Intelligent water conservancy flood prevention and water retaining monitoring and early warning method and system

Abstract

The invention discloses an intelligent water conservancy flood prevention and water retaining monitoring and early warning method and system, and belongs to the technical field of water conservancy, and the method comprises the steps: obtaining hydro meteorological data and video image data in real time, and carrying out the data preprocessing; processing the preprocessed hydro meteorological data and video image data through a water level prediction model constructed based on a long short-term memory network to obtain water level prediction data representing a future water level change trend; performing conjoint analysis on the water level real-time data and the water level prediction data through a machine learning algorithm, and identifying water level abnormal fluctuation characteristics; and processing the water level abnormal fluctuation characteristics based on a preset multi-level early warning threshold judgment rule, and generating early warning information of a corresponding level. According to the technical scheme, the technical means of multi-source heterogeneous data, deep learning water level prediction, machine learning anomaly recognition, multi-level early warning and the like are creatively fused, the technical bottleneck of traditional water conservancy flood prevention monitoring and early warning is broken through, and powerful technical support is provided for flood disaster prevention.

Description

Technical Field

[0001] This invention relates to the field of water conservancy and flood control technology, and in particular to an intelligent water conservancy flood control and water retention monitoring and early warning method and system. Background Technology

[0002] With the intensification of global climate change and the increasing frequency of extreme weather events, floods pose a serious threat to people's lives and property, as well as social and economic stability. Currently, water conservancy projects typically use flood control and water-retaining devices to block floods and prevent them from rushing downstream during the flood season. However, the water-retaining capacity of these devices is limited, which is closely related to their own operational status and water level conditions.

[0003] Traditional flood control monitoring methods mainly rely on manual inspections and decentralized, simple water level observation stations, which suffer from problems such as low monitoring efficiency, poor data real-time performance, limited coverage, and insufficient ability to predict flood trends. Manual inspections are constrained by factors such as terrain, weather, and manpower allocation, making it difficult to achieve uninterrupted monitoring of long-distance water conservancy projects. Most simple water level observation stations only have single-point water level data acquisition capabilities, lacking synchronous monitoring of related factors such as flow rate, velocity, and rainfall. Furthermore, data transmission relies on wired communication lines, which are vulnerable to damage from natural disasters, failing to meet the modern requirements of flood control for timeliness, comprehensiveness, and foresight.

[0004] In recent years, although some water conservancy projects have introduced automated monitoring equipment, most existing systems focus on the collection of data for single hydrological elements. The monitoring equipment operates independently and has not formed an organic data fusion system. For example, some water level monitoring stations can only provide real-time water level values ​​and cannot analyze the correlation between water level changes and causal factors such as rainfall and increased flow. It is difficult to uncover the hidden flood development patterns behind the data, and therefore cannot meet the needs of modern water conservancy for accurate forecasting in flood control and water retention. Summary of the Invention

[0005] In view of the problems that existing technologies for monitoring and early warning of flood control and water retention rely on manual labor and lack data fusion and analysis, the purpose of this invention is to provide an intelligent method for monitoring and early warning of flood control and water retention, so as to at least partially solve the above problems.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] In a first aspect, the present invention provides an intelligent water conservancy flood control and water retention monitoring and early warning method, the method comprising the following steps:

[0008] Real-time acquisition of hydrological and meteorological data and video image data, followed by data preprocessing;

[0009] By processing the preprocessed hydrological and meteorological data and video image data using a water level prediction model built on a long short-term memory network, water level prediction data characterizing future water level change trends can be obtained.

[0010] The real-time water level data and the predicted water level data are jointly analyzed by machine learning algorithms to identify abnormal water level fluctuation characteristics.

[0011] Based on the water level prediction data and the abnormal water level fluctuation characteristics, corresponding level of early warning information is generated through preset multi-level early warning threshold determination rules.

[0012] In some preferred embodiments, the hydrological and meteorological data includes real-time water level data, real-time flow data, real-time flow velocity data, real-time rainfall data, and real-time temperature data, and the video image data includes video image data of water conservancy facility operation and video image data of surrounding flooding.

[0013] In some preferred embodiments, the data preprocessing includes filtering, normalization, and missing value imputation.

[0014] In some preferred embodiments, the water level prediction model constructed based on the long short-term memory network includes an input layer, several hidden layers, and an output layer. The number of neurons in the hidden layer is optimized and determined based on historical data characteristics and prediction accuracy requirements. The activation function is a variant function that can alleviate the gradient vanishing problem to improve the model's ability to capture long-term dependencies.

[0015] In some preferred embodiments, the machine learning algorithm integrates random forest and support vector machine algorithms. Random forest is used to rank the importance of data features and preliminarily screen abnormal feature dimensions. Support vector machine constructs a hyperplane based on the screened features to divide normal and abnormal hydrological states, thereby synergistically improving the accuracy and recall of anomaly identification.

[0016] In some preferred embodiments, the multi-level early warning threshold determination rule includes four levels divided in order of risk severity from low to high: blue warning, yellow warning, orange warning and red warning. Each level corresponds to a preset early warning threshold range, which is determined by comprehensively considering the flood control standards of water conservancy projects, the extent of historical flood inundation, the elevation of the surrounding terrain and real-time weather forecast information.

[0017] In some preferred embodiments, the multi-level early warning threshold determination rule dynamically adjusts the early warning threshold range based on preset emergency situations.

[0018] In some preferred embodiments, the warning information includes, but is not limited to, warning level, time of occurrence, specific location, predicted water level change curve, and expected inundation range. The warning information is stored in a database in a structured data format and supports querying and tracing historical warning information.

[0019] Secondly, the present invention also provides an intelligent water conservancy flood control and water retention monitoring and early warning system, the system comprising:

[0020] The data acquisition module is used to acquire hydrological and meteorological data and video image data in real time.

[0021] The data preprocessing module is used to preprocess the hydrological and meteorological data and the video image data;

[0022] The water level prediction module processes the preprocessed hydrological and meteorological data and the video image data through a water level prediction model built on a long short-term memory network to obtain water level prediction data that characterizes the future water level change trend.

[0023] An anomaly identification module is used to jointly analyze real-time water level data and predicted water level data using machine learning algorithms to identify abnormal water level fluctuation characteristics.

[0024] And a multi-level early warning module, which is used to generate early warning information of corresponding levels based on water level prediction data and abnormal water level fluctuation characteristics, through preset multi-level early warning threshold judgment rules.

[0025] In some preferred embodiments, the system further includes a dynamic adjustment module, which is used to dynamically adjust the warning threshold ranges at each level based on preset emergency situations.

[0026] The beneficial effects of the present invention by adopting the above technical solution are as follows: The technical solution of the present invention innovatively integrates multi-source heterogeneous data, deep learning water level prediction, machine learning anomaly identification and multi-level early warning technologies, breaking through the technical bottleneck of traditional water conservancy flood control monitoring and early warning, and providing strong technical support for flood disaster prevention. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the intelligent water conservancy flood control and water retention monitoring and early warning method in Embodiment 1 of the present invention;

[0028] Figure 2 This is a schematic diagram of the intelligent water conservancy flood control and water retention monitoring and early warning system in Embodiment 2 of the present invention. Detailed Implementation

[0029] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0030] Example 1

[0031] See Figure 1 This is a flowchart illustrating an intelligent water conservancy flood control and water retention monitoring and early warning method provided in an embodiment of the present invention. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose any limitations on this.

[0032] An intelligent water conservancy flood control and water retention monitoring and early warning method, such as Figure 1 As shown, it includes four steps, from step S1 to step S4.

[0033] Step S1. Acquire hydrological and meteorological data and video image data in real time and perform data preprocessing.

[0034] Step S2. Process the preprocessed hydrological and meteorological data and video image data using a water level prediction model built on a long short-term memory network to obtain water level prediction data that characterizes the future water level change trend.

[0035] Step S3. Use machine learning algorithms to jointly analyze real-time water level data and predicted water level data to identify abnormal water level fluctuation characteristics.

[0036] Step S4. Based on the water level prediction data and the characteristics of abnormal water level fluctuations, generate corresponding level of early warning information through the preset multi-level early warning threshold judgment rules.

[0037] In step S1, the hydrological and meteorological data includes real-time water level data, real-time flow data, real-time flow velocity data, real-time rainfall data, and real-time temperature data. These data are collected by corresponding sensors, with at least two of each type of sensor deployed for mutual backup. The video image data includes video image data of the operation of water conservancy facilities and video image data of surrounding flooding.

[0038] For example, high-precision ultrasonic water level gauges can be used to collect real-time water level data. These gauges utilize the time difference of ultrasonic waves propagation in air to accurately measure water level changes in a non-contact manner. They feature a wide measurement range, millimeter-level accuracy, and strong anti-interference capabilities, allowing for real-time detection of subtle fluctuations in water levels at hydraulic engineering sites. Doppler flow meters can collect real-time flow and velocity data. These meters accurately calculate water velocity and flow rate by transmitting and receiving Doppler frequency shift signals. Their measurement accuracy is unaffected by water color or impurity content, making them adaptable to complex hydrological environments. High-sensitivity rain gauges can collect real-time rainfall data. Based on tipping bucket or weighing measurement principles, these sensors accurately measure rainfall intensity and cumulative rainfall with a resolution down to 0.1 mm, providing crucial meteorological input for flood control water level prediction. Finally, temperature and humidity sensors can collect real-time temperature data, monitoring changes in ambient temperature and humidity in real time and assisting in the analysis of the impact of meteorological conditions on evaporation and seepage. By using a high-resolution camera equipped with a zoom lens and infrared night vision, video image data is acquired, thereby capturing the operational status of on-site water conservancy facilities and the surrounding flood situation from all angles, providing visual evidence for intuitive judgment of the flood situation.

[0039] In step S1, the data preprocessing includes filtering, normalization, and missing value imputation.

[0040] To address the issues of noise interference, dimensional differences, and occasional missing values ​​in the hydrological and meteorological data collected by sensors, this embodiment specifically employs the Kalman filter algorithm to smoothly filter out random noise, uses the max-min normalization method to uniformly map different types of data to the [0,1] interval, and utilizes a time series interpolation algorithm to fill in missing data points, ensuring the integrity, accuracy, and consistency of the data input to the subsequent water level prediction model, and improving the operational stability of the subsequent water level prediction model.

[0041] To address issues such as noise interference, insufficient contrast, and unclear target objects in video image data, this embodiment employs image enhancement methods for preprocessing, including illumination compensation and dehazing. Illumination compensation can be achieved through histogram equalization and adaptive histogram equalization (CLAHE), adjusting the contrast and brightness distribution of the video image to maintain good visibility under different lighting conditions. For dehazing, dehazing algorithms (such as dark channel prior-based dehazing methods) can be used to estimate the transmittance and atmospheric light of the video image, recovering a clear video image to improve the recognition of target objects.

[0042] In addition to the above, preprocessing of video images also includes target detection and recognition, thereby detecting targets related to the operational status of water conservancy facilities and the surrounding flood situation, such as the detection and identification of the operational status of flood control and water-blocking devices. Specifically, deep learning-based target detection algorithms (such as YOLO, Faster R-CNN, etc.) can be used. These target detection algorithms can quickly and accurately identify target objects in video image data and provide their location and category information.

[0043] Data preprocessing also includes multi-source data fusion, which aims to combine video image data with hydrological and meteorological data. This provides more comprehensive information and reliable data support for subsequent water level prediction models. For example, combining the location of detected floating objects with water level data can better assess flood risk. Multi-source data fusion can be achieved by establishing a joint coordinate system or time synchronization methods to ensure accurate correspondence between information from different data sources.

[0044] In step S2, the water level prediction model built based on a Long Short-Term Memory (LSTM) network includes an input layer, several hidden layers, and an output layer. The number of neurons in the hidden layers is optimized and determined based on historical data characteristics and prediction accuracy requirements. The activation function is a variant function that can alleviate the gradient vanishing problem to improve the model's ability to capture long-term dependencies.

[0045] LSTM networks effectively address the vanishing and exploding gradient problems of traditional recurrent neural networks (RNNs) when processing long sequence data through a gating mechanism. The input layer receives preprocessed hydrological and meteorological data and video image data. Multiple hidden layer neurons extract deep spatiotemporal features from the input data layer by layer, and the output layer provides a predicted water level for a preset time period (flexibly set to hourly, daily, etc.). In terms of model construction details, extensive experimental verification and parameter tuning have determined the configuration of the number of hidden layer neurons. A variant of the activation function that can adaptively adjust the gradient update amplitude is selected, such as a modified combination of the sigmoid function with a forgetting gate and the tanh function for both the input and output gates. This strengthens the hydrological prediction model's ability to remember long-term dependencies in complex hydrological processes, ensuring that the water level prediction results closely match the actual dynamic changes in water levels, providing a quantitative reference for flood control decisions in advance. The training process of the water level prediction model based on Long Short-Term Memory (LSTM) networks is common knowledge and will not be elaborated upon in this embodiment.

[0046] In step S3, the machine learning algorithm integrates random forest and support vector machine algorithms. Random forest is used to rank the importance of data features and initially screen abnormal feature dimensions. Support vector machine constructs a hyperplane based on the screened features to divide normal and abnormal hydrological states, thus synergistically improving the accuracy and recall of anomaly identification.

[0047] The Random Forest algorithm first performs bootstrap sampling and random feature selection on massive historical monitoring data, constructing multiple decision trees to form a forest. Each decision tree independently votes to output the feature importance ranking, quickly identifying key feature dimensions sensitive to flood risk, such as a sudden increase in rainfall or a sharp rise in upstream flow. The Support Vector Machine (SVM) algorithm takes over the selected key features and, based on the soft margin maximization principle, constructs an optimal hyperplane in the high-dimensional feature space to accurately delineate the boundary between normal and abnormal hydrological states. This enables efficient judgment of abnormal fluctuations in real-time monitoring data. The two algorithms complement each other, leveraging the advantages of Random Forest in feature mining and the strengths of SVM in small sample and nonlinear classification, significantly improving the accuracy and recall of anomaly identification and timely capturing early signals of flood risk.

[0048] By setting step S3, the method provided in this embodiment can promptly detect abnormal fluctuations in real-time water level data, as well as abnormal fluctuations in the difference between real-time water level data and predicted water level data. The former indicates that the current real-time water level data is abnormal compared to previous periods and requires extra attention. The latter indicates that the predicted future water level data may rise rapidly in a short period of time, and even if it does not reach the upper limit of the water level, it still requires extra attention. These situations requiring extra attention will be reflected through the multi-level early warning threshold judgment rules in step S4, and relevant personnel will be reminded to pay attention by providing corresponding level of early warning information.

[0049] It should be noted that the abnormal water level fluctuation characteristics identified in step S3 are not the only basis for early warning. Early warning information will also be triggered when the water level prediction data is high.

[0050] In step S4, the multi-level warning threshold determination rule includes four levels divided in order of risk severity from low to high (in other preferred embodiments, other levels may also be set), namely blue warning, yellow warning, orange warning and red warning. Each level corresponds to a preset warning threshold range, which is determined by comprehensively considering the flood control standards of water conservancy projects, the historical flood inundation range, the surrounding terrain elevation and real-time weather forecast information.

[0051] The water level prediction data in step S2 and the abnormal water level fluctuation characteristics identified in step S3 will be matched with different early warning threshold ranges. Of course, the water level prediction data in step S2 may not have a corresponding early warning threshold range, which means that no early warning is needed, and step S3 may not output the abnormal water level fluctuation characteristics.

[0052] It's easy to understand that the warning threshold ranges in step S4 are not static; they are usually dynamically adjusted based on preset emergency situations. For example, if a water conservancy project is damaged, it indicates a decrease in its water-blocking capacity, so the warning threshold ranges at all levels need to be lowered accordingly. Similarly, if there is a short-term surge in upstream rainfall, it indicates that the local water conservancy project will face significantly increased flood control and water-blocking pressure in the future, so the warning threshold ranges at all levels also need to be lowered accordingly to remind relevant personnel to respond in a timely manner.

[0053] In this embodiment, the early warning information includes, but is not limited to, dimensions such as early warning level, occurrence time, specific location, predicted water level change curve, and expected inundation range. The early warning information is stored in a database in a structured data format to facilitate the query and retrieval of historical early warning information.

[0054] Example 2

[0055] This embodiment provides an intelligent water conservancy flood control and water retention monitoring and early warning system, such as Figure 2 As shown, the system includes a data acquisition module, a data preprocessing module, a water level prediction module, an anomaly identification module, a multi-level early warning module, and a dynamic adjustment module.

[0056] The data acquisition module is used to acquire hydrological and meteorological data and video image data in real time.

[0057] The data preprocessing module is used to preprocess hydrological and meteorological data and video image data.

[0058] The water level prediction module processes preprocessed hydrological and meteorological data and video image data through a water level prediction model built on a long short-term memory network to obtain water level prediction data that characterizes the future water level change trend.

[0059] The anomaly identification module is used to jointly analyze real-time water level data and water level prediction data through machine learning algorithms to identify abnormal water level fluctuation characteristics.

[0060] The multi-level early warning module is used to generate early warning information of corresponding levels based on water level prediction data and abnormal water level fluctuation characteristics, and through preset multi-level early warning threshold judgment rules.

[0061] In this embodiment, the hydrological and meteorological data includes real-time water level data, real-time flow rate data, real-time flow velocity data, real-time rainfall data, and real-time temperature data. These data are collected by corresponding sensors, with at least two of each type of sensor deployed for mutual backup. The video image data includes video image data of the operation of water conservancy facilities and video image data of surrounding flooding.

[0062] For example, high-precision ultrasonic water level gauges can be used to collect real-time water level data. These gauges utilize the time difference of ultrasonic waves propagation in air to accurately measure water level changes in a non-contact manner. They feature a wide measurement range, millimeter-level accuracy, and strong anti-interference capabilities, allowing for real-time detection of subtle fluctuations in water levels at hydraulic engineering sites. Doppler flow meters can collect real-time flow and velocity data. These meters accurately calculate water velocity and flow rate by transmitting and receiving Doppler frequency shift signals. Their measurement accuracy is unaffected by water color or impurity content, making them adaptable to complex hydrological environments. High-sensitivity rain gauges can collect real-time rainfall data. Based on tipping bucket or weighing measurement principles, these sensors accurately measure rainfall intensity and cumulative rainfall with a resolution down to 0.1 mm, providing crucial meteorological input for flood control water level prediction. Finally, temperature and humidity sensors can collect real-time temperature data, monitoring changes in ambient temperature and humidity in real time and assisting in the analysis of the impact of meteorological conditions on evaporation and seepage. By using a high-resolution camera equipped with a zoom lens and infrared night vision, video image data is acquired, thereby capturing the operational status of on-site water conservancy facilities and the surrounding flood situation from all angles, providing visual evidence for intuitive judgment of the flood situation.

[0063] The data acquisition module is directly connected to each sensor and high-resolution camera to acquire hydrological and meteorological data and video image data.

[0064] To address the issues of noise interference, dimensional differences, and occasional missing values ​​in the hydrological and meteorological data collected by sensors, this embodiment employs a Kalman filter algorithm to smoothly filter out random noise, uses a maximum-minimum normalization method to uniformly map different types of data to the [0,1] interval, and utilizes a time series interpolation algorithm to fill in missing data points. This ensures the integrity, accuracy, and consistency of the data input to the subsequent water level prediction model, thereby improving the operational stability of the subsequent water level prediction model.

[0065] To address issues such as noise interference, insufficient contrast, and unclear target objects in video image data, this embodiment employs image enhancement methods, such as illumination compensation and dehazing, in the data preprocessing module. Illumination compensation can be achieved through histogram equalization (HEM) and adaptive histogram equalization (CLAHE), adjusting the contrast and brightness distribution of the video image to maintain good visibility under various lighting conditions. Dehazing can be performed using algorithms (such as dark channel prior-based dehazing methods) to estimate the transmittance and atmospheric light of the video image, recovering a clearer image and improving the recognizability of target objects.

[0066] In addition, the data preprocessing module also performs target detection and recognition on video images to detect targets related to the operation status of water conservancy facilities and the surrounding flood situation. Specifically, it can use deep learning-based target detection algorithms (such as YOLO, Faster R-CNN, etc.). These target detection algorithms can quickly and accurately identify target objects in video image data and provide their location and category information.

[0067] Data preprocessing also includes multi-source data fusion, which aims to combine video image data with hydrological and meteorological data. This provides more comprehensive information and reliable data support for subsequent water level prediction models. For example, combining the location of detected floating objects with water level data can better assess flood risk. Multi-source data fusion can be achieved by establishing a joint coordinate system or time synchronization methods to ensure accurate correspondence between information from different data sources.

[0068] The water level prediction module uses a water level prediction model built on a Long Short-Term Memory (LSTM) network to predict water levels. This model includes an input layer, several hidden layers, and an output layer. The number of neurons in the hidden layers is optimized and determined based on historical data characteristics and prediction accuracy requirements. The activation function is a variant function that can alleviate the gradient vanishing problem to improve the model's ability to capture long-term dependencies.

[0069] LSTM networks effectively address the vanishing and exploding gradient problems of traditional recurrent neural networks (RNNs) when processing long sequence data through a gating mechanism. The input layer receives preprocessed hydrological and meteorological data and video image data. Multiple hidden layer neurons extract deep spatiotemporal features from the input data layer by layer, and the output layer provides a predicted water level for a preset time period (flexibly set to hourly, daily, etc.). In terms of model construction details, extensive experimental verification and parameter tuning have determined the configuration of the number of hidden layer neurons. A variant of the activation function that can adaptively adjust the gradient update amplitude is selected, such as a modified combination of the sigmoid function with a forgetting gate and the tanh function for both the input and output gates. This strengthens the hydrological prediction model's ability to remember long-term dependencies in complex hydrological processes, ensuring that the water level prediction results closely match the actual dynamic changes in water levels, providing a quantitative reference for flood control decisions in advance. The training process of the water level prediction model based on Long Short-Term Memory (LSTM) networks is common knowledge and will not be elaborated upon in this embodiment.

[0070] The machine learning algorithm used in the anomaly detection module combines random forest and support vector machine algorithms. Random forest is used to rank the importance of data features and initially screen the dimensions of abnormal features. Support vector machine constructs a hyperplane based on the screened features to divide normal and abnormal hydrological states, thus synergistically improving the accuracy and recall of anomaly detection.

[0071] The Random Forest algorithm first performs bootstrap sampling and random feature selection on massive historical monitoring data, constructing multiple decision trees to form a forest. Each decision tree independently votes to output the feature importance ranking, quickly identifying key feature dimensions sensitive to flood risk, such as a sudden increase in rainfall or a sharp rise in upstream flow. The Support Vector Machine (SVM) algorithm takes over the selected key features and, based on the soft margin maximization principle, constructs an optimal hyperplane in the high-dimensional feature space to accurately delineate the boundary between normal and abnormal hydrological states. This enables efficient judgment of abnormal fluctuations in real-time monitoring data. The two algorithms complement each other, leveraging the advantages of Random Forest in feature mining and the strengths of SVM in small sample and nonlinear classification, significantly improving the accuracy and recall of anomaly identification and timely capturing early signals of flood risk.

[0072] By setting up an anomaly detection module, the system provided in this embodiment can promptly detect abnormal fluctuations in real-time water level data, as well as abnormal fluctuations in the difference between real-time water level data and predicted water level data. The former indicates that the current real-time water level data is abnormal compared to previous periods and requires extra attention, while the latter indicates that the predicted future water level data may rise rapidly in a short period of time, and even if it does not reach the upper limit of the water level, it still requires extra attention. These are all reflected through a multi-level early warning module, which provides corresponding levels of early warning information to remind relevant personnel to pay attention.

[0073] The multi-level early warning module uses a multi-level early warning threshold determination rule that includes four levels divided in order of risk severity from low to high (other levels may be set in other preferred embodiments), namely blue warning, yellow warning, orange warning and red warning. Each level corresponds to a preset early warning threshold range, which is determined by comprehensively considering the flood control standards of water conservancy projects, the extent of historical flood inundation, the elevation of the surrounding terrain and real-time weather forecast information.

[0074] The water level prediction module and the anomaly detection module will each match different warning threshold ranges for the predicted water level data and the anomaly detection module. Of course, the water level prediction module may also not have a corresponding warning threshold range, which means that no warning is needed. Correspondingly, the anomaly detection module may not have detected any anomaly water level fluctuations.

[0075] It is easy to understand that the warning threshold range in the multi-level warning module is not static. Typically, the system provided in this embodiment also includes a dynamic adjustment module, which is used to dynamically adjust the warning threshold range at each level based on preset emergency situations.

[0076] For example, if a water conservancy project is damaged, it indicates that its water-blocking capacity has decreased, so the warning threshold ranges at all levels need to be lowered accordingly. Or, if there is a short-term surge in upstream rainfall, it indicates that the local water conservancy project will face a significant increase in flood control and water-blocking pressure in the future, so the warning threshold ranges at all levels also need to be lowered accordingly to remind relevant personnel to respond in a timely manner.

[0077] In this embodiment, the early warning information generated by the multi-level early warning module includes, but is not limited to, dimensions such as early warning level, occurrence time, specific location, predicted water level change curve, and expected inundation range. The early warning information is stored in the database in a structured data format to facilitate the query and retrospection of historical early warning information.

[0078] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "front", "rear", etc., indicate the orientation or positional relationship based on the description of the structure of this invention shown in the accompanying drawings. They are only for the convenience of describing this invention and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0079] The terms "first" and "second" in this technical solution are merely designations for corresponding structures that are identical or similar, or that perform similar functions. They do not represent an arrangement of the importance of these structures, nor do they imply any ranking, comparison of size, or other meaning.

[0080] Furthermore, unless otherwise explicitly specified and limited, the terms "installation" and "connection" should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two structures. Those skilled in the art can understand the specific meaning of the above terms in this invention by considering the overall concept of the invention and the specific context of the solution.

[0081] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. An intelligent water conservancy flood control and water retention monitoring and early warning method, characterized in that: The method includes the following steps: Real-time acquisition of hydrological and meteorological data and video image data, followed by data preprocessing; By processing the preprocessed hydrological and meteorological data and video image data using a water level prediction model built on a long short-term memory network, water level prediction data characterizing future water level change trends can be obtained. The real-time water level data and the predicted water level data are jointly analyzed by machine learning algorithms to identify abnormal water level fluctuation characteristics. Based on the water level prediction data and the abnormal water level fluctuation characteristics, corresponding level of early warning information is generated through preset multi-level early warning threshold determination rules.

2. The intelligent water conservancy flood control and water retention monitoring and early warning method according to claim 1, characterized in that: The hydrological and meteorological data includes real-time water level data, real-time flow data, real-time flow velocity data, real-time rainfall data, and real-time temperature data. The video image data includes video image data of water conservancy facility operation and video image data of surrounding flooding.

3. The intelligent water conservancy flood control and water retention monitoring and early warning method according to claim 1, characterized in that: The data preprocessing includes filtering, normalization, and missing value imputation.

4. The intelligent water conservancy flood control and water retention monitoring and early warning method according to claim 1, characterized in that: The water level prediction model based on the Long Short-Term Memory network includes an input layer, several hidden layers, and an output layer. The number of neurons in the hidden layer is optimized and determined based on historical data characteristics and prediction accuracy requirements. The activation function is a variant function that can alleviate the gradient vanishing problem to improve the model's ability to capture long-term dependencies.

5. The intelligent water conservancy flood control and water retention monitoring and early warning method according to claim 1, characterized in that: The machine learning algorithm integrates random forest and support vector machine algorithms. Random forest is used to rank the importance of data features and initially screen abnormal feature dimensions. Support vector machine constructs a hyperplane based on the screened features to divide normal and abnormal hydrological states, thus synergistically improving the accuracy and recall of anomaly identification.

6. The intelligent water conservancy flood control and water retention monitoring and early warning method according to claim 1, characterized in that: The multi-level early warning threshold determination rule includes four levels divided in order of risk severity from low to high: blue warning, yellow warning, orange warning, and red warning. Each level corresponds to a preset early warning threshold range, which is determined by comprehensively considering the flood control standards of water conservancy projects, the extent of historical flood inundation, the elevation of the surrounding terrain, and real-time meteorological forecast information.

7. The intelligent water conservancy flood control and water retention monitoring and early warning method according to claim 5, characterized in that: The multi-level early warning threshold determination rule dynamically adjusts the early warning threshold range based on preset emergency situations.

8. The intelligent water conservancy flood control and water retention monitoring and early warning method according to claim 1, characterized in that: The early warning information includes, but is not limited to, early warning level, time of occurrence, specific location, predicted water level change curve, and expected inundation range. The early warning information is stored in a database in a structured data format and supports querying and tracing back historical early warning information.

9. An intelligent water conservancy flood control and water retention monitoring and early warning system, characterized in that: The system includes: The data acquisition module is used to acquire hydrological and meteorological data and video image data in real time. The data preprocessing module is used to preprocess the hydrological and meteorological data and the video image data; The water level prediction module processes the preprocessed hydrological and meteorological data and the video image data through a water level prediction model built on a long short-term memory network to obtain water level prediction data that characterizes the future water level change trend. An anomaly identification module is used to jointly analyze real-time water level data and predicted water level data using machine learning algorithms to identify abnormal water level fluctuation characteristics. And a multi-level early warning module, which is used to generate early warning information of corresponding levels based on water level prediction data and abnormal water level fluctuation characteristics, through preset multi-level early warning threshold judgment rules.

10. The intelligent water conservancy flood control and water retention monitoring and early warning system according to claim 9, characterized in that: The system also includes a dynamic adjustment module, which is used to dynamically adjust the warning threshold ranges at each level based on preset emergency situations.

Images