Hydrological event liquid level prediction method, device, storage medium and computer equipment

Abstract

The method, device, storage medium and computer equipment for predicting the liquid level of a hydrological event provided by the present application can first obtain the real-time serial data of the hydrological event when predicting the liquid level of a hydrological event to ensure the accuracy of the real-time changes in the data related to the hydrological event; then determine the event type of the real-time serial data and the liquid level prediction model corresponding to the event type, so as to use the model to predict the real-time serial data and obtain the liquid level prediction value, which can improve the accuracy of the model's description of the complex relationship of the hydrological event during the liquid level prediction process, thereby improving the accuracy of the prediction. In order to improve the prediction accuracy of the model, each type of model considers the important influencing factors of its corresponding event type and is trained. For example, the model of drought events is trained using a nonlinear function regression model; the model of target events is trained using a nonlinear autoregressive network model; and the model of depletion flow events is trained using a feedforward neural network model.

Description

Technical Field

[0001] The present application relates to the field of environmental engineering technology, and in particular to a method, device, storage medium and computer equipment for predicting liquid levels of hydrological events. Background Art

[0002] With the rapid urbanization process and the increase in the frequency and magnitude of extreme hydrological events caused by climate change, in order to more effectively respond to extreme weather events and ensure the sustainable use of water resources, urban drainage systems can be used to collect, transmit and treat urban precipitation and sewage, while real-time flood forecasting models can be used to monitor and simulate urban precipitation, water levels and flooding in real time to provide early warning and respond to flood events. The combination of these two systems can help cities better plan, manage and respond to water resources, improve their ability to resist floods, and reduce the impact of floods on residents and infrastructure.

[0003] However, traditional real-time flood forecasting models rely on continuous time series data, and their accuracy is limited. On the one hand, meteorological and climate change introduce additional uncertainties, and the complexity of natural systems causes flood processes to involve many nonlinear and interacting factors, such as topography, soil type, vegetation cover, etc., making it difficult for traditional flood forecasting models to accurately describe these complex relationships, affecting the accuracy of predictions; on the other hand, in order to reduce computational complexity, traditional models usually oversimplify natural systems, ignore some important influencing factors, or simplify the models through parameterization, which leads to a decrease in model accuracy. Therefore, traditional real-time flood forecasting models are difficult to achieve highly accurate prediction results when faced with complex and changeable natural systems. Summary of the invention

[0004] The purpose of this application is to solve at least one of the above-mentioned technical defects, especially the technical defect that the real-time flood forecasting model in the prior art is difficult to obtain highly accurate prediction results when facing complex and changeable natural systems.

[0005] The present application provides a method for predicting a liquid level in a hydrological event, the method comprising:

[0006] Acquire the hydrological event to be predicted and the real-time sequence data of the hydrological event;

[0007] Using a decision tree method to classify the real-time sequence data, obtaining the event type of the hydrological event; the event type includes drought events, target events and depletion flow events;

[0008] Determine the liquid level prediction model corresponding to the event type; wherein the liquid level prediction model for the drought event is obtained by training with a nonlinear function regression model; the liquid level prediction model for the target event is obtained by training with a nonlinear autoregressive network model; the liquid level prediction model for the depletion flow event is obtained by training with a feedforward neural network model;

[0009] The real-time sequence data is input into the liquid level prediction model to obtain a liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model.

[0010] Optionally, obtaining real-time sequence data of the hydrological event includes:

[0011] determining a plurality of satellite images of the hydrological event and rainfall data within a rainfall interval;

[0012] Use image analysis technology to identify regions in each satellite image and obtain the geological features in each satellite image;

[0013] Sorting each satellite image according to the time series, and analyzing each geographical feature according to the sorting result to obtain the geodynamic data of the hydrological event;

[0014] Real-time sequence data of the hydrological event is generated according to the rainfall data and the geoscientific dynamic data.

[0015] Optionally, the classifying the real-time sequence data using a decision tree method to obtain the event type of the hydrological event includes:

[0016] Determine an event classification model; the event classification model is obtained by training a decision tree model;

[0017] The real-time sequence data is input into the event classification model to obtain a prediction result output by the event classification model, and the prediction result is used as the event type of the hydrological event.

[0018] Optionally, determining the event classification model includes:

[0019] Obtain historical sequence data and real event types of historical hydrological events;

[0020] Inputting the historical sequence data into a preset decision tree model to obtain the predicted event type output by the decision tree model;

[0021] Training the decision tree model with the goal of making the predicted event type close to the real event type of the historical hydrological event;

[0022] When the decision tree model meets the preset training conditions, the trained decision tree model is used as an event classification model.

[0023] Optionally, the real event type of the historical hydrological event is obtained, including:

[0024] Filtering cluster feature data from the historical sequence data, and performing standardization processing on the cluster feature data to obtain preprocessed data;

[0025] Performing cluster analysis on the preprocessed data using a K-means clustering algorithm to obtain a cluster center of the preprocessed data;

[0026] Performing feature analysis on the cluster centers, and obtaining the real event types of the historical hydrological events.

[0027] Optionally, determining a liquid level prediction model corresponding to the drought event includes:

[0028] Acquire historical drought data of historical drought events, and divide the historical drought data into a training set and a test set;

[0029] Using the training set to train a preset nonlinear function regression model, and after the nonlinear function regression model meets the preset training conditions, using the trained nonlinear function regression model as an intermediate prediction model;

[0030] Inputting the test set into the intermediate prediction model to obtain the predicted liquid level output by the intermediate prediction model;

[0031] When the difference between the predicted liquid level and the actual liquid level in the historical drought data is less than a preset error, the intermediate prediction model is used as the liquid level prediction model corresponding to the drought event.

[0032] Optionally, determining a liquid level prediction model corresponding to the target event includes:

[0033] Acquire historical target data of historical target events, and divide the historical target data into a training set and a test set using a time series cutting method;

[0034] A cross-validation method is used to select parameters for a preset nonlinear autoregressive network model, and the training set is used to iteratively train the nonlinear autoregressive network model after parameter selection to obtain an intermediate test model;

[0035] The test set is used to perform performance verification on the intermediate test model, and the intermediate test model that passes the verification is used as the liquid level prediction model corresponding to the target event.

[0036] Optionally, determining a liquid level prediction model corresponding to the depletion flow event includes:

[0037] Acquire historical target data of historical depletion flow events, and divide the historical target data into a training set and a test set using a time series cutting method;

[0038] Iteratively training a preset feedforward neural network model using the training set, and adjusting the parameters of the feedforward neural network model using a back propagation algorithm during the training process;

[0039] When the feedforward neural network model meets the preset training conditions, the trained feedforward neural network model is used as an intermediate prediction model;

[0040] The test set is used to verify the performance of the intermediate prediction model, and the verified intermediate prediction model is used as the liquid level prediction model corresponding to the depletion flow event.

[0041] The present application also provides a hydrological event liquid level prediction device, comprising:

[0042] A data acquisition module, used to acquire the hydrological event to be predicted and the real-time sequence data of the hydrological event;

[0043] An event classification module, used to classify the real-time sequence data using a decision tree method to obtain the event type of the hydrological event; the event type includes drought events, target events and depletion flow events;

[0044] A model determination module is used to determine the liquid level prediction model corresponding to the event type; wherein the liquid level prediction model for the drought event is obtained by training with a nonlinear function regression model; the liquid level prediction model for the target event is obtained by training with a nonlinear autoregressive network model; and the liquid level prediction model for the depletion flow event is obtained by training with a feedforward neural network model;

[0045] The liquid level prediction module is used to input the real-time sequence data into the liquid level prediction model to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model.

[0046] The present application also provides a storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the hydrological event liquid level prediction method as described in any of the above embodiments.

[0047] The present application also provides a computer device, comprising: one or more processors, and a memory;

[0048] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the method for predicting the liquid level of a hydrological event as described in any one of the above embodiments are performed.

[0049] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0050] The method, device, storage medium and computer equipment for predicting the liquid level of a hydrological event provided by the present application can first obtain the real-time sequence data of the hydrological event when predicting the liquid level of a hydrological event to ensure the accuracy of the real-time changes in the data related to the hydrological event and improve the accuracy of the model prediction; then the decision tree method can be used to classify the real-time sequence data to obtain the event type of the hydrological event, and then determine the liquid level prediction model corresponding to the event type, so that different types of liquid level prediction models can be used to improve the accuracy of the description of the complex relationship of the hydrological event in the liquid level prediction process, thereby improving the accuracy of the prediction. Among them, each type of liquid level prediction model is trained by considering the important influencing factors of its corresponding event type, so that the prediction accuracy of the model obtained by training is higher. For example, the liquid level prediction model of drought events can be obtained by training using a nonlinear function regression model; the liquid level prediction model of the target event can be obtained by training using a nonlinear autoregressive network model; and the liquid level prediction model of the depletion flow event can be obtained by training using a feedforward neural network model. After obtaining the liquid level prediction model corresponding to the hydrological event, the present application can input the real-time sequence data of the hydrological event into the liquid level prediction model to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model. The present application uses the liquid level prediction model corresponding to the event type of the hydrological event to predict the acquired real-time sequence data, which can reduce the influence of various factors in the complex and changeable natural system, thereby improving the accuracy of the model prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0052] Figure 1 A schematic diagram of a flow chart of a method for predicting a liquid level in a hydrological event provided in an embodiment of the present application;

[0053] Figure 2 A flowchart of a real-time sequence data acquisition process provided in an embodiment of the present application;

[0054] Figure 3 A schematic diagram of a flow chart of a training process of a liquid level prediction model for drought events provided in an embodiment of the present application;

[0055] Figure 4A schematic diagram of a flow chart of a liquid level prediction model training process for a target event provided in an embodiment of the present application;

[0056] Figure 5 A schematic diagram of a flow chart of a liquid level prediction model training process for a depletion flow event provided in an embodiment of the present application;

[0057] Figure 6 A schematic diagram of the structure of a hydrological event liquid level prediction device provided in an embodiment of the present application;

[0058] Figure 7 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0060] Traditional real-time flood forecasting models rely on continuous time series data, and their accuracy is limited. On the one hand, meteorological and climate change introduce additional uncertainties, and the complexity of natural systems causes the flood process to involve many nonlinear and interacting factors, such as topography, soil type, vegetation cover, etc., making it difficult for traditional flood forecasting models to accurately describe these complex relationships, affecting the accuracy of the prediction; on the other hand, in order to reduce computational complexity, traditional models usually oversimplify the natural system, ignore some important influencing factors, or simplify the model through parameterization, which leads to a decrease in the accuracy of the model. Therefore, it is difficult for traditional real-time flood forecasting models to achieve highly accurate prediction results when faced with complex and changeable natural systems.

[0061] Based on this, this application proposes the following technical solutions, see below for details:

[0062] In one embodiment, Figure 1 As shown, Figure 1 A schematic diagram of a flow chart of a method for predicting the liquid level of a hydrological event provided in an embodiment of the present application; the present application also provides a method for predicting the liquid level of a hydrological event, which specifically includes the following:

[0063] S110: Acquire the hydrological event to be predicted and the real-time sequence data of the hydrological event.

[0064] In this step, when it is necessary to predict the liquid level of a hydrological event, the computer device can obtain the real-time serial data of the hydrological event from the relevant system or device, and then predict the liquid level of the hydrological event based on the real-time serial data.

[0065] Among them, the real-time sequence data of the present application may include data such as rainfall time series data and water level time series data generated at the time of occurrence of hydrological events, and may also include geo-related information at the time of rainfall, such as meteorological data, water temperature data and terrain information. To elaborate, meteorological data covers the current weather conditions at the time of rainfall, such as rainfall, temperature, humidity, wind speed, etc.; hydrological data includes parameters related to the hydrological cycle such as river water level, flow and soil moisture; and terrain information may include the topography and geomorphic features of the surface, such as mountains, rivers, lakes, road networks, etc.

[0066] It can be understood that the real-time sequence data obtained by the present application covers the dynamic changes of hydrological events in time and space, ensuring the integrity of the data. This helps computer equipment to better capture the evolution process of the hydrological event when using real-time sequence data to predict the liquid level of hydrological events, including the temporal and spatial distribution of rainfall, the formation of surface runoff, changes in river water levels, etc., thereby reducing the interference of various factors in the complex and changeable natural system, and thus improving the prediction accuracy of the liquid level of hydrological events.

[0067] S120: Classify the real-time sequence data using a decision tree method to obtain the event type of the hydrological event.

[0068] In this embodiment, after obtaining the real-time serial data of the hydrological event through step S110, the computer device can perform feature analysis on the real-time serial data to obtain multiple features of the hydrological event. Then the computer device can adopt a decision tree method and classify the hydrological event according to each feature to obtain its corresponding event type.

[0069] It can be understood that the event types include drought events, target events and depletion flow events. Among them, drought events refer to the phenomenon that dry weather occurs in urban areas within a certain period of time, but the liquid level does not change or the liquid level rises; target events refer to the phenomenon that wet weather occurs in urban areas within a certain period of time, causing the liquid level to change; depletion flow events refer to the phenomenon that dry weather occurs in urban areas within a certain period of time, causing the liquid level to rise.

[0070] In addition, the decision tree method refers to a classification rule method that centrally infers decision tree representations from unordered and irregular data sets. It uses a top-down recursive method to compare attribute values ​​at the internal nodes of the decision tree and judge the branches from the node downward according to different attribute values, and draw conclusions at the leaf nodes of the decision tree. Therefore, a path from the root node to the leaf node corresponds to a rule, and the entire decision tree corresponds to a set of expression rules. In the present application, a computer device can use a decision tree method to classify hydrological events, and generate a tree-like decision path by dividing the various characteristics of the hydrological events in a certain order to achieve the prediction goal.

[0071] S130: Determine a liquid level prediction model corresponding to the event type.

[0072] In this step, after obtaining the event type of the hydrological event through step S120, the computer device can select a liquid level prediction model corresponding to the event type of the hydrological event, and then use the liquid level prediction model to predict the liquid level of the hydrological event.

[0073] It is understandable that, since different types of hydrological events are affected by different factors, the present application can use different modeling methods to predict the liquid level of hydrological events of different event types according to the relationship between the characteristics of different event types and the liquid level. The personalized modeling method in the present application can make the model prediction more flexible, better capture the characteristics of different event types, and thus improve the accuracy of the model prediction.

[0074] For example, for drought events, since nonlinear function regression can better capture various complex relationships and is suitable for situations where there may be nonlinear correlations between liquid level and other factors in drought events, nonlinear function regression can be used to predict liquid level in drought events to more accurately predict the changing trend of liquid level in drought events. For target events, since nonlinear autoregressive networks can take into account the impact of historical data of liquid level itself on the future and are suitable for situations where there is a certain temporal dependency relationship between liquid level in target events, nonlinear autoregressive networks can be used to predict liquid level in target events to better predict the changing trend of liquid level in the future in target events. For depletion flow events, since feedforward neural networks are suitable for processing complex mapping relationships between input and output and are suitable for situations where liquid level is affected by multiple factors in depletion flow events, feedforward neural networks can be used to predict liquid level in depletion flow events to more comprehensively consider the impact of various factors on liquid level, thereby improving the accuracy of prediction.

[0075] Furthermore, after the computer has trained the liquid level prediction model corresponding to each event type, it can also store the trained liquid level prediction model in the computer device, so that when predicting the liquid level of a hydrological event later, the computer device can directly call the pre-stored liquid level prediction model corresponding to the event type of the hydrological event to perform liquid level prediction.

[0076] S140: Input the real-time sequence data into the liquid level prediction model to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model.

[0077] In this step, after determining the liquid level prediction model through step S130, the computer device can input the real-time sequence data of the hydrological event into the liquid level prediction model, and predict the real-time sequence data through the liquid level prediction model, so as to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model.

[0078] Furthermore, since the real-time sequence data covers all relevant data of hydrological events at the time and space levels, when the computer equipment uses the liquid level prediction model to predict the liquid level of hydrological events, it can first select characteristic data related to the depletion flow liquid level prediction from the real-time sequence data, and then use the selected characteristic data as input data of the liquid level prediction model to perform liquid level prediction, which can further improve the prediction efficiency of the liquid level prediction model.

[0079] In the above embodiment, when predicting the liquid level of a hydrological event, the computer device can first obtain the real-time sequence data of the hydrological event to ensure the accuracy of the real-time changes of the data related to the hydrological event and improve the accuracy of the model prediction; then the decision tree method can be used to classify the real-time sequence data to obtain the event type of the hydrological event, and then determine the liquid level prediction model corresponding to the event type, so that different types of liquid level prediction models can be used to improve the accuracy of the description of the complex relationship of the hydrological event in the liquid level prediction process, thereby improving the accuracy of the prediction. Among them, each type of liquid level prediction model is trained by considering the important influencing factors of its corresponding event type, so that the prediction accuracy of the model obtained by training is high. For example, the liquid level prediction model of the drought event can be obtained by training using a nonlinear function regression model; the liquid level prediction model of the target event can be obtained by training using a nonlinear autoregressive network model; and the liquid level prediction model of the depletion flow event can be obtained by training using a feedforward neural network model. After obtaining the liquid level prediction model corresponding to the hydrological event, the computer device can input the real-time sequence data of the hydrological event into the liquid level prediction model to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model. The present application uses the liquid level prediction model corresponding to the event type of the hydrological event to predict the acquired real-time sequence data, which can reduce the impact of various factors in the complex and changeable natural system, thereby improving the accuracy of the model prediction.

[0080] In one embodiment, Figure 2 As shown, Figure 2 A flowchart of a real-time sequence data acquisition process provided in an embodiment of the present application; Figure 2 In step S110, obtaining real-time sequence data of hydrological events may include:

[0081] S111: Determine multiple satellite images of hydrological events and rainfall data within a rainfall interval.

[0082] S112: Using image analysis technology to perform regional recognition on each satellite image to obtain the geological features in each satellite image.

[0083] S113: sorting the satellite images according to the time series, and analyzing the geoscientific features according to the sorting results to obtain geoscientific dynamic data of the hydrological events.

[0084] S114: Generate real-time sequence data of hydrological events based on rainfall data and geoscientific dynamic data.

[0085] In this embodiment, when the computer device obtains the real-time serial data of the hydrological event, it can first determine multiple satellite images of the hydrological event and the rainfall data in the rainfall range, and use image analysis technology to perform regional identification on each satellite image to obtain the geological features in each satellite image. Then, the computer device can sort the satellite images according to the time series, and analyze each geographical feature according to the sorting result to obtain the geological dynamic data of the hydrological event. Finally, the computer device can generate the real-time serial data of the hydrological event based on the rainfall data and the geological dynamic data.

[0086] Specifically, the computer device can first determine the target area of ​​the hydrological event, and obtain satellite images of the target area from a geographic information system or an online map service at a certain time frequency. The computer system can then use image analysis technology, such as remote sensing image classification algorithms, to identify the area covered by the flood in the satellite image, and separate the flood-affected area from the ground feature area through supervised or unsupervised classification methods, and then determine the geoscientific features in the satellite image based on the identification results.

[0087] Among them, the geological features of satellite images can represent terrain and soil characteristics, including water distribution, soil wetness, etc. For example, computer equipment can determine the scope and depth of floods through the distribution of water bodies in satellite images. For example, blue or dark areas in the image indicate the presence of water bodies. After the water flows through, the wetness of the land surface will change. Therefore, the wetness of the soil can be observed through the color and texture of the land surface in the image. Wet soil may appear darker and smoother, in contrast to the surrounding dry land.

[0088] Furthermore, after acquiring satellite images, the computer equipment can first filter out satellite images with high resolution, covering the target area and involving the time of occurrence of hydrological events from various satellite images, and then perform pre-processing operations on the filtered satellite images, including correcting the geometric and radiation distortion of the satellite images, thereby ensuring the accuracy and high quality of the satellite images.

[0089] Furthermore, the computer equipment can perform time series analysis on each pre-processed satellite image, sort each satellite image according to the time series, and compare the various geological features of the satellite images at different time points. In this way, the evolution process of hydrological events can be observed, such as the spread, contraction and impact range changes of floods, water body changes, etc. The computer equipment can then construct a corresponding hydrological event map based on the observed evolution process.

[0090] In order to ensure the accuracy of the data in the hydrological event map, the computer device can also obtain real-time satellite data of hydrological events to verify the hydrological event map. In detail, the computer device can obtain relevant real-time satellite data of hydrological events through sensors such as optical sensors and radar sensors, and extract satellite data with high resolution and appropriate time frequency. Then, the computer device can also perform preprocessing operations on the extracted satellite data, which include but are not limited to atmospheric correction, radiation correction, geometric correction and other operations to ensure the high quality and consistency of satellite data.

[0091] Among them, atmospheric correction refers to eliminating the influence of the atmosphere on the reflectivity of satellite data in order to more accurately reflect the actual situation of the earth's surface. It includes considering factors such as scattering and absorption of the bottom atmosphere, and correcting the changes in the data caused by the atmosphere; radiation correction refers to eliminating the emissivity differences in satellite data caused by factors such as sensor characteristics and changes in solar radiation. This correction can make the data acquired by different sensors at different times have a consistent emissivity level; and geometric correction refers to correcting the geometric distortion in satellite data, including the influence caused by factors such as the curvature of the earth and terrain undulations. Through geometric correction, satellite data can be accurately aligned with the position on the earth's surface in space.

[0092] After obtaining the satellite data after the preprocessing operation, the computer device can use remote sensing image interpretation and classification technology to classify the satellite data and extract the features of each feature information. For example, for hydrological information, the computer device can extract water body characteristics through remote sensing technology, where the water body characteristics can include hydrological parameters such as the water area and water body shape of water bodies such as rivers, lakes and reservoirs; and for terrain information, the computer device can identify the coverage types of different surfaces and their corresponding terrain characteristics through remote sensing analysis of land availability and land cover, where the terrain characteristics can include the elevation of coverage types such as forests, grasslands and urban areas, and terrain parameters such as mountains and valleys. Then the computer device can generate a geoscience information map based on the extracted water body characteristics and terrain characteristics data, and finally use the geoscience information map to compare and verify the hydrological event map on the ground, and then generate more comprehensive geoscience dynamic data based on the verification results.

[0093] It should be noted that computer equipment can regularly acquire new satellite data to update the geological information map in real time, which can ensure the timeliness of geological dynamic data and monitor the dynamic changes of the geographical environment.

[0094] In one embodiment, in step S120, the real-time sequence data is classified using a decision tree method to obtain the event type of the hydrological event, which may include:

[0095] S121: Determine an event classification model; the event classification model is trained using a decision tree model.

[0096] S122: Input the real-time sequence data into the event classification model to obtain the prediction result output by the event classification model, and use the prediction result as the event type of the hydrological event.

[0097] In this embodiment, after obtaining the real-time sequence data of the hydrological event, the computer device can determine the event classification model, and use the event classification model to determine the event type corresponding to the real-time sequence data. Here, the computer device can input the real-time sequence data into the event classification model, obtain the prediction result output by the event classification model, and use the prediction result as the event type of the hydrological event.

[0098] It is understandable that the event classification model of the present application is obtained by training with a decision tree model. When the event classification model is being trained, the computer device can gradually divide the training data according to the selected event features to distinguish different event types to the greatest extent. Here, when dividing the data, the computer device can choose to use information gain, Gini method, etc. as the division criteria, which are not limited here. Then the computer device can perform pruning operations according to the training data, remove unnecessary nodes and branches, and make the decision tree grow to a sufficient depth, while also preventing the model from overfitting. Therefore, after the computer device inputs the real-time sequence data into the event classification model, the prediction result output by the event classification model, i.e., the event type of the hydrological event, can be obtained.

[0099] In one embodiment, determining the event classification model in step S121 may include:

[0100] S1211: Obtain historical sequence data and actual event types of historical hydrological events.

[0101] S1212: Input the historical sequence data into a preset decision tree model to obtain the predicted event type output by the decision tree model.

[0102] S1213: Train the decision tree model with the goal of predicting event types close to the actual event types of historical hydrological events.

[0103] S1214: When the decision tree model meets the preset training conditions, the trained decision tree model is used as an event classification model.

[0104] In this embodiment, the computer device can obtain historical sequence data and real event types of historical hydrological events, and input the historical sequence data into a preset decision tree model to obtain the predicted event type output by the decision tree model. The computer device can then train the decision tree model with the goal of making the predicted event type close to the real event type of the historical hydrological event, and after the decision tree model meets the preset training conditions, the trained decision tree model is used as an event classification model.

[0105] Specifically, when determining the event classification model, the computer device can first obtain the relevant characteristic data of the historical hydrological events, including characteristic data such as precipitation, water level, soil moisture, etc., and then filter out the characteristic data for decision tree classification from these characteristic data, and perform standardization on the selected characteristic data, remove its missing values ​​and outliers, and obtain historical sequence data to ensure the high quality of the characteristic data. Then the computer device can select the corresponding semantic segmentation model as the preset decision tree model for improvement and training. During the training process, the historical sequence data can be used as a training sample, and the corresponding sample label can be marked in the training sample. The sample label is the real event type of the historical sequence data. After the historical sequence data is input into the preset decision tree model, the predicted event type output by the decision tree model can be obtained. Then, the computer device can train the decision tree model with the goal of predicting the event type close to the real event type of the historical hydrological event, and when the decision tree model meets the preset training conditions, the trained decision tree model is used as the event classification model.

[0106] Furthermore, during the training process of the decision tree model, the computer device can also use new training samples to verify the performance of the iteratively trained decision tree model, so as to evaluate the generalization ability of the decision tree model on new data. When evaluating the generalization ability of the model, confusion matrix, accuracy, recall rate and other indicators can be used for evaluation, and there is no restriction here.

[0107] Furthermore, after the computer obtains the event classification model through training, it can also store the trained event classification model in the computer device, so that when subsequently identifying the event type of a hydrological event, the computer device can directly call the pre-stored event classification model to identify the event type of the hydrological event.

[0108] In one embodiment, obtaining the real event type of the historical hydrological event in step S1211 may include:

[0109] S2111: Filter out cluster feature data from historical sequence data, and standardize the cluster feature data to obtain preprocessed data.

[0110] S2112: Perform cluster analysis on the preprocessed data using the K-means clustering algorithm to obtain the cluster center of the preprocessed data.

[0111] S2113: Perform feature analysis on cluster centers to obtain the real event types of historical hydrological events.

[0112] In this embodiment, when obtaining the real event type of historical hydrological events, the computer device can first filter out cluster feature data from the historical sequence data, and standardize the cluster feature data to obtain preprocessed data. Then, the K-means clustering algorithm can be used to perform cluster analysis on the preprocessed data to obtain the cluster center of the preprocessed data. Finally, the computer device can perform feature analysis on the cluster center to obtain the real event type of the historical hydrological event.

[0113] It should be noted that the historical sequence data includes multiple feature data, each of which has a consistent timestamp, thereby ensuring the correlation between the feature data and improving the accuracy of subsequent data analysis.

[0114] Specifically, when the computer device analyzes the real event type based on the historical sequence data, it can first filter out the cluster feature data from the historical sequence data. The cluster feature data may include time series data such as precipitation and water level. The filtered feature data can be used to distinguish different event types. Then the computer device can standardize the cluster feature data into preprocessed data. The standardization process here can keep the data with different features on the same scale, thereby avoiding some of the feature data from having too much influence on the clustering results.

[0115] In the cluster analysis process, the computer device can determine the number of clusters (K value) according to the distribution of the preprocessed data, that is, the number of clusters into which the preprocessed data is divided; then the computer device can randomly select K data points as the initial cluster centers, use these points as representatives of each cluster, and assign each data point to the nearest cluster by calculating the distance from each data point to each cluster center, and then update the center of each cluster and take the average value of all data points in the cluster; then the computer device can repeatedly iterate and update the cluster center until the change of the cluster center is less than a predetermined threshold, or when the set maximum number of iterations is reached, the update of the cluster center is terminated; finally, the computer device can perform feature analysis on the latest cluster center of the preprocessed data. Among them, each cluster represents an event type, so the computer device can obtain the real event type of the historical hydrological event based on the feature analysis results of the cluster center.

[0116] In one embodiment, Figure 3 As shown, Figure 3A schematic diagram of a flow chart of a training process of a liquid level prediction model for drought events provided in an embodiment of the present application; Figure 3 In step S130, determining the liquid level prediction model corresponding to the drought event may include:

[0117] S1311: Obtain historical drought data of historical drought events, and divide the historical drought data into a training set and a test set.

[0118] S1312: Train the preset nonlinear function regression model using the training set, and after the nonlinear function regression model meets the preset training conditions, use the trained nonlinear function regression model as an intermediate prediction model.

[0119] S1313: Input the test set into the intermediate prediction model to obtain the predicted liquid level output by the intermediate prediction model.

[0120] S1314: When the difference between the predicted liquid level and the actual liquid level in the historical drought data is less than a preset error, the intermediate prediction model is used as the liquid level prediction model corresponding to the drought event.

[0121] In this embodiment, when determining the liquid level prediction model corresponding to the drought event, the computer device may first obtain historical drought data of the historical drought event, and divide the historical drought data into a training set and a test set. Then, the computer device may use the training set to train the preset nonlinear function regression model, and after the nonlinear function regression model meets the preset training conditions, the trained nonlinear function regression model is used as the intermediate prediction model. After obtaining the intermediate prediction model, the computer device may also input the test set into the intermediate prediction model to obtain the predicted liquid level output by the intermediate prediction model. When the difference between the predicted liquid level and the actual liquid level in the historical drought data is less than the preset error, the intermediate prediction model is used as the liquid level prediction model corresponding to the drought event.

[0122] Specifically, when training the liquid level prediction model corresponding to a drought event, the computer device can first obtain a data set of historical drought events, which includes characteristic data such as precipitation, soil moisture, and temperature during the drought period. Then, characteristic data related to drought liquid level prediction can be extracted from the data set and standardized or normalized to form historical drought data to ensure that the selected characteristic data has the ability to explain the liquid level changes, so that the model can converge faster and improve the prediction performance. Then the computer device can divide the historical drought data into a training set and a test set in a ratio of 80 / 20 or 70 / 30.

[0123] It should be noted that the computer equipment can determine the real liquid level in the historical drought data, and the determination of the real liquid level here matches the specific definition and measurement method of the liquid level.

[0124] After obtaining the training data, the computer device can select an appropriate nonlinear function regression model for improvement and training according to the nonlinear relationship between the liquid level and the characteristic data in the drought event. The nonlinear function regression model here can select vector machine regression (SVR), Gaussian process regression (GPR) or neural network models, which are not limited here. Then the computer device can use the training set to train the selected nonlinear function regression model. During the training process, the parameters of the model can be adjusted according to the training data to minimize the error of the liquid level prediction.

[0125] Furthermore, after the nonlinear function regression model is trained using the training set to obtain an intermediate prediction model, the computer device can verify the performance of the model on the test set, evaluate the accuracy and generalization ability of the model by comparing the difference between the liquid level predicted by the model and the actual liquid level, and use the intermediate prediction model as the liquid level prediction model corresponding to the drought event when the difference between the predicted liquid level and the actual liquid level in the historical drought data is less than a preset error.

[0126] In one embodiment, Figure 4 As shown, Figure 4 A schematic diagram of a flow chart of a liquid level prediction model training process for a target event provided in an embodiment of the present application; Figure 4 In step S130, determining the liquid level prediction model corresponding to the target event may include:

[0127] S1321: Obtain historical target data of historical target events, and use a time series cutting method to divide the historical target data into a training set and a test set.

[0128] S1322: Using a cross-validation method to select parameters for a preset nonlinear autoregressive network model, and using a training set to iteratively train the nonlinear autoregressive network model after parameter selection to obtain an intermediate test model.

[0129] S1323: Use the test set to verify the performance of the intermediate test model, and use the verified intermediate test model as the liquid level prediction model corresponding to the target event.

[0130] In this embodiment, when determining the liquid level prediction model corresponding to the target event, the computer device can first obtain the historical target data of the historical target event, and use the time series cutting method to divide the historical target data into a training set and a test set. Then, the computer device can use the cross-validation method to select parameters for the preset nonlinear autoregressive network model, and use the training set to iteratively train the nonlinear autoregressive network model after parameter selection to obtain an intermediate test model; finally, the computer device can use the test set to verify the performance of the intermediate test model, and use the verified intermediate test model as the liquid level prediction model corresponding to the target event.

[0131] Specifically, when training the liquid level prediction model corresponding to the target event, the computer device can first obtain a data set of historical target events, wherein the target event can include flood events and non-flood events. In order to better capture the precursors and changes when the target event occurs, the present application can select a data set of historical flood events as the data set of historical target events, which includes characteristic data such as precipitation, soil moisture, and water level during the flood. Then the computer device can use a time series cutting method to divide the historical target data into a training set and a test set, and dynamically divide the data set to adapt to model training at different time points.

[0132] After obtaining the training data, the computer device can select a suitable nonlinear autoregressive network model for improvement and training based on the dynamic nonlinear relationship between the liquid level and the characteristic data in the target event, such as the E-NARX structure model. The E-NARX structure model here may include a nonlinear activation function, the number of network layers, etc., which are not limited here. Then the computer device can use the training set to train the selected nonlinear autoregressive network model. During the training process, the computer device can adjust the hyperparameters of the nonlinear autoregressive network model and use the cross-validation method to select the parameters. Among them, the hyperparameter adjustment may include but is not limited to selecting the appropriate lag order and the number of hidden layer neurons.

[0133] Furthermore, after the computer equipment uses the training set to train the intermediate test model, the test set can also be used to verify the performance of the intermediate test model. The training process is consistent with the training process of the intermediate prediction model of drought events, and will not be elaborated here.

[0134] In one embodiment, Figure 5 As shown, Figure 5 A schematic diagram of a flow chart of a liquid level prediction model training process for a depletion flow event provided in an embodiment of the present application; Figure 5 In step S130, determining the liquid level prediction model corresponding to the depletion flow event may include:

[0135] S1331: Obtain historical target data of historical exhaustion flow events, and use a time series cutting method to divide the historical target data into a training set and a test set.

[0136] S1332: Perform iterative training on a preset feedforward neural network model using the training set, and adjust the parameters of the feedforward neural network model using a back propagation algorithm during the training process.

[0137] S1333: When the feedforward neural network model meets the preset training conditions, the trained feedforward neural network model is used as an intermediate prediction model.

[0138] S1334: Use the test set to verify the performance of the intermediate prediction model, and use the verified intermediate prediction model as the liquid level prediction model corresponding to the depletion flow event.

[0139] In this embodiment, when determining the liquid level prediction model corresponding to the depletion flow event, the computer device can first obtain the historical target data of the historical depletion flow event, and use the time series cutting method to divide the historical target data into a training set and a test set, and then use the training set to iteratively train the preset feedforward neural network model, and adjust the parameters of the feedforward neural network model through the back propagation algorithm during the training process. When the feedforward neural network model meets the preset training conditions, the computer device can use the trained feedforward neural network model as an intermediate prediction model; finally, the computer device can use the test set to perform performance verification on the intermediate prediction model, and use the verified intermediate prediction model as the liquid level prediction model corresponding to the depletion flow event.

[0140] Specifically, when training the liquid level prediction model corresponding to the depletion flow event, the computer device can first obtain a data set of historical depletion flow events, which includes characteristic data such as precipitation, flow rate, and water level during the depletion flow. Then, characteristic data related to the depletion flow liquid level prediction can be extracted from the data set, such as water level and flow rate changes at historical moments, and the selected characteristic data can be standardized or normalized to form historical depletion flow data to ensure that the historical depletion flow data has the ability to explain the liquid level changes, so that the model can converge faster and improve the prediction performance. Then the computer device can use the time series cutting method to divide the historical drought data into a training set and a test set.

[0141] After obtaining the training data, the computer device can select a suitable feedforward neural network structure model for improvement and training based on the multi-factor influence relationship between the liquid level and the characteristic data in the depletion flow event. The feedforward neural network structure model here includes parameters such as the number of neuron layers and the number of neurons in each layer. In this application, the FFNN model can be selected as the feedforward neural network structure model. Then the computer device can use the training set to train the selected feedforward neural network structure model. During the training process, the computer device can adjust the weights and biases of the neural network through the back propagation algorithm to minimize the error of the liquid level prediction.

[0142] Furthermore, after the computer equipment uses the training set to train the intermediate prediction model, it can also use the test set to verify the performance of the intermediate prediction model. The training process is consistent with the training process of the intermediate prediction model of drought events, and will not be elaborated here.

[0143] The following is a description of a hydrological event liquid level prediction device provided in an embodiment of the present application. The hydrological event liquid level prediction device described below and the hydrological event liquid level prediction method described above can be referenced to each other.

[0144] In one embodiment, Figure 6 As shown, Figure 6 A schematic diagram of the structure of a hydrological event liquid level prediction device provided in an embodiment of the present application; the present application also provides a hydrological event liquid level prediction device, including a data acquisition module 210, an event classification module 220, a model determination module 230 and a liquid level prediction module 240, specifically including the following:

[0145] The data acquisition module 210 is used to acquire the hydrological event to be predicted and the real-time sequence data of the hydrological event.

[0146] The event classification module 220 is used to classify the real-time sequence data using a decision tree method to obtain event types of hydrological events; event types include drought events, target events, and depletion flow events.

[0147] The model determination module 230 is used to determine the liquid level prediction model corresponding to the event type; wherein the liquid level prediction model for drought events is obtained by training with a nonlinear function regression model; the liquid level prediction model for target events is obtained by training with a nonlinear autoregressive network model; and the liquid level prediction model for depletion flow events is obtained by training with a feedforward neural network model.

[0148] The liquid level prediction module 240 is used to input the real-time sequence data into the liquid level prediction model to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model.

[0149] In the above embodiment, when predicting the liquid level of a hydrological event, the real-time sequence data of the hydrological event can be obtained first to ensure the accuracy of the real-time changes of the data related to the hydrological event and improve the accuracy of the model prediction; then the decision tree method can be used to classify the real-time sequence data to obtain the event type of the hydrological event, and then determine the liquid level prediction model corresponding to the event type, so that different types of liquid level prediction models can be used to improve the accuracy of the description of the complex relationship of the hydrological event in the liquid level prediction process, thereby improving the accuracy of the prediction. Among them, each type of liquid level prediction model is trained by considering the important influencing factors of its corresponding event type, so that the prediction accuracy of the model obtained by training is high. For example, the liquid level prediction model of the drought event can be obtained by training using a nonlinear function regression model; the liquid level prediction model of the target event can be obtained by training using a nonlinear autoregressive network model; and the liquid level prediction model of the depletion flow event can be obtained by training using a feedforward neural network model. After obtaining the liquid level prediction model corresponding to the hydrological event, the present application can input the real-time sequence data of the hydrological event into the liquid level prediction model to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model. The present application uses the liquid level prediction model corresponding to the event type of the hydrological event to predict the acquired real-time sequence data, which can reduce the impact of various factors in the complex and changeable natural system, thereby improving the accuracy of the model prediction.

[0150] In one embodiment, the data acquisition module 210 may include:

[0151] The image acquisition submodule is used to determine multiple satellite images of hydrological events and rainfall data within the rainfall interval.

[0152] The image recognition submodule is used to use image analysis technology to perform regional recognition on each satellite image to obtain the geological features in each satellite image.

[0153] The image analysis submodule is used to sort the satellite images in time series and analyze the geographical features according to the sorting results to obtain the geodynamic data of the hydrological events.

[0154] The data generation submodule is used to generate real-time sequence data of hydrological events based on rainfall data and geodynamic data.

[0155] In one embodiment, the event classification module 220 may include:

[0156] The classification model determination submodule is used to determine the event classification model; the event classification model is trained using a decision tree model.

[0157] The type prediction submodule is used to input the real-time sequence data into the event classification model, obtain the prediction result output by the event classification model, and use the prediction result as the event type of the hydrological event.

[0158] In one embodiment, the classification model determination submodule may include:

[0159] The event data acquisition unit is used to acquire the historical sequence data and real event types of historical hydrological events.

[0160] The event type prediction unit is used to input the historical sequence data into a preset decision tree model to obtain the predicted event type output by the decision tree model.

[0161] The classification model training unit is used to train the decision tree model with the goal of predicting event types close to the actual event types of historical hydrological events.

[0162] The classification model generating unit is used to use the trained decision tree model as an event classification model when the decision tree model meets the preset training conditions.

[0163] In one embodiment, the event data acquisition unit may include:

[0164] The data processing subunit is used to filter out cluster feature data from historical sequence data and perform standardization on the cluster feature data to obtain preprocessed data.

[0165] The cluster analysis subunit is used to perform cluster analysis on the preprocessed data using the K-means clustering algorithm to obtain the cluster center of the preprocessed data.

[0166] The feature analysis subunit is used to perform feature analysis on the cluster centers to obtain the real event types of historical hydrological events.

[0167] In one embodiment, the model determination module 230 may include:

[0168] The first data division submodule is used to obtain historical drought data of historical drought events and divide the historical drought data into a training set and a test set.

[0169] The first model training submodule is used to train a preset nonlinear function regression model using a training set, and after the nonlinear function regression model meets preset training conditions, use the trained nonlinear function regression model as an intermediate prediction model.

[0170] The first model prediction submodule is used to input the test set into the intermediate prediction model to obtain the predicted liquid level output by the intermediate prediction model.

[0171] The first model generation submodule is used to use the intermediate prediction model as the liquid level prediction model corresponding to the drought event when the difference between the predicted liquid level and the actual liquid level in the historical drought data is less than a preset error.

[0172] In one embodiment, the model determination module 230 may further include:

[0173] The second data division submodule is used to obtain historical target data of historical target events and divide the historical target data into a training set and a test set using a time series cutting method.

[0174] The second model training submodule is used to select parameters for a preset nonlinear autoregressive network model using a cross-validation method, and iteratively train the nonlinear autoregressive network model after parameter selection using a training set to obtain an intermediate test model.

[0175] The second model generation submodule is used to verify the performance of the intermediate test model using the test set, and use the verified intermediate test model as the liquid level prediction model corresponding to the target event.

[0176] In one embodiment, the model determination module 230 may further include:

[0177] The third data partitioning submodule is used to obtain historical target data of historical depletion flow events, and divide the historical target data into a training set and a test set using a time series cutting method.

[0178] The third data training submodule is used to perform iterative training on a preset feedforward neural network model using a training set, and to adjust the parameters of the feedforward neural network model through a back propagation algorithm during the training process.

[0179] The third intermediate model determination submodule is used to use the trained feedforward neural network model as the intermediate prediction model when the feedforward neural network model meets the preset training conditions.

[0180] The third model generation submodule is used to verify the performance of the intermediate prediction model using the test set, and use the verified intermediate prediction model as the liquid level prediction model corresponding to the depletion flow event.

[0181] In one embodiment, the present application also provides a storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the hydrological event liquid level prediction method as described in any of the above embodiments.

[0182] In one embodiment, the present application also provides a computer device having computer-readable instructions stored therein. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the method for predicting the liquid level of a hydrological event as described in any one of the above embodiments.

[0183] Indicatively, Figure 7 As shown, Figure 7 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. The computer device 300 may be provided as a server. Figure 7 The computer device 300 includes a processing component 302, which further includes one or more processors, and a memory resource represented by a memory 301 for storing instructions executable by the processing component 302, such as an application. The application stored in the memory 301 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 302 is configured to execute instructions to perform the hydrological event liquid level prediction method of any of the above embodiments.

[0184] The computer device 300 may further include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate based on an operating system stored in the memory 301, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM, or the like.

[0185] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0186] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0187] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0188] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the liquid level of a hydrological event, characterized in that: The method comprises: Acquire the hydrological event to be predicted and the real-time sequence data of the hydrological event; The real-time sequence data is classified by using a decision tree method to obtain the event type of the hydrological event; the event type includes drought events, target events and depletion flow events; the target event indicates that the liquid level changes due to wet weather in an urban area within a certain period of time; the depletion flow event indicates that the liquid level rises due to dry weather in an urban area within a certain period of time; Determine the liquid level prediction model corresponding to the event type; wherein the liquid level prediction model for the drought event is obtained by training with a nonlinear function regression model; the liquid level prediction model for the target event is obtained by training with a nonlinear autoregressive network model; the liquid level prediction model for the depletion flow event is obtained by training with a feedforward neural network model; The real-time sequence data is input into the liquid level prediction model to obtain a liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model.

2. The method for predicting the liquid level of a hydrological event according to claim 1, characterized in that: Acquiring real-time serial data of the hydrological event, including: determining a plurality of satellite images of the hydrological event and rainfall data within a rainfall interval; Use image analysis technology to identify regions in each satellite image and obtain the geological features in each satellite image; Sorting each satellite image according to the time series, and analyzing each geographical feature according to the sorting result to obtain the geodynamic data of the hydrological event; Real-time sequence data of the hydrological event is generated according to the rainfall data and the geoscientific dynamic data.

3. The method for predicting the liquid level of a hydrological event according to claim 1, characterized in that: The method of using a decision tree method to classify the real-time sequence data to obtain the event type of the hydrological event includes: Determine an event classification model; the event classification model is obtained by training a decision tree model; The real-time sequence data is input into the event classification model to obtain a prediction result output by the event classification model, and the prediction result is used as the event type of the hydrological event.

4. The method for predicting the liquid level of a hydrological event according to claim 3, characterized in that: The determining of the event classification model comprises: Obtain historical sequence data and real event types of historical hydrological events; Inputting the historical sequence data into a preset decision tree model to obtain the predicted event type output by the decision tree model; Training the decision tree model with the goal of making the predicted event type close to the real event type of the historical hydrological event; When the decision tree model meets the preset training conditions, the trained decision tree model is used as an event classification model.

5. The method for predicting the liquid level of a hydrological event according to claim 4, characterized in that: The real event types of the historical hydrological events include: Filtering cluster feature data from the historical sequence data, and performing standardization processing on the cluster feature data to obtain preprocessed data; Performing cluster analysis on the preprocessed data using a K-means clustering algorithm to obtain a cluster center of the preprocessed data; Performing feature analysis on the cluster centers, and obtaining the real event types of the historical hydrological events.

6. The method for predicting the liquid level of a hydrological event according to claim 1, characterized in that: The step of determining a liquid level prediction model corresponding to the drought event comprises: Acquire historical drought data of historical drought events, and divide the historical drought data into a training set and a test set; Using the training set to train a preset nonlinear function regression model, and after the nonlinear function regression model meets the preset training conditions, using the trained nonlinear function regression model as an intermediate prediction model; Inputting the test set into the intermediate prediction model to obtain the predicted liquid level output by the intermediate prediction model; When the difference between the predicted liquid level and the actual liquid level in the historical drought data is less than a preset error, the intermediate prediction model is used as the liquid level prediction model corresponding to the drought event.

7. The method for predicting the liquid level of a hydrological event according to claim 1, characterized in that: The step of determining the liquid level prediction model corresponding to the target event includes: Acquire historical target data of historical target events, and divide the historical target data into a training set and a test set using a time series cutting method; A cross-validation method is used to select parameters for a preset nonlinear autoregressive network model, and the training set is used to iteratively train the nonlinear autoregressive network model after parameter selection to obtain an intermediate test model; The test set is used to perform performance verification on the intermediate test model, and the intermediate test model that passes the verification is used as the liquid level prediction model corresponding to the target event.

8. The method for predicting the liquid level of a hydrological event according to claim 1, characterized in that: Determining the liquid level prediction model corresponding to the depletion flow event includes: Acquire historical target data of historical depletion flow events, and divide the historical target data into a training set and a test set using a time series cutting method; Iteratively training a preset feedforward neural network model using the training set, and adjusting the parameters of the feedforward neural network model using a back propagation algorithm during the training process; When the feedforward neural network model meets the preset training conditions, the trained feedforward neural network model is used as an intermediate prediction model; The test set is used to verify the performance of the intermediate prediction model, and the verified intermediate prediction model is used as the liquid level prediction model corresponding to the depletion flow event.

9. A hydrological event liquid level prediction device, characterized in that: include: A data acquisition module, used to acquire the hydrological event to be predicted and the real-time sequence data of the hydrological event; An event classification module, used to classify the real-time sequence data using a decision tree method to obtain the event type of the hydrological event; the event type includes drought events, target events and depletion flow events; The target event indicates that the liquid level changes due to wet weather in a certain period of time in the urban area; the depletion flow event indicates that the liquid level rises due to dry weather in a certain period of time in the urban area; A model determination module is used to determine the liquid level prediction model corresponding to the event type; wherein the liquid level prediction model for the drought event is obtained by training with a nonlinear function regression model; the liquid level prediction model for the target event is obtained by training with a nonlinear autoregressive network model; and the liquid level prediction model for the depletion flow event is obtained by training with a feedforward neural network model; The liquid level prediction module is used to input the real-time sequence data into the liquid level prediction model to obtain the liquid level prediction value corresponding to the hydrological event output by the liquid level prediction model.

10. A storage medium, characterized in that: The storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the method for predicting the liquid level of a hydrological event as described in any one of claims 1 to 8.

11. A computer device, characterized in that: include: one or more processors, and memory; The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the method for predicting the liquid level of a hydrological event as described in any one of claims 1 to 8 are performed.

Images