A precipitation prediction method and system based on artificial intelligence algorithm and knowledge graph

Abstract

The present invention discloses a precipitation prediction method and system based on artificial intelligence algorithm and knowledge graph, which belongs to the field of weather forecast technology, and includes the following steps: constructing a multimodal data container, inputting different types of meteorological data into the multimodal data container according to their own structural characteristics to obtain multimodal data, performing spatiotemporal alignment, data cleaning and preprocessing, and data sequence segmentation on the multimodal data; constructing a knowledge graph about precipitation prediction; creating a multimodal precipitation prediction model based on artificial intelligence algorithm, and intelligently correcting the predicted precipitation data. The present invention not only makes full use of different types of multimodal meteorological data, improves the accuracy of precipitation prediction through a multimodal precipitation prediction method, but also constructs a precipitation prediction knowledge graph, which can intelligently correct the results of model prediction to reduce the uncertainty of the artificial intelligence algorithm and improve the accuracy and reliability of model prediction.

Description

Technical Field

[0001] The present invention belongs to the technical field of weather forecasting, and specifically relates to a precipitation prediction method and system based on artificial intelligence algorithm and knowledge graph. Background Art

[0002] In recent years, precipitation has increased significantly around the world. Whether in urban transportation or in agricultural systems, or even in water resources planning and flood warning, precipitation forecasting plays an important role. At present, the use of meteorological big data to predict near-falling precipitation is a hot spot and difficulty in domestic disaster prevention and reduction. Prediction of near-falling precipitation with high temporal and spatial resolution has important value and social significance.

[0003] From the perspective of the development of precipitation forecasting, at the beginning, meteorologists mainly used the weather map method. This method has a long history and mature theory, but it is relatively subjective and is a qualitative analysis. With the continuous development of meteorology and computer technology, numerical prediction methods have gradually developed. It is mainly a method to predict precipitation changes by solving the atmospheric dynamics equations. This method is also a method that has been used more in practice in recent years, but this method has certain limitations when applied. First, the prediction accuracy of this method in the short term (especially within 1-2 hours) is often relatively low. This is mainly because the results of short-term predictions are more dependent on the initial values ​​of the atmospheric dynamics equations, and the current atmospheric information obtained is limited, so it is difficult to accurately estimate the initial values ​​of this equation. Secondly, it requires the cooperation of supercomputers, which are relatively large in size and require expensive computing and maintenance costs. In addition, in terms of timeliness, the time spent on numerical models is often very long. From data acquisition to assimilation to forecasting, it often takes several hours to calculate the results. For specific fields, the forecast may have lost its timeliness.

[0004] In recent years, as the types and quantity of meteorological data have increased exponentially, combining artificial intelligence technology with precipitation prediction has become an inevitable trend. It can mine useful information from massive multimodal meteorological big data, discover climate characteristics and atmospheric movement, and then make more accurate predictions of precipitation. However, artificial intelligence algorithms have certain uncertainties and may have certain errors in the prediction of some rare scenarios. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a precipitation prediction method and system based on artificial intelligence algorithm and knowledge graph to solve the problems existing in the above-mentioned background technology.

[0006] The present invention is implemented as follows: a precipitation prediction method based on artificial intelligence algorithm and knowledge graph, the method comprising the following steps:

[0007] Construct a multimodal data container, input different types of meteorological data into the multimodal data container according to their own structural characteristics to obtain multimodal data, perform spatiotemporal alignment, data cleaning and preprocessing, and data sequence segmentation on the multimodal data;

[0008] Construct a knowledge graph about precipitation prediction;

[0009] Create a multimodal precipitation prediction model based on artificial intelligence algorithms, specifically: build a multimodal radar echo prediction model, integrate various historical meteorological data in time and space, and use a deep neural network model to predict future radar sequence diagrams; build a radar precipitation intelligent conversion model, combine geographic location data, and use adversarial generative networks to convert radar data and precipitation to each other, so that the predicted radar sequence diagram can be converted into precipitation data; based on the constructed precipitation prediction knowledge graph, use neural network algorithms for knowledge reasoning to intelligently correct the predicted precipitation data;

[0010] Classification and visualization of corrected precipitation data.

[0011] As a further solution of the present invention: the step of performing spatiotemporal alignment on the multimodal data specifically includes:

[0012] Merge and align historical radar echo sequence data with a time error of less than 2 minutes with satellite cloud image data; perform bilinear interpolation on weather station data, fuse the data obtained after bilinear interpolation with radar data and satellite data, and perform spatial matching and alignment of radar data, satellite data, weather station data, and altitude data according to longitude and latitude;

[0013] The radar sequence data and the geographic location big data are temporally and spatially aligned with the precipitation data, wherein the geographic location big data includes latitude and longitude data and altitude data.

[0014] As a further solution of the present invention: the step of constructing a knowledge graph about precipitation prediction specifically includes:

[0015] Access precipitation-related data, including unstructured and semi-structured precipitation data from public websites and research, and structured precipitation data from radar and satellites;

[0016] For unstructured precipitation data and semi-structured precipitation data, a convolutional neural network based on the attention mechanism is used to extract entities, relationships and attributes for precipitation prediction to obtain precipitation information. The entities include longitude and latitude, altitude, time, temperature, wind direction, humidity and precipitation, etc.

[0017] The structured precipitation data and the extracted precipitation information are fused, and the fusion process includes precipitation entity disambiguation and alignment;

[0018] The extracted precipitation information is further processed, and the relationship between longitude and latitude, altitude, time, temperature, wind direction and humidity and precipitation is analyzed using a machine learning correlation analysis algorithm. That is, given the specific values ​​of meteorological factors, the maximum and minimum precipitation in the area are obtained;

[0019] Based on the extracted precipitation information, a precipitation prediction knowledge graph is constructed, and the precipitation prediction knowledge graph is stored in the graph database.

[0020] As a further solution of the present invention: the step of constructing a multi-modal radar echo prediction model specifically includes:

[0021] A multimodal radar echo prediction model consisting of a Data Integration module, a CNN+LSTM module, a DS+LSTM module, a US+LSTM module, a CNN+LSTM module, and a Data Generation module is created. To better predict precipitation in the central area, the multimodal radar echo prediction model uses meteorological data of a 512km*512km area and radar images of a 256km*256km prediction center area. The multimodal radar echo prediction model is used to fuse historical multimodal meteorological big data, which includes radar data, satellite cloud images, latitude and longitude data, altitude data, temperature, wind direction, and humidity. Multiple neural network modules are used for training and learning to accurately predict future high-resolution radar echo images. The loss function used by the multimodal radar echo prediction model is:

[0022]

[0023] in, Represents the radar chart data predicted by the model, y k Represents the real radar chart data, ω k and μ k Indicates the weighted values ​​for radar charts of different intensities;

[0024] The multimodal radar echo prediction model is evaluated using the evaluation function, which is:

[0025]

[0026] The Value evaluation result value is between -1 and 1. The larger the Value evaluation result value, the better the effect. M is the length of the predicted radar chart sequence, and Value is i is the score of a single radar chart, namely:

[0027]

[0028] Among them, L is the predicted category, N is the total number of samples, ωj is the weight of the jth category, p(R i , T j ) represents the total number of pixels of the predicted category i whose true category is j, p(R i ) represents the total number of pixels predicted as category i, p(T j ) represents the total number of pixels whose true category is j.

[0029] As a further solution of the present invention: the step of constructing the radar precipitation intelligent conversion model specifically includes: constructing a conversion model including four modules: R2P generator, P2R generator, P discriminator and R discriminator, combining the longitude and latitude and altitude geographic location data, using historical radar data and precipitation data, training and optimizing the conversion model on the GPU cluster, and obtaining the radar precipitation intelligent conversion model; the loss function used by the radar precipitation intelligent conversion model is:

[0030] Loss = Loss1 + Loss2 + Loss3

[0031] in:

[0032]

[0033]

[0034]

[0035] Another object of the present invention is to provide a precipitation prediction system based on artificial intelligence algorithm and knowledge graph, the system comprising:

[0036] The multi-model data container is used to input different types of meteorological data into the multi-modal data container according to their own structural characteristics to obtain multi-modal data, and to perform spatiotemporal alignment, data cleaning and preprocessing, and data sequence segmentation on the multi-modal data;

[0037] Precipitation prediction knowledge graph construction module, used to construct the knowledge graph about precipitation prediction;

[0038] A multimodal precipitation prediction and correction module based on artificial intelligence algorithms is used to create a multimodal precipitation prediction model based on artificial intelligence algorithms, specifically including: a radar echo prediction model construction unit, which is used to construct a multimodal radar echo prediction model, integrate various historical meteorological data in time and space, and use a deep neural network model to predict future radar sequence diagrams; an intelligent conversion model construction unit, which is used to construct a radar precipitation intelligent conversion model, which combines geographic location data and uses an adversarial generation network to convert radar data and precipitation to each other, so that the predicted radar sequence diagram can be converted into precipitation data; an intelligent correction unit, which uses a neural network algorithm to perform knowledge reasoning based on the constructed precipitation prediction knowledge graph to intelligently correct the predicted precipitation data; and

[0039] Regional precipitation visualization module, used to classify and visualize the corrected precipitation data.

[0040] Beneficial effects: The present invention proposes a precipitation prediction method and system based on artificial intelligence algorithm and knowledge graph, which not only makes full use of different types of multimodal meteorological data, improves the accuracy of precipitation prediction through the multimodal precipitation prediction method, solves the problems existing in the prior art, but also constructs a precipitation prediction knowledge graph, which can intelligently correct the results of model prediction to reduce the uncertainty of artificial intelligence algorithm and further improve the accuracy and reliability of model prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions and embodiments of the present invention, the drawings used in the technical solutions and embodiments are briefly introduced below.

[0042] Figure 1 This is a flowchart of a precipitation prediction method based on artificial intelligence algorithm and knowledge graph.

[0043] Figure 2 A flowchart for constructing a knowledge graph for precipitation prediction in an embodiment of the present invention.

[0044] Figure 3 Schematic diagram of the structure of a multi-modal radar echo prediction model in an embodiment of the present invention.

[0045] Figure 4 Schematic diagram of the structure of the radar precipitation intelligent conversion model in an embodiment of the present invention.

[0046] Figure 5 This is an architectural diagram of a precipitation prediction system based on artificial intelligence algorithms and knowledge graphs. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0048] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0049] like Figure 1 As shown, an embodiment of the present invention provides a precipitation prediction method based on an artificial intelligence algorithm and a knowledge graph, and the method includes the following steps:

[0050] S100, constructing a multimodal data container, inputting different types of meteorological data into the multimodal data container according to their own structural characteristics to obtain multimodal data, and performing spatiotemporal alignment, data cleaning and preprocessing, and data sequence segmentation on the multimodal data;

[0051] S200, build a knowledge graph about precipitation prediction, use natural language processing technology to extract precipitation-related knowledge from public websites and research to obtain precipitation-related information, integrate precipitation-related information with precipitation data collected by radar, satellites and meteorological stations, and use deep learning algorithms to build a precipitation prediction knowledge graph;

[0052] S300, creates a multimodal precipitation prediction model based on artificial intelligence algorithms, specifically:

[0053] S301, build a multi-modal radar echo prediction model, integrate various historical meteorological data in time and space, and use a deep neural network model to predict future radar sequence graphs;

[0054] S302, constructing a radar precipitation intelligent conversion model, combining geographic location data, and using a generative adversarial network to convert radar data and precipitation to each other, so that the predicted radar sequence graph can be converted into precipitation data;

[0055] S303, based on the constructed precipitation prediction knowledge graph, using a neural network algorithm to perform knowledge reasoning and intelligently correct the predicted precipitation data;

[0056] S400, classifies and visualizes the corrected precipitation data to facilitate observation and comparative analysis by users. For more intuitive observation, the precipitation is classified into 8 categories, namely 0, (0,2.5], (2.5,5], (5,10], (10,25], (25,50], (50,100], (100,+∞) (unit: mm / h). Then, based on the latitude and longitude data, the precipitation data is displayed on the map.

[0057] In an embodiment of the present invention, before obtaining multimodal meteorological data related to precipitation, it is necessary to obtain meteorological data in the same area within a period of time, including historical radar echo sequence data, satellite cloud map data, meteorological station data (temperature, humidity, wind direction, etc.), geographic location big data (including latitude and longitude data, altitude data, etc.), precipitation data, etc., and save these data in a multimodal data container according to categories. Then, it is necessary to align the acquired multimodal data in time and space, so that the processed data can be used to build a multimodal radar echo prediction model and a radar precipitation intelligent conversion model; in addition, it is necessary to clean and preprocess all data, introduce a smoothing factor to remove objects from historical data, and then use two-dimensional wavelet transform to remove noise, and use bilinear interpolation to fill the default data. In addition, some abnormal sequences or sequences with too many missing values ​​are eliminated to prevent these noises from interfering with the model. Finally, all preprocessed data are stored in the data container for subsequent modeling.

[0058] In the embodiment of the present invention, the step of performing spatiotemporal alignment on the multimodal data specifically includes:

[0059] S101, merge and align the historical radar echo sequence data with a time error of less than 2 minutes with the satellite cloud image data; perform bilinear interpolation on the meteorological station data, fuse the data obtained after the bilinear interpolation processing with the radar data and the satellite data, and spatially match and align the radar data, satellite data, meteorological station data and altitude data according to the longitude and latitude; it should be noted that in terms of time, the time interval of the historical radar echo sequence data is generally 6 minutes, and the time interval of the satellite cloud image data is generally 5 minutes. Based on the radar image time, the data with the closest time distance (error less than 2 minutes) between the two are integrated; secondly, the time interval of the meteorological station data (temperature, humidity, wind direction, etc.) is generally 1 hour, which can be fused with the radar and satellite data by performing bilinear interpolation; in terms of space, the radar data, satellite data, meteorological station data (temperature, humidity, wind direction, etc.) and altitude data are spatially aligned according to the longitude and latitude, and the fused data can be used to construct a multimodal radar echo prediction model.

[0060] S102, aligning the radar sequence data and the geographic location big data with the precipitation data in time and space, wherein the geographic location big data includes latitude and longitude data and altitude data. Similarly, the time interval of the precipitation data currently obtained is generally 1 hour. Based on this data, the radar sequence data and the geographic location big data (including latitude and longitude data, altitude data, etc.) and precipitation data are aligned in time and space, and the fused data can be used to build a radar precipitation intelligent conversion model.

[0061] like Figure 2As shown, in an embodiment of the present invention, the step of constructing a knowledge graph about precipitation prediction specifically includes:

[0062] S201, obtain precipitation-related data, including unstructured precipitation data and semi-structured precipitation data obtained from public websites and research, and structured precipitation data obtained from radar and satellite;

[0063] S202, for unstructured precipitation data and semi-structured precipitation data, extracting entities, relationships and attributes of precipitation prediction using a convolutional neural network based on an attention mechanism to obtain extracted precipitation information, wherein the entities include longitude and latitude, altitude, time, temperature, wind direction, humidity and precipitation, etc.;

[0064] S203, fusing the structured precipitation data and the extracted precipitation information, wherein the fusion process includes precipitation entity disambiguation and alignment.

[0065] S204, further processing the extracted precipitation information, using a machine learning correlation analysis algorithm to analyze the relationship between longitude and latitude, altitude, time, temperature, wind direction and humidity and precipitation, that is, given the specific values ​​of the meteorological factors, obtaining the maximum precipitation and the minimum precipitation in the area;

[0066] S205: construct a precipitation prediction knowledge graph based on the extracted precipitation information, and store the precipitation prediction knowledge graph in a graph database.

[0067] In an embodiment of the present invention, first, unstructured data and semi-structured data are obtained from public web pages and papers through crawlers and other technologies, and the pages and data are parsed, for example: "From 4 to 5 pm on July 20, 2021, the cumulative precipitation in Zhengzhou Station reached 201.9 mm in one hour, setting a record for single-hour precipitation in large, medium and small cities in the world", and precipitation-related structured data are obtained from radar satellites and other products. Knowledge extraction is performed on unstructured and semi-structured precipitation data, and first entity extraction is performed, such as longitude, latitude, temperature, altitude, wind direction, humidity, precipitation and other precipitation-related entities. Due to the limited data obtained, fine-tuning is mainly based on the pre-trained transfer model in natural language processing, and the field with rich training expectations is used to help complete entity extraction. Secondly, relationships and attributes are extracted from these entities. Here, a convolutional neural network based on the attention mechanism is mainly used to extract relationships and attributes in units of combined sentence vectors. For example, the knowledge extracted for an area is: longitude: 109.809000, latitude: 40.657000, altitude: 1069m, time: June 1, 2021, temperature: 10 degrees Celsius, wind direction: level 3, precipitation: 20mm / h (note that this is just for a brief explanation of the extraction results, and the data may have certain deviations).

[0068] It is also necessary to use structured data obtained from radar satellites and other products, combined with previously extracted knowledge and meteorological experts' experience, to fuse knowledge, including entity disambiguation and alignment, such as "precipitation", "rainfall", "rainfall amount", "rainfall amount", "precipitation amount", etc. are all unified as "precipitation amount". Here, the biLSTM neural network is mainly used, and entity alignment is regarded as the similarity comparison between strings, and finally the cosine similarity is calculated as the matching probability of the entity. Finally, the extracted knowledge is further processed. Using the machine learning correlation analysis algorithm, the relationship between other meteorological factors such as longitude and latitude, altitude, time, temperature, wind direction, humidity and precipitation is analyzed, that is, given the specific value of the meteorological factor, the maximum and minimum precipitation in the area can be obtained. Construct a knowledge graph for precipitation prediction, use an inference engine based on the RDF model, construct the extracted data into an ontology in RDF format, import it into the graph database, and complete the construction and storage of the knowledge graph.

[0069] like Figure 3 As shown, in an embodiment of the present invention, the step of constructing a multi-modal radar echo prediction model specifically includes:

[0070] S3011, create a multimodal radar echo prediction model consisting of Data Integration module, CNN+LSTM module, DS+LSTM module, US+LSTM module, CNN+LSTM module and Data Generation module. Since the atmospheric clouds are constantly moving, in order to better predict the precipitation in the central area, the model increases the receptive field of the observation data, that is, using the meteorological data of the 512km*512km area and the radar image of the 256km*256km prediction center area; the multimodal radar echo prediction model is used to integrate the historical multimodal meteorological big data, which includes radar data, satellite cloud images, latitude and longitude data, altitude data, temperature, wind direction and humidity, and uses multiple neural network modules for training and learning to make more accurate predictions of future high-resolution radar echo images; the loss function used by the multimodal radar echo prediction model is:

[0071]

[0072] in, Represents the radar chart data predicted by the model, y k Represents the real radar chart data, ω k and μ k Indicates the weighting values ​​for radar charts of different intensities.

[0073] S3012, using an evaluation function to evaluate the multi-modal radar echo prediction model, the evaluation function is:

[0074]

[0075] The Value evaluation result value is between -1 and 1. The larger the Value evaluation result value, the better the effect. M is the length of the predicted radar chart sequence, and Value is i is the score of a single radar chart, namely:

[0076]

[0077] Among them, L is the predicted category, N is the total number of samples, ω j is the weight of the jth category, p(R i , T j ) represents the total number of pixels of the predicted category i whose true category is j, p(R i ) represents the total number of pixels predicted as category i, p(T j ) represents the total number of pixels of true category j;

[0078] In an embodiment of the present invention, the processed sequence data in the data container is segmented according to demand, that is, based on the meteorological data of the previous 3 hours, the radar sequence data of the next 3 hours is predicted, and the data is divided into a training set, a validation set and a test set in a ratio of 7:1:2. First, the input data is the multimodal meteorological big data of the previous 3 hours, and the time interval is 6 minutes, that is, -180min, -174min, ..., 0min. The multimodal meteorological big data includes radar data, satellite data, temperature, wind direction, longitude and latitude, and altitude, which are input into each module. The Data Integration module mainly adopts the data fusion technology in deep learning to fuse radar data, satellite cloud images, longitude and latitude data, altitude data, temperature, wind direction, and humidity. First, in order to prevent the model from exploding or disappearing in the future, the hyperbolic tangent function is used to normalize the data of each dimension:

[0079]

[0080] Then, data fusion is performed using technologies such as Space-to-depth, Mean-pooling, Center-crop, Conv (convolution), and Max-pooling. The CNN+LSTM module is an encoder-decoder framework as a whole. Here, time and space are mainly fused. The convolution calculation is introduced into the LSTM calculation inside the network to extract the spatiotemporal correlation of meteorological data. If the feature extraction is more sufficient and the sample size is sufficient, the CNN+LSTM module can stack several layers of networks, but since the samples we obtained are limited, in order to prevent the samples from overfitting, the number of layers M can be set to 1. The DS+LSTM module mainly uses downsampling technology to fully extract the deep features of historical meteorological data in various dimensions. The US+LSTM module mainly uses upsampling technology to predict the future radar sequence data features using the extracted atmospheric motion features. The CNN+LSTM module + DataGeneration module mainly generates the radar sequence data of the original area as accurately as possible based on the extracted radar sequence data features. Finally, the Data Generation module outputs the radar sequence diagram for the next 3 hours, with a time interval of 30 minutes, that is, a prediction every half an hour.

[0081] In this embodiment of the present invention, the loss function used by the entire model is:

[0082]

[0083] in, Represents the radar chart data predicted by the model, y k Represents the real radar chart data, ω k and μ k Represents the weighted value of radar images of different intensities. Using the multimodal meteorological big data of the previous 3 hours, this improved model is trained on the GPU cluster until the model reaches the optimal value. The following evaluation function is used to evaluate the above optimal model:

[0084]

[0085] The Value evaluation result range is between -1 and 1, and the larger the value, the better the effect.

[0086] like Figure 4 As shown, the present invention proposes an artificial intelligence conversion model, which can regard radar images and precipitation images as two image domains. It is based on a generative adversarial network and includes four modules: an R2P generator, a P2R generator, a P discriminator, and an R discriminator. It combines geographic location data such as longitude and latitude, altitude, and uses historical radar data and precipitation data to train and optimize the model on a GPU cluster to obtain a model that can intelligently convert radar and precipitation.

[0087] In the embodiment of the present invention, first, radar data, geographic location big data (including latitude and longitude data, altitude data, etc.) and precipitation data are obtained, and set as {(X 1 , X 1 , Y 1 ), (X 2 , Z 2 , Y 2 ),...,(X n , Z n , Y n )}), divide the data into training set, validation set and test set in the ratio of 7:1:2. Then, a deep generative network is constructed, which can be regarded as consisting of two adversarial networks as a whole, including four modules: R2P generator: mainly adopts the encoder-decoder framework, using real radar data, combined with geographic location data such as longitude, latitude and altitude, and encodes its features through the encoder encoder, and then uses the decoder decoder to generate the corresponding precipitation data. P discriminator: mainly used to distinguish whether it is fake precipitation data of the R2P generator or real precipitation data. P2R generator: similar to the R2P generator structure, but its main function is to combine the real precipitation data with geographic location data such as longitude, latitude and altitude, encode its features through the encoder encoder, and then use the decoder decoder to generate the corresponding radar data. R discriminator: mainly used to distinguish whether it is fake radar data of the P2R generator or real radar data. The model first uses real radar data, combined with longitude, latitude and altitude data, that is, (X i , Z i ), i = 1, 2, ..., n, input to the R2P generator to generate the corresponding precipitation data Y i ′, i = 1, 2, ..., n. Then, the P discriminator is used to generate Y i ′ and the real Y i Similarly, using real precipitation data, combined with longitude, latitude and altitude data (Y i , Z i ), i = 1, 2, ..., n, input to the P2R generator to generate the corresponding radar data X i ′, i = 1, 2, ..., n. Then, use the R discriminator to generate X i and the real X i Make a judgment.

[0088] In this embodiment of the present invention, the loss function used by the entire model is:

[0089] Loss = Loss1 + Loss2 + Loss3

[0090] in:

[0091]

[0092]

[0093]

[0094] After adversarial training optimization on the GPU cluster, the R2P generator and P2R generator finally obtain a set of optimal parameters, which can flexibly transform radar data and precipitation data by combining latitude, longitude and altitude. The obtained R2P generator is the nearby precipitation conversion model, which can convert radar data into precipitation data.

[0095] In the embodiment of the present invention, based on the constructed precipitation prediction knowledge graph, a deep learning algorithm is used for knowledge reasoning to intelligently correct the obtained precipitation data. Using the data and relationships obtained in the knowledge graph obtained in S200, the maximum and minimum values ​​of regional precipitation can be predicted based on meteorological factors such as longitude and latitude, altitude, time, temperature, wind direction, and humidity through a deep learning prediction model. Furthermore, the precipitation data obtained in S32 is intelligently corrected according to the precipitation threshold predicted by the algorithm, further improving the accuracy and reliability of precipitation prediction.

[0096] In the embodiment of the present invention, the precipitation data is visualized for more intuitive observation. The predicted and corrected precipitation is classified into 8 categories, namely 0, (0, 2.5], (2.5, 5], (5, 10], (10, 25], (25, 50], (50, 100], (100, +∞) (unit: mm / h). Then, based on the latitude and longitude data, the precipitation data is displayed on the map for easy observation and analysis.

[0097] like Figure 5 As shown, the embodiment of the present invention also discloses a precipitation prediction system based on artificial intelligence algorithm and knowledge graph, and the system includes:

[0098] The multi-model data container is used to input different types of meteorological data into the multi-modal data container according to their own structural characteristics to obtain multi-modal data, and to perform spatiotemporal alignment, data cleaning and preprocessing, and data sequence segmentation on the multi-modal data;

[0099] The precipitation prediction knowledge graph construction module is used to construct the knowledge graph about precipitation prediction.

[0100] A multimodal precipitation prediction and correction module based on artificial intelligence algorithms is used to create a multimodal precipitation prediction model based on artificial intelligence algorithms, specifically including: a radar echo prediction model construction unit, which is used to construct a multimodal radar echo prediction model, integrate various historical meteorological data in time and space, and use a deep neural network model to predict future radar sequence diagrams; an intelligent conversion model construction unit, which is used to construct a radar precipitation intelligent conversion model, which combines geographic location data and uses an adversarial generation network to convert radar data and precipitation to each other, so that the predicted radar sequence diagram can be converted into precipitation data; an intelligent correction unit, which uses a neural network algorithm to perform knowledge reasoning based on the constructed precipitation prediction knowledge graph to intelligently correct the predicted precipitation data; and

[0101] Regional precipitation visualization module, used to classify and visualize the corrected precipitation data.

[0102] During use, the system first needs to obtain real-time data from radar, satellite, weather station and other dimensions of a certain area A, and then divide this area into n small areas of 500*500km. After that, the intelligent precipitation prediction system is used to predict the future precipitation of these n small areas through various modules. Finally, these small areas are merged and visualized to observe and analyze the future precipitation situation.

[0103] The above only describes in detail the preferred embodiments of the present invention, which is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0104] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0105] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0106] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the disclosure in the specification and examples. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

Claims

1. A precipitation prediction method based on artificial intelligence algorithm and knowledge graph, characterized in that: The method comprises the following steps: Construct a multimodal data container, input different types of meteorological data into the multimodal data container according to their own structural characteristics to obtain multimodal data, perform spatiotemporal alignment, data cleaning and preprocessing, and data sequence segmentation on the multimodal data; Construct a knowledge graph about precipitation prediction; Create a multimodal precipitation prediction model based on artificial intelligence algorithms, specifically: build a multimodal radar echo prediction model, integrate various historical meteorological data in time and space, and use a deep neural network model to predict future radar sequence diagrams; build a radar precipitation intelligent conversion model, combine geographic location data, and use adversarial generative networks to convert radar data and precipitation to each other, so that the predicted radar sequence diagram can be converted into precipitation data; based on the constructed precipitation prediction knowledge graph, use neural network algorithms for knowledge reasoning to intelligently correct the predicted precipitation data; Classify and visualize the corrected precipitation data; The step of performing spatiotemporal alignment on the multimodal data specifically includes: Merge and align historical radar echo sequence data with a time error of less than 2 minutes with satellite cloud image data; perform bilinear interpolation on weather station data, fuse the data obtained after bilinear interpolation with radar data and satellite data, and perform spatial matching and alignment of radar data, satellite data, weather station data, and altitude data according to longitude and latitude; Performing spatiotemporal alignment of radar sequence data and geographic location big data with precipitation data, wherein the geographic location big data includes latitude and longitude data and altitude data; The step of constructing a knowledge graph about precipitation prediction specifically includes: Access precipitation-related data, including unstructured and semi-structured precipitation data from public websites and research, and structured precipitation data from radar and satellites; For unstructured precipitation data and semi-structured precipitation data, a convolutional neural network based on an attention mechanism is used to extract entities, relationships, and attributes for precipitation prediction to obtain precipitation information. The entities include latitude and longitude, altitude, time, temperature, wind direction, humidity, and precipitation. The structured precipitation data and the extracted precipitation information are fused, and the fusion process includes precipitation entity disambiguation and alignment; The extracted precipitation information is further processed, and the relationship between longitude and latitude, altitude, time, temperature, wind direction and humidity and precipitation is analyzed using a machine learning correlation analysis algorithm. That is, given the specific values ​​of meteorological factors, the maximum and minimum precipitation in the region are obtained; Based on the extracted precipitation information, a precipitation prediction knowledge graph is constructed, and the precipitation prediction knowledge graph is stored in a graph database; The step of constructing a multi-modal radar echo prediction model specifically includes: A multimodal radar echo prediction model consisting of a Data Integration module, a CNN+LSTM module, a DS+LSTM module, a US+LSTM module, a CNN+LSTM module, and a Data Generation module is created. To better predict precipitation in the central area, the multimodal radar echo prediction model uses meteorological data of a 512km*512km area and radar images of a 256km*256km prediction center area. The multimodal radar echo prediction model is used to fuse historical multimodal meteorological big data, which includes radar data, satellite cloud images, latitude and longitude data, altitude data, temperature, wind direction, and humidity. Multiple neural network modules are used for training and learning to accurately predict future high-resolution radar echo images. The loss function used by the multimodal radar echo prediction model is: in, Radar chart data representing the model's predictions, represents the real radar chart data, and Indicates the weighted values ​​for radar charts of different intensities; The multimodal radar echo prediction model is evaluated using the evaluation function, which is: The evaluation result value is between -1 and 1. The larger the evaluation result value, the better the effect. is the predicted radar chart sequence length, is the score of a single radar chart, namely: in, is the predicted category, is the total number of samples, For the The weight of each category, Indicates the predicted category is The true category of the pixel point is The total number of pixels, Indicates that the prediction is a category The total number of pixels, The true category is The total number of pixels.

2. According to claim 1, a precipitation prediction method based on artificial intelligence algorithm and knowledge graph is characterized in that: The step of constructing the radar precipitation intelligent conversion model specifically includes: constructing a conversion model including four modules: R2P generator, P2R generator, P discriminator and R discriminator; combining the latitude, longitude and altitude geographic location data, using historical radar data and precipitation data, training and optimizing the conversion model on a GPU cluster, and obtaining a radar precipitation intelligent conversion model; the loss function used by the radar precipitation intelligent conversion model is: in: 。 3. A precipitation prediction system based on artificial intelligence algorithm and knowledge graph, used to implement the method of claim 1 or 2, characterized in that: The system comprises: The multi-model data container is used to input different types of meteorological data into the multi-modal data container according to their own structural characteristics to obtain multi-modal data, and to perform spatiotemporal alignment, data cleaning and preprocessing, and data sequence segmentation on the multi-modal data; Precipitation prediction knowledge graph construction module, used to construct the knowledge graph about precipitation prediction; A multimodal precipitation prediction and correction module based on artificial intelligence algorithms is used to create a multimodal precipitation prediction model based on artificial intelligence algorithms, specifically including: a radar echo prediction model construction unit, which is used to construct a multimodal radar echo prediction model, integrate various historical meteorological data in time and space, and use a deep neural network model to predict future radar sequence diagrams; an intelligent conversion model construction unit, which is used to construct a radar precipitation intelligent conversion model, which combines geographic location data and uses an adversarial generation network to convert radar data and precipitation to each other, so that the predicted radar sequence diagram can be converted into precipitation data; an intelligent correction unit, which uses a neural network algorithm to perform knowledge reasoning based on the constructed precipitation prediction knowledge graph to intelligently correct the predicted precipitation data; and Regional precipitation visualization module, used to classify and visualize the corrected precipitation data.

Images