Wind speed measurement method and system based on bistatic laser wind finding radar with separate transmitting and receiving

Abstract

The invention discloses a wind speed measurement method and system based on a bistatic laser wind finding radar with separate transceiving, the wind speed in the atmosphere is accurately measured by transmitting a laser beam with a specific wavelength and using the Doppler frequency shift principle, the system integrates a plurality of modules, and the system is simple in structure and convenient to operate. Comprising an environmental parameter acquisition module, a Doppler frequency shift correction module, a signal processing module and an intelligent layout optimization module. By collecting environmental parameters in real time, the system dynamically adjusts Doppler frequency shift correction factors to compensate errors caused by environmental changes; the multi-stage signal processing module improves the signal-to-noise ratio and accuracy of wind speed measurement through multi-scale feature extraction, depth separable convolution and adaptive noise suppression technologies; the intelligent layout optimization module dynamically adjusts the position configuration of the base station based on a depth Q network algorithm, and ensures the measurement precision and stability in a complex environment; the method significantly improves the precision and reliability of wind speed measurement, and is suitable for the fields of wind power plant optimization layout, meteorological monitoring, aerospace flight plans and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of atmospheric wind speed measurement, and in particular to a wind speed measurement method and system based on a dual-base laser wind measurement radar with separate transmitters and receivers. Background Art

[0002] With the increasing demand for high-precision wind speed measurement in fields such as wind power generation, meteorological monitoring, and aerospace, the traditional single-base laser wind radar system has been unable to meet the needs of these fields due to the limitations of measurement accuracy and anti-interference ability in complex environments. In the prior art, Doppler laser radar calculates wind speed by measuring the frequency shift of the scattered signal of particulate matter in the atmosphere. However, environmental factors (such as air pressure, temperature, humidity, etc.) have a significant impact on the Doppler frequency shift, and these changes can lead to large measurement errors. In addition, the existing radar system's base station layout optimization method lacks flexibility and is difficult to adapt to wind field changes in complex environments. Therefore, a dual-base laser wind radar system that can adjust correction parameters in real time and optimize the base station layout is needed to improve the accuracy and reliability of measurement. Summary of the invention

[0003] Technical purpose: In view of the shortcomings of existing laser wind measurement technology, the present invention discloses a wind speed measurement method and system based on a dual-base laser wind measurement radar with separate transmitters and receivers. By collecting environmental parameter data, dynamically adjusting Doppler frequency shift correction parameters, optimizing base station layout, and using multi-level signal processing technology, the accuracy and signal-to-noise ratio of wind speed measurement are improved.

[0004] Technical solution: To achieve the above technical objectives, the present invention adopts the following technical solution:

[0005] A wind speed measurement method based on a bistatic laser wind measuring radar with separate transmitter and receiver, comprising the following steps:

[0006] S1, initializing the positions of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar, and collecting environmental parameter data in real time through multiple distributed sensor nodes, the environmental parameter data including air pressure, temperature and humidity, and transmitting the environmental parameter data to the central processing unit for preprocessing;

[0007] S2. Use a dual-base laser wind measurement radar to obtain Doppler frequency shift data of the wind measurement area, and calculate the initial wind speed of the wind measurement area based on the Doppler frequency shift data;

[0008] S3, calculating an adaptive correction factor based on the environmental parameter data, and using the adaptive correction factor to correct the Doppler frequency shift data to obtain corrected wind speed data;

[0009] S4, performing multi-stage processing on the corrected wind speed data, the multi-stage processing including multi-scale feature extraction, depth-separable convolution and adaptive noise suppression to improve the signal-to-noise ratio and measurement accuracy of the wind speed data;

[0010] S5, construct a state vector, and input the state vector into the trained deep Q network, the state vector includes the wind speed data after multi-level processing, the current environmental parameter data, and the current positions of the transmitting base station and the receiving base station;

[0011] S6. Calculate the Q value of the possible action using the trained deep Q network based on the state vector, select the action with the largest Q value, and adjust the position of the transmitting base station and the receiving base station according to the action with the largest Q value. The action includes adjusting the direction and amplitude of the position of the transmitting base station and the receiving base station.

[0012] S7. At the adjusted positions of the transmitting base station and the receiving base station, the wind speed is remeasured using a dual-base laser wind measuring radar to obtain new Doppler frequency shift data, and new wind speed is calculated based on the new Doppler frequency shift data, and new multi-level processed wind speed data is obtained after correction and multi-level processing;

[0013] S8. Repeat steps S5 to S7 to continuously optimize the positions of the transmitting base station and the receiving base station and the wind speed measurement results until a preset measurement accuracy is achieved or the system is stable.

[0014] The present invention also provides a wind speed measurement system based on a bistatic laser wind measuring radar with separate transmitters and receivers, which is used to implement the wind speed measurement method based on a bistatic laser wind measuring radar with separate transmitters and receivers as described above, comprising:

[0015] a central processing unit for receiving and processing environmental parameter data transmitted from a plurality of sensor nodes;

[0016] An environmental parameter acquisition module is used to collect environmental parameter data in real time through multiple sensor nodes, and the sensor nodes transmit the data to a central processing unit through a wireless communication network or a wired communication network;

[0017] A bistatic laser wind measuring radar, including a transmitting base station and a receiving base station, for transmitting a laser beam of a specific wavelength to a target area, and receiving laser light scattered by particles in the atmosphere of the wind measuring area to obtain Doppler frequency shift data. The bistatic laser wind measuring radar is connected to a central processing unit to transmit the Doppler frequency shift data to the central processing unit for processing;

[0018] A Doppler frequency shift correction module is used to calculate an adaptive correction factor based on environmental parameter data, and to correct the Doppler frequency shift data according to the adaptive correction factor to obtain corrected wind speed data. The Doppler frequency shift correction module is connected to the central processing unit;

[0019] Intelligent layout optimization module, which is used to dynamically optimize the location configuration of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar according to the corrected wind speed data and environmental parameter data through the deep Q network algorithm;

[0020] The multi-level signal processing module is used to perform multi-level processing on the corrected wind speed data to improve the signal-to-noise ratio and measurement accuracy of the signal. The multi-level signal processing module is connected to the central processing unit.

[0021] Beneficial effects: The wind speed measurement method and system based on a bistatic laser wind measuring radar with separate transmitter and receiver provided by the present invention have the following beneficial effects:

[0022] 1. The present invention takes into account complex terrain and changeable meteorological conditions to ensure stable operation even in harsh environments. By integrating multiple functional modules and intelligent algorithms, the system can respond to environmental changes in real time and provide continuous and stable wind measurement capabilities. The present invention introduces a Doppler frequency shift correction module and uses an adaptive correction factor to dynamically compensate for the Doppler frequency shift error caused by environmental changes, thereby effectively improving the accuracy of wind speed measurement. This dynamic correction mechanism enables the system to maintain high-precision measurements under various complex meteorological conditions.

[0023] 2. The multi-level signal processing module of the present invention improves the signal-to-noise ratio of the wind speed signal through multi-scale feature extraction, deep separable convolution and adaptive noise suppression, and can effectively suppress various noise interferences from the environment, which not only improves the quality of the measurement data, but also reduces the influence of noise on the measurement results, ensuring the reliability of the data; by using the deep separable convolution technology in the multi-level signal processing module, the amount of calculation in the feature processing of the wind speed signal is greatly reduced, which not only improves the processing speed, but also maintains the high-precision feature extraction capability, which is suitable for wind measurement applications with high real-time requirements.

[0024] 3. The intelligent layout optimization module of the present invention adopts a deep Q network algorithm, which can dynamically optimize the base station location configuration according to real-time wind speed data and environmental parameters. This adaptive optimization mechanism ensures the optimal performance of the system under different terrain and environmental conditions, allowing the wind measuring radar to quickly adapt to the changing wind field characteristics and minimize measurement errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0026] Figure 1 The overall flow chart of the wind speed measurement method based on the bistatic laser wind measuring radar with separate transmitter and receiver of the present invention;

[0027] Figure 2The overall architecture diagram of the wind speed measurement system based on the bistatic laser wind measurement radar with separate transmitter and receiver in the present invention;

[0028] Figure 3 It is a flow chart of the training process of the deep Q network of the present invention;

[0029] Figure 4 The following is a working flow chart of the Doppler frequency shift correction module of the present invention;

[0030] Figure 5 It is a block diagram of the deep Q network algorithm of the intelligent layout optimization module of the present invention;

[0031] Figure 6 It is a processing flow chart of the multi-stage signal processing module of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be described more clearly and completely below by way of a preferred embodiment in conjunction with the accompanying drawings, but the present invention is not limited to the scope of the embodiment.

[0033] like Figure 1 As shown, the present invention provides a wind speed measurement method based on a bistatic laser wind measuring radar with separate transmitter and receiver, comprising the following steps:

[0034] S1. Initialize the positions of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar, and collect environmental parameter data in real time through multiple distributed sensor nodes. The environmental parameter data includes air pressure, temperature and humidity, and transmit the environmental parameter data to the central processing unit for preprocessing.

[0035] In one embodiment, the sensor nodes are arranged according to the terrain characteristics of the wind measurement area to ensure that the measurement area can be fully covered. Each sensor node is integrated with a pressure sensor, a temperature sensor, and a humidity sensor for real-time collection of environmental parameter data in the wind measurement area. Each sensor node collects environmental parameter data at a preset time interval, such as once per second. The collected environmental parameter data is transmitted to the central processing unit via a wireless communication protocol such as LoRa, Wi-Fi, or a wired communication protocol such as RS485. The central processing unit pre-processes the environmental parameter data, including outlier detection and data smoothing. Outlier detection can use the Z-score method to identify outliers in the data, and data smoothing can use the moving average method to eliminate data fluctuations in a short period of time to ensure the accuracy and stability of the data.

[0036] S2. Use a dual-base laser wind measurement radar to obtain Doppler frequency shift data in the wind measurement area, and calculate the initial wind speed in the wind measurement area based on the Doppler frequency shift data.

[0037] In one embodiment, the calculation formula for calculating the initial wind speed in the wind measurement area based on the Doppler frequency shift data is as follows:

[0038]

[0039] Among them, v s represents the initial wind speed, δ represents the angle between the initial wind speed and the bisector of the bistatic angle, β represents the bistatic angle, that is, the angle between the direction of the laser beam emitted by the transmitting base station and the direction of the laser received by the receiving base station. In the configuration of the bistatic laser wind measuring radar, the transmitting base station and the receiving base station are usually physically separated, that is, the transmitting base station and the receiving base station are respectively set in different places, and λ represents the laser wavelength.

[0040] S3, such as Figure 4 As shown, an adaptive correction factor is calculated based on environmental parameter data, and the Doppler frequency shift data is corrected using the adaptive correction factor to obtain corrected wind speed data.

[0041] In one embodiment, the calculation formula of the adaptive correction factor is:

[0042]

[0043] Among them, AP, ΔT and ΔH represent the difference between the actual air pressure, temperature and humidity and the standard air pressure, temperature and humidity respectively, AP = PP 0 , ΔT=TT 0 , ΔH=HH 0 , where P, T and H represent the actual air pressure, temperature and humidity respectively, P 0 、T 0 and H 0 Represents standard atmospheric pressure, temperature and humidity respectively, k 1 , k 2 and k 3 It represents the correction coefficient, which reflects the influence of each environmental variable on the measurement results and can be obtained by fitting the experimental data.

[0044] The Doppler frequency shift data is corrected using the adaptive correction factor, and the calculation formula for the corrected wind speed data is as follows:

[0045] v r =v s ×α

[0046] Where α represents the adaptive correction factor.

[0047] S4. Perform multi-level processing on the corrected wind speed data, including multi-scale feature extraction, depth-separable convolution and adaptive noise suppression to improve the signal-to-noise ratio and measurement accuracy of the wind speed data.

[0048] In one embodiment, Figure 6 As shown, the specific implementation process is as follows:

[0049] Multi-scale feature extraction: The goal is to extract the features of wind speed data at different time scales. Use multi-scale convolution kernels, such as 3x3, 5x5, 7x7, etc., to perform convolution processing on the corrected wind speed data to extract multi-scale features. The convolution operation formula is:

[0050]

[0051] Among them, W k represents the convolution kernel matrix of different scales, and x represents the input corrected wind speed data matrix.

[0052] Depth-wise separable convolution processing: Perform depth-wise separable convolution on the corrected wind speed data after feature extraction to reduce computational complexity and improve computational efficiency. The specific formula is:

[0053] F sep (x) = P(W d *x)

[0054] Among them, W d Represents the depth convolution kernel, and P represents the point-by-point convolution operation.

[0055] Adaptive noise suppression: Through the adaptive noise suppression layer, the noise estimation network is used to predict the noise level in the corrected wind speed data, and the output signal is dynamically adjusted to suppress the noise in the wind speed data and improve the signal-to-noise ratio. The specific calculation formula is as follows:

[0056] F out =F sep (x)·(1-NEN(x))

[0057] Among them, F out represents the final output signal, and NEN(x) represents the noise level output by the noise estimation network.

[0058] S5. Construct a state vector s and input the state vector s into the trained deep Q network. The state vector s includes wind speed data after multi-level processing, current environmental parameter data, and the current positions of the transmitting base station and the receiving base station.

[0059] S6. Calculate the Q value of the possible action based on the state vector using the trained deep Q network, select the action with the largest Q value, and adjust the positions of the transmitting base station and the receiving base station according to the action with the largest Q value. The action includes adjusting the direction and amplitude of the positions of the transmitting base station and the receiving base station.

[0060] In one embodiment, Figure 5 As shown, the structure of the deep Q network is:

[0061] Input layer: Input the current location (x, y, z coordinates) of the transmitting base station and the receiving base station, multi-level processed wind speed data, wind field historical data and environmental parameter data. The spatial coordinates of the transmitting base station and the receiving base station are dynamically calculated by the deep Q network algorithm. The initial location of the transmitting base station and the receiving base station can be set according to the preset location or using GPS positioning. As the data is collected and analyzed, the deep Q network algorithm continuously updates and optimizes the location of the base station to adapt to the real-time changing wind field characteristics.

[0062] Hidden layer: two fully connected layers, each containing 256 neurons, using ReLU activation function;

[0063] Output layer: Outputs the Q value of each possible action and determines the optimal location of the transmitting base station and the receiving base station.

[0064] In one embodiment, Figure 3 As shown, the specific steps of deep Q network training are as follows:

[0065] S51, initializing the positions, wind speed data and environmental parameter data of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar, constructing a main Q network and a target Q network, and randomly initializing the weight parameters of the two networks.

[0066] Build the main Q network, which is a deep neural network responsible for approximating the Q-value function The input of the network is the current state vector, and the output is the Q value of each action a selected in the current state vector, which initializes the weight parameters of the main Q network. It is usually initialized using a random distribution such as the normal distribution or the uniform distribution.

[0067] Construct a target Q network, which has the same structure as the main Q network, but the weight parameters of the target Q network are Initialize the weight parameters of the target Q network without updating them with each training step is the weight parameter of the main Q network Same initial value.

[0068] S52: Create an experience replay buffer for storing state transition data obtained during the training process, and initialize a priority buffer, wherein the state transition data includes a four-tuple of a current state vector, an action selected in the current state vector, an immediate reward, and a next state vector.

[0069] Experience replay buffer: Create an experience replay buffer to store the state transition data obtained during the training process, namely the experience samples (s, a, r, s'), including the current state vector, the action selected in the current state vector, the immediate reward, and the next state vector. Set the capacity of the experience replay buffer to 1000000, for example. When the experience samples in the experience replay buffer reach the upper limit of the capacity, the new experience samples will overwrite the oldest samples.

[0070] Priority Buffer: Create a priority buffer to store the priority P(i) of each experience sample.

[0071] Current state vector s: the current state of the system, including environmental parameter data, current wind speed data, and the location coordinates of the transmitting and receiving base stations;

[0072] Action a selected in the current state vector: adjustment operation of the base station position, such as changes in the base station position coordinates in the x, y and z directions;

[0073] Instant reward r: The reward function is related to the wind speed measurement error, which is usually the negative value of the measurement error. The goal is to minimize the error. Therefore, the smaller the measurement error, the higher the reward. The specific calculation formula is:

[0074] r=-measurement error+λ 1 SNR-λ 2 Energy consumption

[0075] Among them, the measurement error represents the error of wind speed measurement under the current base station configuration, SNR represents the signal-to-noise ratio, energy consumption represents the energy consumption of executing the action, and λ 1 and λ 2 Represents the weight parameter, which is generally obtained through experiments or prior experience.

[0076] S53, according to the current state vector, using the main Q network through ∈ explore -The greedy strategy selects an action and adjusts the positions of the transmitting and receiving base stations according to the selected action, obtains the immediate reward and the next state vector, and stores the corresponding state transition data in the experience replay buffer as an experience sample.

[0077] In one embodiment, at the beginning of training, a high exploration probability ∈ explore For example, let ∈ explore = 1 to encourage the deep Q network to explore different combinations of states and actions, according to ∈ explore -The greedy strategy randomly selects an action a under each current state vector s to collect initial experience, executes action a to interact with the environment to obtain an immediate reward r and observes the next state vector s', and stores the four-tuple (s, a, r, s') into the experience replay buffer;

[0078] The above process is repeated until the experience replay buffer accumulates enough experience samples, usually 10%-20% of the buffer capacity.

[0079] S54, calculating the TD error of each experience sample, calculating the priority of each experience sample according to the TD error, and storing the priority value in the priority buffer.

[0080] For each newly stored experience sample, calculate its TD error. The specific calculation formula is as follows:

[0081]

[0082] Among them, ρ represents the TD error of the experience sample, r represents the immediate reward, γ represents the discount factor, s' represents the next state vector, that is, it represents the next state entered by the dual-base laser wind measurement radar after executing the action a selected by the current state vector under the current state vector s, such as adjusting the positions of the transmitting base station and the receiving base station, a' represents all possible actions at the next state vector s', that is, it represents one of all possible actions that the dual-base laser wind measurement radar can take under the next state vector s'. The action can be further adjustment of the base station position, change of measurement strategy, optimization of transmission and reception directions, etc., s represents the state vector, a represents the action selected in the current state vector, represents the weight parameter of the main Q network, represents the weight parameter of the target Q network, Represents the maximum value of the Q value of all possible actions a' at the next state vector s';

[0083] γ represents the discount factor, which is used to calculate the long-term optimization goal of wind speed measurement and base station position adjustment strategy. It represents the impact of future wind speed measurement errors, signal-to-noise ratio (SNR), energy consumption and other indicators on current decisions. For example, when the dual-base laser wind measurement radar adjusts the position of the transmitting base station and the receiving base station, the strategy adopted will not only affect the current measurement accuracy and signal-to-noise ratio, but also the accuracy and energy consumption of a series of future measurement steps. By setting an appropriate discount factor γ, the deep Q network can balance the current measurement accuracy and the accuracy of future measurement results when optimizing the base station position strategy.

[0084] If a higher discount factor is chosen, such as γ = 0.9 or γ = 0.99, the deep Q network will attach more importance to future measurements, which means that it will try to find a long-term optimal position adjustment strategy, possibly sacrificing some immediate measurement accuracy for better long-term results.

[0085] If a lower discount factor is chosen, such as γ = 0.1 or γ = 0.2, the deep Q network will focus more on optimizing the immediate measurement results, emphasizing high accuracy and low error in current measurements, and less on potential future rewards.

[0086] The TD error is used to calculate the priority of each experience sample. The specific calculation formula is:

[0087] P(i)=|ρ i |+∈ prior

[0088] Among them, P(i) represents the priority of the i-th experience sample, ρ i represents the TD error of the i-th empirical sample, ∈ prior A small constant that prevents the priority from being zero, such as 0.01.

[0089] For example, when training the main Q network, the total number of training steps and the update frequency of the target Q network can be set. For example, the total number of steps is set to 100,000 steps, and the update frequency is set to update once every 100,000 steps. In each step of training, according to the current state vector and ∈ explore - Greedy strategy selects an action and generates a random number w between 0 and 1. If w<∈ explore , that is, in the exploration phase, a random action is selected. If w ≥ ∈ explore , that is, in the utilization stage, the action corresponding to the maximum Q value predicted by the main Q network is selected. As the training proceeds, ∈ explore Gradually decrease, for example, from 1 to 0.1, to balance the relationship between exploration and exploitation.

[0090] S55. Randomly extract a small batch of experience samples from the experience replay buffer according to the priority, and calculate the target Q value for each experience sample.

[0091] The calculation formula of the empirical sample target Q value is:

[0092]

[0093] Among them, r represents the immediate reward, γ represents the discount factor, and s′ represents the next state vector. Represents the weight parameters of the target Q network.

[0094] S56, calculating the loss function of the main Q network, and minimizing the loss function through the back propagation algorithm, updating the weight parameters of the main Q network, and copying the weight parameters of the main Q network to the weight parameters of the target Q network every fixed number of steps.

[0095] The loss function of the main Q network is calculated as:

[0096]

[0097] in, represents the loss function of the main Q network, i represents the index of the small batch experience sample, N represents the number of small batch experience samples, and y i represents the target Q value of the i-th experience sample, s i represents the current state vector of the i-th experience sample, a i Represents the action selected by the i-th experience sample in the current state vector.

[0098] Use the back propagation algorithm to calculate the loss function with respect to the main Q network weight parameters The gradient of the main Q network is updated according to the calculated gradient, such as StochasticGradient Descent or Adam optimizer. To minimize the loss function

[0099] Every fixed number of steps, for example 10,000 steps, the weight parameters of the main Q network are Update to target Q network weight parameters To ensure the stability of training, the parameter update formula is as follows:

[0100]

[0101] Among them, τ represents the soft update parameter, and its value is usually very small, such as 0.0001.

[0102] S57, repeat steps S53 to S26 until the training obtains a preset number of steps, the loss function converges to a desired threshold, or the performance of the main Q network reaches the target requirement.

[0103] S7. At the adjusted positions of the transmitting base station and the receiving base station, the wind speed is remeasured using a dual-base laser wind measurement radar to obtain new Doppler frequency shift data, and new wind speed is calculated based on the new Doppler frequency shift data. After correction and multi-stage processing, new multi-stage processed wind speed data is obtained.

[0104] S8. Repeat steps S5 to S7 to continuously optimize the positions of the transmitting base station and the receiving base station and the wind speed measurement results until a preset measurement accuracy is achieved or the system is stable.

[0105] Based on the output results of Deep Q, the base station position is adjusted in time through mechanical control devices such as servo motors or stepper motors to ensure the optimal configuration of the measurement system.

[0106] In one embodiment, the corrected and processed wind speed data is analyzed to generate a wind speed distribution diagram showing the wind speed changes in different areas. The distribution characteristics and change trends of the wind speed are determined by statistical analysis methods (such as mean and variance analysis).

[0107] In one embodiment, a wind speed prediction model is constructed using historical data and real-time measurement data. The prediction model uses machine learning algorithms (such as regression analysis, time series prediction, etc.) to predict wind speed changes over a period of time in the future. The training and verification process of the model takes into account a variety of environmental factors and historical data patterns to ensure the accuracy and reliability of the prediction results.

[0108] The measurement and prediction results are used for applications such as wind farm layout optimization, meteorological monitoring, and the formulation and adjustment of aerospace flight plans. Based on the analysis results of wind speed data, the system can provide optimization suggestions to improve the operational efficiency and safety of the system.

[0109] like Figure 2 As shown, the present invention also provides a wind speed measurement system based on a bistatic laser wind measuring radar with separate transmitters and receivers, which is used to implement the wind speed measurement method based on a bistatic laser wind measuring radar with separate transmitters and receivers as described above, and specifically includes the following modules:

[0110] The central processing unit is used to receive and process environmental parameter data transmitted from multiple sensor nodes and execute algorithms for data analysis and optimization control. The central processing unit is the core control module of the system and is responsible for coordinating the work of various modules, including data preprocessing, correction factor calculation, base station location optimization and signal processing.

[0111] In one embodiment, the central processing unit receives data from the environmental parameter acquisition module, and performs outlier detection and data smoothing to ensure the accuracy and stability of the input data; calculates an adaptive correction factor based on the environmental parameter data to correct the Doppler frequency shift data; dynamically optimizes the base station location configuration based on the corrected wind speed data and environmental parameters through a deep Q network algorithm; and controls the multi-level signal processing module to perform data processing to improve the signal-to-noise ratio and measurement accuracy.

[0112] The environmental parameter acquisition module is used to collect environmental parameter data in real time through multiple sensor nodes. The environmental parameter data includes air pressure, temperature and humidity. The sensor nodes are arranged at different positions in the wind measurement area to ensure the comprehensiveness and accuracy of data acquisition. The sensor nodes transmit the data to the central processing unit through a wireless communication network or a wired communication network.

[0113] The dual-base laser wind measuring radar includes a transmitting base station and a receiving base station, which are used to transmit a laser beam of a specific wavelength to a target area and receive laser scattered by particles in the atmosphere of the wind measuring area to obtain Doppler frequency shift data. The dual-base laser wind measuring radar includes a transmitting base and a receiving base, wherein the transmitting base is responsible for transmitting the laser beam, and the receiving base is responsible for measuring the Doppler frequency shift rate of the scattered laser. The dual-base laser wind measuring radar is connected to a central processing unit and is used to transmit the Doppler frequency shift data to the central processing unit for processing.

[0114] A Doppler frequency shift correction module, wherein the correction factor calculation unit is used to calculate an adaptive correction factor based on the environmental parameter data, and apply the adaptive correction factor to the Doppler frequency shift data for correction to obtain corrected wind speed data. The Doppler frequency shift correction module is connected to the central processing unit. The correction process of the Doppler frequency shift data includes calculating the radial velocity and adjusting the line of sight velocity to improve the measurement accuracy.

[0115] The intelligent layout optimization module is used to dynamically optimize the position configuration of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar according to the corrected wind speed data and environmental parameter data through the deep Q network algorithm. The optimization includes adjusting the spatial coordinates of the base station to minimize the measurement error and improve the measurement accuracy. The module can dynamically adjust the base station layout according to the real-time wind field data and environmental parameters to ensure the optimal performance of the system under different environmental conditions.

[0116] The multi-level signal processing module is used to perform multi-level processing on the corrected wind speed data. The processing includes a multi-scale feature extraction unit, a depth-separable convolution unit and an adaptive noise suppression unit to improve the signal-to-noise ratio and measurement accuracy of the signal. The multi-level signal processing module is connected to the central processing unit.

[0117] The multi-scale feature extraction unit is used to extract multi-scale features in the corrected wind speed data, the depth-separable convolution unit is used to reduce the computational complexity when processing the extracted multi-scale features, and the adaptive noise suppression unit is used to suppress environmental noise and measurement noise in the wind speed data, thereby improving the signal-to-noise ratio and measurement accuracy of the wind speed data.

[0118] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A wind speed measurement method based on a bistatic laser wind radar with separate transmitter and receiver, characterized in that: The following steps are involved: S1, initializing the positions of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar, and collecting environmental parameter data in real time through multiple distributed sensor nodes, the environmental parameter data including air pressure, temperature and humidity, and transmitting the environmental parameter data to the central processing unit for preprocessing; S2. Use a dual-base laser wind measurement radar to obtain Doppler frequency shift data of the wind measurement area, and calculate the initial wind speed of the wind measurement area based on the Doppler frequency shift data; S3, calculating an adaptive correction factor based on the environmental parameter data, and using the adaptive correction factor to correct the Doppler frequency shift data to obtain corrected wind speed data; S4, performing multi-stage processing on the corrected wind speed data, the multi-stage processing including multi-scale feature extraction, depth-separable convolution and adaptive noise suppression to improve the signal-to-noise ratio and measurement accuracy of the wind speed data; S5, construct a state vector, and input the state vector into the trained deep Q network, the state vector includes the wind speed data after multi-level processing, the current environmental parameter data, and the current positions of the transmitting base station and the receiving base station; S6. Calculate the Q value of the possible action using the trained deep Q network based on the state vector, select the action with the largest Q value, and adjust the position of the transmitting base station and the receiving base station according to the action with the largest Q value. The action includes adjusting the direction and amplitude of the position of the transmitting base station and the receiving base station. S7. At the adjusted positions of the transmitting base station and the receiving base station, the wind speed is remeasured using a dual-base laser wind measuring radar to obtain new Doppler frequency shift data, and new wind speed is calculated based on the new Doppler frequency shift data, and new multi-level processed wind speed data is obtained after correction and multi-level processing; S8. Repeat steps S5 to S7 to continuously optimize the positions of the transmitting base station and the receiving base station and the wind speed measurement results until the preset measurement accuracy is reached or the system is stable, and output the wind speed measurement results.

2. The method for measuring wind speed based on a bistatic laser wind radar with separate transmitter and receiver according to claim 1, characterized in that: The specific steps for training the deep Q network are as follows: S51, initializing the positions, wind speed data and environmental parameter data of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar, constructing a main Q network and a target Q network, and randomly initializing the weight parameters of the two networks; S52, creating an experience replay buffer for storing state transition data obtained during the training process, and initializing a priority buffer, wherein the state transition data includes a four-tuple of a current state vector, an action selected in the current state vector, an immediate reward, and a next state vector; S53, according to the current state vector, using the main Q network through ∈ explore -The greedy strategy selects an action and adjusts the positions of the transmitting and receiving base stations according to the selected action, obtains the immediate reward and the next state vector, and stores the corresponding state transition data in the experience playback buffer as an experience sample; S54, calculating the TD error of each experience sample, calculating the priority of each experience sample according to the TD error, and storing the priority value in a priority buffer; S55, randomly extracting a small batch of experience samples from the experience playback buffer according to the priority, and calculating the target Q value for each experience sample; S56, calculating the loss function of the main Q network, and minimizing the loss function through the back propagation algorithm, updating the weight parameters of the main Q network, and copying the weight parameters of the main Q network to the weight parameters of the target Q network every fixed number of steps; S57, repeat steps S53 to S56 until the training obtains a preset number of steps, the loss function converges to a desired threshold, or the performance of the main Q network reaches the target requirement.

3. The method for measuring wind speed based on a bistatic laser wind radar with separate transmitter and receiver according to claim 2, characterized in that: The calculation formula of the TD error of the empirical sample is: Among them, ρ represents the TD error of the experience sample, r represents the immediate reward, which is related to the wind speed measurement error and is the negative value of the wind speed measurement error. The smaller the wind speed measurement error, the higher the reward. γ represents the discount factor, which is used to calculate the long-term optimization goal of wind speed measurement and the position adjustment strategy of the transmitting base station and the receiving base station. s represents the state vector, including the wind speed data after multi-level processing, the current environmental parameter data, and the current position of the transmitting base station and the receiving base station. a represents the action selected in the current state vector, that is, the adjustment operation of the position of the transmitting base station and the receiving base station. s' represents the next state vector, that is, the next state entered by the dual-base laser wind measurement radar after executing the action a selected by the current state vector (such as adjusting the position of the transmitting base station and the receiving base station) under the current state vector s. a' represents all possible actions at the next state vector s'. represents the weight parameter of the main Q network, represents the weight parameter of the target Q network, Represents the maximum value of the Q value of all possible actions a' at the next state vector s'; The calculation formula for the experience sample priority is: P(i)=|δ i |+∈ prior Among them, P(i) represents the priority of the i-th experience sample, δ i represents the TD error of the i-th empirical sample, ∈ prior A small constant representing protection against priority zero.

4. The method for measuring wind speed based on a bistatic laser wind radar with separate transmitter and receiver according to claim 2, characterized in that: The calculation formula of the empirical sample target Q value is: Among them, r represents the immediate reward, γ represents the discount factor, s' represents the next state vector, Represents the weight parameters of the target Q network.

5. The method for measuring wind speed based on a bistatic laser wind radar with separate transmitter and receiver according to claim 2, characterized in that: The loss function of the main Q network is calculated as: in, represents the loss function of the main Q network, i represents the index of the small batch experience sample, N represents the number of small batch experience samples, and y i represents the target Q value of the i-th experience sample, s i represents the current state vector of the i-th experience sample, a i Represents the action selected by the i-th experience sample in the current state vector.

6. The method for measuring wind speed based on a bistatic laser wind radar with separate transmitter and receiver according to claim 1, characterized in that: The calculation formula for the initial wind speed in the wind measurement area based on Doppler frequency shift data is as follows: Among them, v s represents the initial wind speed, δ represents the angle between the initial wind speed and the bistatic angle bisector, β represents the bistatic angle, and λ represents the laser wavelength.

7. The method for measuring wind speed based on a bistatic laser wind radar with separate transmitter and receiver according to claim 1, characterized in that: The calculation formula of the adaptive correction factor is: Wherein, α represents the adaptive correction factor, ΔP, ΔT and ΔH represent the difference between the actual air pressure, temperature and humidity and the standard air pressure, temperature and humidity respectively, P0, T0 and H0 represent the standard air pressure, temperature and humidity respectively, k1, k2 and k3 represent the correction coefficients; The Doppler frequency shift data is corrected using the adaptive correction factor, and the calculation formula for the corrected wind speed data is as follows: v r =v s ×α。 8. A wind speed measurement system based on a bistatic laser wind radar with separate transmitters and receivers, used to implement a wind speed measurement method based on a bistatic laser wind radar with separate transmitters and receivers as claimed in any one of claims 1 to 7, characterized in that: include: a central processing unit for receiving and processing environmental parameter data transmitted from a plurality of sensor nodes; An environmental parameter acquisition module is used to collect environmental parameter data in real time through multiple sensor nodes, and the sensor nodes transmit the data to a central processing unit through a wireless communication network or a wired communication network; A bistatic laser wind measuring radar, including a transmitting base station and a receiving base station, for transmitting a laser beam of a specific wavelength to a target area, and receiving laser light scattered by particles in the atmosphere of the wind measuring area to obtain Doppler frequency shift data. The bistatic laser wind measuring radar is connected to a central processing unit to transmit the Doppler frequency shift data to the central processing unit for processing; A Doppler frequency shift correction module is used to calculate an adaptive correction factor based on environmental parameter data, and to correct the Doppler frequency shift data according to the adaptive correction factor to obtain corrected wind speed data. The Doppler frequency shift correction module is connected to the central processing unit; Intelligent layout optimization module, which is used to dynamically optimize the location configuration of the transmitting base station and the receiving base station in the dual-base laser wind measurement radar according to the corrected wind speed data and environmental parameter data through the deep Q network algorithm; The multi-level signal processing module is used to perform multi-level processing on the corrected wind speed data to improve the signal-to-noise ratio and measurement accuracy of the signal. The multi-level signal processing module is connected to the central processing unit.

9. A wind speed measurement system based on a bistatic laser wind radar with separate transmitter and receiver according to claim 8, characterized in that: The Doppler frequency shift correction module comprises a correction factor calculation unit, wherein the correction factor calculation unit is used to calculate an adaptive correction factor based on the environmental parameter data, and apply the adaptive correction factor to the correction of the Doppler frequency shift data.

10. The wind speed measurement system based on a bistatic laser wind radar with separate transmitter and receiver according to claim 8, characterized in that: The multi-stage signal processing module includes a multi-scale feature extraction unit, a depth-separable convolution unit and an adaptive noise suppression unit. The multi-scale feature extraction unit is used to extract multi-scale features in the corrected wind speed data, the depth-separable convolution unit is used to reduce the computational complexity when processing the extracted multi-scale features, and the adaptive noise suppression unit is used to suppress environmental noise and measurement noise in the wind speed data, thereby improving the signal-to-noise ratio and measurement accuracy of the wind speed data.

Images