needs to go through the sigmoid function. represents the weight matrix, is the module output at the previous moment, [ , ] represents the concatenation of two matrices. Represents bias.

[0062] The memory gate determines the current input The unit status The unit state preserved to the current moment The calculation formula is shown in formula (3) and formula (4):

[0063] (3);

[0064] (4);

[0065] The updated cell state is shown in formula (5):

[0066] (5);

[0067] in, is the input gate, is the current cell state, is the unit state at the previous moment, The current input cell state determines what information can be added to the cell state.

[0068] The calculation formula of the output gate is shown in formulas (6) and (7):

[0069] (6);

[0070] (7);

[0071] in, is the output gate, It is the module output at the current moment.

[0072] In one embodiment, inputting the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model includes: determining a future preset time point at which prediction is required; determining a calculation method for the initial field data and boundary condition data consistent with the grid of the area to be predicted based on the future preset time point; and inputting the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model in a manner set by the calculation method.

[0073] By means of the long and short-term memory correction forecasting method, device and processor of the ice thickness of the wind turbine blades, the forecasting area to be forecasted where the numerical forecast of freezing rain and icing is required is determined; the forecasting area to be forecasted is divided into equidistant grids; initial field data and boundary condition data for numerical forecasting of freezing rain and icing are obtained, the initial field data refers to the data on each grid point at the time when the numerical model starts integration, and the boundary condition data refers to the data in the boundary area of ​​the numerical model; the initial field data and boundary condition data are interpolated according to the grid of the forecasting area to make the grid of the initial field data and boundary condition data consistent with the grid of the forecasting area; the initial field data and boundary condition data consistent with the grid of the forecasting area are input into the numerical model to drive the numerical model to integrate forward, and a first freezing rain and icing value of the wind turbine blades in the forecasting area at a preset time point in the future is obtained; the first freezing rain and icing value is input into the long and short-term memory neural network; a second freezing rain and icing value obtained after the long and short-term memory neural network corrects the first freezing rain and icing value is obtained, and the second freezing rain and icing value is determined as the freezing rain and icing value of the wind turbine blades in the forecasting area at the preset time point in the future. Through the above technical solution, a scientific and reasonable method can be used to guide power grid companies to optimize power generation scheduling to reduce freezing rain and icing, and improve the accuracy of numerical prediction of freezing rain and icing on wind turbines.

[0074] In one embodiment, Figure 2 As shown, a long- and short-term memory correction and prediction device for ice thickness of wind turbine blades is provided, comprising:

[0075] The area selection module 201 is used to determine the forecast area where the freezing rain and icing numerical forecast needs to be carried out.

[0076] The data processing module 202 is used to divide the area to be forecasted into equidistant grids; obtain the initial field data and boundary condition data for numerical prediction of freezing rain and icing, the initial field data refers to the data on each grid point at the time when the numerical model starts integration, and the boundary condition data refers to the data in the boundary area of ​​the numerical model; interpolate the initial field data and boundary condition data according to the grid of the area to be forecasted, so that the grids of the initial field data and boundary condition data are consistent with the grids of the area to be forecasted.

[0077] The first prediction module 203 is used to input the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model to drive the numerical model to integrate forward and obtain the first freezing rain ice coverage value of the wind turbine blades in the area to be predicted at a preset time point in the future.

[0078] The second prediction module 204 is used to input the first freezing rain and icing value into the long short-term memory neural network; obtain the second freezing rain and icing value obtained after the long short-term memory neural network corrects the first freezing rain and icing value, and determine the second freezing rain and icing value as the freezing rain and icing value of the wind turbine blades in the area to be forecasted at a preset time point in the future.

[0079] In one embodiment, the long short-term memory neural network includes multiple ones, and each long short-term memory neural network predicts data at different time scales. The second prediction module 204 is also used to obtain a second freezing rain and ice coverage value obtained after each long short-term memory neural network corrects the first freezing rain and ice coverage value; weighted sum the multiple second freezing rain and ice coverage values ​​according to the weight of each long short-term memory neural network to obtain a third freezing rain and ice coverage value; determine the third freezing rain and ice coverage value as the freezing rain and ice coverage value of the wind turbine blades in the area to be predicted at a preset time point in the future.

[0080] In one embodiment, the time scale includes at least one of a seasonal scale, a monthly scale and a daily scale, wherein the long short-term memory neural network predicts the seasonal scale data by predicting the freezing rain and icing values ​​of the wind turbine blades in the area to be forecasted in the next three months based on the input data; the long short-term memory neural network predicts the monthly scale data by predicting the freezing rain and icing values ​​of the wind turbine blades in the area to be forecasted in the next month based on the input data; the long short-term memory neural network predicts the daily scale data by predicting the freezing rain and icing values ​​of the wind turbine blades in the area to be forecasted tomorrow based on the input data.

[0081] In one embodiment, the long-short-term memory correction and prediction device 200 for the ice coating thickness of wind turbine blades also includes a training module (not shown in the figure), which is used to obtain multiple historical freezing rain icing values ​​corresponding to different time scales, and the historical freezing rain icing values ​​are historical prediction values ​​obtained by inputting initial field data and boundary condition data into a numerical model during a historical time period; the historical freezing rain icing values ​​of different time scales are respectively input into different long-short-term memory neural networks to obtain a long-short-term memory neural network that can perform predictions for data of each time scale.

[0082] In one embodiment, when training the long short-term memory neural network, the input data and output data of the long short-term memory neural network are as shown in formula (1):

[0083] (1);

[0084] Among them, Y is the predicted value of freezing rain and icing on wind turbine blades output by the long short-term memory neural network, S is the season type of the year, M is the monthly mark, D is the date mark, Y1 is the predicted data for the current moment in the numerical mode, Y- is the predicted data for the previous moment in the numerical mode, and Re is the predicted value for the previous moment output by the long short-term memory neural network.

[0085] In one embodiment, the first prediction module 203 is also used to determine a preset future time point at which prediction is required; determine a parameterization scheme for initial field data and boundary condition data consistent with the grid of the area to be predicted based on the preset future time point; and input the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model in a manner set by the parameterization scheme.

[0086] An embodiment of the present invention provides a processor, which is used to run a program, wherein the program executes the long- and short-term memory correction and prediction method of ice thickness on wind turbine blades when running.

[0087] An embodiment of the present invention provides a storage medium having a program stored thereon, and when the program is executed by a processor, the long- and short-term memory correction and prediction method of ice coating thickness of a wind turbine blade is implemented.

[0088] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 3 As shown. The computer device includes a processor A01, a network interface A02, a display screen A04, an input device A05 and a memory (not shown in the figure) connected through a system bus. Among them, the processor A01 of the computer device is used to provide computing and control capabilities. The memory of the computer device includes an internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A06. The network interface A02 of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor A01, a long-short-term memory correction prediction method for the ice thickness of a wind turbine blade is implemented. The display screen A04 of the computer device can be a liquid crystal display or an electronic ink display, and the input device A05 of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0089] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0090] The embodiment of the present application provides a device, the device includes a processor, a memory, and a program stored in the memory and executable on the processor, and the processor implements the following steps when executing the program: determining an area to be predicted where numerical prediction of freezing rain and icing is required; dividing the area to be predicted into equidistant grids; obtaining initial field data and boundary condition data for numerical prediction of freezing rain and icing, the initial field data refers to data on each grid point at the time when the numerical model starts integration, and the boundary condition data refers to data in the boundary area of ​​the numerical model; interpolating the initial field data and the boundary condition data according to the grid of the area to be predicted, so that The grid of the initial field data and the boundary condition data is consistent with the grid of the area to be forecasted; the initial field data and the boundary condition data consistent with the grid of the area to be forecasted are input into the numerical model to drive the numerical model to integrate forward, and obtain a first freezing rain and icing value of the wind turbine blades in the area to be forecasted at a preset time point in the future; the first freezing rain and icing value is input into the long short-term memory neural network; the second freezing rain and icing value obtained by correcting the first freezing rain and icing value by the long short-term memory neural network is obtained, and the second freezing rain and icing value is determined as the freezing rain and icing value of the wind turbine blades in the area to be forecasted at the preset time point in the future.

[0091] In one embodiment, a long short-term memory correction and prediction method for ice coverage thickness of wind turbine blades comprises a plurality of long short-term memory neural networks, each of which predicts data at different time scales; obtaining a second freezing rain ice coverage value obtained after the long short-term memory neural network corrects the first freezing rain ice coverage value, and determining the second freezing rain ice coverage value as the freezing rain ice coverage value of the wind blades in the area to be predicted at a preset future time point, comprising: obtaining the second freezing rain ice coverage value obtained after each long short-term memory neural network corrects the first freezing rain ice coverage value; weightedly summing the plurality of second freezing rain ice coverage values ​​according to the weight of each long short-term memory neural network to obtain a third freezing rain ice coverage value; and determining the third freezing rain ice coverage value as the freezing rain ice coverage value of the wind blades in the area to be predicted at a preset future time point.

[0092] In one embodiment, a long short-term memory correction forecasting method for wind turbine blade ice thickness comprises a time scale including at least one of a seasonal scale, a monthly scale and a daily scale, wherein the long short-term memory neural network predicts seasonal scale data, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of wind turbine blades in the area to be forecasted in the next three months based on the input data; the long short-term memory neural network predicts monthly scale data, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of wind turbine blades in the area to be forecasted in the next month based on the input data; the long short-term memory neural network predicts daily scale data, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of wind turbine blades in the area to be forecasted tomorrow based on the input data.

[0093] In one embodiment, a long short-term memory correction prediction method for wind turbine blade ice coating thickness includes a step of training a long short-term memory neural network, the training step including: obtaining a plurality of historical freezing rain icing values ​​corresponding to different time scales, the historical freezing rain icing values ​​being historical prediction values ​​obtained by inputting initial field data and boundary condition data into a numerical model within a historical time period; and inputting the historical freezing rain icing values ​​of different time scales into different long short-term memory neural networks, respectively, to obtain a long short-term memory neural network that can perform predictions for data of each time scale.

[0094] In one embodiment, in a long short-term memory correction prediction method for wind turbine blade ice thickness, when training a long short-term memory neural network, the input data and output data of the long short-term memory neural network are as shown in formula (1):

[0095] (1);

[0096] Among them, Y is the predicted value of freezing rain and icing on wind turbine blades output by the long short-term memory neural network, S is the season type of the year, M is the monthly mark, D is the date mark, Y1 is the predicted data for the current moment in the numerical mode, Y- is the predicted data for the previous moment in the numerical mode, and Re is the predicted value for the previous moment output by the long short-term memory neural network.

[0097] In one embodiment, a long-short-term memory correction prediction method for ice thickness on wind turbine blades, inputting initial field data and boundary condition data consistent with the grid of the area to be predicted into a numerical model includes: determining a future preset time point required for prediction; determining a parameterization scheme for the initial field data and boundary condition data consistent with the grid of the area to be predicted according to the future preset time point; and inputting the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model in a manner set by the parameterization scheme.

[0098] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0099] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0101] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0102] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0103] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0104] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0105] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0106] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for long- and short-term memory correction and prediction of ice thickness of wind turbine blades, characterized in that: include: Determine the forecast areas where numerical forecast of freezing rain and icing is needed; Dividing the area to be forecasted into equidistant grids; Acquire initial field data and boundary condition data for numerical prediction of freezing rain and icing, wherein the initial field data refers to data at each grid point at the time when the numerical model starts integration, and the boundary condition data refers to data in the boundary area of ​​the numerical model; interpolating the initial field data and the boundary condition data according to the grid of the area to be forecasted, so that the grids of the initial field data and the boundary condition data are consistent with the grid of the area to be forecasted; Inputting initial field data and boundary condition data consistent with the grid of the area to be forecasted into the numerical model to drive the numerical model to integrate forward and obtain the first freezing rain ice coverage value of the wind turbine blades in the area to be forecasted at a preset time point in the future; Inputting the first freezing rain ice coverage value into a long short-term memory neural network; A second freezing rain and icing value obtained by correcting the first freezing rain and icing value by the long short-term memory neural network is obtained, and a third freezing rain and icing value obtained by weighted summation of the second freezing rain and icing values ​​is determined as the freezing rain and icing value of the wind turbine blades in the area to be forecasted at the future preset time point.

2. The long- and short-term memory correction and prediction method for ice thickness of wind turbine blades according to claim 1 is characterized in that: The long short-term memory neural network includes multiple ones, and each long short-term memory neural network predicts data of different time scales; The step of obtaining a second freezing rain icing value obtained by correcting the first freezing rain icing value by the long short-term memory neural network, and determining a third freezing rain icing value obtained by weighted summing the second freezing rain icing values ​​as the freezing rain icing value of the wind turbine blades in the to-be-forecasted area at the future preset time point includes: Obtaining a second freezing rain icing value obtained by correcting the first freezing rain icing value through each long short-term memory neural network; weighted summing the plurality of second freezing rain icing values ​​according to the weight of each long short-term memory neural network to obtain a third freezing rain icing value; The third freezing rain icing value is determined as the freezing rain icing value of the wind turbine blades in the to-be-forecasted area at the future preset time point.

3. The long- and short-term memory correction and prediction method for ice thickness of wind turbine blades according to claim 2 is characterized in that: The time scale includes at least one of a seasonal scale, a monthly scale and a daily scale, wherein the long short-term memory neural network predicts the data on the seasonal scale, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of the wind turbine blades in the area to be forecasted in the next three months based on the input data; the long short-term memory neural network predicts the data on the monthly scale, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of the wind turbine blades in the area to be forecasted in the next month based on the input data; the long short-term memory neural network predicts the data on the daily scale, which means that the long short-term memory neural network predicts the freezing rain and icing values ​​of the wind turbine blades in the area to be forecasted tomorrow based on the input data.

4. The long- and short-term memory correction and prediction method for ice thickness of wind turbine blades according to claim 2 is characterized in that: The method further comprises a step of training the long short-term memory neural network, wherein the training step comprises: Acquire a plurality of historical freezing rain and icing values ​​corresponding to different time scales, wherein the historical freezing rain and icing values ​​are historical prediction values ​​obtained by inputting initial field data and boundary condition data into a numerical model during a historical time period; The historical freezing rain and ice cover values ​​at different time scales are input into different long short-term memory neural networks respectively to obtain a long short-term memory neural network that can make predictions for the data at each time scale.

5. The long- and short-term memory correction and prediction method for ice thickness of wind turbine blades according to claim 4 is characterized in that: When the long short-term memory neural network is trained, the input data and output data of the long short-term memory neural network are as shown in formula (1): （1）； Among them, Y is the predicted value of freezing rain and icing on the wind turbine blades output by the long short-term memory neural network, S is the season type of the year, M is the monthly mark, D is the date mark, Y1 is the predicted data for the current moment in the numerical mode, Y- is the predicted data for the previous moment in the numerical mode, and Re is the predicted value for the previous moment output by the long short-term memory neural network.

6. The long- and short-term memory correction and prediction method for ice thickness of wind turbine blades according to claim 5 is characterized in that: The long short-term memory neural network includes a forget gate , Memory Gate and output gate ; Among them, the forget gate determines the unit state at the previous moment The unit state preserved to the current moment The calculation formula is shown in formula (2): （2）； in, represents the output of the forget gate, Represents the sigmoid layer. Each data needs to pass through the sigmoid function. represents the weight matrix, is the module output at the previous moment, [ , ] represents the concatenation of two matrices. represents bias; The memory gate determines the The unit status The unit state preserved to the current moment The calculation formula is shown in formula (3) and formula (4): （3）； （4）； The updated cell state is shown in formula (5): （5）； in, is the input gate, is the current cell state, is the unit state at the previous moment, The current input cell status; The calculation formula of the output gate is shown in formulas (6) and (7): （6）； （7）； in, is the output gate, It is the module output at the current moment.

7. The long- and short-term memory correction and prediction method for ice thickness of wind turbine blades according to claim 1 is characterized in that: The step of inputting the initial field data and boundary condition data consistent with the grid of the area to be forecasted into the numerical model comprises: Determine the preset future time point for which the forecast needs to be made; Determine a parameterization scheme for initial field data and boundary condition data consistent with a grid of the area to be forecasted according to the future preset time point; The initial field data and boundary condition data consistent with the grid of the area to be forecasted are input into the numerical model in a manner set by the parameterization scheme.

8. A long- and short-term memory correction and prediction device for ice thickness of wind turbine blades, characterized in that: include: The area selection module is used to determine the forecast area where the freezing rain and ice numerical forecast needs to be carried out; A data processing module is used to divide the area to be forecasted into equidistant grids; obtain initial field data and boundary condition data for numerical prediction of freezing rain and icing, wherein the initial field data refers to data on each grid point at the time when the numerical model starts integration, and the boundary condition data refers to data in the boundary area of ​​the numerical model; interpolate the initial field data and the boundary condition data according to the grid of the area to be forecasted, so that the grids of the initial field data and the boundary condition data are consistent with the grids of the area to be forecasted; A first prediction module is used to input the initial field data and boundary condition data consistent with the grid of the area to be predicted into the numerical model to drive the numerical model to integrate forward to obtain the first freezing rain ice coverage value of the wind turbine blades in the area to be predicted at a preset time point in the future; A second prediction module, used for inputting the first freezing rain ice coverage value into a long short-term memory neural network; A second freezing rain and icing value obtained by correcting the first freezing rain and icing value by the long short-term memory neural network is obtained, and a third freezing rain and icing value obtained by weighted summation of the second freezing rain and icing values ​​is determined as the freezing rain and icing value of the wind turbine blades in the area to be forecasted at the future preset time point.

9. A processor, characterized in that: The method is configured to execute the long-short term memory correction prediction method of ice thickness on wind turbine blades according to any one of claims 1 to 7.

10. A machine-readable storage medium having instructions stored thereon, characterized in that: When the instruction is executed by a processor, the processor is configured to execute the long- and short-term memory correction prediction method for ice thickness on wind turbine blades according to any one of claims 1 to 7.

Images