Method, system and apparatus for temporarily replacing a failed laser wind lidar

Abstract

The application discloses a kind of temporarily replace the method, system and device of failure laser wind radar, comprising: detection module, for analyzing the output data of radar, whether laser radar occurs failure based on analysis result is judged, if failure output logic 0, otherwise output logic 1;Data prediction module is used to predict the wind speed, wind direction, temperature, air pressure and humidity five kinds of meteorological information of wind farm according to historical data and meteorological source data;Data switching module is used to when output logic changes from 0 to 1, data source is switched from simulation system data source to laser wind radar device data source, when logic value changes from 1 to 0, data source is switched from laser wind radar device data source to simulation system data source, when logic value does not change, keep original state.The application can realize temporarily replace failure laser wind radar.

Description

Technical Field

[0001] This invention relates to the field of laser wind radar, and in particular to a method, system, and apparatus for temporarily replacing a faulty laser wind radar. Background Technology

[0002] Due to the dynamic nature of wind resources, offshore wind farms typically install laser wind-measuring radar devices to measure wind resource conditions in order to accurately assess the power output of renewable energy sources. A laser wind-measuring radar is essentially a small weather station, capable of monitoring multiple meteorological data points in real time, such as wind speed, wind direction, temperature, air pressure, and humidity. As offshore wind farms increasingly locate in deeper waters, a malfunction in the laser radar can lead to the loss of all meteorological measurements during the malfunction period if personnel cannot arrive promptly for repairs, directly impacting the assessment of the wind farm's power output. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, and apparatus for temporarily replacing a faulty laser wind measuring radar, aiming to solve the problem of temporarily replacing a faulty laser wind measuring radar.

[0004] This invention provides a method for temporarily replacing a faulty laser wind-measuring radar, comprising:

[0005] S1. Analyze the output data of the radar through the detection module, and determine whether the lidar has malfunctioned based on the analysis results. If a malfunction occurs, output logic 0; otherwise, output logic 1.

[0006] S2. The data prediction module predicts five types of meteorological information of the wind farm, namely wind speed, wind direction, temperature, air pressure and humidity, based on historical data and meteorological source data.

[0007] S3. When the output logic changes from 0 to 1, the data source is switched from the simulation system data source to the laser wind measuring radar device data source. When the logic value changes from 1 to 0, the data source is switched from the laser wind measuring radar device data source to the simulation system data source. When the logic value does not change, the original state is maintained.

[0008] The present invention also provides a system for temporarily replacing a faulty laser wind-measuring radar, comprising:

[0009] The detection module is used to analyze the output data of the radar and determine whether the lidar has malfunctioned based on the analysis results. If a malfunction occurs, it outputs logic 0; otherwise, it outputs logic 1.

[0010] The data prediction module is used to predict five types of meteorological information of wind farms, namely wind speed, wind direction, temperature, air pressure and humidity, based on historical data and meteorological source data.

[0011] The data switching module is used to switch the data source from the simulation system data source to the laser wind measuring radar device data source when the output logic changes from 0 to 1, and to switch the data source from the laser wind measuring radar device data source to the simulation system data source when the logic value changes from 1 to 0, and to maintain the original state when the logic value does not change.

[0012] This invention also provides an apparatus for temporarily replacing a faulty laser wind-measuring radar, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the above method.

[0013] This invention also provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor, implements the steps of the above-described method.

[0014] Using the embodiments of the present invention, a temporary replacement for a faulty laser wind measuring radar can be achieved.

[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a method for temporarily replacing a faulty laser wind-measuring radar according to an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of a temporary replacement faulty laser wind measuring radar system according to an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of a device for temporarily replacing a faulty laser wind-measuring radar according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Method Implementation Examples

[0022] According to embodiments of the present invention, a method for temporarily replacing a faulty laser wind-measuring radar is provided. Figure 1 This is a flowchart of a method for temporarily replacing a faulty laser wind-measuring radar according to an embodiment of the present invention, as follows: Figure 1 As shown, it specifically includes:

[0023] S1. Analyze the output data of the radar through the detection module, and determine whether the lidar has malfunctioned based on the analysis results. If a malfunction occurs, output logic 0; otherwise, output logic 1.

[0024] S2. The data prediction module predicts five types of meteorological information of the wind farm, namely wind speed, wind direction, temperature, air pressure and humidity, based on historical data and meteorological source data.

[0025] S3. When the output logic changes from 0 to 1, the data source is switched from the simulation system data source to the laser wind measuring radar device data source. When the logic value changes from 1 to 0, the data source is switched from the laser wind measuring radar device data source to the simulation system data source. When the logic value does not change, the original state is maintained.

[0026] The data prediction module predicts five types of meteorological information for wind farms—wind speed, wind direction, temperature, air pressure, and humidity—based on historical data and meteorological source data. Specifically, it uses a conditional autoregressive flow model to model the multidimensional data distribution of high-resolution wind farm meteorological information and learns the distribution through sampling to recursively predict data for future moments.

[0027] The conditional autoregressive flow model comprises a compression layer, a reversible convolutional layer, an affine coupling layer, a separation layer, and a periodic fitting layer connected in sequence. Meteorological information is normalized and then input into the compression layer. The compression layer compresses high-resolution and low-resolution meteorological information and extracts salient features before inputting them into the reversible convolutional layer. The reversible convolutional layer shuffles and reassembles information from different dimensions before inputting it into the affine coupling layer. The affine coupling layer obtains the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions based on affine functions. The affine coupling layer is used for collaborative prediction across all dimensions. The separation layer is used to restore latent noise variables and guide low-resolution meteorological information. The periodic fitting layer ensures that the generated wind farm meteorological data can capture the true periodic variation patterns. The condition of the conditional autoregressive flow model is based on the forward and backward bidirectional propagation of information from the neighborhood content of low-resolution meteorological forecast data, thereby obtaining maximum fluidity and effectively guiding the establishment of multi-dimensional high-resolution wind farm meteorological data.

[0028] The affine coupling layer, based on affine functions, obtains the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions. Specifically, in the first stage, the input and output of the four dimensions of wind direction o, temperature t, humidity h, and air pressure p are fixed, and an affine transformation is performed on the wind speed s dimension. The formula for the affine transformation is: Where the scaling function w s Translation function w e The neural network model to be learned is as follows: In the second stage, the input and output of the four dimensions of wind speed, temperature, humidity, and air pressure are kept constant, and an affine transformation is performed on the wind direction dimension; in the third stage, the input and output of the four dimensions of wind speed, wind direction, humidity, and air pressure are kept constant, and an affine transformation is performed on the temperature dimension; in the fourth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and air pressure are kept constant, and an affine transformation is performed on the humidity dimension; in the fifth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and humidity are kept constant, and an affine transformation is performed on the air pressure dimension; all affine transformations are reversible processes. In this way, the interaction relationship between different dimensions is obtained through five stages of cyclical transformation. The interaction relationship is used to promote the collaborative prediction of the entire dimension.

[0029] The specific implementation method is as follows:

[0030] The system mainly consists of three modules: a laser wind measurement radar fault detection module, a data simulation and prediction module, and a laser wind measurement radar data transmission channel switching module.

[0031] Laser wind measuring radar fault detection module: By judging the changes in the output data of the radar before and after, if multiple data such as air pressure, wind speed, wind direction and temperature do not change within 10 minutes, it can be determined that the laser wind measuring radar has failed and outputs a logic value of 0; otherwise, it outputs a logic value of 1.

[0032] The data switching module changes the data source based on the output result of the logical value. When the logical value changes from 0 to 1, the data source switches from the simulation system data source to the laser wind measuring radar device data source. When the logical value changes from 1 to 0, the data source switches from the laser wind measuring radar device data source to the simulation system data source. When the logical value does not change, the original state is maintained.

[0033] The data simulation and prediction module needs to simultaneously simulate five types of meteorological data for the wind farm in the future, including wind speed, wind direction, temperature, air pressure, and humidity, based on historical data and meteorological source forecast data. These five types of data are closely related, but the interdependence between them is difficult to describe through explicit mathematical formulas. At the same time, although the multidimensional space composed of historical data sequences of the five types of indicators contains a lot of prior knowledge, the complex multidimensional distribution cannot be described by empirical distribution expressions.

[0034] Currently, the highest resolution scale for meteorological sources in weather forecasting is 25 kilometers, which is relatively coarse. However, the wind turbine layout range of wind farms is generally between 3 and 5 kilometers, requiring much more precise meteorological change information. To reconstruct the true variability of wind farm data from existing coarse meteorological source information, the actual meteorological data variables of wind farms can be assumed to consist of two parts: low-frequency meteorological source data variables and high-frequency components. Due to the randomness of meteorological data changes, and the need to learn information from five dimensions—wind speed, wind direction, temperature, air pressure, and humidity—this invention does not directly learn the functional mapping relationship between low-frequency meteorological source data and high-frequency wind farm data. Instead, it uses a conditional autoregressive flow model to model the multidimensional data distribution of high-frequency wind farms, and then iteratively extrapolates future data by sampling the learned distribution.

[0035] Assuming the noise distribution of the training and test data is identical, the conditional autoregressive flow model designed in this invention is expressed as: P(y_t|x_t-1,x_t,x_t+1). The overall framework of the model is as follows: During the training phase, it is an encoding process. Guided by historical low-frequency meteorological source data sequences, high-frequency meteorological information is encoded into noise latent variables following a certain distribution through a reversible neural network. During the inference phase, it transforms into a decoding process. Under the influence of ambiguous future meteorological source data, the original noise latent variables are decoded into detailed future wind farm meteorological content. The parameters of the conditional autoregressive flow model are updated using the likelihood function maximization and gradient descent method. This model can effectively solve the problem of learning and predicting multi-dimensional data.

[0036] The conditional autoregressive flow model consists of an activation layer, an affine layer, a compression layer, and a separation layer. First, the data is normalized. The compression layer combines high-resolution meteorological information with low-resolution meteorological information, while the separation layer does the opposite. Given the internal interactions among the five variables—wind speed, wind direction, temperature, air pressure, and humidity—this invention designs an affine function for dimensional interaction to explore the relationship between each dimension and the others. In the first stage, the input and output of the four dimensions—wind direction (o), temperature (t), humidity (h), and air pressure (p)—are fixed. An affine transformation is performed on the wind speed (s) dimension, with the formula: o,t,h,p=o,t,h,p; s(i)=w s (o,t,h,p)*s(i-1)+w e (o,t,h,p); where the scaling function is w s Translation function w e This is the neural network model to be learned. In the second stage, the input and output of the four dimensions of wind speed, temperature, humidity, and air pressure are kept constant, and an affine transformation is performed on the wind direction dimension. In the third stage, the input and output of the four dimensions of wind speed, wind direction, humidity, and air pressure are kept constant, and an affine transformation is performed on the temperature dimension. In the fourth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and air pressure are kept constant, and an affine transformation is performed on the humidity dimension. In the fifth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and humidity are kept constant, and an affine transformation is performed on the air pressure dimension. All affine transformations are reversible processes. Through these five stages of cyclical transformations, the interaction relationships between different dimensions are obtained. These interaction relationships can effectively promote the collaborative prediction of the entire dimension.

[0037] It is generally accepted that changes in meteorological conditions follow the Markov property, meaning that the meteorological condition at the next moment is determined only by the meteorological condition at the current moment. Based on the Markov property, the meteorological condition at each moment should be closely related to the meteorological conditions at the previous and next moments, but independent of the conditions at other non-adjacent moments. Therefore, a bidirectional flow structure of time lag and time lead is introduced to guide the reconstruction of high-resolution wind farm meteorological information. The output feature of the time lag flow is modeled as: h_i = f_z(x_t, x_t-1, h_i-1), with information propagating from the lag direction. The output feature of the time lead flow is modeled as: h_i = f_c(x_t, x_t+1, h_i+1), with information propagating from the lead direction. The output feature values ​​of the lead and lag flows are fused and concatenated to form low-resolution meteorological source information containing the current moment and adjacent moments, which is used as a condition to update the high-resolution wind farm meteorological data at the target moment. The network weights of f_c and f_z at each time point are shared, but the weights of the autoregressive flow model are not shared, but rather interconnected. In this way, the information from each high-resolution wind farm can fully interact with the low-resolution information from neighboring meteorological sources.

[0038] Considering the daily variation characteristics of weather over time, generally speaking, from the early morning of the first day to the early morning of the second day, the temperature gradually rises from low to high and then gradually falls from high to low in a cyclical process. Similarly, from the early morning of the first day to the early morning of the second day, the humidity gradually falls from high to low and then gradually rises from low to high in a cyclical process. Therefore, the historical data in these two dimensions should exhibit a cyclical trend. However, since these trends are mainly influenced by sunrise and sunset, and the times of sunrise and sunset are dynamic (sunrise is earlier and sunset is later in summer, and sunrise is later and sunset is earlier in winter), this invention introduces a variable period fitting layer. The time base vector is set to t=12 hours, with sunrise at 6 am and sunset at 6 pm. Simultaneously, 12 different time periods are constructed based on the 12 months of the year. This layer approximates the annual temperature and humidity cycle by weighted superposition of 12 sine and cosine functions, y=Σa_icos(it) +b_isin(it). a and b are calculated using a nonlinear neural network model. The final model can fully learn the dynamic periodic variation of the sequence, thereby guiding the accurate establishment of meteorological information for wind farms.

[0039] During normal radar data operation without failure, the data is transmitted to the data center via a reverse isolation device for training without prediction. When the radar fails, prediction is performed but training is not conducted.

[0040] This invention proposes a mapping relationship between the probability distribution of low-resolution meteorological source data, the distribution of noise latent variables, and the probability distribution of high-resolution lidar meteorological data. This mapping relationship is modeled using a reversible conditional autoregressive flow model. The autoregressive flow employs a reversible multidimensional alternating dependent affine transformation. Low-resolution flow information propagates from two different directions, leading and lagging, which can fully explore the continuity and differences of features. At the same time, the periodic variation of sunrise and sunset is also learned through trigonometric transformation.

[0041] System or device embodiment 1

[0042] According to embodiments of the present invention, a system for temporarily replacing a faulty laser wind-measuring radar is provided. Figure 2 This is a schematic diagram of a system for temporarily replacing a faulty laser wind-measuring radar according to an embodiment of the present invention, as shown below. Figure 2 As shown, it specifically includes:

[0043] The detection module is used to analyze the output data of the radar and determine whether the lidar has malfunctioned based on the analysis results. If a malfunction occurs, it outputs logic 0; otherwise, it outputs logic 1.

[0044] The data prediction module is used to predict five types of meteorological information of wind farms, namely wind speed, wind direction, temperature, air pressure and humidity, based on historical data and meteorological source data.

[0045] The data switching module is used to switch the data source from the simulation system data source to the laser wind measuring radar device data source when the output logic changes from 0 to 1, and to switch the data source from the laser wind measuring radar device data source to the simulation system data source when the logic value changes from 1 to 0, and to maintain the original state when the logic value does not change.

[0046] The data prediction module is specifically used to: model the multidimensional data distribution of high-resolution wind farm meteorological information using a conditional autoregressive flow model, and to recursively predict data for future times by sampling and learning the good distribution.

[0047] The conditional autoregressive flow model comprises a compression layer, a reversible convolutional layer, an affine coupling layer, a separation layer, and a periodic fitting layer connected in sequence. Normalized meteorological information is input into the compression layer. The compression layer compresses high-resolution and low-resolution meteorological information and extracts salient features before inputting them into the reversible convolutional layer. The reversible convolutional layer shuffles and reassembles information from different dimensions before inputting it into the affine coupling layer. The affine coupling layer uses affine functions to obtain the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions. The affine coupling layer is used for collaborative prediction across all dimensions. The separation layer is used to restore latent noise variables and guide low-resolution meteorological information. The periodic fitting layer ensures that the generated wind farm meteorological data captures the true periodic variation patterns. The condition of the conditional autoregressive flow model is based on the forward and backward bidirectional propagation of information from the neighborhood content of low-resolution meteorological forecast data, thereby obtaining maximum fluidity and effectively guiding the establishment of multi-dimensional high-resolution wind farm meteorological data.

[0048] The affine coupling layer, based on affine functions, obtains the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions. Specifically, in the first stage, the input and output of the four dimensions of wind direction o, temperature t, humidity h, and air pressure p are fixed, and an affine transformation is performed on the wind speed s dimension. The formula for the affine transformation is: Where the scaling function w s Translation function w eThe neural network model to be learned is as follows: In the second stage, the input and output of the four dimensions of wind speed, temperature, humidity, and air pressure are kept constant, and an affine transformation is performed on the wind direction dimension; in the third stage, the input and output of the four dimensions of wind speed, wind direction, humidity, and air pressure are kept constant, and an affine transformation is performed on the temperature dimension; in the fourth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and air pressure are kept constant, and an affine transformation is performed on the humidity dimension; in the fifth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and humidity are kept constant, and an affine transformation is performed on the air pressure dimension; all affine transformations are reversible processes. In this way, the interaction relationship between different dimensions is obtained through five stages of cyclical transformation. The interaction relationship is used to promote the collaborative prediction of the entire dimension.

[0049] The embodiments of the present invention are system embodiments corresponding to the above method embodiments. The specific operation of each module can be understood by referring to the description of the method embodiments, and will not be repeated here.

[0050] Device Example 1

[0051] This invention provides a device for temporarily replacing a faulty laser wind-measuring radar, such as... Figure 3 As shown, it includes: a memory 30, a processor 32, and a computer program stored on the memory 30 and executable on the processor 32. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0052] Device Example 2

[0053] This invention provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor 32, implements the steps described in the above method embodiments.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions to the technical solutions of the embodiments of the present invention do not cause the essence of the corresponding technical solutions to deviate from the scope of the present solution.

Claims

1. A system for temporarily replacing a faulty laser wind-measuring radar, characterized in that, include: The detection module is used to analyze the output data of the radar and determine whether the lidar has malfunctioned based on the analysis results. If a malfunction occurs, it outputs logic 0; otherwise, it outputs logic 1. The data prediction module is used to predict five types of meteorological information of wind farms, namely wind speed, wind direction, temperature, air pressure and humidity, based on historical data and meteorological source data. The data switching module is used to switch the data source from the simulation system data source to the laser wind measuring radar device data source when the output logic changes from 0 to 1, and to switch the data source from the laser wind measuring radar device data source to the simulation system data source when the logic value changes from 1 to 0, and to maintain the original state when the logic value does not change. The data prediction module is specifically used to: model the multidimensional data distribution of high-resolution wind farm meteorological information using a conditional autoregressive flow model, and to recursively predict data for future times by sampling and learning the good distribution. The conditional autoregressive flow model comprises a compression layer, a reversible convolutional layer, an affine coupling layer, a separation layer, and a periodic fitting layer connected in sequence. Meteorological information is normalized and then input into the compression layer. The compression layer compresses high-resolution and low-resolution meteorological information and extracts significant features before inputting them into the reversible convolutional layer. The reversible convolutional layer shuffles and reassembles information from different dimensions before inputting it into the affine coupling layer. The affine coupling layer obtains the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions based on affine functions. The affine coupling layer is used for collaborative prediction across all dimensions. The separation layer is used to restore latent noise variables and guide low-resolution meteorological information. The periodic fitting layer ensures that the generated wind farm meteorological data can capture the true periodic variation patterns. The condition of the conditional autoregressive flow model is based on the forward and backward bidirectional propagation of information from the neighborhood content of low-resolution meteorological forecast data, thereby obtaining maximum fluidity and effectively guiding the establishment of multi-dimensional high-resolution wind farm meteorological data. The expression of the conditional autoregressive flow model is: P(y_t|x_t-1,x_t,x_t+1). The overall framework of the model is: the training phase is an encoding process. Under the guidance of historical low-frequency meteorological source data sequences, high-frequency meteorological information is encoded into noise latent variables that follow a certain distribution through a reversible neural network. The reasoning stage transforms into a decoding process, where, under the influence of fuzzy meteorological data from future sources, the original noise latent variables are decoded into detailed meteorological content for future wind farms.

2. The system according to claim 1, characterized in that, The affine coupling layer, based on affine functions, obtains the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions. Specifically, in the first stage, the input and output of the four dimensions of wind direction o, temperature t, humidity h, and air pressure p are fixed, and an affine transformation is performed on the wind speed s dimension. The formula for the affine transformation is as follows: Where the scaling function w s Translation function w e The neural network model to be learned is as follows: In the second stage, the input and output of the four dimensions of wind speed, temperature, humidity, and air pressure are kept constant, and an affine transformation is performed on the wind direction dimension; in the third stage, the input and output of the four dimensions of wind speed, wind direction, humidity, and air pressure are kept constant, and an affine transformation is performed on the temperature dimension; in the fourth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and air pressure are kept constant, and an affine transformation is performed on the humidity dimension; in the fifth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and humidity are kept constant, and an affine transformation is performed on the air pressure dimension; all affine transformations are reversible processes. In this way, the interaction relationship between different dimensions is obtained through five stages of cyclical transformation. The interaction relationship is used to promote the collaborative prediction of the entire dimension.

3. A method for temporarily replacing a faulty laser wind-measuring radar, characterized in that, include: S1. Analyze the output data of the radar through the detection module, and determine whether the lidar has malfunctioned based on the analysis results. If a malfunction occurs, output logic 0; otherwise, output logic 1. S2. The data prediction module predicts five types of meteorological information of the wind farm, namely wind speed, wind direction, temperature, air pressure and humidity, based on historical data and meteorological source data. S3. When the output logic changes from 0 to 1, the data source is switched from the simulation system data source to the laser wind radar device data source. When the logic value changes from 1 to 0, the data source is switched from the laser wind radar device data source to the simulation system data source. When the logic value does not change, the original state is maintained. The method of predicting five types of meteorological information of wind farms—wind speed, wind direction, temperature, air pressure, and humidity—based on historical data and meteorological source data through the data prediction module specifically includes: using a conditional autoregressive flow model to model the multidimensional data distribution of high-resolution wind farm meteorological information, and recursively inferring data for future moments by sampling and learning the good distribution; The conditional autoregressive flow model comprises a compression layer, a reversible convolutional layer, an affine coupling layer, a separation layer, and a periodic fitting layer connected in sequence. Meteorological information is normalized and then input into the compression layer. The compression layer compresses high-resolution and low-resolution meteorological information and extracts significant features before inputting them into the reversible convolutional layer. The reversible convolutional layer shuffles and reassembles information from different dimensions before inputting it into the affine coupling layer. The affine coupling layer obtains the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions based on affine functions. The affine coupling layer is used for collaborative prediction across all dimensions. The separation layer is used to restore latent noise variables and guide low-resolution meteorological information. The periodic fitting layer ensures that the generated wind farm meteorological data can capture the true periodic variation patterns. The condition of the conditional autoregressive flow model is based on the forward and backward bidirectional propagation of information from the neighborhood content of low-resolution meteorological forecast data, thereby obtaining maximum fluidity and effectively guiding the establishment of multi-dimensional high-resolution wind farm meteorological data. The expression for the conditional autoregressive flow model is: P(y_t|x_t-1,x_t,x_t+1). The overall framework of the model is as follows: In the training phase, it is an encoding process. Guided by historical low-frequency meteorological source data sequences, high-frequency meteorological information is encoded into noise latent variables that follow a certain distribution through a reversible neural network. In the inference phase, it transforms into a decoding process. Under the influence of future fuzzy meteorological source data, the original noise latent variables are decoded into future refined wind farm meteorological content.

4. The method according to claim 3, characterized in that, The affine coupling layer, based on affine functions, obtains the relationship between wind speed, wind direction, temperature, air pressure, and humidity in one dimension and other dimensions. Specifically, in the first stage, the input and output of the four dimensions of wind direction o, temperature t, humidity h, and air pressure p are fixed, and an affine transformation is performed on the wind speed s dimension. The formula for the affine transformation is as follows: Where the scaling function w s Translation function w e The neural network model to be learned is as follows: In the second stage, the input and output of the four dimensions of wind speed, temperature, humidity, and air pressure are kept constant, and an affine transformation is performed on the wind direction dimension; in the third stage, the input and output of the four dimensions of wind speed, wind direction, humidity, and air pressure are kept constant, and an affine transformation is performed on the temperature dimension; in the fourth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and air pressure are kept constant, and an affine transformation is performed on the humidity dimension; in the fifth stage, the input and output of the four dimensions of wind speed, wind direction, temperature, and humidity are kept constant, and an affine transformation is performed on the air pressure dimension; all affine transformations are reversible processes. In this way, the interaction relationship between different dimensions is obtained through five stages of cyclical transformation. The interaction relationship is used to promote the collaborative prediction of the entire dimension.

5. A device for temporarily replacing a faulty laser wind-measuring radar, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for temporarily replacing a faulty laser wind-measuring radar as described in any one of claims 3 to 4.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an implementation program for information transmission, which, when executed by a processor, implements the steps of the method for temporarily replacing a faulty laser wind-measuring radar as described in any one of claims 3 to 4.

Images