Multi-fan matrix-oriented space wind speed modeling prediction method and system

Abstract

The invention relates to the technical field of wind speed prediction, in particular to a space wind speed modeling prediction method and system for a multi-fan matrix, and the modeling prediction method comprises the steps: building a three-row and three-column initial measurement fan matrix, respectively setting a plurality of measurement points in front of a plurality of set mapping positions on a central fan at equal intervals, the wind speed mean value and the turbulence degree corresponding to the measuring points are obtained under different fan rotating speed combinations of the initial measuring fan matrix, a measuring point-wind speed-turbulence degree sample library is constructed and used for training a multi-fan matrix prediction model, and the multi-fan matrix prediction model can predict the wind speed and the turbulence degree of any position in the k-row k-column multi-fan matrix; according to the modeling prediction system, the prediction modeling method is applied, enough fan rotating speed-wind speed distribution sample data are obtained within finite time, a high-precision multi-fan matrix prediction model is trained and obtained, the multi-fan matrix prediction model is applied to larger-scale multi-fan matrix configuration, and the applicability is high.

Description

Technical Field

[0001] The present invention relates to the technical field of wind speed prediction, and in particular to a spatial wind speed modeling and prediction method and system for a multi-fan matrix. Background Art

[0002] In the application of a multi-fan wind wall or a fan array, if it is necessary to obtain the wind field distribution formed by different gear speeds, it is necessary to debug the speed of each fan one by one and measure the wind speed at the corresponding position. However, when there are a large number of fans in the multi-fan wind wall or fan matrix and the fan speeds are adjustable in multiple ways, the test volume of the wind field distribution is huge and time-consuming. The wind speed distribution is not simply determined by the superposition of the speeds of all fans, but is dominated by the cooperative action of the fans in a specific area. When it is necessary to obtain the wind speed at a specific position in the wind field, the measurement efficiency is low. Summary of the Invention

[0003] The purpose of the present invention is to provide a spatial wind speed modeling and prediction method and system for a multi-fan matrix to solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solutions:

[0005] A spatial wind speed modeling and prediction method for a multi-fan matrix includes the following steps:

[0006] Build an initial measurement fan matrix with three rows and three columns, where the initial measurement fan matrix includes a central fan and 8 peripheral fans surrounding the central fan;

[0007] Adjust the speed of each fan in the initial measurement fan matrix, set a number of measuring points at a preset distance in front of multiple set mapping positions on the central fan of the initial measurement fan matrix, obtain the average wind speed and turbulence intensity of the measuring points, and form a measuring point - wind speed - turbulence intensity sample database;

[0008] Build a ConvLSTM–Transformer deep learning model, divide the measuring point - wind speed - turbulence intensity sample database into a training set and a validation set, input the sample data of the training set into the deep learning model for training, the deep learning model outputs a training prediction result, compare the training prediction result with the validation set, optimize according to the comparison result, and output a multi-fan matrix prediction model corresponding to the wind speed of any measuring point directly in front of any fan in the measurement fan matrix;

[0009] In a multi - fan matrix with k rows and k columns, the fan facing the target prediction point is used as the prediction fan, where k > 3. The three - row and three - column fan matrix centered on the prediction fan is used as the current measurement fan matrix. In the multi - fan matrix prediction model, the fan speed of each fan in the current measurement fan matrix and the distance from the target prediction point to the prediction fan are input to obtain the wind speed and turbulence intensity of the target prediction point.

[0010] Preferably, each fan in the initial measurement fan matrix and the multi - fan matrix is set with m kinds of fan speeds, and the total number of fan speed combinations of the initial measurement fan matrix is m 3×3 species.

[0011] Preferably, in the initial measurement fan matrix, each fan is labeled and encoded in row - column order. A number of measurement points are set at equal intervals in front of a set position on the central fan and numbered. The mean wind speed and turbulence intensity of each measurement point within the set acquisition period t are obtained and recorded. Among them, multiple set mapping positions include the center of the central fan, and set mapping positions are arranged in a cross shape centered on the center of the central fan around the central fan.

[0012] Preferably, for the mean wind speed and turbulence intensity of n wind speed values collected by each measurement point within the set acquisition period t, the mean wind speed The calculation formula is:

[0013]

[0014] The wind speed standard deviation σ u The calculation formula is:

[0015]

[0016] The turbulence intensity TI calculation formula is:

[0017]

[0018] Among them, u i represents the wind speed value obtained at a certain moment within the current acquisition period t.

[0019] Preferably, for the construction of the measurement point - wind speed - turbulence intensity sample library: after normalizing the mean wind speed and turbulence intensity of each measurement point within the set acquisition period t, a measurement point - wind speed - turbulence intensity sample of the measurement point under the set fan speed combination of the initial measurement fan matrix is formed. The measurement point - wind speed - turbulence intensity samples corresponding to all measurement points in the initial measurement fan matrix under different fan speed combinations are integrated and stored in the measurement point - wind speed - turbulence intensity sample library in the form of a structured data table.

[0020] Preferably, for the multi - fan matrix prediction model to obtain the wind speed and turbulence intensity of the target prediction point, the following steps are included:

[0021] Taking the bottom fan endpoint of the multi-fan matrix as the origin, with the length direction of the multi-fan matrix as the x-axis, the width direction of the multi-fan matrix as the y-axis, and the normal direction of the bottom fan endpoint of the multi-fan matrix as the z-axis, and each fan being 1 unit, a three-dimensional coordinate system is constructed, and the real-world coordinates of the target prediction point are set as (x, y, z).

[0022] Map the target prediction point onto the multi-fan matrix, mark the fan where the mapped point is located as the predicted fan, take the three-row and three-column fan matrix centered on the predicted fan as the current measurement fan matrix, and extract the fan speeds of each fan in the current measurement fan matrix.

[0023] Input the distance z of the target prediction point relative to the multi-fan matrix and the fan speeds of the current measurement fan matrix into the multi-fan matrix prediction model.

[0024] Output the predicted wind speed and predicted turbulence intensity of the target prediction point.

[0025] A spatial wind speed modeling and prediction system for a multi-fan matrix, which is applied to the above-mentioned spatial wind speed modeling and prediction method for a multi-fan matrix, includes:

[0026] The data acquisition and management module is configured to collect the wind speed values of several measurement points in the wind field generated by the initial measurement fan matrix regulated by the monitoring fan regulation system at a set acquisition period t, calculate the wind speed mean and turbulence intensity, integrate the measurement point - wind speed - turbulence intensity samples corresponding to all measurement points in the initial measurement fan matrix under different fan speed combinations in the form of a structured data table, and construct a measurement point - wind speed - turbulence intensity sample database for storage.

[0027] The fan regulation module is configured to regulate the fan speeds of the initial measurement fan matrix and the multi-fan matrix.

[0028] The multi-fan matrix prediction model module is configured to input the training set and validation set divided from the measurement point - wind speed - turbulence intensity sample database into the constructed deep learning model of ConvLSTM–Transformer, and train and optimize it to obtain the multi-fan matrix prediction model.

[0029] The multi-fan matrix prediction model application module is configured to obtain the corresponding predicted fan and the current measurement fan matrix according to the set target prediction point, input the distance from the target prediction point to the predicted fan and the fan speeds of the current measurement fan matrix into the multi-fan matrix prediction model, and output the predicted wind speed and predicted turbulence intensity of the target prediction point.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] A spatial wind speed modeling and prediction method for a multi-fan matrix provided by the present invention sets various fan speed combinations on an initial measurement fan matrix, sets multiple set mapping positions on the central fan of the initial measurement fan matrix, and sets several measurement points at equal intervals in front of the multiple set mapping positions respectively. The several measurement points in the wind field generated by the initial measurement fan matrix are fully monitored to obtain the wind speed mean value and turbulence intensity, construct a database covering the wind speed mean value and turbulence intensity of all set measurement points under various different fan speed combinations, and use a deep learning model of ConvLSTM–Transformer for training and optimization to obtain a multi-fan matrix prediction model, realizing accurate prediction of the wind speed and turbulence intensity at any position in a larger multi-fan matrix and effectively reducing the test volume of the multi-fan matrix.

[0032] A spatial wind speed modeling and prediction system for a multi-fan matrix provided by the present invention is applied to a spatial wind speed modeling and prediction method for a multi-fan matrix, can obtain sufficient fan speed - wind speed distribution sample data within a limited time, train and obtain a high-precision wind wall wind speed prediction model, and widely apply the wind wall wind speed prediction model to a multi-fan matrix. Description of the Drawings

[0033] Figure 1 It is a schematic flow chart of a spatial wind speed modeling and prediction method for a multi-fan matrix according to an embodiment of the present invention;

[0034] Figure 2 is Figure 1 a schematic flow chart of the S4 process in

[0035] Figure 3 It is a schematic structural diagram of an initial measurement fan matrix with three rows and three columns according to an embodiment of the present invention;

[0036] Figure 4 It is a schematic structural diagram of a multi-fan matrix according to an embodiment of the present invention;

[0037] Figure 5 It is a schematic connection diagram of a spatial wind speed modeling and prediction system for a multi-fan matrix according to an embodiment of the present invention.

[0038] Reference numerals: 1, central fan; 2, peripheral fan; 3, multi-fan matrix. Detailed Embodiments

[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] In the description of the following invention, it should be noted that the orientation or positional relationship indicated by terms such as "upper", "lower", "left", "right", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation. The term "connection" only represents the connection between devices and has no special meaning.

[0041] In addition, the technical fields and installation methods involved in the embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0042] Specific embodiments: Please refer to Figures 1-4 , a spatial wind speed modeling and prediction method for a multi-fan matrix, including the following steps:

[0043] S1: Build an initial measurement fan matrix with three rows and three columns. The initial measurement fan matrix includes a central fan 1 and 8 peripheral fans 2 surrounding the central fan 1.

[0044] Specifically, the initial measurement fan matrix numbers each fan in order of rows and columns from row1-col1 to row3-col3.

[0045] S2: Adjust the fan speed of each fan in the initial measurement fan matrix. At a preset distance in front of multiple set mapping positions on the central fan of the initial measurement fan matrix, set a number of measurement points respectively. Obtain the mean wind speed and turbulence intensity of the measurement points to form a measurement point - wind speed - turbulence intensity sample database.

[0046] Specifically, multiple set mapping positions are set on the central fan 1 of the initial measurement fan matrix. Among them, the multiple set mapping positions include the center of the central fan, and with the center of the central fan as the origin, set mapping positions are arranged around the central fan. The multiple set mapping positions are distributed in a cross shape. In this embodiment, the multiple set mapping positions are arranged in three rows and three columns. A number of measurement points are set at equal distances directly in front of the multiple set mapping positions. The measurement points are encoded respectively in the format of r-c-d, where r represents the row number of the set mapping position corresponding to each measurement point, c represents the column number of the set mapping position corresponding to each measurement point, and d represents the distance between each measurement point and the initial measurement fan matrix. Each measurement point corresponds to a wind speed value and a turbulence intensity value. A number of anemometer sensors are installed at a number of measurement points respectively to detect all wind speed data of the measurement points within a collection period t.

[0047] Among them, each fan in the initial measurement fan matrix is set with m fan speeds. In this embodiment, m = 6, and six fan speeds are set, namely 0, 20, 40, 60, 80, and 100. Both the central fan 1 and the peripheral fan 2 are driven by an adjustable motor. The total number of fan speed combinations in the initial measurement fan matrix is m 3×3 kinds. Record the mean wind speed and turbulence intensity measured at each measurement point position in the wind fields generated by various fan speed combinations during the current acquisition period t, and calculate the mean wind speed and turbulence intensity.

[0048] Among them, the mean wind speed The calculation formula is:

[0049]

[0050] Among them, u i represents the wind speed value obtained at a certain moment during the current acquisition period t, and n represents the number of wind speed values collected during the current acquisition period t.

[0051] The standard deviation of wind speed σ u The calculation formula is:

[0052]

[0053] The calculation formula for turbulence intensity TI is:

[0054]

[0055] Specifically, the fan matrix generates a wind field under the currently set fan speed combination. The current acquisition period t is set to 1 minute. A number of anemometer sensors respectively collect n wind speed values at each measurement point. Calculate the mean wind speed and turbulence intensity according to the n wind speed values, and perform normalization processing on the mean wind speed and turbulence intensity to form a measurement point - wind speed - turbulence intensity sample at the set fan speed combination of the initial measurement fan matrix; construct a measurement point - wind speed - turbulence intensity sample library, integrate the measurement point - wind speed - turbulence intensity samples corresponding to all measurement points under different fan speed combinations in the form of a structured data table, and store them in the measurement point - wind speed - turbulence intensity sample library.

[0056] S3: Build a ConvLSTM–Transformer deep learning model, divide the measurement point - wind speed - turbulence intensity sample database into a training set and a validation set, input the sample data of the training set into the deep learning model for training, the deep learning model outputs the training prediction results, compare the training prediction results with the validation set, optimize according to the comparison results, and output a multi - fan matrix prediction model corresponding to the wind speed and turbulence intensity at any measurement point directly in front of any fan in the prediction fan matrix.

[0057] Specifically, a deep learning model of ConvLSTM–Transformer is built. The measurement point - wind speed - turbulence intensity sample database is divided into a training set and a validation set. The training set is input into the deep learning model for training. The deep learning model outputs the training prediction results, which are compared with the validation set. According to the comparison results, verification and optimization are carried out, and a multi - fan matrix prediction model is output.

[0058] Among them, the training set inputs the measurement point - wind speed - turbulence intensity samples. The ConvLSTM layer extracts spatio - temporal features and outputs a feature sequence. The Transformer layer processes the output of ConvLSTM to capture global time dependencies. The fully - connected layer maps to the predicted target wind speed value. The training prediction results output by the deep learning model are compared with the validation set. The RMSE (root mean square error) and MAE (mean absolute error) are respectively used to evaluate the multi - fan matrix prediction model. The error range is set to 0 - 5%. If the error requirements are met, the multi - fan matrix prediction model can be obtained. If the error requirements are not met, other measurement points are set on the initial measurement matrix for a small amount of measurement, and the temperature, air pressure data, etc. at the measurement points are used as additional input features to correct or fine - tune the model parameters and optimize the multi - fan matrix prediction model.

[0059] S4: In the k - row and k - column multi - fan matrix, the fan facing the target prediction point is used as the prediction fan, where k > 3. The three - row and three - column fan matrix centered on the prediction fan is used as the current measurement fan matrix. The fan speeds of each fan in the current measurement fan matrix and the distance from the target prediction point to the prediction fan are input into the multi - fan matrix prediction model to obtain the wind speed at the target prediction point. The steps are as follows:

[0060] S41: Taking the bottom - fan endpoint of the multi - fan matrix 3 as the origin, the length direction of the multi - fan matrix 3 as the x - axis, the width direction of the multi - fan matrix 3 as the y - axis, and the normal direction of the bottom - fan endpoint of the multi - fan matrix 3 as the z - axis, a three - dimensional coordinate system is constructed, and the real - world coordinates of the target prediction point are set as (x, y, z).

[0061] S42: The target prediction point is mapped onto the multi - fan matrix 3. The fan where the mapped point is located is marked as the prediction fan. The three - row and three - column fan matrix centered on the prediction fan is used as the current measurement fan matrix, and the fan speeds of each fan in the current measurement fan matrix are extracted.

[0062] Specifically, the target prediction point is any point in the space directly in front of the multi-fan matrix 3. A perpendicular line is drawn from the target prediction point to the multi-fan matrix 3, and the fan corresponding to the foot of the perpendicular is used as the prediction fan. Since the length and width of each fan are fixed, the length and width of the position where the prediction fan is located can be obtained. The x and y values of the target prediction point are respectively compared with the length and width of the position where the prediction fan is located. According to the comparison results, it is checked whether there are 8 circumferential fans around the prediction fan. If there are, the current measurement fan matrix can be directly formed, and the fan speeds of each fan in the current measurement fan matrix are extracted;

[0063] If there are not 8 circumferential fans around the prediction fan, virtual fans are set at the positions where the circumferential fans are missing to indirectly form the current measurement fan matrix, and the fan speeds of each fan in the current measurement fan matrix are extracted, where the rotational speed of the virtual fan is set to 0.

[0064] S43: Input the distance z of the target prediction point relative to the multi-fan matrix 3 and the fan speeds of the current measurement fan matrix into the multi-fan matrix prediction model;

[0065] S44: Output the predicted wind speed and predicted turbulence intensity of the target prediction point;

[0066] A wind speed sensor is set at the target prediction point to calculate the mean measured wind speed and measured turbulence intensity at the target prediction point, and compare them with the predicted wind speed and predicted turbulence intensity. The differences between the mean measured wind speed and the predicted wind speed and between the measured turbulence intensity and the predicted turbulence intensity are both within the error range of 0-5%, to verify the prediction accuracy of the multi-fan matrix prediction model.

[0067] Please refer to Figure 5 , a spatial wind speed modeling and prediction system for a multi-fan matrix, which is applied to the above-mentioned spatial wind speed modeling and prediction method for a multi-fan matrix, and includes:

[0068] A data acquisition and management module, which acquires the wind speed values of several measuring points in the wind field generated by the initial measurement fan matrix regulated by the monitoring fan regulation system at a set fan speed combination within a set acquisition period t, calculates the mean wind speed and turbulence intensity, integrates the sample of measuring point - wind speed - turbulence intensity corresponding to all measuring points in the initial measurement fan matrix under different fan speed combinations in the form of a structured data table, and constructs a measuring point - wind speed - turbulence intensity sample database for storage;

[0069] A fan regulation module, which regulates the fan speeds of the initial measurement fan matrix and the multi-fan matrix 3; among them, the fan regulation system is connected to the regulating motor, and the fan regulation system can perform single-point regulation on the central fan 1 or the peripheral fan 2 in the initial measurement fan matrix respectively.

[0070] The multi-fan matrix prediction model module divides the measurement point - wind speed - turbulence intensity sample database into a training set and a validation set, inputs the established deep learning model of ConvLSTM–Transformer, trains and optimizes it to obtain a multi-fan matrix prediction model;

[0071] The multi-fan matrix prediction model application module sets the three-dimensional spatial coordinates of the target prediction point, obtains the prediction fan corresponding to the target prediction point and the current measured fan matrix, inputs the distance from the target prediction point to the prediction fan and the fan speeds of the current measured fan matrix into the multi-fan matrix prediction model, and outputs the predicted wind speed and predicted turbulence intensity of the target prediction point.

Claims

1. A spatial wind speed modeling and prediction method for a multi-fan matrix, characterized in that: It includes the following steps: Build an initial measurement fan matrix with three rows and three columns. The initial measurement fan matrix includes a central fan and 8 peripheral fans surrounding the central fan; Adjust the rotational speed of each fan in the initial measurement fan matrix. At a preset distance in front of multiple set mapping positions on the central fan of the initial measurement fan matrix, set a number of measurement points respectively. Obtain the mean wind speed and turbulence intensity of the measurement points to form a measurement point - wind speed - turbulence intensity sample database; Build a ConvLSTM–Transformer deep learning model. Divide the measurement point - wind speed - turbulence intensity sample database into a training set and a validation set. Input the sample data of the training set into the deep learning model for training. The deep learning model outputs a training prediction result. Compare the training prediction result with the validation set and optimize according to the comparison result to output a multi - fan matrix prediction model corresponding to the wind speed of any measurement point directly in front of any fan in the measurement fan matrix; In the k - row and k - column multi - fan matrix, the fan directly opposite the target prediction point is used as the prediction fan, where k > 3. And take the three - row and three - column fan matrix centered on the prediction fan as the current measurement fan matrix. Input the rotational speed of each fan in the current measurement fan matrix and the distance from the target prediction point to the prediction fan into the multi - fan matrix prediction model to obtain the wind speed and turbulence intensity of the target prediction point.

2. The spatial wind speed modeling and prediction method for a multi-fan matrix according to claim 1, characterized in that: Each fan in the initial measurement fan matrix and the multi-fan matrix is set with m fan speeds, and the total number of fan speed combinations of the initial measurement fan matrix is m 3×3 species.

3. A spatial wind speed modeling and prediction method for a multi-fan matrix according to claim 1, characterized in that: In the initial measurement fan matrix, each fan is labeled and encoded in the order of rows and columns. A number of measurement points are set at equal intervals in front of the set positions on the central fan respectively. Obtain the mean wind speed and turbulence intensity of each measurement point within the set acquisition period t and record them. Among them, the multiple set mapping positions include the center of the central fan, and with the center of the central fan as the origin, a cross - shaped distribution is arranged around the central fan to set the mapping positions.

4. The spatial wind speed modeling and prediction method for a multi-fan matrix according to claim 3, characterized in that: The average wind speed and turbulence intensity of the n wind speed values collected at each measurement point within the set collection period t, the average wind speed The calculation formula is as follows: Standard deviation of wind speed σ u The calculation formula is as follows: The calculation formula for turbulence intensity TI is: Among them, u i represents the wind speed value obtained at a certain moment within the current acquisition period t.

5. A spatial wind speed modeling and prediction method for a multi-fan matrix according to claim 4, characterized in that: The construction of the measurement point - wind speed - turbulence intensity sample library: After normalizing the mean wind speed and turbulence intensity of each measurement point within the set acquisition period t, form a measurement point - wind speed - turbulence intensity sample of the measurement point under the set fan rotational speed combination of the initial measurement fan matrix. The measurement point - wind speed - turbulence intensity samples corresponding to all measurement points in the initial measurement fan matrix under different fan rotational speed combinations are integrated and stored in the measurement point - wind speed - turbulence intensity sample library in the form of a structured data table.

6. The spatial wind speed modeling and prediction method for a multi-fan matrix according to claim 1, characterized in that: The multi - fan matrix prediction model obtains the wind speed and turbulence intensity of the target prediction point, including the following steps: Taking the bottom fan end point of the multi - fan matrix as the origin, with the length direction of the multi - fan matrix as the x - axis, the width direction of the multi - fan matrix as the y - axis, and the normal direction of the bottom fan end point of the multi - fan matrix as the z - axis, construct a three - dimensional coordinate system. Set the real - world coordinates of the target prediction point as (x, y, z); Map the target prediction point onto the multi - fan matrix. Mark the fan where the mapped point is located as the prediction fan. Take the three - row and three - column fan matrix centered on the prediction fan as the current measurement fan matrix, and extract the rotational speed of each fan in the current measurement fan matrix; Input the distance z of the target prediction point relative to the multi - fan matrix and the rotational speed of the fans in the current measurement fan matrix into the multi - fan matrix prediction model; Output the predicted wind speed and predicted turbulence intensity of the target prediction point.

7. A spatial wind speed modeling and prediction system for a multi-fan matrix, which is applied to the method for modeling and predicting the spatial wind speed of a multi-fan matrix according to any one of claims 1-6 above, characterized in that: Including: A data acquisition and management module, configured to collect the wind speed values of several measurement points in the wind field generated by the initial measurement fan matrix of the monitoring fan control system at a set fan speed combination within a set acquisition period t, calculate the average wind speed and turbulence intensity, integrate the measurement point - wind speed - turbulence intensity samples corresponding to all measurement points in the initial measurement fan matrix under different fan speed combinations in the form of a structured data table, and construct a measurement point - wind speed - turbulence intensity sample database for storage; A fan control module, configured to control the fan speed of the initial measurement fan matrix and the fan speed of the multi - fan matrix; A multi - fan matrix prediction model module, configured to input the training set and validation set divided from the measurement point - wind speed - turbulence intensity sample database into the constructed deep - learning model of ConvLSTM–Transformer, and train and optimize it to obtain a multi - fan matrix prediction model; A multi - fan matrix prediction model application module, configured to obtain the corresponding prediction fan and the current measurement fan matrix according to the set target prediction point, input the distance from the target prediction point to the prediction fan and the fan speed of the current measurement fan matrix into the multi - fan matrix prediction model, and output the predicted wind speed and predicted turbulence intensity of the target prediction point.

Images