Space-time three-dimensional neural network calculation method based on flow field distribution characteristics of wind farm

Abstract

This invention discloses a spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms, belonging to the field of wind power technology. The invention utilizes a communication system to acquire position data of the anemometer tower, wind turbine, and anemometer, as well as corresponding wind speed and direction data, at a set time resolution. Based on the spatial positional relationship between the anemometer tower, wind turbine, and anemometer, and the spatiotemporal correlation between wind speed and direction, a spatiotemporal three-dimensional neural network calculation model of the wind farm flow field is constructed. Based on this model, a training strategy for the spatiotemporal three-dimensional neural network calculation model is proposed. This solves the problem of grid instability and unsafe operation caused by unreasonable switching of distributed photovoltaic power generation systems in existing technologies. Specifically, based on the structure of the spatiotemporal three-dimensional neural network model, a mathematical calculation formula considering the spatiotemporal distribution characteristics of neurons is proposed, providing a new calculation method for the accurate calculation of the spatiotemporal distribution characteristics of wind farm flow fields.

Description

Technical Field

[0001] This invention belongs to the field of wind power technology, specifically relating to a spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms. Background Technology

[0002] With the national goals of "carbon peaking" and "carbon neutrality," wind power has been developed and utilized on a large scale in recent years. Due to the inherent randomness and uncertainty of wind power generation, accurate real-time calculation of the spatiotemporal distribution characteristics of the wind farm flow field is of great help for wind farm power prediction and optimized operation. However, for wind farms in complex terrain, the flow field distribution characteristics are affected by factors such as incoming wind speed and direction, terrain, and the positional relationship between wind turbines, making it extremely difficult to accurately calculate the flow field distribution characteristics of wind farms in complex terrain.

[0003] Currently, the calculation of wind farm flow field distribution characteristics mainly employs three methods: field measurement, numerical simulation, and wind tunnel experiments. Field measurement requires extensive measuring equipment and time, and is typically used in scientific research, with limited application in practical engineering. Numerical simulation combines wind farm topographic and climatic information, using numerical simulation software to calculate the flow field distribution characteristics. Numerical simulation has advantages such as low computational cost and the ability to calculate wind farm flow field distribution characteristics under all operating conditions, but it suffers from low computational accuracy and long computation time. Wind tunnel experiments involve compressing a field prototype to a certain scale and then measuring the wind farm flow field distribution characteristics through wind tunnel experiments. Wind tunnel experiments have relatively low measurement costs, but the wind tunnel environment differs from the real environment, leading to significant discrepancies between wind tunnel measurement results and the actual flow field distribution.

[0004] The above three methods for calculating the flow field distribution of wind farms can all calculate the distribution characteristics of the wind farm flow field to a certain extent, but none of these three methods can describe the spatiotemporal variation characteristics of the wind farm flow field, nor can they accurately calculate the spatiotemporal distribution characteristics of the wind farm flow field. Summary of the Invention

[0005] The purpose of this invention is to address the problem of grid instability and insecurity caused by unreasonable switching of distributed photovoltaic power generation systems in existing technologies. It proposes a spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms. Based on the spatial positional relationship between the anemometer tower, wind turbine, and wind speed meter, and the spatiotemporal correlation between wind speed and direction, a spatiotemporal three-dimensional neural network model that can accurately describe the flow field distribution characteristics of wind farms is constructed. A training strategy for the constructed spatiotemporal three-dimensional neural network model is proposed. The constructed spatiotemporal three-dimensional neural network model is used to achieve accurate calculation of the spatiotemporal distribution characteristics of the wind farm flow field.

[0006] To address the aforementioned technical problems, this invention provides a spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms, comprising the following steps:

[0007] S1. Use a communication system to acquire the location data of the wind measurement tower, wind turbine and wind speed measuring instrument, as well as the wind speed and direction data at the corresponding locations, at a set time resolution.

[0008] S2. Based on the spatial positional relationship between the wind measurement tower, wind turbine and anemometer and the spatiotemporal correlation between wind speed and wind direction, construct a spatiotemporal three-dimensional neural network calculation model of the wind farm flow field.

[0009] S21. Determine the structure of the spatiotemporal three-dimensional neural network model based on the spatial positional relationship between the wind farm's wind measurement tower, wind turbine, and wind speed measuring instrument, and the spatiotemporal correlation between wind speed and wind direction.

[0010] S22. Based on the spatiotemporal three-dimensional neural network model structure, construct a spatiotemporal three-dimensional neural network calculation model for the wind farm flow field;

[0011] S3. Based on the spatiotemporal three-dimensional neural network calculation model of the wind farm flow field, a training strategy for the spatiotemporal three-dimensional neural network calculation model is proposed.

[0012] Preferably, the spatiotemporal three-dimensional neural network model structure includes several neurons, each neuron having time-varying and spatial correlation characteristics in its input and output, and each neuron is connected to several neurons around it.

[0013] Preferably, the input of the neuron is specifically represented as follows:

[0014] ;

[0015] In the formula, The coordinate position in the spatiotemporal 3D neural network model is... neurons in The wind speed and direction information is entered in real time; express The coordinate position is where the current flows from the directly connected neuron at any given time. The wind speed and direction information of neurons; among them, Component representation From the coordinate position at time The coordinates of the neuron inflow are The wind speed and direction information of the neurons, and the meanings of other wind speed and direction components can be deduced in the same way; The directly connected neurons and their coordinate positions are The connection weights between neurons represent the connection weights between neurons. The wind speed and direction information output by the directly connected neurons at any given time corresponds to the coordinate position. The degree to which wind speed and direction information input into neurons has an impact. The value of is in the range [0,1]; where Indicates in The time coordinate position is Neuron and coordinate position The meaning of the connection weights of neurons is similar to that of other connection weight components; here each neuron is directly connected to its 6 surrounding neurons, so the number of wind speed and direction information inputs to each neuron is 6, and the number of connection weights between each neuron and its surrounding neurons is also 6. The spatial distance between directly connected neurons is represented by the coordinate position. The greater the spatial distance, the smaller the influence of directly connected neurons on that neuron. Indicates in The wind speed and direction information output by the directly connected neurons at any given time corresponds to the coordinate position. The degree of influence of wind speed and direction information input into neurons is The bias value at time.

[0016] Preferably, the spatiotemporal three-dimensional neural network computation model is specifically represented as follows:

[0017] ;

[0018] In the formula, The coordinate position in the spatiotemporal 3D neural network model is neurons in The wind speed and direction information is output in real time. The coordinate position in the spatiotemporal 3D neural network model is... neurons in The wind speed and direction information is entered in real time.

[0019] Preferably, the training strategy of the spatiotemporal stereo neural network computation model includes a calculation strategy for the weight adjustment of the spatiotemporal stereo neural network model, specifically expressed as follows:

[0020] ;

[0021] In the formula, for The time coordinate position is Actual measured values ​​of wind speed and direction at the location; The learning rate for the weights ranges from [0,1]. For neuron connection weights in The amount of adjustment at any given time.

[0022] As a preferred approach, the weight update formula for the spatiotemporal stereo neural network is obtained based on the adjustment amount of the weights in the spatiotemporal stereo neural network.

[0023] As a preferred embodiment, the weight update formula for the spatiotemporal stereo neural network is specifically expressed as follows:

[0024] ;

[0025] In the formula, The momentum coefficient is used in the weighting process, and its value ranges from [0,1].

[0026] Preferably, the training strategy of the spatiotemporal three-dimensional neural network calculation model further includes: calculating the wind farm flow field spatiotemporal three-dimensional neural network model to make predictions and comparing them with the actual measurement results. If the difference between the two meets the requirements, the connection weights of the current wind farm flow field spatiotemporal three-dimensional neural network model are saved. If the difference between the two does not meet the requirements, the connection weights of each neuron in the wind farm flow field spatiotemporal three-dimensional neural network model are adjusted according to the calculation strategy of the weight adjustment amount of the spatiotemporal three-dimensional neural network model and the weight update formula of the spatiotemporal three-dimensional neural network model. Then, the spatiotemporal distribution characteristics of the wind farm flow field are recalculated until the difference between the calculated value of the spatiotemporal three-dimensional neural network model and the actual measurement value meets the requirements or the model training number is reached.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] 1. Based on the spatial positional relationship between the wind measurement tower, wind turbine, and anemometer, and the spatiotemporal correlation between wind speed and direction, this scheme constructs a spatiotemporal three-dimensional neural network calculation model of the wind farm flow field.

[0029] Existing neural network models are mostly two-dimensional and cannot accurately describe the three-dimensional spatiotemporal distribution characteristics of wind farm flow fields. Therefore, this invention proposes a spatiotemporal three-dimensional neural network model for calculating the spatiotemporal distribution characteristics of wind farm flow fields, based on these characteristics. This model considers both the temporal correlation between data transmissions between neurons and the spatial distance between neurons. Based on the structure of this model, mathematical formulas considering the spatiotemporal distribution characteristics of neurons are proposed, providing a new method for the accurate calculation of the spatiotemporal distribution characteristics of wind farm flow fields.

[0030] 2. Based on the spatiotemporal three-dimensional neural network calculation model of the wind farm flow field, this solution proposes a training strategy for the spatiotemporal three-dimensional neural network calculation model to achieve convergence of the spatiotemporal three-dimensional neural network calculation model.

[0031] The connection weights of individual neurons in the spatiotemporal three-dimensional neural network computational model of wind farm flow field contain both temporal and spatial parameters. Conventional neuron weight adjustment strategies are not applicable to adjusting neuron weights in spatiotemporal three-dimensional neural network models. Therefore, this invention proposes a weight adjustment strategy for spatiotemporal three-dimensional neural network models based on the model structure and neuron input-output characteristics. This strategy fully considers the influence of temporal factors and spatial location relationships on weight changes, constructing a mathematical model for neuron weight adjustment that incorporates spatiotemporal factors. This achieves rapid convergence of neuron weights, and the correctness of the convergence is verified through numerical examples. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the system flow of the present invention;

[0033] Figure 2 This is a schematic diagram of the spatiotemporal three-dimensional neural network computation model of the present invention;

[0034] Figure 3 This is a schematic diagram of the device of the present invention;

[0035] Figure 4 This is a virtual wind farm turbine layout diagram of the present invention;

[0036] Figure 5 This is a schematic diagram of the calculation results of the spatiotemporal distribution of the wind farm flow field according to the present invention;

[0037] Figure 6 This is a schematic diagram of the turbine layout of a wind farm according to the present invention;

[0038] Figure 7 This is a schematic diagram of the wind field distribution characteristics of a wind farm under wind speed of 8.5 m / s and wind direction of 270° according to an embodiment of the present invention.

[0039] Figure 8 This is a schematic diagram comparing the calculated wind speed and the measured wind speed at the location of the fourth row of wind turbines, calculated by the spatiotemporal three-dimensional neural network model under an incoming wind speed of 8.5 m / s and a wind direction of 270°, according to an embodiment of the present invention.

[0040] Figure 9 This is a schematic diagram of the wind field distribution characteristics of a wind farm under wind speed of 8.5 m / s and wind direction of 222° according to an embodiment of the present invention.

[0041] Figure 10 This is a schematic diagram comparing the calculated wind speed and the measured wind speed at the location of the 8th row of wind turbines, calculated by the spatiotemporal three-dimensional neural network model under an incoming wind speed of 8.5 m / s and a wind direction of 222°, according to an embodiment of the present invention. Detailed Implementation

[0042] Example 1: As Figure 1 As shown, the spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms includes:

[0043] (1) Obtain wind speed and direction data from the wind measurement tower, wind turbine and wind speed meter. The time resolution of wind speed and direction data acquisition is 1 minute. The dedicated communication network for wind farms usually adopts optical fiber communication, but 5G or other communication networks can also be used for communication.

[0044] (2) Based on the spatial relationship between the wind measurement tower, wind turbine and anemometer and the spatiotemporal correlation between wind speed and wind direction, a spatiotemporal three-dimensional neural network model of the wind farm flow field is constructed. The construction process of the spatiotemporal three-dimensional neural network model is as follows:

[0045] 1) Based on the spatial relationships between the wind farm's anemometer tower, wind turbine, and wind speed meter, and the spatiotemporal correlation between wind speed and direction, determine the structure of the spatiotemporal three-dimensional neural network model as follows: Figure 2 As shown. In Figure 2 In this model, each neuron represents a point in space. The input to each neuron represents the wind speed and direction flowing into that point, and the output of each neuron represents the wind speed and direction flowing out of that point. Neurons are used to construct the spatiotemporal three-dimensional neural network model because they can accurately describe the nonlinear mapping relationship between the output data of other neurons and the output data of this neuron. This corresponds to the influence of wind speed and direction at other points in the wind farm flow field on the wind speed and direction at a specific point. To accurately calculate the spatiotemporal distribution characteristics of the wind farm flow field, this invention utilizes neurons to construct a spatiotemporal three-dimensional neural network model, thereby achieving precise calculation of the spatiotemporal distribution characteristics of the wind farm flow field.

[0046] 2) By Figure 2 From the spatiotemporal three-dimensional neural network model structure, it can be seen that the magnitude of wind speed and direction flowing into each neuron is related to the distance between directly connected neurons, and also to the temporal variation characteristics of wind speed and direction. Figure 2 It is known that the minimum number of neurons directly connected to each neuron is 6 (up, down, front, back, left, and right), and more neurons at different positions can be considered as needed (such as upper left, lower left, upper front, lower front, etc.). For ease of explanation, this invention selects 6 neurons directly connected to each neuron for calculation. Therefore, the wind speed and direction data flowing into a neuron comes from the 6 neurons directly connected to it; simultaneously, the wind speed and direction information output from that neuron also flows to the 6 neurons directly connected to it. Based on the correlation between the wind speed and direction at each neuron's location and the wind speed and direction of the other directly connected neurons, the input expression for each neuron is constructed as follows:

[0047] (1)

[0048] The coordinate position in the spatiotemporal 3D neural network model is... neurons in The total wind speed and direction values ​​flowing in at any given time. express The coordinate position is where the current flows from the directly connected neuron at any given time. The wind speed and direction information of neurons; among them, Component representation From the coordinate position at time The coordinates of the neuron inflow are The wind speed and direction information of neurons, and the meanings of other wind speed and direction components can be deduced similarly. The directly connected neurons and their coordinate positions are The connection weights between neurons represent the connection weights between neurons. The wind speed and direction information output by the directly connected neurons at any given time corresponds to the coordinate position. The degree to which wind speed and direction information input into neurons has an impact. The value of is between [0, 1]; where Indicates in The time coordinate position is Neuron and coordinate position The connection weights of neurons, and the meanings of other connection weight components, follow the same logic. Each neuron is directly connected to its six surrounding neurons, so each neuron receives six wind speed and direction inputs, and each neuron also has six connection weights with its surrounding neurons. In the spatiotemporal three-dimensional neural network model, neurons are... The bias value at time.

[0049] The spatial distance between directly connected neurons is represented by the coordinate position. The greater the spatial distance, the smaller the influence of directly connected neurons on that neuron.

[0050] 3) In order to accurately describe each neuron in To investigate the correlation between the wind speed and direction information input at any given time and the wind speed and direction information output by the neuron, this invention uses the hyperbolic tangent function to describe the quantitative relationship between the input and output of the neuron in the spatiotemporal three-dimensional neural network. The transfer function expression between the input and output of the neuron in the spatiotemporal three-dimensional neural network is shown in equation (2).

[0051] (2)

[0052] The coordinate position in the spatiotemporal 3D neural network model is neurons in The wind speed and direction information is output in real time. The coordinate position in the spatiotemporal 3D neural network model is... neurons in The wind speed and direction information is entered in real time.

[0053] (3) Based on the constructed spatiotemporal three-dimensional neural network calculation model of wind farm flow field, a strategy for adjusting the weights of the spatiotemporal three-dimensional neural network model is proposed to achieve rapid convergence of the spatiotemporal three-dimensional neural network calculation model.

[0054] 1) For each neuron in the spatiotemporal three-dimensional neural network model, when the difference between the wind speed and direction value calculated by the neuron at a certain moment and the actual measured wind speed and direction value is greater than a given threshold, it is necessary to adjust the weights connected to the neuron so that the difference between the wind speed and direction value calculated by the neuron and the actual measured wind speed and direction value is less than the given threshold.

[0055] 2) According to the principle of adjusting the weights of a neural network, the adjustment amount of the neuron weights is related to factors such as the magnitude of the neuron's calculation error, the derivative of the neuron's transfer function with respect to the weights, the spatial distance between neurons, the amount of input to the neuron, and the weight learning rate. Thus, the formula for adjusting the weights of a neuron is shown in equation (3).

[0056] (3)

[0057] In equation (3), Indicates the coordinate position as neurons in The weight adjustment amount at any given time; These are the actual measured wind speed and direction values; The learning rate is the weight value, and the learning rate ranges from [0, 1]. For neurons in The input at time step; the derivative of the neuron's transfer function with respect to the weights is... .

[0058] 3) The number of other neurons directly connected to the neuron is 6. Based on the coordinate positions of these other neurons... Based on the positional relationship of the neurons, the formula for calculating the adjustment amount of the connection weight of the neurons is shown in equation (4).

[0059] (4)

[0060] In equation (4), for The time coordinate position is Actual measured values ​​of wind speed and direction at the location; The learning rate for the weights ranges from [0, 1]. For neuron connection weights in The amount of adjustment at any given time.

[0061] 4) After calculating the magnitude of the neuron weight adjustment, the connection weights of each neuron can be updated according to equation (5). In the process of updating neuron weights, a momentum coefficient is introduced to balance the quantitative relationship between the original weights and the weight adjustment. The update formula for the weights of the spatiotemporal three-dimensional neural network is shown in equation (5).

[0062] (5)

[0063] In equation (5), The momentum coefficient is used in the weighting process, and its value ranges from [0, 1].

[0064] The connection weights of individual neurons in the spatiotemporal three-dimensional neural network computational model of wind farm flow field contain both temporal and spatial parameters. Conventional neuron weight adjustment strategies are not applicable to adjusting neuron weights in spatiotemporal three-dimensional neural network models. Therefore, this invention proposes a weight adjustment strategy for spatiotemporal three-dimensional neural network models based on the model structure and neuron input-output characteristics. This strategy fully considers the influence of temporal factors and spatial location relationships on weight changes, constructing a mathematical model for neuron weight adjustment that incorporates spatiotemporal factors. This achieves rapid convergence of neuron weights, and the correctness of the convergence is verified through numerical examples.

[0065] On the other hand, this application also proposes a wind farm flow field spatiotemporal distribution characteristic calculation device based on spatiotemporal three-dimensional neural network, which consists of a data acquisition and transmission system, a wind farm flow field spatiotemporal three-dimensional neural network model, a software operating system and related supporting software, hardware equipment and other components.

[0066] The data acquisition and transmission system is used to acquire the location data of the wind measurement tower, wind turbine and anemometer, as well as the wind speed and direction data at that location.

[0067] A spatiotemporal three-dimensional neural network model for wind farm flow field is proposed to calculate the spatiotemporal three-dimensional neural network model for wind farm flow field based on the spatial positional relationship between the wind measuring tower, wind turbine and anemometer and the spatiotemporal correlation between wind speed and wind direction.

[0068] Software operating systems and related supporting software and hardware devices are relatively conventional existing carriers, so they will not be discussed in detail here.

[0069] In addition, this application proposes a method for constructing a wind farm flow field spatiotemporal distribution characteristic calculation device based on a spatiotemporal three-dimensional neural network. This device comprises a data acquisition and transmission system, a wind farm flow field spatiotemporal three-dimensional neural network model, a software operating system and related supporting software, and hardware equipment. The structure of the wind farm flow field spatiotemporal distribution characteristic calculation device based on a spatiotemporal three-dimensional neural network is as follows: Figure 3 As shown, the specific steps are as follows:

[0070] 1) Use a dedicated communication network or 5G network to transmit the wind speed and direction data from the wind turbine SCADA system, the wind speed and direction data from the meteorological tower, and the wind speed and direction data from the wind speed measuring instrument to the support platform where the wind farm flow field spatiotemporal three-dimensional neural network model is located.

[0071] 2) The spatiotemporal neural network model of the wind farm flow field calculates the spatiotemporal distribution characteristics of the wind farm flow field based on the historical wind speed and direction data on the support platform and the real-time wind speed and direction data transmitted.

[0072] 3) Calculate the difference between the calculated value and the actual measured value of the wind farm flow field spatiotemporal three-dimensional neural network model. If the difference meets the requirements, save the connection weights of the current wind farm flow field spatiotemporal three-dimensional neural network model. When the difference between the calculated value and the actual measured value of the spatiotemporal three-dimensional neural network model does not meet the requirements, adjust the connection weights of each neuron in the wind farm flow field spatiotemporal three-dimensional neural network model according to formulas (3), (4) and (5), and then recalculate the spatiotemporal distribution characteristics of the wind farm flow field until the difference between the calculated value and the actual measured value meets the requirements or reaches the number of training times of the model.

[0073] Example 2: In the case study analysis, two cases are used to verify the correctness and feasibility of the present invention.

[0074] The first example is a virtual wind farm containing 9 wind turbines. According to the People's Republic of China Energy Industry Standard "Technical Specification for Micro-Site Selection of Wind Farm Engineering" NB / T 10103-2018, the row spacing of wind turbines should not be less than 3 times the rotor diameter, and the column spacing should not be less than 5 times the rotor diameter. Therefore, in this example, the north-south spacing (row spacing) between wind turbines is set to 4 times the rotor diameter (4D), and the east-west spacing (column spacing) is set to 7 times the rotor diameter (7D). Figure 4 As shown.

[0075] Figure 5 This is a diagram showing the wind farm flow field distribution characteristics calculated according to the present invention. Figure 5The calculation results show that when the incoming wind speed passes through the wind turbine, the wind speed decreases and a wake wind speed is formed. As the wake wind speed propagates downstream, the wake wind speed gradually recovers. This is consistent with the flow field distribution characteristics of a real wind farm, indicating that the method proposed in this patent can correctly calculate the flow field distribution characteristics of a wind farm.

[0076] Figure 6 , Figure 7 , Figure 8 , Figure 9 and Figure 10 The diagram shows the turbine layout and calculation results for a wind farm. Figure 6 The turbine layout of a wind farm is as follows: the spacing between turbines along the 270° direction (east-west direction) is 7 times the rotor diameter (7D), and the spacing along the 222° direction (northeast direction) is 9.4 times the rotor diameter (9.4D).

[0077] Figure 7 The wind farm flow field distribution characteristics are calculated by a spatiotemporal three-dimensional neural network model under an incoming wind speed of 8.5 m / s and a wind direction of 270°. From... Figure 7 As can be seen, the wind speed distribution at the first row of wind turbines is 8.5 m / s, which is consistent with the actual flow field distribution of the wind farm. This is because the first row of wind turbines is not affected by the wakes of other wind turbines, and the wind speed at its location is the incoming wind speed (8.5 m / s). For the second row and subsequent rows of wind turbines, due to the influence of the wakes of the preceding wind turbines, their flow field velocity is significantly lower than the incoming wind speed. This is consistent with the actual distribution characteristics of the wind farm flow field, indicating that the spatiotemporal three-dimensional neural network model can correctly calculate the spatiotemporal distribution characteristics of the wind farm flow field. Figure 8 This section compares the calculated wind speed at the location of the fourth row of wind turbines, obtained from a spatiotemporal three-dimensional neural network model, with the measured wind speed under an incoming wind speed of 8.5 m / s and a wind direction of 270°. Figure 8 It can be seen that the maximum error between the calculated wind speed and the measured wind speed by the spatiotemporal three-dimensional neural network model is 0.102 m / s, indicating that the spatiotemporal three-dimensional neural network model can accurately calculate the flow field distribution characteristics of the wind farm.

[0078] Figure 9 The flow field distribution characteristics of a wind farm at a wind speed of 8.5 m / s and a wind direction of 222° are shown. Figure 9 It can be seen that when the incoming wind direction is 222°, the first and last wind turbines are not affected by the wakes of other wind turbines, and the wind speed at their locations is the incoming wind speed (8.5 m / s). For wind turbines at other locations, due to the influence of the wakes of the preceding wind turbines, their flow field velocity is significantly lower than the incoming wind speed. This is consistent with the actual distribution characteristics of the wind farm flow field, indicating that the spatiotemporal three-dimensional neural network model can correctly calculate the spatiotemporal distribution characteristics of the wind farm flow field. Figure 10The 8th row, calculated by the spatiotemporal three-dimensional neural network model under an incoming wind speed of 8.5 m / s and a wind direction of 222°, is shown in the figure. Figure 6 A comparison of the calculated and measured wind speeds at the location of the wind turbine. Figure 10 It can be seen that the maximum error between the calculated wind speed and the measured wind speed by the spatiotemporal three-dimensional neural network model is 0.126 m / s, which further proves that the spatiotemporal three-dimensional neural network model can accurately calculate the flow field distribution characteristics of the wind farm.

Claims

1. A spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms, characterized in that, Includes the following steps: S1. Use a communication system to acquire the location data of the wind measurement tower, wind turbine and wind speed measuring instrument, as well as the wind speed and direction data at the corresponding locations, at a set time resolution. S2. Based on the spatial positional relationship between the wind measurement tower, wind turbine and anemometer and the spatiotemporal correlation between wind speed and wind direction, construct a spatiotemporal three-dimensional neural network calculation model of the wind farm flow field. S21. Determine the structure of the spatiotemporal three-dimensional neural network model based on the spatial positional relationship between the wind farm's wind measurement tower, wind turbine, and wind speed measuring instrument, and the spatiotemporal correlation between wind speed and wind direction. The spatiotemporal three-dimensional neural network model structure includes several neurons, each neuron's input and output have time-varying characteristics and spatial correlation characteristics, and each neuron is connected to several neurons around it; The input to a neuron is specifically represented as follows: ； In the formula, The coordinate position in the spatiotemporal 3D neural network model is... neurons in The wind speed and direction information is entered in real time; express The flow of data from directly connected neurons into the coordinate position at any given time is... The wind speed and direction information of neurons; among them, Component representation From the coordinate position at time The coordinates of the neuron inflow are The wind speed and direction information of the neurons, and the meanings of other wind speed and direction components can be deduced in the same way; The directly connected neurons and their coordinate positions are The connection weights between neurons represent the connection weights between neurons. The wind speed and direction information output by the directly connected neurons at any given time corresponds to the coordinate position. The degree of influence of wind speed and direction information input into the neurons; among which Indicates in The time coordinate position is Neuron and coordinate position The meanings of the connection weights of neurons, and other connection weight components, can be deduced similarly. The spatial distance between directly connected neurons is represented by the coordinate position. The degree of influence on neurons; Indicates in The wind speed and direction information output by the directly connected neurons at any given time corresponds to the coordinate position. The degree of influence of wind speed and direction information input into neurons is The offset value at time; S22. Based on the spatiotemporal three-dimensional neural network model structure, construct a spatiotemporal three-dimensional neural network calculation model for the wind farm flow field; S3. Based on the spatiotemporal three-dimensional neural network calculation model of the wind farm flow field, a training strategy for the spatiotemporal three-dimensional neural network calculation model is proposed.

2. The spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms according to claim 1, characterized in that, The spatiotemporal three-dimensional neural network computation model is specifically represented as follows: ； In the formula, The coordinate position in the spatiotemporal 3D neural network model is neurons in The wind speed and direction information is output in real time. The coordinate position in the spatiotemporal 3D neural network model is... neurons in The wind speed and direction information is entered in real time.

3. The spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms according to claim 2, characterized in that, The training strategy for the spatiotemporal stereo neural network computation model includes a strategy for calculating the weight adjustment of the spatiotemporal stereo neural network model, specifically expressed as follows: ； In the formula, for The time coordinate position is Actual measured values ​​of wind speed and direction at the location; The learning rate for the weights ranges from [0,1]. For neuron connection weights in The amount of adjustment at any given time.

4. The spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms according to claim 3, characterized in that, Based on the adjustment amount of the weights in the spatiotemporal 3D neural network, the weight update formula of the spatiotemporal 3D neural network is obtained.

5. The spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of a wind farm according to claim 4, characterized in that, The specific formula for weight update in a spatiotemporal 3D neural network is as follows: ； In the formula, This is the momentum coefficient during the weighting process, and its value ranges from [0,1].

6. The spatiotemporal three-dimensional neural network calculation method based on the flow field distribution characteristics of wind farms according to claim 1, characterized in that, The training strategy of the spatiotemporal three-dimensional neural network computational model further includes: predicting the wind farm flow field using the spatiotemporal three-dimensional neural network model and comparing it with the actual measurement results. If the difference between the two meets the requirements, the connection weights of the current wind farm flow field spatiotemporal three-dimensional neural network model are saved. If the difference between the two does not meet the requirements, the connection weights of each neuron in the wind farm flow field spatiotemporal three-dimensional neural network model are adjusted according to the calculation strategy of the weight adjustment amount of the spatiotemporal three-dimensional neural network model and the weight update formula of the spatiotemporal three-dimensional neural network. Then, the spatiotemporal distribution characteristics of the wind farm flow field are recalculated until the difference between the calculated value of the spatiotemporal three-dimensional neural network model and the actual measurement value meets the requirements or the model training number is reached.

Images