Radar echo prediction method and device based on potential diffusion model

Abstract

The invention provides a radar echo prediction method and device based on a potential diffusion model, and relates to the technical field of radar echo prediction, and the method comprises the steps: inputting the environment grid data of a target region and a live echo sequence of a pixel space into the potential diffusion model; through a multi-layer convolutional neural network in the condition encoder, feature extraction is carried out on the environment grid data in sequence according to the size from large to small, and environment condition features of the target area are generated; the real-time echo sequence of the pixel space is converted into a real-time echo low-dimensional dense vector of a hidden space through an echo encoder, and noise adding processing is carried out; under the action of the real-time echo low-dimensional dense vectors of the hidden space and environmental condition characteristics, the real-time echo low-dimensional dense vectors after noise adding are sequentially processed through a plurality of noise reduction structures connected in series in a noise reduction model, and then are converted into a prediction echo sequence of a pixel space through an echo decoder; the technical problem that in the prior art, a radar echo prediction result is not high in accuracy is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of radar echo prediction, and in particular to a radar echo prediction method and device based on a potential diffusion model. Background Art

[0002] Currently, radar echo prediction results are commonly used to estimate precipitation, and radar echo extrapolation methods include traditional advection methods, deterministic deep learning methods, and image generation model methods.

[0003] The inventors have found that in practical applications, the traditional advection method predicts the future echo field by translating the current radar echo field in a certain direction. Since it does not take into account the evolution of the precipitation system, the intensity of the precipitation results estimated by this method is too high. Deterministic deep learning methods are applied to radar echo prediction, which can better capture the precipitation area, but due to the lack of sensitivity to changes in precipitation intensity, the prediction results obtained by this method are relatively vague. The image generation model method does not take into account the actual factors that have an important impact on the evolution of radar echoes, resulting in limited accuracy of the prediction results of image extrapolation. Summary of the invention

[0004] The purpose of the present invention is to provide a radar echo prediction method and device based on a potential diffusion model to alleviate the technical problem of low accuracy of radar echo prediction results in the prior art.

[0005] In a first aspect, the present invention provides a radar echo prediction method based on a potential diffusion model, comprising: Preprocessing radar data and environmental data of the target area to obtain a live echo sequence in pixel space and a plurality of environmental grid data, and inputting the environmental grid data and the live echo sequence in pixel space into a potential diffusion model; wherein the potential diffusion model is composed of an echo encoder, a noise addition model, a noise reduction model, a conditional encoder and an echo decoder; By using a multi-layer convolutional neural network in the conditional encoder, feature extraction is performed on each of the environmental grid data in order from large to small scales to generate environmental condition features of the target area; The echo encoder converts the live echo sequence in the pixel space into a low-dimensional dense vector of the live echo in the latent space; Inputting the live echo low-dimensional dense vector into the noise adding model, and outputting the noisy live echo low-dimensional dense vector through sequential processing of a plurality of serially connected noise adding structures in the noise adding model; Under the action of the low-dimensional dense vector of the live echo in the latent space and the characteristics of the environmental conditions, the low-dimensional dense vector of the live echo after adding noise is processed in sequence by a plurality of serially connected noise reduction structures in the noise reduction model to output a predicted low-dimensional dense vector of the echo; The predicted echo low-dimensional dense vector is converted into a predicted echo sequence in pixel space through the echo decoder.

[0006] In an optional embodiment, the environmental grid data includes terrain information, satellite observation data, model forecast data, spatial information and time information; the multi-layer convolutional neural network includes multiple feature extraction branches; each feature extraction branch is composed of a convolution layer, an activation function and a maximum pooling layer; The step of extracting features from each of the environmental grid data in order from large to small scales through a multi-layer convolutional neural network in the conditional encoder to generate environmental condition features of the target area includes: Inputting terrain information of a first preset scale into a first feature extraction branch for processing, and outputting a first feature extraction result; The first feature extraction result and the satellite observation data of the second preset scale are superimposed and input into the second feature extraction branch for processing, and the second feature extraction result is output; The second feature extraction result and the model forecast data of the third preset scale are superimposed and input into the third feature extraction branch for processing, and the third feature extraction result is output; The third feature extraction result and the time information and space information of the fourth preset scale are superimposed to obtain the environmental condition characteristics of the target area.

[0007] In an optional embodiment, under the action of the low-dimensional dense vector of the live echo in the latent space and the characteristics of the environmental conditions, the step of sequentially processing the low-dimensional dense vector of the live echo after adding noise through a plurality of serially connected noise reduction structures in the noise reduction model to output a predicted echo low-dimensional dense vector comprises: Inputting the low-dimensional dense vector of the actual echo in the latent space, the environmental condition feature and the low-dimensional dense vector of the actual echo after adding noise into the first denoising structure of the denoising model; The first denoising structure performs denoising on the noisy live echo low-dimensional dense vector under the influence of the live echo low-dimensional dense vector in the latent space and the environmental condition characteristics, and outputs a first denoised live echo low-dimensional dense vector; Repeat the following process until each noise reduction structure in the noise reduction model is traversed and the first noise reduction actual echo low-dimensional dense vector outputted is used as the predicted echo low-dimensional dense vector, including: Inputting the low-dimensional dense vector of the actual echo in the latent space, the environmental condition feature and the first denoised actual echo low-dimensional dense vector into the second denoising structure; The second denoising structure performs denoising processing on the first denoised live echo low-dimensional dense vector under the influence of the live echo low-dimensional dense vector in the latent space and the environmental condition characteristics, and outputs a second denoised live echo low-dimensional dense vector; The second denoised live echo low-dimensional dense vector is used as a new first denoised live echo low-dimensional dense vector.

[0008] In an optional embodiment, the method further comprises: The difference between each current frame and the previous frame of the predicted echo low-dimensional dense vector within the echo range of the target area is weighted summed to determine the loss function of the potential diffusion model.

[0009] In an optional embodiment, the loss function can be calculated by the following formula:

[0010] in, is the tensor of the previous frame in the predicted echo low-dimensional dense vector, is the tensor of the current frame in the predicted echo low-dimensional dense vector, is the number of image frames in the predicted echo low-dimensional dense vector, is the weight of the i-th interval of the current frame within the echo range of the target area, is the weight of the current frame when the j-th image frame is the current frame.

[0011] In an optional embodiment, the method further comprises: The precipitation in the target area is estimated using the predicted echo sequence in the pixel space.

[0012] In an optional embodiment, the step of preprocessing the radar data and the environmental data of the target area to obtain a live echo sequence in a pixel space and a plurality of environmental grid data includes: Filtering outliers and interference noise in radar data of the target area, and then slicing according to a preset scale to generate a live echo sequence in a pixel space of a first preset scale; The environmental data of the target area is sliced ​​according to the corresponding preset resolution to generate environmental grid data of multiple preset scales.

[0013] In a second aspect, the present invention provides a radar echo prediction device based on a potential diffusion model, comprising: A preprocessing module preprocesses the radar data and environmental data of the target area to obtain a live echo sequence in pixel space and a plurality of environmental grid data, and inputs the environmental grid data and the live echo sequence in pixel space into a potential diffusion model; wherein the potential diffusion model is composed of an echo encoder, a noise addition model, a noise reduction model, a conditional encoder and an echo decoder; A first feature extraction module, which extracts features from each of the environmental grid data in descending order of scale through a multi-layer convolutional neural network in the conditional encoder to generate environmental condition features of the target area; A second feature extraction module converts the live echo sequence in the pixel space into a live echo low-dimensional dense vector in the latent space via the echo encoder; A noise adding module, inputting the low-dimensional dense vector of the live echo into the noise adding model, and outputting the low-dimensional dense vector of the live echo after the noise is added through sequential processing of a plurality of serially connected noise adding structures in the noise adding model; A denoising module, under the action of the low-dimensional dense vector of the live echo in the latent space and the characteristics of the environmental conditions, sequentially processes the low-dimensional dense vector of the live echo after adding noise through a plurality of serially connected denoising structures in the denoising model, and outputs a predicted low-dimensional dense vector of the echo; A decoding module converts the predicted echo low-dimensional dense vector into a predicted echo sequence in pixel space through the echo decoder.

[0014] In a third aspect, the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and capable of running on the processor, wherein the processor implements a method as described in any one of the aforementioned embodiments when executing the program.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed, the method described in any one of the aforementioned embodiments is implemented.

[0016] The embodiment of the present invention provides a radar echo prediction method and device based on a potential diffusion model. First, each environmental factor that affects the radar echo is encoded to generate conditional features, and then the pixel space is transferred to a low-dimensional dense latent space through the echo encoder and echo decoder in the potential diffusion model for diffusion to achieve richer feature learning; based on the environmental features and low-dimensional dense vector features, the low-dimensional dense vector of the noisy actual echo after the noise processing using the potential diffusion model is denoised to obtain a more accurate predicted echo low-dimensional dense vector, which is then converted into a predicted echo sequence in the pixel space.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A flow chart of a radar echo prediction method based on a potential diffusion model provided by an embodiment of the present invention; Figure 2 A schematic diagram of a potential diffusion model architecture provided by an embodiment of the present invention; Figure 3 A schematic diagram of a conditional encoder architecture provided by an embodiment of the present invention; Figure 4 A schematic diagram of comparing the true value and predicted value of a deterministic model provided in an embodiment of the present invention; Figure 5 A schematic diagram of comparing the true value and predicted value of an image extrapolation model provided by an embodiment of the present invention; Figure 6 A schematic diagram of comparing the true value and predicted value of a potential diffusion model provided by an embodiment of the present invention; Figure 7 A schematic diagram of functional modules of a radar echo prediction device based on a potential diffusion model provided by an embodiment of the present invention; Figure 8 A schematic diagram of the hardware architecture of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] At present, the inventor has found that the radar echo prediction method based on image extrapolation lacks an accurate description of the movement and generation and disappearance characteristics of precipitation clouds. The movement of precipitation clouds is not only strongly correlated with conditions such as wind speed, wind direction, and terrain, but its generation and disappearance process is also closely related to humidity, temperature, and whether strong convection occurs. These environmental factors have an important impact on the evolution of radar echoes, and traditional image extrapolation methods often ignore these factors, resulting in limited accuracy and reliability of prediction results. In the prediction of long-term radar echoes, due to the large time span, external environmental factors may change significantly. For example, during the echo movement, the wind force may change, resulting in changes in the echo movement speed. These changes have an important impact on the movement and generation and disappearance characteristics of precipitation clouds. Methods based on image extrapolation often cannot effectively integrate these external environmental characteristics, resulting in reduced prediction accuracy.

[0023] Based on this, an embodiment of the present invention provides a radar echo prediction method and device based on a potential diffusion model, which achieves more accurate radar echo prediction by considering environmental factors that affect radar echoes and a richer learning expression of the potential diffusion model in latent space diffusion.

[0024] To facilitate understanding of this embodiment, a radar echo prediction method based on a potential diffusion model disclosed in an embodiment of the present invention is first introduced in detail. This method can be applied to intelligent control devices such as host computers, servers, and controllers.

[0025] Figure 1 A flow chart of a radar echo prediction method based on a potential diffusion model provided by an embodiment of the present invention.

[0026] Reference Figure 1 As shown, the method may include the following steps: Step S102, preprocessing the radar data and environmental data of the target area to obtain a live echo sequence in pixel space and a plurality of environmental grid data, and inputting the environmental grid data and the live echo sequence in pixel space into a potential diffusion model; The target area can be understood as the actual geographical area range of the radar echo result to be predicted and estimated; the potential diffusion model provided by the embodiment of the present invention is a trained model; the potential diffusion model is composed of an echo encoder, a noise addition model, a noise reduction model, a conditional encoder and an echo decoder, such as Figure 2 As shown; each component in the potential diffusion model performs steps S104-step S112 according to Figure 2 as shown in .

[0027] Step S104, through the multi-layer convolutional neural network in the conditional encoder, feature extraction is performed on each environmental grid data in order from large to small scale to generate environmental condition features of the target area.

[0028] Here, the conditional encoder uses a multi-layer convolutional neural network to encode environmental mesh data of different scales, thereby achieving feature extraction from multi-source environmental mesh data.

[0029] Step S106: converting the live echo sequence in the pixel space into a low-dimensional dense vector of the live echo in the latent space through an echo encoder.

[0030] The structure of the echo encoder is not limited, and models such as VAE or VQGAN can be used, such as Figure 2 As shown, the embodiment of the present invention is explained by taking VAE Encoder as an example; the echo encoder is a single-channel single-frame radar data encoder, and the encoder extracts features from the input high-dimensional vector and converts it into a feature map sequence of a low-dimensional dense vector, namely, a live echo low-dimensional dense vector; for example, the size of each image frame in the live echo sequence that limits the input pixel space is 512x512, and the number of channels is 1; the image frame size in the output live echo low-dimensional dense vector is 64x64, and the number of channels is 4.

[0031] Step S108, inputting the low-dimensional dense vector of the live echo into the noisy model, and outputting the noisy low-dimensional dense vector of the live echo through sequential processing of multiple serially connected noisy structures in the noisy model.

[0032] For example, after the real-time echo low-dimensional dense vector is processed by the first noisy structure for noise addition, it is input into the second noisy structure in a serial order to continue the noise addition process. The above process is repeated until each noisy structure in the noisy model performs the noise addition operation and terminates, and outputs the noisy real-time echo low-dimensional dense vector.

[0033] Step S110, under the influence of the low-dimensional dense vector of the actual echo in the latent space and the characteristics of the environmental conditions, the low-dimensional dense vector of the actual echo after adding noise is processed in sequence through multiple serially connected noise reduction structures in the noise reduction model, and a predicted echo low-dimensional dense vector is output.

[0034] The denoising model is used to realize the extrapolation prediction of radar echoes that integrate external environmental characteristics. Specifically, based on the powerful generation capability of the latent diffusion model and the conditional encoder, the correct encoding of conditional information is realized, and the conditional injection of the DiT Denoise denoising structure in the diffusion model is realized. In the denoising stage of the latent diffusion model, environmental factors such as time, space, terrain, satellite observations, and model forecasts are fully integrated as conditions to achieve accurate and efficient radar echo prediction.

[0035] Step S112: The predicted echo low-dimensional dense vector is converted into a predicted echo sequence in pixel space through an echo decoder.

[0036] The echo decoder VAE Decoder is a single-channel single-frame radar data decoder, which is used to restore the low-dimensional latent space vector to continuous frames of radar echo in the high-dimensional pixel space.

[0037] In a preferred embodiment of practical application, each environmental factor that affects the radar echo is first encoded to generate conditional features, and then the pixel space is transferred to a low-dimensional dense latent space for diffusion through the echo encoder and echo decoder in the latent diffusion model to achieve richer feature learning; based on the environmental features and low-dimensional dense vector features, the low-dimensional dense vector of the actual echo after the noise processing using the latent diffusion model is denoised to obtain a more accurate predicted echo low-dimensional dense vector, which is then converted into a predicted echo sequence in the pixel space; the radar echo prediction result obtained in this way has a higher accuracy.

[0038] In some embodiments, the radar data is quality controlled, screened and normalized by data preprocessing, and the environmental condition data is standardized; illustratively, step S102 includes: In step 1.1, outliers and interference noises in the radar data of the target area are filtered out, and then sliced ​​according to a preset scale to generate a live echo sequence in a pixel space of a first preset scale.

[0039] The radar data is sliced ​​into 512x512 scales to meet the input requirements of the potential diffusion model to ensure the consistency and processability of the radar data. The clutter in the radar echo is then quality controlled to reduce the impact of noise on subsequent analysis. In practical applications, filters or other signal processing components can be used to identify and remove outliers. At the same time, small echoes in the radar data are also filtered out to reduce interference and noise in the data and improve the quality of the data. Finally, the data is normalized so that the data has a uniform scale, which helps the model learn and generalize better.

[0040] Specifically, for the slice data connected to the radar echo combined reflectivity puzzle, the radar data is a continuous radar observation sequence with an image frame scale of 512*512, and the small echoes below 5dBz are filtered out and normalized.

[0041] In step 1.2, the environmental data of the target area is sliced ​​according to the corresponding preset resolution to generate environmental grid data of multiple preset scales.

[0042] Here, the terrain information, satellite observations, and model forecast data in the environmental data are interpolated to form standardized grid data with a resolution of 0.01 degrees, 0.02 degrees, and 0.4 degrees, respectively. The data is sliced ​​according to the same area as the radar, and the time and space location information is expanded and mapped onto the grid.

[0043] Specifically, the terrain information is interpolated onto a 4 km grid and converted into slices of 512*512 scale, the satellite observation data is interpolated onto an 8 km grid and converted into slices of 256*256 scale, the model forecast data is interpolated onto a 16 km grid and converted into slices of 128*128 scale, and the time and space are mapped into 64*64 slices to ensure that the slices of radar data and external environment data are aligned.

[0044] In practical applications, environmental grid data includes terrain information, satellite observation data, model forecast data, spatial information and temporal information; Figure 3 As shown in the figure, the multi-layer convolutional neural network includes multiple feature extraction branches; each feature extraction branch is composed of a convolution layer, an activation function and a maximum pooling layer; the conditional encoder can realize feature extraction and learning of environmental factors such as time, space, terrain, satellite observation, model forecast, etc. at different scales; Exemplarily, step S104 may be implemented by the following steps, specifically including: In step 2.1), the terrain information of the first preset scale 512x512 is input into the first feature extraction branch for processing, and the first feature extraction result is output.

[0045] Among them, the convolution layer Conv, activation function Relu and maximum pooling layer MaxPooling of the first feature extraction branch are processed to output the first feature extraction result.

[0046] In step 2.2, the first feature extraction result is superimposed with the satellite observation data of the second preset scale 256x256 and then input into the second feature extraction branch for processing, and the second feature extraction result is output.

[0047] Here, the processing process of the second feature extraction branch is similar to that of the first feature extraction branch, and will not be repeated here.

[0048] In step 2.3), the second feature extraction result and the pattern forecast data of the third preset scale 128x128 are superimposed and input into the third feature extraction branch for processing, and the third feature extraction result is output.

[0049] Similarly, the processing process of the third feature extraction branch is similar to that of the first feature extraction branch, and will not be repeated here.

[0050] In step 2.4), the third feature extraction result is superimposed with the time information and spatial information of the fourth preset scale 64×64 to obtain the environmental condition characteristics of the target area.

[0051] It can be understood that the terrain information is first extracted through the convolution layer (Conv), and then the ReLU activation function is applied to increase the nonlinear expression ability of the network, and then the feature dimension is reduced and important features are extracted through the maximum pooling layer (MaxPooling). Then, the satellite observation data, model forecast data and spatiotemporal feature information are fused in turn to form the final conditional feature (Condition Feature), which provides rich information reference for subsequent prediction tasks. The multi-convolution conditional encoder fuses multiple environmental condition data at different scales to form an environmental feature extraction network, and the output feature size of the network is 64×64.

[0052] In some embodiments, the noise reduction step S110 of the embodiment of the present invention may specifically include the following steps: In step 3.1), the low-dimensional dense vector of the actual echo in the latent space, the environmental condition characteristics and the low-dimensional dense vector of the actual echo after noise addition are input into the first denoising structure of the denoising model.

[0053] In step 3.2, the first denoising structure performs denoising on the low-dimensional dense vector of the actual echo after adding noise under the influence of the low-dimensional dense vector of the actual echo in the latent space and the characteristics of the environmental conditions, and outputs a first denoised low-dimensional dense vector of the actual echo.

[0054] Step 3.3), repeat the following steps 3.3.1) to 3.3.3) until each denoising structure in the denoising model is traversed and the first denoised actual echo low-dimensional dense vector outputted is used as the predicted echo low-dimensional dense vector, including: Step 3.3.1), input the low-dimensional dense vector of the actual echo in the latent space, the environmental condition characteristics and the first denoised actual echo low-dimensional dense vector into the second denoising structure; Step 3.3.2), the second denoising structure performs denoising on the first denoised actual echo low-dimensional dense vector under the influence of the actual echo low-dimensional dense vector in the latent space and the environmental condition characteristics, and outputs a second denoised actual echo low-dimensional dense vector; In step 3.3.3), the second denoised live echo low-dimensional dense vector is used as the new first denoised live echo low-dimensional dense vector.

[0055] It should be noted that the construction of the denoising model in the embodiment of the present invention is to accurately predict and denoise radar echo data. By using the Transformer structure inherent in the multi-layer DIT (Diffusion Transformer) module, the radar echo data can be gradually denoised. In each reasoning process, the latent space vector representation of the actual radar echo sequence is used as the conditional input, and the conditional features generated by the conditional encoder Condition Encoder are fused. These features are derived from terrain information, satellite observation data, model forecast data, and time and space information, providing the model with environmental and background information as the basis for radar echo extrapolation, so that the model can accurately predict the generation and movement of radar echoes. The output result of the model is the vector representation of the predicted radar echo sequence in the latent space, that is, the predicted echo low-dimensional dense vector.

[0056] Before the patch embedding of the DIT model, the latent space vector and environmental feature vector of the live radar echo sequence are fused. The number of layers of the DIT module is 6 to 12, which can take into account both the generation effect and the inference speed. The input is an echo sequence of a period of time (such as 1 hour), and the output is a subsequent echo sequence. Considering the timeliness and generation effect of short-term forecasts, the number of autoregressive iterations of the model is set between 50 and 100 for the best. MSELoss is used for training, and the echo encoder and echo decoder are frozen during training.

[0057] On the basis of the aforementioned embodiments, the ability to predict large and extreme echo values ​​can be achieved based on the loss-weighted tuning training step. In the deep learning model training for radar echo prediction, due to the uneven distribution of radar echo strengths, common problems include a large proportion of echoes of 10-20dbz, while echoes of 60-70dbz rarely appear. This uneven distribution may cause the prediction effect of the model to be biased towards the average. As the forecast time increases, the size of the echo will become average, and the echo area will spread out, resulting in the lack of forecasts for extreme values. The embodiment of the present invention uses a new loss function setting to ensure that the prediction results of the potential diffusion model can be more accurate and will not miss extreme data; the method also includes: The difference between each current frame and the previous frame of the predicted echo low-dimensional dense vector within the echo range of the target area is weighted summed to determine the loss function of the potential diffusion model.

[0058] The loss function can be calculated by the following formula:

[0059] in, To predict the echo tensor of the previous frame in the low-dimensional dense vector, To predict the tensor of the current frame in the low-dimensional dense vector of the echo, To predict the number of image frames in the low-dimensional dense vector of echoes, is the weight of the i-th interval of the current frame within the echo range of the target area, is the weight of the current frame when the j-th image frame is the current frame.

[0060] In the application process, the weight of the interval is defined inversely according to the probability of the echo appearing in different ranges, and this weighted relationship is "amplified" after the time series becomes longer, so that the model will pay more attention to those extreme echo values ​​that are less common but critical to forecasting skills during the application process, thereby improving the overall forecasting performance. This method helps to improve the model's forecasting ability for extreme weather events, especially in the radar echo prediction of severe convective weather.

[0061] For the latent diffusion model involved in the above application process, it is necessary to freeze the echo encoder, echo decoder and conditional feature encoder during training, and focus on the rest of the training network. The weighted weight values ​​in the above loss function are not static, but are dynamically adjusted according to the shape and effect of the predicted echo during training. This dynamic adjustment strategy can more finely control the training process to optimize model performance. The setting of the initial weight value should take into account the possibility of echoes of different sizes, and perform reverse weighting to ensure that the model can pay enough attention to those echo types that appear less frequently but are equally important in the early stages of training. In this way, the latent diffusion model can learn various echo features more evenly, thereby providing more accurate and reliable prediction results in practical applications.

[0062] In practical applications, the potential diffusion model of radar echo prediction that integrates environmental conditions has achieved the integration of environmental factors in echo prediction and made more accurate predictions. The prediction result, that is, the predicted echo sequence in pixel space, can be used to more accurately estimate the precipitation in the target area.

[0063] Figure 4 The figure is a schematic diagram of the comparison between the echo prediction and the true echo value of a deterministic model; the top sequence is the true echo value of the next hour, and the bottom sequence is the echo prediction value of the next hour; Figure 5 The figure is a schematic diagram of the comparison between the echo prediction and the echo true value output by an image extrapolation model that does not integrate external conditions. The top sequence is the echo true value in the next hour, and the bottom sequence is the echo prediction value in the next hour. It can be seen that the predicted echoes output by the above two models are not consistent with the actual echo generation and disappearance compared with the echo true value.

[0064] Compared with the traditional deterministic model, the embodiment of the present invention has the ability to predict extreme values ​​and echo details. Based on the above two advantages, the embodiment of the present invention can output the prediction results of both extreme values ​​and details, and the radar echo sequence with small deviation of the large value position, such as Figure 6 The figure shows the radar echo prediction sequence at 08:00 on May 1, 2024 and 19:00 on August 20, 2024 Beijing time; the actual echo in the previous hour is the input of the potential diffusion model, and the echo prediction in the next hour is output; the top sequence is the true value of the echo in the next hour, and the bottom sequence is the predicted value of the echo in the next hour; it can be seen that the potential diffusion model provided in the embodiment of the present invention can accurately predict extreme values ​​and echo details, and has good forecasting skills.

[0065] In some embodiments, Figure 7 As shown, an embodiment of the present invention provides a radar echo prediction device based on a potential diffusion model, comprising: A preprocessing module preprocesses the radar data and environmental data of the target area to obtain a live echo sequence in pixel space and a plurality of environmental grid data, and inputs the environmental grid data and the live echo sequence in pixel space into a potential diffusion model; wherein the potential diffusion model is composed of an echo encoder, a noise addition model, a noise reduction model, a conditional encoder and an echo decoder; A first feature extraction module, which extracts features from each of the environmental grid data in descending order of scale through a multi-layer convolutional neural network in the conditional encoder to generate environmental condition features of the target area; A second feature extraction module converts the live echo sequence in the pixel space into a live echo low-dimensional dense vector in the latent space via the echo encoder; A noise adding module, inputting the low-dimensional dense vector of the live echo into the noise adding model, and outputting the low-dimensional dense vector of the live echo after the noise is added through sequential processing of a plurality of serially connected noise adding structures in the noise adding model; A denoising module, under the action of the low-dimensional dense vector of the live echo in the latent space and the characteristics of the environmental conditions, sequentially processes the low-dimensional dense vector of the live echo after adding noise through a plurality of serially connected denoising structures in the denoising model, and outputs a predicted low-dimensional dense vector of the echo; A decoding module converts the predicted echo low-dimensional dense vector into a predicted echo sequence in pixel space through the echo decoder.

[0066] Furthermore, the environmental grid data includes terrain information, satellite observation data, model forecast data, spatial information and time information; the multi-layer convolutional neural network includes multiple feature extraction branches; each feature extraction branch is composed of a convolution layer, an activation function and a maximum pooling layer; the first feature extraction module is specifically used to input the terrain information of a first preset scale into the first feature extraction branch for processing, and output a first feature extraction result; superimpose the first feature extraction result and the satellite observation data of a second preset scale and input them into the second feature extraction branch for processing, and output the second feature extraction result; superimpose the second feature extraction result and the model forecast data of a third preset scale and input them into the third feature extraction branch for processing, and output the third feature extraction result; superimpose the third feature extraction result and the time information and spatial information of a fourth preset scale to obtain the environmental condition characteristics of the target area.

[0067] Furthermore, the denoising module is specifically used to input the low-dimensional dense vector of the actual echo in the latent space, the environmental condition characteristics and the low-dimensional dense vector of the actual echo after noise addition into the first denoising structure of the denoising model; the first denoising structure performs denoising on the low-dimensional dense vector of the actual echo after noise addition under the influence of the low-dimensional dense vector of the actual echo in the latent space and the environmental condition characteristics, and outputs a first denoised low-dimensional dense vector of the actual echo; and the following processing is repeatedly performed until each denoising structure in the denoising model is traversed and the first denoising real echo low-dimensional dense vector output is obtained. The method comprises: inputting the actual echo low-dimensional dense vector in the latent space, the environmental condition characteristics and the first denoised actual echo low-dimensional dense vector into a second denoising structure; performing denoising on the first denoised actual echo low-dimensional dense vector under the influence of the actual echo low-dimensional dense vector in the latent space and the environmental condition characteristics through the second denoising structure, and outputting a second denoised actual echo low-dimensional dense vector; and using the second denoised actual echo low-dimensional dense vector as a new first denoised actual echo low-dimensional dense vector.

[0068] Furthermore, the device is also used to perform weighted summation of the difference between each current frame and the previous frame of the predicted echo low-dimensional dense vector within the echo range of the target area to determine the loss function of the potential diffusion model.

[0069] Furthermore, the loss function can be calculated by the following formula:

[0070] in, is the tensor of the previous frame in the predicted echo low-dimensional dense vector, is the tensor of the current frame in the predicted echo low-dimensional dense vector, is the number of image frames in the predicted echo low-dimensional dense vector, is the weight of the i-th interval of the current frame within the echo range of the target area, is the weight of the current frame when the j-th image frame is the current frame.

[0071] Furthermore, the device is also used to estimate the precipitation in the target area using the predicted echo sequence in the pixel space.

[0072] Furthermore, the preprocessing module is specifically used to filter out outliers and interference noise in the radar data of the target area, and then slice it according to a preset scale to generate a live echo sequence in the pixel space of a first preset scale; slice the environmental data of the target area according to a corresponding preset resolution to generate environmental grid data of multiple preset scales.

[0073] An embodiment of the present invention provides an electronic device for implementing an electronic device. In this embodiment, the electronic device may be, but is not limited to, a personal computer (PC), a notebook computer, a monitoring device, a server, or other computer device with analysis and processing capabilities.

[0074] As an exemplary embodiment, see Figure 8 The electronic device 110 includes a communication interface 111, a processor 112, a memory 113 and a bus 114. The processor 112, the communication interface 111 and the memory 113 are connected via the bus 114. The memory 113 is used to store a computer program that supports the processor 112 to execute the method. The processor 112 is configured to execute the program stored in the memory 113.

[0075] The machine-readable storage medium mentioned in this article can be any electronic, magnetic, optical or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, the machine-readable storage medium can be: RAM (Radom Access Memory), volatile memory, non-volatile memory, flash memory, storage drive (such as hard disk drive), any type of storage disk (such as CD, DVD, etc.), or similar storage medium, or a combination thereof.

[0076] The non-volatile medium may be a non-volatile memory, a flash memory, a storage drive (such as a hard drive), any type of storage disk (such as a CD, DVD, etc.), or a similar non-volatile storage medium, or a combination thereof.

[0077] It can be understood that the specific operation methods of each functional module in this embodiment can refer to the detailed description of the corresponding steps in the above method embodiment, and will not be repeated here.

[0078] The computer-readable storage medium provided in the embodiment of the present invention stores a computer program. When the computer program code is executed, the method described in any of the above embodiments can be implemented. For specific implementation, please refer to the method embodiment, which will not be described in detail here.

[0079] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0080] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0081] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0082] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the aforementioned embodiments, those of ordinary skill in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed by the present invention, or can easily conceive of changes, or make equivalent replacements for some of the technical features therein. Such modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention.

Claims

1. A radar echo prediction method based on a potential diffusion model, characterized in that: include: Preprocessing radar data and environmental data of the target area to obtain a live echo sequence in pixel space and a plurality of environmental grid data, and inputting the environmental grid data and the live echo sequence in pixel space into a potential diffusion model; wherein the potential diffusion model is composed of an echo encoder, a noise addition model, a noise reduction model, a conditional encoder and an echo decoder; By using a multi-layer convolutional neural network in the conditional encoder, feature extraction is performed on each of the environmental grid data in order from large to small scales to generate environmental condition features of the target area; The echo encoder converts the live echo sequence in the pixel space into a low-dimensional dense vector of the live echo in the latent space; Inputting the live echo low-dimensional dense vector into the noise adding model, and outputting the noisy live echo low-dimensional dense vector through sequential processing of a plurality of serially connected noise adding structures in the noise adding model; Under the action of the low-dimensional dense vector of the live echo in the latent space and the characteristics of the environmental conditions, the low-dimensional dense vector of the live echo after adding noise is processed in sequence by a plurality of serially connected noise reduction structures in the noise reduction model to output a predicted low-dimensional dense vector of the echo; The predicted echo low-dimensional dense vector is converted into a predicted echo sequence in pixel space through the echo decoder.

2. The method according to claim 1, characterized in that The environmental grid data includes terrain information, satellite observation data, model forecast data, spatial information and time information; the multi-layer convolutional neural network includes multiple feature extraction branches; Each feature extraction branch consists of a convolutional layer, an activation function, and a maximum pooling layer; The step of extracting features from each of the environmental grid data in order from large to small scales through a multi-layer convolutional neural network in the conditional encoder to generate environmental condition features of the target area includes: Inputting terrain information of a first preset scale into a first feature extraction branch for processing, and outputting a first feature extraction result; The first feature extraction result and the satellite observation data of the second preset scale are superimposed and input into the second feature extraction branch for processing, and the second feature extraction result is output; The second feature extraction result and the model forecast data of the third preset scale are superimposed and input into the third feature extraction branch for processing, and the third feature extraction result is output; The third feature extraction result and the time information and space information of the fourth preset scale are superimposed to obtain the environmental condition characteristics of the target area.

3. The method according to claim 1, characterized in that Under the action of the low-dimensional dense vector of the live echo in the latent space and the characteristics of the environmental conditions, the step of sequentially processing the low-dimensional dense vector of the live echo after adding noise through a plurality of serially connected noise reduction structures in the noise reduction model to output a predicted low-dimensional dense vector of the echo comprises: Inputting the low-dimensional dense vector of the actual echo in the latent space, the environmental condition feature and the low-dimensional dense vector of the actual echo after adding noise into the first denoising structure of the denoising model; The first denoising structure performs denoising on the noisy live echo low-dimensional dense vector under the influence of the live echo low-dimensional dense vector in the latent space and the environmental condition characteristics, and outputs a first denoised live echo low-dimensional dense vector; Repeat the following process until each noise reduction structure in the noise reduction model is traversed and the first noise reduction actual echo low-dimensional dense vector outputted is used as the predicted echo low-dimensional dense vector, including: Inputting the low-dimensional dense vector of the actual echo in the latent space, the environmental condition feature and the first denoised actual echo low-dimensional dense vector into the second denoising structure; The second denoising structure performs denoising processing on the first denoised live echo low-dimensional dense vector under the influence of the live echo low-dimensional dense vector in the latent space and the environmental condition characteristics, and outputs a second denoised live echo low-dimensional dense vector; The second denoised live echo low-dimensional dense vector is used as a new first denoised live echo low-dimensional dense vector.

4. The method according to claim 3, characterized in that The method further comprises: The difference between each current frame and the previous frame of the predicted echo low-dimensional dense vector within the echo range of the target area is weighted summed to determine the loss function of the potential diffusion model.

5. The method according to claim 4, characterized in that The loss function can be calculated by the following formula: in, is the tensor of the previous frame in the predicted echo low-dimensional dense vector, is the tensor of the current frame in the predicted echo low-dimensional dense vector, is the number of image frames in the predicted echo low-dimensional dense vector, is the weight of the i-th interval of the current frame within the echo range of the target area, is the weight of the current frame when the j-th image frame is the current frame.

6. The method according to claim 1, characterized in that The method further comprises: The precipitation in the target area is estimated using the predicted echo sequence in the pixel space.

7. The method according to claim 1, characterized in that The step of preprocessing radar data and environmental data of the target area to obtain a live echo sequence in pixel space and multiple environmental grid data includes: Filtering outliers and interference noise in radar data of the target area, and then slicing according to a preset scale to generate a live echo sequence in a pixel space of a first preset scale; The environmental data of the target area is sliced ​​according to the corresponding preset resolution to generate environmental grid data of multiple preset scales.

8. A radar echo prediction device based on a potential diffusion model, characterized in that: include: A preprocessing module preprocesses the radar data and environmental data of the target area to obtain a live echo sequence in pixel space and a plurality of environmental grid data, and inputs the environmental grid data and the live echo sequence in pixel space into a potential diffusion model; wherein the potential diffusion model is composed of an echo encoder, a noise addition model, a noise reduction model, a conditional encoder and an echo decoder; A first feature extraction module, which extracts features from each of the environmental grid data in descending order of scale through a multi-layer convolutional neural network in the conditional encoder to generate environmental condition features of the target area; A second feature extraction module converts the live echo sequence in the pixel space into a live echo low-dimensional dense vector in the latent space via the echo encoder; A noise adding module, inputting the low-dimensional dense vector of the live echo into the noise adding model, and outputting the low-dimensional dense vector of the live echo after the noise is added through sequential processing of a plurality of serially connected noise adding structures in the noise adding model; A denoising module, under the action of the low-dimensional dense vector of the live echo in the latent space and the characteristics of the environmental conditions, sequentially processes the low-dimensional dense vector of the live echo after adding noise through a plurality of serially connected denoising structures in the denoising model, and outputs a predicted low-dimensional dense vector of the echo; A decoding module converts the predicted echo low-dimensional dense vector into a predicted echo sequence in pixel space through the echo decoder.

9. An electronic device, characterized in that: The method comprises a memory, a processor, and a program stored in the memory and capable of being run on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 7 is implemented.

Images