A Deep Learning-Based Lake Hydrological Monitoring Method and System

Abstract

This invention relates to the field of hydrological monitoring technology and provides a deep learning-based method and system for lake hydrological monitoring. The method includes the following steps: classifying historical hydrological data based on hydrological anomaly types; determining strongly correlated hydrological parameters corresponding to each anomaly type and establishing historical variation curves; collecting real-time hydrological data and predicting its variation based on deep learning and climate conditions to obtain predicted variation curves; splitting the predicted variation curves into several sub-curves based on the strongly correlated hydrological parameters corresponding to each anomaly type; calculating the similarity between the predicted variation sub-curves and their corresponding historical variation curves, and determining potential anomaly types based on the similarity values. This invention calculates only using variation curves of strongly correlated hydrological parameters, eliminating interference from other parameter curves, and considering the interaction of multiple strongly correlated hydrological parameters, thus accurately identifying hydrological anomaly types.

Description

Technical Field

[0001] This invention relates to the field of water condition monitoring technology, specifically to a lake water condition monitoring method and system based on deep learning. Background Technology

[0002] With the acceleration of global climate change and urbanization, lakes, as important water resource reserves and ecosystems, are experiencing increasingly complex and far-reaching changes in their hydrological conditions. Lake hydrological conditions not only concern the secure supply of water resources but also directly impact the surrounding ecological environment, agricultural irrigation, industrial production, and residents' lives. Traditional lake hydrological monitoring mainly relies on manual patrols and periodic sampling and analysis. This method is not only time-consuming and labor-intensive but also struggles to achieve real-time and comprehensive monitoring of hydrological conditions. Furthermore, traditional methods determine anomalies based on the values ​​of single water body parameters, while hydrological anomalies are often caused by multiple parameters, such as eutrophication and aquatic ecological abnormalities. Traditional methods struggle to accurately distinguish between different types of hydrological anomalies, leading to inaccurate early warning information and an inability to take timely and effective countermeasures. Moreover, they fail to fully utilize useful information from historical data and lack adaptability to complex hydrological changes. Therefore, a deep learning-based lake hydrological monitoring method and system are needed to address these problems. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a lake water condition monitoring method and system based on deep learning, so as to solve the problems existing in the above-mentioned background technology.

[0004] This invention is implemented as follows: a deep learning-based method for monitoring lake water conditions, the method comprising the following steps:

[0005] Historical hydrological data are classified based on the types of hydrological anomalies, and each historical hydrological data set contains time-series data of various hydrological parameters.

[0006] Analyze the historical hydrological data for each category, determine the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and determine the historical change curves, which contain the change curves of all strongly correlated hydrological parameters.

[0007] Real-time hydrological data is collected, and predictions are made based on deep learning and climate conditions to obtain predicted change curves. The predicted change curves contain the change curves of all hydrological parameters.

[0008] Based on the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, the predicted change curve is split into several predicted change curve sub-graphs.

[0009] The similarity between the predicted change curve subplot and the corresponding historical change curve plot is calculated, and the potential anomaly type is determined based on the similarity value.

[0010] Early warning information and monitoring plans are generated based on the types of potential anomalies.

[0011] As a further aspect of the present invention, the step of determining the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly specifically includes:

[0012] All historical hydrological data in the same category are extracted and grouped according to hydrological parameters, so that each group of data is time series data of the same hydrological parameter;

[0013] Based on time series data, curves of corresponding hydrological parameters are plotted, and the similarity between any two curves in each group is calculated sequentially to obtain the mean similarity value.

[0014] The similarity mean of the hydrological parameters is used to determine whether the hydrological parameter is a strongly correlated hydrological parameter of the corresponding hydrological anomaly type.

[0015] As a further aspect of the present invention, the step of predicting real-time hydrological data based on deep learning and climate conditions specifically includes:

[0016] Define an input window where each sample is all feature data from a certain period in the past. The feature data includes hydrological parameter values, climate data, and statistical characteristics.

[0017] Define a prediction window, where the label for each sample is the hydrological parameter information after a certain time period in the past, delayed by N hours.

[0018] All samples are divided into training set, validation set and test set according to time sequence;

[0019] The training set data is input into the LSTM model, and the loss function is minimized through the backpropagation algorithm and optimizer, and the model parameters are updated iteratively.

[0020] Early stopping is performed using the validation set to monitor the model's performance on the validation set. When the loss on the validation set no longer decreases, training is stopped and the model parameters of the best-performing validation set group are restored.

[0021] The final performance of the LSTM model is evaluated based on the test set, and the evaluation metrics include root mean square error, mean absolute error, and coefficient of determination.

[0022] As a further aspect of the present invention, the step of calculating the similarity between the predicted change curve sub-graph and the corresponding historical change curve graph specifically includes:

[0023] The curve color is determined according to the type of hydrological parameter, and all change curves in the predicted change curve sub-plot and the historical change curve plot are colored.

[0024] The predicted change curve subplot and the historical change curve plot are compared using DTW distance calculation. The distance values ​​calculated by DTW are then normalized to obtain similarity values.

[0025] Summarize all similarity values ​​obtained from the predicted change curve subplot, and take the largest one as the final similarity value.

[0026] As a further aspect of the present invention, the step of determining the potential anomaly type based on similarity values ​​specifically includes:

[0027] Determine whether the similarity value of each predicted change curve subplot is greater than the similarity threshold. If it is greater, the corresponding hydrological anomaly type is identified as a potential anomaly type.

[0028] By summarizing the potential anomaly types identified through several predicted change curve subplots, an anomaly type set is obtained.

[0029] As a further aspect of the present invention, the step of generating early warning information and monitoring schemes based on potential anomaly types specifically includes:

[0030] Anomaly levels are determined based on similarity values, and flood warning information is generated based on potential anomaly types and anomaly levels.

[0031] Based on the anomaly level, determine the monitoring frequency, monitoring range, and new monitoring points for the corresponding strongly correlated hydrological parameters;

[0032] Determine the water situation treatment plan based on the type and level of potential anomalies.

[0033] Another objective of this invention is to provide a deep learning-based lake hydrological monitoring system, the system comprising:

[0034] The historical hydrological data classification module is used to classify historical hydrological data based on the type of hydrological anomaly. Each historical hydrological data set contains time-series data of various hydrological parameters.

[0035] The strongly correlated parameter module is used to analyze historical hydrological data for each category, determine the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and determine the historical change curves, which contain the change curves of all strongly correlated hydrological parameters.

[0036] The real-time hydrological prediction module is used to collect real-time hydrological data, predict the real-time hydrological data based on deep learning and climate conditions, and obtain the predicted change curve, which contains the change curves of all hydrological parameters.

[0037] The prediction map splitting module is used to split the prediction change curve map based on the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and obtain several prediction change curve sub-maps.

[0038] The Potential Anomaly Type Module is used to calculate the similarity between the predicted change curve subplot and the corresponding historical change curve plot, and determine the potential anomaly type based on the similarity value.

[0039] The anomaly warning and monitoring module is used to generate warning information and monitoring plans based on the types of potential anomalies.

[0040] As a further aspect of the present invention, the strongly correlated parameter module includes:

[0041] The hydrological parameter grouping unit is used to extract and group all historical hydrological data in the same category according to hydrological parameters, so that each group of data is time series data of the same hydrological parameter;

[0042] The similarity mean determination unit is used to draw curves of corresponding hydrological parameters based on time series data, and calculate the similarity between any two curves in each group in turn to obtain the average similarity.

[0043] The hydrological parameter determination unit is used to determine whether a hydrological parameter is a strongly correlated hydrological parameter of the corresponding hydrological anomaly type based on the mean similarity value of the hydrological parameters.

[0044] As a further aspect of the present invention, the real-time hydrological prediction module includes:

[0045] The input window definition unit is used to define the input window. Each sample is all the feature data of a certain period of time in the past. The feature data includes hydrological parameter values, climate data and statistical features.

[0046] The prediction window definition unit is used to define the prediction window. The label for each sample is the hydrological parameter information after a certain time period in the past, delayed by N hours.

[0047] The set classification unit is used to divide all samples into training set, validation set and test set in chronological order;

[0048] The model learning unit is used to input training set data into the LSTM model, minimize the loss function through the backpropagation algorithm and optimizer, and iteratively update the model parameters.

[0049] The model optimization unit is used to perform early stopping using the validation set, monitor the model's performance on the validation set, stop training when the validation set loss no longer decreases, and restore the model parameters of the best-performing group on the validation set.

[0050] The model evaluation unit is used to evaluate the final performance of the LSTM model based on the test set. The evaluation metrics include root mean square error, mean absolute error, and coefficient of determination.

[0051] As a further aspect of the present invention, the potential anomaly type module includes:

[0052] The curve color addition unit is used to determine the curve color according to the type of hydrological parameter, and to color all the change curves in the predicted change curve subplot and the historical change curve plot.

[0053] The similarity calculation unit is used to perform DTW distance calculation between the predicted change curve subplot and the historical change curve plot, and to normalize the distance value calculated by DTW to obtain the similarity value.

[0054] The similarity value determination unit is used to summarize all similarity values ​​obtained through the predicted change curve subplot, and take the largest one as the final similarity value.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] By analyzing historical hydrological data, strongly correlated hydrological parameters are identified for each type of hydrological anomaly. Based on these parameters, the predicted change curve is divided into several sub-curves. The similarity between these sub-curves and their corresponding historical curves is calculated, and the potential anomaly type is determined based on the similarity value. In this way, calculations are performed solely using the change curves of strongly correlated hydrological parameters, eliminating interference from other parameter curves. Furthermore, since the graph contains multiple curves, a single calculation allows for comparison of multiple curves. By comprehensively considering the interactions and influences of various strongly correlated hydrological parameters, the types of hydrological anomalies can be quickly and accurately identified. Attached Figure Description

[0057] Figure 1 This is a flowchart of a deep learning-based lake hydrological monitoring method.

[0058] Figure 2 This is a flowchart for determining strongly correlated hydrological parameters in a deep learning-based lake hydrological monitoring method.

[0059] Figure 3 This is a flowchart illustrating the prediction of real-time water level data in a deep learning-based lake water level monitoring method.

[0060] Figure 4 This is a flowchart illustrating similarity calculation in a deep learning-based lake hydrological monitoring method.

[0061] Figure 5This is a flowchart for identifying potential anomaly types in a deep learning-based lake hydrological monitoring method.

[0062] Figure 6 This is a flowchart illustrating the generation of early warning information and monitoring schemes in a deep learning-based lake hydrological monitoring method.

[0063] Figure 7 This is a schematic diagram of a lake hydrological monitoring system based on deep learning. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0065] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0066] like Figure 1 As shown in the figure, this embodiment of the invention provides a lake hydrological monitoring method based on deep learning, the method comprising the following steps:

[0067] S100 classifies historical hydrological data based on the type of hydrological anomaly, with each historical hydrological data set containing time-series data of various hydrological parameters;

[0068] S200 analyzes historical hydrological data for each category, determines the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and determines historical change curves, which contain the change curves of all strongly correlated hydrological parameters.

[0069] S300 collects real-time hydrological data, and predicts the real-time hydrological data based on deep learning and climate conditions to obtain a predicted change curve. The predicted change curve contains the change curves of all hydrological parameters.

[0070] S400: Based on the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, the predicted change curve is split into several predicted change curve sub-graphs.

[0071] S500 calculates the similarity between the predicted change curve subplot and the corresponding historical change curve plot, and determines the potential anomaly type based on the similarity value.

[0072] S600 generates early warning information and monitoring plans based on potential anomaly types.

[0073] In this embodiment of the invention, historical hydrological data is first retrieved. Each historical hydrological data set contains time-series data of various hydrological parameters, including pH, dissolved oxygen, transparency, water temperature, water level, flow velocity, etc. Each historical hydrological data set is also marked with a hydrological anomaly type, including eutrophication, aquatic ecosystem anomalies, sediment content anomalies, and water level anomalies. Then, the historical hydrological data is categorized according to the hydrological anomaly type, with each category containing data corresponding to the same anomaly type. The historical hydrological data for each category is then analyzed sequentially to determine the strongly correlated hydrological parameters corresponding to each anomaly type. Strongly correlated hydrological parameters are those that clearly reflect or cause the specific hydrological anomaly type. Historical change curves are then plotted based on these determined strongly correlated hydrological parameters, with one historical change curve reflecting the change curves of all strongly correlated hydrological parameters. Additionally, real-time hydrological data and climate conditions are collected, and deep learning methods are used to predict the real-time hydrological data, resulting in a predicted change curve. One predicted change curve contains the change curves of all hydrological parameters. Then, based on the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, the predicted change curve is split into several sub-curves. For example, if there are 10 types of hydrological anomalies, 10 sub-curves will be generated. Each sub-curve contains the change curves of all the corresponding strongly correlated hydrological parameters. Next, the similarity between the predicted change curve sub-curves and the corresponding historical change curves is calculated. The potential anomaly type is determined based on the similarity value. In this way, calculations are performed only using the change curves of strongly correlated hydrological parameters, eliminating interference from other parameter curves. Furthermore, since the graph contains multiple curves, a single similarity calculation comprehensively considers the interaction and influence of multiple strongly correlated hydrological parameters, enabling rapid and accurate identification of hydrological anomaly types. Finally, early warning information and monitoring plans are generated based on the potential anomaly types, allowing for timely warnings and improving monitoring efficiency.

[0074] like Figure 2 As shown in the preferred embodiment of the present invention, the step of determining the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly specifically includes:

[0075] S201, extract and group all historical hydrological data in the same category according to hydrological parameters, so that each group of data is time series data of the same hydrological parameter;

[0076] S202, Based on time series data, plot the curves of corresponding hydrological parameters, calculate the similarity between any two curves in each group, and obtain the average similarity value;

[0077] S203. Determine whether a hydrological parameter is a strongly correlated hydrological parameter of the corresponding hydrological anomaly type based on the mean similarity of the hydrological parameters.

[0078] In this embodiment of the invention, to eliminate interference from other parameters and automatically extract strongly correlated hydrological parameters, all historical hydrological data within the same category are extracted and grouped according to hydrological parameters, ensuring that each group contains time-series data of the same hydrological parameter. For example, if there are 20 hydrological parameters, there will be 20 groups of data. Then, a curve graph corresponding to the hydrological parameter is plotted for each group of data. For example, if there are 50 historical hydrological data points in the category, each group will have 50 curve graphs. Next, the 50 curve graphs are paired, and the similarity between any two curve graphs in each group is calculated. Many similarity results are obtained, and the average similarity is calculated. Then, it is determined whether the average similarity is greater than a set value. If it is greater, it indicates that the 50 curves for that hydrological parameter have similar trends, and that the hydrological parameter is a strongly correlated hydrological parameter corresponding to the hydrological anomaly type; if it is less, it is not a strongly correlated hydrological parameter.

[0079] like Figure 3 As shown, in a preferred embodiment of the present invention, the step of predicting real-time hydrological data based on deep learning and climate conditions specifically includes:

[0080] S301 defines the input window. Each sample is all the feature data of a certain period in the past. The feature data includes hydrological parameter values, climate data, and statistical features.

[0081] S302, Define the prediction window, where the label for each sample is the hydrological parameter information after a certain time period in the past, delayed by N hours;

[0082] S303, divide all samples into training set, validation set and test set in chronological order;

[0083] S304: Input the training set data into the LSTM model, minimize the loss function through the backpropagation algorithm and optimizer, and iteratively update the model parameters;

[0084] S305 uses the validation set for early stopping, monitors the model's performance on the validation set, and stops training when the validation set loss no longer decreases, restoring the model parameters of the best-performing validation set group.

[0085] S306 evaluates the final performance of the LSTM model based on the test set, with evaluation metrics including root mean square error, mean absolute error, and coefficient of determination.

[0086] In this embodiment of the invention, when making predictions based on deep learning, it is first necessary to retrieve historical hydrological data and corresponding climate data, and preprocess the data, including handling missing values, outlier handling, and data standardization / normalization. Then, the data is transformed into "sample-label" pairs required by the deep learning model. An input window is defined, where each sample represents all feature data from a past period, including hydrological parameter values, climate data, and statistical features such as mean and variance. A prediction window is also defined, where the label for each sample is the hydrological parameter situation N hours after a past period, where N is a constant, for example, 6 hours. Next, the data is divided into a training set (e.g., the earliest 70% of data), a validation set (the middle 15% of data), and a test set (the latest 15% of data) in chronological order. The chronological order must not be disrupted to avoid data corruption (predicting the past using future information). Then, a model architecture is selected. Since time-series data is being processed, LSTM (Long Short-Term Memory) networks can effectively capture time dependencies. Mean squared error or mean absolute error is used as the loss function to minimize the difference between the predicted and actual values. The training data is then fed into the LSTM model, and the loss function is minimized using the backpropagation algorithm and optimizer, iteratively updating the model parameters. Early stopping is performed using the validation set to monitor the model's performance on the validation set. When the validation set loss stops decreasing or even starts to increase (indicating overfitting), training is stopped, and the model parameters of the best-performing validation set are restored. Finally, the final performance of the LSTM model is evaluated using a comprehensive set of metrics based on the test set.

[0087] like Figure 4 As shown, in a preferred embodiment of the present invention, the step of calculating the similarity between the predicted change curve sub-plot and the corresponding historical change curve plot specifically includes:

[0088] S501, determine the curve color according to the type of hydrological parameters, and color all the change curves in the predicted change curve sub-plot and the historical change curve plot.

[0089] S502, perform DTW distance calculation between the predicted change curve subplot and the historical change curve plot, and normalize the distance value calculated by DTW to obtain a similarity value.

[0090] S503, summarize all similarity values ​​obtained through the predicted change curve subplot, and take the largest one as the final similarity value.

[0091] In this embodiment of the invention, to achieve more accurate similarity matching, a curve color is defined for each type of hydrological parameter, and all change curves in the predicted change curve sub-plot and the historical change curve plot are then colored. Next, the predicted change curve sub-plot and the historical change curve plot are compared using the Time-Divergence Written (DTW) distance calculation. The DTW distance value is then normalized to obtain a similarity value, for example: similarity = 1 / (1 + DTW distance). The DTW algorithm uses dynamic programming to find the optimal nonlinear alignment path between two sequences, automatically compensating for offsets or local deformations on the time axis (such as the time lag of abnormal peaks), significantly improving the accuracy of similarity calculation and the robustness of anomaly detection. Finally, all similarity values ​​obtained from the predicted change curve sub-plot are summarized, and the largest one is taken as the final similarity value for subsequent calculations.

[0092] like Figure 5 As shown, in a preferred embodiment of the present invention, the step of determining the potential anomaly type based on similarity values ​​specifically includes:

[0093] S504, determine whether the similarity value of each predicted change curve subplot is greater than the similarity threshold. When it is greater, the corresponding hydrological anomaly type is identified as a potential anomaly type.

[0094] S505, summarize the potential anomaly types determined by several predicted change curve subplots to obtain an anomaly type set.

[0095] In this embodiment of the invention, after the final similarity value of each predicted change curve sub-graph is determined, it is determined whether the similarity value is greater than a similarity threshold. The similarity threshold is a pre-set fixed value. When it is greater than the threshold, the potential anomaly type can be determined. Then, the potential anomaly types determined by the 10 predicted change curve sub-graphs are summarized. When more than one potential anomaly type is determined, it indicates that a compound anomaly situation will occur.

[0096] like Figure 6 As shown in the preferred embodiment of the present invention, the step of generating early warning information and monitoring schemes based on potential anomaly types specifically includes:

[0097] S601, determine the anomaly level based on similarity values, and generate flood warning information based on potential anomaly type and anomaly level;

[0098] S602, determine the monitoring frequency, monitoring range and new monitoring points of the corresponding strongly correlated hydrological parameters according to the anomaly level;

[0099] S603, determine the water situation treatment plan based on the potential anomaly type and anomaly level.

[0100] In this embodiment of the invention, an anomaly level table is established, with each level corresponding to a range of similar values. This allows for direct determination of the anomaly level based on similar values, thereby generating flood warning information. Furthermore, the monitoring frequency, monitoring range, and new monitoring points for corresponding strongly correlated flood parameters are determined based on the anomaly level. Higher levels indicate higher risk, requiring higher monitoring frequencies, larger monitoring ranges, and greater density of monitoring points. Finally, flood control solutions are determined based on the potential anomaly type and its level. Multiple solutions are pre-defined for each anomaly type, with each solution applicable to different levels.

[0101] like Figure 7 As shown in the figure, this embodiment of the invention also provides a lake hydrological monitoring system based on deep learning, the system comprising:

[0102] The historical hydrological classification module 100 is used to classify historical hydrological data based on the type of hydrological anomaly. Each historical hydrological data contains time-series data of various hydrological parameters.

[0103] The strongly correlated parameter module 200 is used to analyze historical hydrological data for each category, determine the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and determine the historical change curves, which contain the change curves of all strongly correlated hydrological parameters.

[0104] The real-time hydrological prediction module 300 is used to collect real-time hydrological data, predict the real-time hydrological data based on deep learning and climate conditions, and obtain a predicted change curve. The predicted change curve contains the change curves of all hydrological parameters.

[0105] The prediction map splitting module 400 is used to split the prediction change curve map based on the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and obtain several prediction change curve sub-maps.

[0106] The Potential Anomaly Type Module 500 is used to calculate the similarity between the predicted change curve subplot and the corresponding historical change curve plot, and determine the potential anomaly type based on the similarity value.

[0107] The anomaly warning and monitoring module 600 is used to generate warning information and monitoring plans based on the types of potential anomalies.

[0108] In this embodiment of the invention, the strongly correlated parameter module 200 includes:

[0109] The hydrological parameter grouping unit is used to extract and group all historical hydrological data in the same category according to hydrological parameters, so that each group of data is time series data of the same hydrological parameter;

[0110] The similarity mean determination unit is used to draw curves of corresponding hydrological parameters based on time series data, and calculate the similarity between any two curves in each group in turn to obtain the average similarity.

[0111] The hydrological parameter determination unit is used to determine whether a hydrological parameter is a strongly correlated hydrological parameter of the corresponding hydrological anomaly type based on the mean similarity value of the hydrological parameters.

[0112] In this embodiment of the invention, the real-time hydrological prediction module 300 includes:

[0113] The input window definition unit is used to define the input window. Each sample is all the feature data of a certain period of time in the past. The feature data includes hydrological parameter values, climate data and statistical features.

[0114] The prediction window definition unit is used to define the prediction window. The label for each sample is the hydrological parameter information after a certain time period in the past, delayed by N hours.

[0115] The set classification unit is used to divide all samples into training set, validation set and test set in chronological order;

[0116] The model learning unit is used to input training set data into the LSTM model, minimize the loss function through the backpropagation algorithm and optimizer, and iteratively update the model parameters.

[0117] The model optimization unit is used to perform early stopping using the validation set, monitor the model's performance on the validation set, stop training when the validation set loss no longer decreases, and restore the model parameters of the best-performing group on the validation set.

[0118] The model evaluation unit is used to evaluate the final performance of the LSTM model based on the test set. The evaluation metrics include root mean square error, mean absolute error, and coefficient of determination.

[0119] In this embodiment of the invention, the potential anomaly type module 500 includes:

[0120] The curve color addition unit is used to determine the curve color according to the type of hydrological parameter, and to color all the change curves in the predicted change curve subplot and the historical change curve plot.

[0121] The similarity calculation unit is used to perform DTW distance calculation between the predicted change curve subplot and the historical change curve plot, and to normalize the distance value calculated by DTW to obtain the similarity value.

[0122] The similarity value determination unit is used to summarize all similarity values ​​obtained through the predicted change curve subplot, and take the largest one as the final similarity value.

[0123] The above description only details the preferred embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0124] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0125] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0126] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A lake hydrological monitoring method based on deep learning, characterized in that, The method includes the following steps: Historical hydrological data are classified based on the types of hydrological anomalies, and each historical hydrological data set contains time-series data of various hydrological parameters. Analyze the historical hydrological data for each category, determine the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and determine the historical change curves. The historical change curves only contain the change curves of all strongly correlated hydrological parameters. Real-time hydrological data is collected, and predictions are made based on deep learning and climate conditions to obtain predicted change curves. The predicted change curves contain the change curves of all hydrological parameters. Based on the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, the predicted change curve is split into several predicted change curve sub-graphs. The similarity between the predicted change curve subplot and the corresponding historical change curve plot is calculated, and the potential anomaly type is determined based on the similarity value. Early warning information and monitoring plans are generated based on the types of potential anomalies.

2. The lake hydrological monitoring method based on deep learning according to claim 1, characterized in that, The step of determining the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly specifically includes: All historical hydrological data in the same category are extracted and grouped according to hydrological parameters, so that each group of data is time series data of the same hydrological parameter; Based on time series data, curves of corresponding hydrological parameters are plotted, and the similarity between any two curves in each group is calculated sequentially to obtain the mean similarity value. The similarity mean of the hydrological parameters is used to determine whether the hydrological parameter is a strongly correlated hydrological parameter of the corresponding hydrological anomaly type.

3. The lake hydrological monitoring method based on deep learning according to claim 1, characterized in that, The steps for predicting real-time hydrological data based on deep learning and climate conditions specifically include: Define an input window where each sample is all feature data from a certain period in the past. The feature data includes hydrological parameter values, climate data, and statistical characteristics. Define a prediction window, where the label for each sample is the hydrological parameter information after a certain time period in the past, delayed by N hours. All samples are divided into training set, validation set and test set according to time sequence; The training set data is input into the LSTM model, and the loss function is minimized through the backpropagation algorithm and optimizer, and the model parameters are updated iteratively. Early stopping is performed using the validation set to monitor the model's performance on the validation set. When the loss on the validation set no longer decreases, training is stopped and the model parameters of the best-performing validation set group are restored. The final performance of the LSTM model is evaluated based on the test set, and the evaluation metrics include root mean square error, mean absolute error, and coefficient of determination.

4. The lake hydrological monitoring method based on deep learning according to claim 1, characterized in that, The step of calculating the similarity between the predicted change curve sub-plot and the corresponding historical change curve plot specifically includes: The curve color is determined according to the type of hydrological parameter, and all change curves in the predicted change curve sub-plot and the historical change curve plot are colored. The predicted change curve subplot and the historical change curve plot are compared using DTW distance calculation. The distance values ​​calculated by DTW are then normalized to obtain similarity values. Summarize all similarity values ​​obtained from the predicted change curve subplot, and take the largest one as the final similarity value.

5. The lake hydrological monitoring method based on deep learning according to claim 4, characterized in that, The step of determining the potential anomaly type based on similarity values ​​specifically includes: Determine whether the final similarity value corresponding to each predicted change curve subplot is greater than the similarity threshold. If it is greater, the corresponding hydrological anomaly type is identified as a potential anomaly type. By summarizing the potential anomaly types identified through several predicted change curve subplots, an anomaly type set is obtained.

6. The lake hydrological monitoring method based on deep learning according to claim 1, characterized in that, The steps for generating early warning information and monitoring plans based on potential anomaly types specifically include: Anomaly levels are determined based on similarity values, and flood warning information is generated based on potential anomaly types and anomaly levels. Based on the anomaly level, determine the monitoring frequency, monitoring range, and new monitoring points for the corresponding strongly correlated hydrological parameters; Determine the water situation treatment plan based on the type and level of potential anomalies.

7. A lake hydrological monitoring system based on deep learning, characterized in that, The system includes: The historical hydrological data classification module is used to classify historical hydrological data based on the type of hydrological anomaly. Each historical hydrological data set contains time-series data of various hydrological parameters. The strongly correlated parameter module is used to analyze historical hydrological data for each category, determine the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and determine the historical change curve. The historical change curve only contains the change curves of all strongly correlated hydrological parameters. The real-time hydrological prediction module is used to collect real-time hydrological data, predict the real-time hydrological data based on deep learning and climate conditions, and obtain the predicted change curve, which contains the change curves of all hydrological parameters. The prediction map splitting module is used to split the prediction change curve map based on the strongly correlated hydrological parameters corresponding to each type of hydrological anomaly, and obtain several prediction change curve sub-maps. The Potential Anomaly Type Module is used to calculate the similarity between the predicted change curve subplot and the corresponding historical change curve plot, and determine the potential anomaly type based on the similarity value. The anomaly warning and monitoring module is used to generate warning information and monitoring plans based on the types of potential anomalies.

8. The lake hydrological monitoring system based on deep learning according to claim 7, characterized in that, The strongly correlated parameter module includes: The hydrological parameter grouping unit is used to extract and group all historical hydrological data in the same category according to hydrological parameters, so that each group of data is time series data of the same hydrological parameter; The similarity mean determination unit is used to draw curves of corresponding hydrological parameters based on time series data, and calculate the similarity between any two curves in each group in turn to obtain the average similarity. The hydrological parameter determination unit is used to determine whether a hydrological parameter is a strongly correlated hydrological parameter of the corresponding hydrological anomaly type based on the mean similarity value of the hydrological parameters.

9. The lake hydrological monitoring system based on deep learning according to claim 7, characterized in that, The real-time hydrological prediction module includes: The input window definition unit is used to define the input window. Each sample is all the feature data of a certain period of time in the past. The feature data includes hydrological parameter values, climate data and statistical features. The prediction window definition unit is used to define the prediction window. The label for each sample is the hydrological parameter information after a certain time period in the past, delayed by N hours. The set classification unit is used to divide all samples into training set, validation set and test set in chronological order; The model learning unit is used to input training set data into the LSTM model, minimize the loss function through the backpropagation algorithm and optimizer, and iteratively update the model parameters. The model optimization unit is used to perform early stopping using the validation set, monitor the model's performance on the validation set, stop training when the validation set loss no longer decreases, and restore the model parameters of the best-performing group on the validation set. The model evaluation unit is used to evaluate the final performance of the LSTM model based on the test set. The evaluation metrics include root mean square error, mean absolute error, and coefficient of determination.

10. The lake hydrological monitoring system based on deep learning according to claim 7, characterized in that, The potential anomaly type module includes: The curve color addition unit is used to determine the curve color according to the type of hydrological parameter, and to color all the change curves in the predicted change curve subplot and the historical change curve plot. The similarity calculation unit is used to perform DTW distance calculation between the predicted change curve subplot and the historical change curve plot, and to normalize the distance value calculated by DTW to obtain the similarity value. The similarity value determination unit is used to summarize all similarity values ​​obtained through the predicted change curve subplot, and take the largest one as the final similarity value.

Images