Multi-information source fusion ballistic wind identification method
Through the method of multi-information source fusion, combining the thermal wind formula, SPSS, ELM and Elman networks, a multi-model integrated wind field model is constructed, which solves the problem of insufficient timeliness and accuracy of wind field modeling in the existing technology, and realizes artillery combat Fast and accurate wind speed acquisition in the process reduces cost and noise impact.
Patent Information
- Application Number
- CN202411091917.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-09
AI Technical Summary
The existing wind farm modeling methods are insufficient in artillery combat, and are sensitive to data noise and outliers, which cannot meet the requirement of quickly and accurately obtaining the full ballistic wind speed.
Using multi-information source fusion method, through technologies such as thermal wind formula, SPSS analysis, extreme learning machine (ELM) and recurrent neural network (Elman), a multi-model integrated wind field model is constructed, combining atmospheric dynamics and statistics to reduce the Noise influences, improving the applicability of the time and space domains.
It achieves rapid, accurate and low-cost acquisition of wind direction and wind speed, improves the accuracy and applicability of wind field prediction in artillery operations, reduces costs, and is suitable for the fast and accurate combat needs of artillery.
Smart Images

Figure CN118917205A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of exterior ballistics, and in particular relates to a ballistic wind identification method by fusing multiple information sources. Background Art
[0002] Ignoring muzzle disturbances, wind direction and wind speed have become the primary factors affecting the range and accuracy of artillery (accounting for 70%). In current practical projects, the means of obtaining wind direction and wind speed are mainly physical measurement methods represented by sounding balloons and wind profile radars. The above physical measurement methods have low timeliness, limited use, and high costs. The detection position cannot cover the entire ballistic area and cannot meet the requirements of future war operations. With the development of technologies such as meteorology, statistics, and artificial intelligence, building wind field models has become the mainstream technical means to quickly and accurately obtain wind direction and wind speed. According to different modeling principles, wind field modeling methods can be roughly divided into numerical methods of atmospheric dynamics, statistical modeling methods, and data-driven methods.
[0003] The numerical method of atmospheric dynamics constructs a physical model of wind speed and direction through the time and space integration of conservation equations such as mass, momentum, heat, water vapor and aerosol. Although it can efficiently simulate wind field changes, its model is based on a large number of assumptions and has high computational complexity. The statistical modeling rule is based on historical meteorological data and fits the time and space variation patterns of wind direction and wind speed. The wind field model obtained by this method reflects the variation patterns of the average direction and average wind speed, and due to the errors caused by statistics itself, it is not fully applied to historical meteorological data. The data-driven rule is based on good generalization performance such as neural networks to construct mapping relationships and realize wind field prediction, but the prediction accuracy of this method is too dependent on the number and quality of samples.
[0004] From the perspective of research content, existing modeling methods mainly focus on the changing laws of wind fields in the time domain. However, in actual combat, the flight time of artillery is short and the spatial span is large. The time interval span of existing wind field models is too large for artillery combat. From the perspective of the number of models, a single model is currently used to predict wind fields. A single model is sensitive to the distribution and feature selection of data and is easily affected by data noise and outliers. From the perspective of the type of data used, the wind direction and wind speed within the entire trajectory range have both vertical spatial distribution characteristics under different atmospheric boundary layer types and time series change characteristics. Existing studies mostly use other meteorological data to predict wind field changes, which are not closely combined with the characteristics of the artillery itself. Summary of the invention
[0005] The purpose of the present invention is to provide a ballistic wind identification method integrating multiple information sources, so as to provide a fast, accurate and low-cost way to obtain wind direction and wind speed for artillery combat.
[0006] The technical solution to achieve the purpose of the present invention is: a multi-information source fusion ballistic wind identification method, comprising the following steps:
[0007] Step (1): Use the thermal wind formula to construct wind field model I;
[0008] Step (2): Use SPSS to analyze historical data and construct wind field model II;
[0009] Step (3): Using data-driven approach, construct wind farm model III based on extreme learning machine ELM;
[0010] Step (4): The wind direction and wind speed data obtained by the wind field model in steps (1) to (3) and the actual wind direction and wind speed are taken as new data sets, normalized and preprocessed, and then the process proceeds to step (5);
[0011] Step (5): Based on the stacking concept, the recursive neural network Elman learning method is used to learn the multi-information source data preprocessed in step (4) to establish the final wind field model and identify the ballistic wind.
[0012] Furthermore, the wind field model I established in step (1) includes a wind speed model I and a wind direction model I;
[0013] Wind speed model I is as follows:
[0014]
[0015] Where u is the wind speed, h is the height, h1 is the sea level height, u1 is the wind speed at sea level, z0 is the roughness parameter, u g is the zonal wind speed, v g is the radial wind speed, f is the geostrophic parameter, k1 is the turbulence coefficient, ω is the geostrophic angular velocity, Λ is the local dimension, and f * For position potential;
[0016] Wind direction model I is as follows:
[0017] Within 100m near the ground: wind direction is consistent with average wind speed
[0018] Upper boundary layer 100m~1500m:
[0019]
[0020] Free atmosphere 1500m~20000m:
[0021]
[0022] where ψ is the wind direction,
[0023] Furthermore, step (2) preprocessing of historical meteorological data: selecting meteorological wind data of the target point within n days in history, obtaining the average values of wind direction and wind speed data, and obtaining a wind direction and wind speed change curve; constructing a wind direction and wind speed model by using SPSS segment fitting according to the concave and convex points on the wind direction and wind speed change curve;
[0024] Wind speed model II is as follows:
[0025]
[0026] Wind direction model II is as follows:
[0027]
[0028] Furthermore, the wind field model III constructed in step (3) is:
[0029] When given N sets of training samples For the ELM with the model structure of ILO, where I, L, and O are the number of neurons in the input layer, hidden layer, and output layer respectively, and the activation function is h(x), the output of ELM is:
[0030]
[0031]
[0032]
[0033]
[0034] in is the connection weight matrix between the hidden layer and the output layer, H L×N is the hidden layer output matrix, and b L×N is a randomly generated structural parameter matrix;
[0035] Optimize through training Make the network output as close to the actual output as possible, that is
[0036]
[0037] Based on the least squares criterion, the connection weight matrix between the hidden layer and the output layer is
[0038]
[0039] In the formula H L×N The Moor-Penrose generalized inverse matrix.
[0040] Furthermore, step (3) specifically includes the following steps:
[0041] Step (31): Data normalization: Use the min-max normalization formula to preprocess the data:
[0042]
[0043] where r * is the normalized data, r min is the minimum value of the data, r max For data;
[0044] Step (32): Construct the ELM network model: determine the number of neurons in the input layer I, the number of neurons in the output layer O, the number of neurons in the hidden layer L, and the activation function h(x):
[0045] Number of neurons in the input layer I
[0046] ELM is used to construct a mapping relationship between ballistic data and meteorological wind. The input layer neurons are used to record the ballistic data. The ballistic data includes the artillery range x, elevation y, side deviation z, and the velocity component v in the corresponding direction. x ,v y ,v z and the velocity azimuth v θ ; Therefore, the number of neurons in the ELM input layer is I = 7;
[0047] The number of neurons in the output layer is O
[0048] The ELM output layer is used to output the ballistic wind data that needs to be identified: wind speed u and wind direction ψ, so the number of neurons in the ELM output layer is O = 2;
[0049] The number of neurons in the hidden layer is L
[0050] Combining the exponential growth strategy with the dichotomy method, the range of L is [L min ,L max ]=[2 j-1 ,2 j ], the binary method continuously narrows the interval range until the final number of hidden layer neurons L is determined;
[0051] Activation function h(x)
[0052] The trajectory data is mapped to the [0,1] interval after normalization, and the Sigmoid function is selected as the activation function. The expression of the Sigmoid function is as follows:
[0053]
[0054] Furthermore, in step (5), the Elman neural network consists of an input layer, a hidden layer, a receiving layer and an output layer; after passing through the hidden layer, the signal is connected to the receiving layer to memorize the output value of the hidden layer unit at the previous moment. Assuming that the number of nodes in the input layer, hidden layer, receiving layer and output layer are R, G, G and M respectively, the nonlinear state expression of the Elman neural network is described as:
[0055] x(t)=f(ω1x c (t)+ω2p(t-1))
[0056] x c (t) = x(t-1)
[0057] y t =g(ω3x(t))
[0058] Where:
[0059] t——the tth moment; x(t)——the output of the hidden layer;
[0060] p——input layer input; x c (t)——output of receiving layer;
[0061] ω1——weight from receiving layer to hidden layer; y(t)——output of output layer;
[0062] ω2——weight from input layer to hidden layer; g(x)——transfer function of output layer;
[0063] ω3——weight from hidden layer to output layer; f(x)——activation function of hidden layer.
[0064] Furthermore, the number of input layer neurons R in the Elman neural network in step (5) is the dimension corresponding to the data, R=3; the Elman output layer neurons are used to output the final result, M=1; the number of hidden layer / continuing layer neurons G is determined by combining the exponential growth strategy with the dichotomy method; the activation function f(x) adopts the Sigmoid function.
[0065] Furthermore, the Elman neural network uses the error index function E to obtain the error function of the data preprocessed in step (4); the network weights of the Elman neural network are updated, and after the update, it is determined whether the training stop condition is met, specifically:
[0066] The error index function E is:
[0067]
[0068] make but:
[0069]
[0070] Taking the partial derivative of function E with respect to weight ω3, we can obtain:
[0071]
[0072] make Then there is
[0073]
[0074] Taking the partial derivative of function E with respect to ω1, we can obtain:
[0075]
[0076] in:
[0077]
[0078] Without counting x c (t) for weight ω 1jl In case of dependency, then:
[0079]
[0080] According to the theory of gradient descent, each weight is corrected to
[0081]
[0082]
[0083]
[0084] Among them, η1, η2, and η3 represent the learning steps of weights ω1, ω2, and ω3 respectively.
[0085] Compared with the prior art, the present invention has the following significant advantages:
[0086] The present invention integrates atmospheric dynamics, statistics and exterior ballistics, and proposes a multi-information source ballistic wind identification method; the method is low-cost, and the use of multi-model integration can effectively avoid the influence of noise in historical data on identification accuracy. The obtained meteorological wind data is more suitable for artillery combat in the time domain and space domain. BRIEF DESCRIPTION OF THE DRAWINGS
[0087] Figure 1 It is a flow chart of the ballistic wind identification method of the present invention.
[0088] Figure 2 This is a wind speed data diagram used in the present invention.
[0089] Figure 3 This is the wind direction data diagram used in the present invention.
[0090] Figure 4 The network structure diagram of ELM in the present invention
[0091] Figure 5 Flow chart of the hidden layer determination method based on dichotomy and exponential growth strategy adopted by the present invention
[0092] Figure 6 This is the network structure diagram of Elman in the present invention
[0093] Figure 7 This is the workflow diagram of Elman in the present invention.
[0094] Figure 8 This is a diagram of wind speed identification results of the present invention.
[0095] Fig. 9 This is the wind direction identification result diagram of the present invention DETAILED DESCRIPTION
[0096] The present invention is further described in detail below in conjunction with the accompanying drawings.
[0097] The ballistic wind identification method of the present invention by fusion of multiple information sources comprises the following steps:
[0098] Step 1: Use the thermal wind formula to construct wind field model I;
[0099] Step 1.1 Wind speed model I
[0100]
[0101] Where u is the wind speed, h is the height, h1 is the sea level height, u1 is the wind speed at sea level, z0 is the roughness parameter, u g is the zonal wind speed, v g is the radial wind speed, f is the geostrophic parameter, k1 is the turbulence coefficient, ω is the geostrophic angular velocity, Λ is the local dimension, and f * For potential.
[0102] Step 1.2 Wind direction model I
[0103] Near the ground layer (within 100m): wind direction is consistent with average wind speed
[0104] Upper boundary layer (100m~1500m):
[0105]
[0106] Free atmosphere (1500m~20000m):
[0107]
[0108] where ψ is the wind direction,
[0109] Step 2: Use SPSS to analyze historical data and build wind field model II;
[0110] Step 2.1 Preprocessing of historical meteorological data: Select the meteorological wind data of the target point within the past 120 days, and obtain the average wind direction and wind speed data. The specific wind direction and wind speed change curve is as follows: Figure 2 Figure 3 shown.
[0111] Step 2.2 According to Figure 2 , Figure 3 Concave and convex points, using SPSS segmented fitting, to construct the wind direction and wind speed model as follows
[0112] Wind speed model II:
[0113]
[0114] Wind direction model II:
[0115]
[0116] Step 3: Using data-driven approach, build wind farm model III based on extreme leaning machine (ELM);
[0117] When given N sets of training samples For an ELM with an ILO model structure (I, L, O are the number of neurons in the input layer, hidden layer, and output layer, respectively) and an activation function of h(x), the output of ELM is:
[0118]
[0119]
[0120]
[0121]
[0122] in is the connection weight matrix between the hidden layer and the output layer, H L×N is the hidden layer output matrix, and b L×N is a randomly generated structural parameter matrix;
[0123] Optimize through training Make the network output as close to the actual output as possible, that is
[0124]
[0125] Based on the least squares criterion, the connection weight matrix between the hidden layer and the output layer is
[0126]
[0127] In the formula H L×N The Moor-Penrose generalized inverse matrix.
[0128] Step 3.1 Data normalization: Ballistic data and meteorological data have different dimensions. To eliminate the influence of the indicator dimension, the min-max normalization formula is used to preprocess the data.
[0129]
[0130] where r * is the normalized data, r min is the minimum value of the data, r max For data
[0131] Step 3.2 Build the ELM network model ( Figure 4 ): ELM is a single hidden layer feedforward neural network. When building its network model, it is necessary to determine the number of neurons in the input layer I, the number of neurons in the output layer O, the number of neurons in the hidden layer L, and the activation function h(x)
[0132] 1) Number of neurons in the input layer I
[0133] In the present invention, ELM is used to construct a mapping relationship between ballistic data and meteorological wind, and the input layer neurons are used to record the ballistic data, which includes the artillery range x, elevation y, lateral deviation z, and the velocity component v in the corresponding direction. x ,v y ,v z and the velocity azimuth v θ . Therefore, the number of neurons in the ELM input layer is I=7.
[0134] 2) Number of neurons in the output layer O
[0135] The ELM output layer is used to output the ballistic wind data required for identification in the present invention: wind speed u and wind direction ψ, so the number of neurons in the ELM output layer is O=2.
[0136] 3) Number of neurons in the hidden layer L
[0137] Selecting an appropriate number of hidden layer neurons L can effectively improve the generalization ability of the model. The present invention combines the exponential growth strategy with the dichotomy method to determine the number of hidden layer neurons L. The value range of L is [L min ,L max ]=[2j-1 ,2 j ], the binary method continuously narrows the interval range until the final number of hidden layer neurons L is determined. The method for determining the number of hidden layer nodes is as follows: Figure 5 As shown. Where floor(·) represents the floor rounding function, Error max is the maximum allowable error.
[0138] 4) Activation function h(x)
[0139] The activation function can introduce nonlinear characteristics into the network, so that the network can approximate any other nonlinear function. The trajectory data is mapped to the interval [0,1] after min-max normalization. In this interval, the Sigmoid function has good nonlinear characteristics, so the Sigmoid function is selected as the activation function. In addition, since the structural parameters of the ELM do not need to be adjusted based on gradient descent, there is no gradient vanishing problem when using the Sigmoid function as the activation function. The expression of the Sigmoid function is as follows
[0140]
[0141] Step 4: The wind direction and wind speed data obtained in steps 1 to 3 above and the actual wind direction and wind speed are used as new data sets, and after preprocessing using formula (12), the process proceeds to step 5;
[0142] Step 5: Based on the stacking concept, the recursive neural network Elman is used to learn the multi-information source data in step 4 to establish the final wind field model and identify the ballistic wind.
[0143] Elman neural network has four parts: input layer, hidden layer, receiving layer and output layer (structure diagram as shown in Figure 6 As shown). After passing through the hidden layer, the signal is not transmitted back to the input end, but is connected to the receiving layer to remember the output value of the hidden layer unit at the previous moment. The learning and memory mode of internal feedback increases the adaptive time-varying characteristics of the network and the system characteristics that reflect the dynamic process. Assuming that the number of nodes in the input layer, hidden layer, receiving layer, and output layer are R, G, G, and M respectively, the nonlinear state expression of the Elman neural network can be described as:
[0144] x(t)=f(ω1x c (t)+ω2p(t-1)) (14)
[0145] x c (t) = x(t-1) (15)
[0146] y t =g(ω3x(t)) (16)
[0147] Where:
[0148] t——the tth moment; x(t)——the output of the hidden layer;
[0149] p——input layer input; x c (t)——output of receiving layer;
[0150] ω1——weight from receiving layer to hidden layer; y(t)——output of output layer;
[0151] ω2——weight from input layer to hidden layer; g(x)——transfer function of output layer; ω3——weight from hidden layer to output layer; f(x)——activation function of hidden layer. Then the error index function E is:
[0152]
[0153] make but:
[0154]
[0155] Taking the partial derivative of function E with respect to weight ω3, we can obtain:
[0156]
[0157] make Then there is
[0158]
[0159] Taking the partial derivative of function E with respect to ω1, we can obtain:
[0160]
[0161] in:
[0162]
[0163] Without counting x c (t) for weight ω 1jl In case of dependency, then:
[0164]
[0165] According to the theory of gradient descent, each weight is corrected to
[0166]
[0167]
[0168]
[0169] Among them, η1, η2, and η3 represent the learning steps of weights ω1, ω2, and ω3 respectively.
[0170] Elman's workflow diagram is as follows Figure 7 shown.
[0171] Step 5.1: Build the Elman network model: When building the Elman network model, you need to determine the number of neurons in the input layer R, the number of neurons in the output layer M, the number of neurons in the hidden layer G, the number of neurons in the receiving layer G, and the activation function f(x).
[0172] 1) Number of neurons in the input layer R
[0173] In the present invention, the Elman input layer neurons are used to receive the data in step 4, so the value of R is the dimension corresponding to the data, that is, R=3;
[0174] 2) Number of neurons in the output layer M
[0175] The neurons in the Elman output layer are used to output the final result, so M = 1;
[0176] 3) Number of neurons in hidden layer / connecting layer G
[0177] The method for determining the number of neurons in the hidden layer and the receiving layer is the same as step 3.2 3)
[0178] 5) Activation function f(x)
[0179] The activation function selection method is the same as step 3.2 4)
[0180] Step 5.2: Calculate the error function: Obtain the error function for the data given in step 4 according to formula (17);
[0181] Step 5.3: Update network weights: Update network weights according to equations (25) to (27);
[0182] Step 5.4 determines whether the training stop condition is met. If it is met, go to step 5.5. If not, return to step 5.2.
[0183] Step 5.6 Data testing: Randomly sample data for testing to verify the generalization performance of the model; (this step is optional);
[0184] Step 5.7 Experimental test: Given the trajectory data of a target point, use the model to obtain the predicted wind direction and wind speed curve, and compare it with the actual wind direction and wind speed curve. (Specifically, Figure 8 , 9 shown).
Claims
1. A ballistic wind identification method based on multi-information source fusion, characterized in that: The steps include: Step (1): Use the thermal wind formula to construct wind field model I; Step (2): Use SPSS to analyze historical data and construct wind field model II; Step (3): Using data-driven approach, construct wind farm model III based on extreme learning machine ELM; Step (4): The wind direction and wind speed data obtained by the wind field model in steps (1) to (3) and the actual wind direction and wind speed are taken as new data sets, normalized and preprocessed, and then the process proceeds to step (5); Step (5): Based on the stacking concept, the recursive neural network Elman learning method is used to learn the multi-information source data preprocessed in step (4) to establish the final wind field model and identify the ballistic wind.
2. The method according to claim 1, characterized in that The wind field model I established in step (1) includes a wind speed model I and a wind direction model I; Wind speed model I is as follows: Where u is the wind speed, h is the height, h1 is the sea level height, u1 is the wind speed at sea level, z0 is the roughness parameter, u g is the zonal wind speed, v g is the radial wind speed, f is the geostrophic parameter, k1 is the turbulence coefficient, ω is the geostrophic angular velocity, Λ is the local dimension, and f * For position potential; Wind direction model I is as follows: Within 100m near the ground: wind direction is consistent with average wind speed Upper boundary layer 100m~1500m: Free atmosphere 1500m~20000m: where ψ is the wind direction, 3. The method according to claim 2, characterized in that Step (2) Preprocessing of historical meteorological data: Select the meteorological wind data of the target point within n days in history, obtain the average value of wind direction and wind speed data, and obtain the wind direction and wind speed change curve; according to the concave and convex points on the wind direction and wind speed change curve, use SPSS segmented fitting to construct a wind direction and wind speed model; Wind speed model II is as follows: Wind direction model II is as follows:
4. The method according to claim 3, characterized in that Wind field model III constructed in step (3): When given N sets of training samples For the ELM with the model structure of ILO, where I, L, and O are the number of neurons in the input layer, hidden layer, and output layer, respectively, and the activation function is h(x), the output of ELM is in is the connection weight matrix between the hidden layer and the output layer, H L×N is the hidden layer output matrix, and b L×N is a randomly generated structural parameter matrix; Optimize through training Make the network output as close to the actual output as possible, that is Based on the least squares criterion, the connection weight matrix between the hidden layer and the output layer is In the formula H L×N The Moor-Penrose generalized inverse matrix.
5. The method according to claim 4, characterized in that Step (3) specifically includes the following steps: Step (31): Data normalization: Use the min-max normalization formula to preprocess the data: where r * is the normalized data, r min is the minimum value of the data, r max For data; Step (32): Construct the ELM network model: determine the number of neurons in the input layer I, the number of neurons in the output layer O, the number of neurons in the hidden layer L, and the activation function h(x): Number of neurons in the input layer I ELM is used to construct a mapping relationship between ballistic data and meteorological wind. The input layer neurons are used to record the ballistic data. The ballistic data includes the artillery range x, elevation y, side deviation z, and the velocity component v in the corresponding direction. x ,v y ,v z and the velocity azimuth v θ ; Therefore, the number of neurons in the ELM input layer I = 7; The number of neurons in the output layer is O The ELM output layer is used to output the ballistic wind data that needs to be identified: wind speed u and wind direction ψ, so the number of neurons in the ELM output layer is O = 2; The number of neurons in the hidden layer is L Combining the exponential growth strategy with the dichotomy method, the range of L is [L min ,L max ]=[2 j-1 ,2 j ], the binary method continuously narrows the interval range until the final number of hidden layer neurons L is determined; Activation function h(x) The trajectory data is mapped to the [0,1] interval after normalization, and the Sigmoid function is selected as the activation function. The expression of the Sigmoid function is as follows:
6. The method according to claim 5, characterized in that In step (5), the Elman neural network consists of an input layer, a hidden layer, a receiving layer and an output layer; after passing through the hidden layer, the signal is connected to the receiving layer, which is used to memorize the output value of the hidden layer unit at the previous moment. Assuming that the number of nodes in the input layer, hidden layer, receiving layer and output layer are R, G, G and M respectively, the nonlinear state expression of the Elman neural network is described as: x(t)=f(ω1x c (t)+ω2p(t-1)) x c (t)=x(t-1) y t =g(ω3x(t)) Where: t——the tth moment; x(t)——the output of the hidden layer; p——input layer input; x c (t)——output of receiving layer; ω1——weight from receiving layer to hidden layer; y(t)——output of output layer; ω2——weight from input layer to hidden layer; g(x)——transfer function of output layer; ω3——weight from hidden layer to output layer; f(x)——activation function of hidden layer.
7. The method according to claim 6, characterized in that The number of input layer neurons R in the Elman neural network in step (5) is the dimension corresponding to the data, R = 3; the Elman output layer neurons are used to output the final result, M = 1; the number of hidden layer / continuing layer neurons G is determined by combining the exponential growth strategy with the dichotomy method; the activation function f(x) adopts the Sigmoid function.
8. The method according to claim 7, characterized in that The Elman neural network uses the error index function E to obtain the error function of the data preprocessed in step (4); the network weights of the Elman neural network are updated, and after the update, it is determined whether the training stop condition is met, specifically: The error index function E is: make but: Taking the partial derivative of function E with respect to weight ω3, we can obtain: make Then there is Taking the partial derivative of function E with respect to ω1, we can obtain: in: Without counting x c (t) for weight ω 1jl In case of dependency, then: According to the theory of gradient descent, each weight is corrected to Among them, η1, η2, and η3 represent the learning steps of weights ω1, ω2, and ω3 respectively.
Citation Information
Patent Citations
Non-stationary pulse wind speed high-precision prediction method based on EEMD-ELM
CN105138782A
Wind speed prediction method based on integration of negative correlation learning and regularization extreme learning machine
CN111178616A
Wind resource prediction method and system based on integrated extreme learning machine
CN113344270A
Wind speed and wind direction prediction method
CN113779892A
Guidance cartridge high-altitude wind speed and wind direction estimation method based on GA-BP neural network
CN116956717A
Cited By
Method and device for selecting passive load shedding trajectory of spacecraft before launching and electronic equipment
CN120124193A
