Hydrological station supervision method and device

Abstract

The application discloses a hydrological station supervision method and equipment, and relates to the technical field of computers, which comprises the following steps: receiving hydrological element data reported by each hydrological station, obtaining future rainfall information of the corresponding hydrological station according to the data, integrating the two types of information to generate an input data set of a hydrological decision network model, the model comprising a local bidirectional model and a global unidirectional model; the input data set is first subjected to local bidirectional model operation to generate a local feature sequence, then the local feature sequence is input into the global unidirectional model, and finally, the decision results of the gate switch of each hydrological station are output and fed back to the corresponding hydrological station. In this way, the local bidirectional model can mine the bidirectional correlation features of the surrounding data at each time step, the global unidirectional model can grasp the overall time sequence law, the combination of the two can greatly improve the decision accuracy, avoid the subjectivity and hysteresis of traditional experience judgment, and the whole process is efficiently connected to improve the dispatching response speed and enhance the supervision initiative and predictability.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method and equipment for monitoring hydrological stations. Background Technology

[0002] Hydrological stations are sites used to observe, record, and analyze hydrological elements, forming the foundation of the hydrological monitoring network. Data from hydrological stations supports water resource management, disaster prevention, ecological protection, and engineering construction. If hydrological stations lack effective supervision, data gaps and decision-making errors can easily occur, leading to resource crises. Traditional hydrological station supervision has long relied heavily on manual labor in data collection and scheduling decision-making. For example, hydrological data requires regular manual visits to stations to copy or summarize records reported from various stations, and the timeliness of the data is constrained by the manual operation cycle. Furthermore, gate scheduling decisions based on this reliance on staff judgment limit the speed and accuracy of decision-making. Summary of the Invention

[0003] This invention provides a method and equipment for monitoring hydrological stations, which at least solves the problems of poor timeliness, delayed decision-making response, and low accuracy caused by reliance on human experience in related technologies.

[0004] This invention provides a hydrological station monitoring method for a data center, comprising:

[0005] Receive hydrological data reported by various hydrological stations;

[0006] Based on the aforementioned hydrological data, future rainfall information for each hydrological station is obtained;

[0007] The hydrological element data from each hydrological station and the future rainfall information are integrated to generate the input dataset for the hydrological decision network model; the hydrological decision network model includes a local bidirectional model and a global unidirectional model.

[0008] The input dataset is fed into the local bidirectional model for computation to generate a local feature sequence.

[0009] The local feature sequence is used as the input parameter of the global one-way model. After processing by the global one-way model, the decision result of the corresponding hydrological station gate opening and closing is output, and the decision result is fed back to the corresponding hydrological station.

[0010] The present invention also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described hydrological station monitoring methods.

[0011] The hydrological station monitoring method provided by this invention first receives hydrological element data reported by each hydrological station; based on the hydrological element data, it obtains future rainfall information for each hydrological station; then, it integrates the hydrological element data and future rainfall information from each hydrological station to generate the input dataset for a hydrological decision-making network model. This ensures the comprehensiveness and timeliness of the model's input data, laying a data foundation for accurate decision-making. The hydrological decision-making network model includes a local bidirectional model and a global unidirectional model; the input dataset is input into the local bidirectional model for computation, generating a local feature sequence; finally, the local feature sequence is used as the input parameter for the global unidirectional model, which processes the data and outputs the decision result for the corresponding hydrological station's gate opening and closing, and then feeds the decision result back to the corresponding hydrological station. The local bidirectional model can deeply explore the bidirectional correlation characteristics of surrounding hydrological data at each time step, while the global unidirectional model can grasp the overall temporal change pattern. The combination of the two greatly improves the accuracy of gate opening and closing decisions, effectively avoiding the subjectivity and lag of relying on experience judgment in traditional supervision. At the same time, the decision results can be quickly fed back to the corresponding hydrological station, realizing efficient connection of the entire process from data collection and analysis to dispatching instructions. This not only improves the response speed of hydrological station gate dispatching, but also enhances the initiative and foresight of watershed hydrological supervision, providing strong technical support for flood control and disaster reduction, and optimal allocation of water resources.

[0012] In addition, the present invention also provides corresponding electronic equipment for hydrological station monitoring methods, which has the same or corresponding technical features as the hydrological station monitoring methods mentioned above, and has the same effect. Attached Figure Description

[0013] To more clearly illustrate the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart of a hydrological station monitoring method provided in an embodiment of the present invention;

[0015] Figure 2 This is a schematic diagram of the structure of the hydrological station monitoring device provided in an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present invention.

[0017] It should be noted that, in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., used in this invention are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0018] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] The specific application environment architecture or specific hardware architecture on which the implementation of hydrological station monitoring methods depends is described here.

[0020] The embodiments of the present invention provide a method for monitoring hydrological stations, and the method is described in detail in conjunction with the execution flow of the method for monitoring hydrological stations. Figure 1 A flowchart of the hydrological station monitoring method provided in the embodiments of the present invention is shown below. Figure 1 As shown, this method is used in a data center and includes:

[0021] S101, Receive hydrological data reported by each hydrological station.

[0022] It should be noted that in this invention, the data center and each hydrological station can be connected via a network communication protocol. This network communication protocol can be WebSocket (a network protocol that enables full-duplex communication over a single Transmission Control Protocol connection). That is, this invention can utilize WebSocket to establish a persistent connection and full-duplex communication between the hydrological station and the data center. Besides WebSocket communication, other real-time or near-real-time communication methods can also be used, and are not limited here.

[0023] Each hydrological station is equipped with sensors and cameras. In this invention, each hydrological station can initiate a program to acquire sensor data and a program to calculate water levels based on water level images captured by cameras; at set time intervals, the data collected by sensors, the water level data calculated from water level images, and related images are processed together; the processed data and images are periodically reported to the data center. The data center can receive hydrological element data such as sensor data, water level images, and their corresponding water level data at the same time.

[0024] Furthermore, in specific implementation, in the above-mentioned hydrological station monitoring method provided in the embodiments of the present invention, after receiving the hydrological element data reported by each hydrological station in step S101, it may further include: after judging the accuracy of the water level data in the hydrological element data, feeding back the judgment result to the corresponding hydrological station so that the corresponding hydrological station updates the water level data according to the judgment result; after the hydrological station updates the local water level data, updating the water level data reported by the corresponding hydrological station.

[0025] In practice, since the data collected by hydrological stations may not be entirely accurate, this invention supports data center staff in judging the accuracy of water level data through water gauge images. Specifically, it can determine whether the water level data needs to be updated based on the water level data in the sensor data and the water level data calculated from the water gauge image. If the water level data in the sensor data and the water level data calculated from the water gauge image are consistent, it is determined that no update is needed; if they are inconsistent, it is determined that an update is needed. The judgment result is returned to the hydrological station that needs to update for data correction, avoiding the problem of inaccurate water level data caused by relying solely on water level sensor and water gauge data without manual verification.

[0026] S102. Based on hydrological data, obtain future rainfall information for each hydrological station.

[0027] It should be noted that hydrological element data is the foundation for reflecting the current hydrological state, while rainfall is the core input variable driving subsequent hydrological changes. This invention can establish a predictive mapping relationship between combinations of hydrological elements and future rainfall based on a dataset of the correspondence between historical hydrological element data and rainfall, using statistical models (such as time series analysis) or machine learning models (such as regression models and time series prediction models). Furthermore, based on this predictive mapping relationship and combined with real-time hydrological element data, future rainfall information for each hydrological station can be obtained.

[0028] S103. Integrate hydrological element data from various hydrological stations and future rainfall information to generate the input dataset for the hydrological decision network model; the hydrological decision network model includes a local bidirectional model and a global unidirectional model.

[0029] In implementation, this invention integrates real-time hydrological data from various hydrological stations with future rainfall information to form a multi-dimensional, forward-looking input dataset, which serves as the input to a hydrological decision-making network model. The hydrological decision-making network model is used to obtain the decision results for opening and closing hydrological station gates. The hydrological decision-making network model can consist of a local bidirectional model and a global unidirectional model. The local bidirectional model is a machine learning model architecture that utilizes both forward and backward data features within a defined local window at the current position when processing sequence data. The local bidirectional model can be a local bidirectional LSTM (Long Short-Term Memory) model. The global unidirectional model is a model that uses the entire sequence information unidirectionally for prediction when processing sequence data. The global unidirectional model can be a global unidirectional LSTM model.

[0030] S104. Input the input dataset into the local bidirectional model for computation to generate a local feature sequence.

[0031] It should be noted that the local bidirectional LSTM model can utilize its bidirectional computational characteristics to mine local hydrological correlations before and after each time step in the input data. This invention allows the integrated hydrological data to be input into a local bidirectional LSTM model, which can mine hydrological correlation information before and after each time step, and after computation, form a sequence that reflects local data characteristics, i.e., a local feature sequence.

[0032] S105. The local feature sequence is used as the input parameter of the global one-way model. After processing by the global one-way model, the decision result of the corresponding hydrological station gate opening and closing is output, and the decision result is fed back to the corresponding hydrological station.

[0033] It should be noted that the global unidirectional LSTM model can grasp the overall trend of changes in data from various hydrological stations within a watershed through its unidirectional time series modeling capabilities. This invention allows local feature sequences to be input into the global unidirectional LSTM model, which can then combine these local details to grasp the overall hydrological change patterns, process the data to obtain the decision results for the opening and closing of gates at each hydrological station, and then feed the results back to the corresponding hydrological station.

[0034] In the hydrological station monitoring method provided in this embodiment of the invention, hydrological element data reported by each hydrological station is first received; based on the hydrological element data, future rainfall information of each hydrological station is obtained; then, the hydrological element data and future rainfall information of each hydrological station are integrated to generate the input dataset of the hydrological decision network model. This ensures the comprehensiveness and timeliness of the model input data, laying a data foundation for accurate decision-making. The hydrological decision network model includes a local bidirectional model and a global unidirectional model; the input dataset is input into the local bidirectional model for computation to generate a local feature sequence; finally, the local feature sequence is used as the input parameter of the global unidirectional model, and after processing by the global unidirectional model, the decision result of the corresponding hydrological station gate opening and closing is output, and the decision result is fed back to the corresponding hydrological station. The local bidirectional model can deeply explore the bidirectional correlation characteristics of surrounding hydrological data at each time step, while the global unidirectional model can grasp the overall temporal change pattern. The combination of the two greatly improves the accuracy of gate opening and closing decisions, effectively avoiding the subjectivity and lag of relying on experience judgment in traditional supervision. At the same time, the decision results can be quickly fed back to the corresponding hydrological station, realizing efficient connection of the entire process from data collection and analysis to dispatching instructions. This not only improves the response speed of hydrological station gate dispatching, but also enhances the initiative and foresight of watershed hydrological supervision, providing strong technical support for flood control and disaster reduction, and optimal allocation of water resources.

[0035] Furthermore, in specific implementation, in the hydrological station monitoring method provided in the embodiments of the present invention, the data center can deploy a server containing a clustered version of the hydrological management platform, and each hydrological station can deploy an edge server containing a wireless communication component and a standalone version of the hydrological management platform; each hydrological station establishes a network communication link with the data center through the wireless communication component; the clustered version of the hydrological management platform is used to manage each standalone version of the hydrological management platform.

[0036] It should be noted that the standalone hydrological management platform is a software system deployed and running on a standalone computer for collecting, processing, analyzing, storing, and displaying hydrological data. Its core characteristic is localized operation; data processing and system operation are primarily limited to a single device, not relying on a network environment or a multi-terminal collaborative server architecture. At each hydrological station, users can log in to the standalone hydrological management platform and complete two configurations: first, inputting various on-site monitoring sensors (covering water level, flow velocity, rainfall, etc.) and defining the sensor data parsing rules (including data field mapping relationships, such as the byte positions corresponding to water level data; data type identifiers; data correction parameters, such as scaling factors; and units of measurement); second, for water level monitoring sensors, associating them with the corresponding video cameras at the monitoring points, configuring basic camera information (including real-time flow address, monitoring area range, etc.), and establishing a visual calculation association mechanism for water level values ​​based on the water level image data collected by the camera.

[0037] The clustered hydrological management platform is a software system deployed based on a server cluster architecture for the comprehensive management of large-scale, high-concurrency hydrological data. Its core feature is the collaborative operation of multiple servers forming a cluster, overcoming the performance bottlenecks of single-machine or single-server systems and meeting the high requirements for data processing, storage, access, and system stability in complex hydrological monitoring scenarios such as regional and watershed levels. After configuring sensors and cameras on the standalone hydrological management platform, registering them with the clustered version enables WebSocket communication between the hydrological station and the data center, simultaneously synchronizing sensor and camera information to the data center.

[0038] In this invention, the data center servers deploy a clustered hydrological management platform, while the edge servers at each hydrological station are equipped with wireless communication components and standalone platforms. The wireless communication components can be 4G / 5G wireless communication modules or other communicable modules, such as BeiDou. Each hydrological station establishes communication with the data center through the wireless communication components, and the clustered platform is responsible for the unified management of all standalone platforms. The data center servers possess artificial intelligence capabilities and can run a hydrological decision-making network model to predict whether to open or close the gates. Because deciding whether to open or close the gates at a hydrological station requires consideration of multiple factors, such as water level, flow velocity, evaporation, and rainfall, and these data exhibit significant time dependencies, this invention employs a hydrological decision-making network model composed of a local bidirectional LSTM model and a global unidirectional LSTM model. This model effectively captures long-term and short-term dependencies, achieving efficient and stable predictions.

[0039] Accordingly, step S101 receives hydrological element data reported by each hydrological station, which may specifically include: receiving hydrological element data reported by each hydrological station through the stand-alone hydrological management platform at set time intervals; the hydrological element data includes time-series data collected by sensors containing water level, flow velocity, evaporation and rainfall, water level data calculated based on water gauge images monitored by cameras and the corresponding water gauge images.

[0040] In practice, each hydrological station reports hydrological data to the data center at set time intervals through a standalone hydrological management platform. This includes time-series data such as water level and flow velocity collected by sensors, as well as water level readings and corresponding images calculated by cameras. This multi-source data acquisition method integrates real-time sensor data and incorporates image verification, improving the integrity and reliability of the data and providing a more reliable foundation for subsequent analysis and decision-making.

[0041] It should be noted that the local bidirectional model of this invention can consider only the influence of water level, flow velocity, evaporation, and rainfall parameters within a short period of time. This avoids the problems of computational burden and storage space occupied by intermediate calculation results caused by using a global bidirectional model, while also incorporating the impact of future precipitation on the results. The global unidirectional model can effectively consider long-term time series characteristics such as water level, flow velocity, and evaporation, accurately capturing the overall temporal evolution of watershed hydrology and providing a basis for gate opening and closing decisions that aligns with macroscopic trends.

[0042] Furthermore, in specific implementation, in the hydrological station monitoring method provided in the embodiments of the present invention, step S104 inputs the input dataset into the local bidirectional model for calculation to generate a local feature sequence, which may specifically include: aligning the relevant data in the input dataset according to the set time step to form a multi-dimensional original sequence with a unified time axis; inputting the multi-dimensional original sequence into the local bidirectional model for calculation to generate a local feature sequence.

[0043] In practice, this invention can align relevant data in the input dataset according to a set time step to form a multi-dimensional original sequence with a unified time axis. ,in The time step can be set to 5 minutes to indicate the sequence length. The vector consists of water level, flow velocity, evaporation, and rainfall. This vector is then input into a local bidirectional model to generate a local feature sequence. This approach first addresses the temporal dimension differences of multi-source data through time alignment, ensuring data correlation within the same temporal framework and laying a solid foundation for subsequent analysis. The subsequent model computation then more accurately uncovers the local data correlations before and after each time point, improving the effectiveness of feature extraction.

[0044] Furthermore, in specific implementation, in the above steps, the multi-dimensional original sequence is input into the local bidirectional model for calculation to generate a local feature sequence. Specifically, this may include: extracting data segments of a set time period before and after each time step in the multi-dimensional original sequence as a local analysis window; using the local bidirectional model to extract the forward and backward features within the local analysis window, merging them to obtain the local features corresponding to each time step; and integrating the local features corresponding to each time step to generate a local feature sequence.

[0045] In practice, this invention can extract data from the original multi-dimensional sequence before and after each time step, within a set time period, as a local analysis window. For example, for each time step... Take data from one hour before and after the current timeframe, i.e., a fixed window W=12 before and after the current timeframe, and extract the subsequence as follows: Then, a bidirectional model is applied. The local bidirectional model extracts and merges forward and backward features, and then integrates the local features from each time step to generate a sequence. This approach focuses on data in key time periods through windows, and combined with the bidirectional model's mining of correlations between time periods, it can accurately capture local time series details, making the feature sequence more closely match the dynamic correlation characteristics of hydrological data, and improving the targeting and accuracy of subsequent analysis.

[0046] Furthermore, in specific implementation, in the above steps, the forward and backward features within the local analysis window are extracted using a local bidirectional model and merged to obtain the local features corresponding to each time step. Specifically, this may include: inputting the data corresponding to the local analysis window into the local bidirectional model, performing calculations along the forward and backward time axes respectively, extracting the forward and backward hidden states corresponding to each time step to obtain the forward and backward features within the local analysis window; merging the forward and backward features within the local analysis window to obtain the local features corresponding to each time step.

[0047] In practice, this invention can input local analysis window data into a local bidirectional model, perform bidirectional operations along the time axis, and extract the forward hidden state at each time step. and backward hidden state As corresponding features, these two types of features are then combined to obtain the local features at each time step. The local feature sequences for all time steps are: This bidirectional extraction and fusion method can simultaneously capture the temporal correlation of data (such as the mutual influence between water level at a certain moment and rainfall and flow rate in the preceding and following periods), allowing local features to more comprehensively reflect the dynamic relationship of time series and providing more accurate local detail support for subsequent global analysis.

[0048] Furthermore, in specific implementation, in the hydrological station monitoring method provided in the embodiments of the present invention, step S104, while inputting the input dataset into the local bidirectional model for calculation, may also include: activating an attention mechanism based on key event detection; setting corresponding benchmark intervals for hydrological element parameters according to historical data of each water station using statistical methods; calculating the deviations of each parameter at each time step within the local analysis window from the corresponding benchmark interval, and determining the coefficients for adjusting the weights of the forward and backward hidden states corresponding to each time step based on the distance between each time step and the current time step; using the adjusted forward hidden state as the forward feature within each local analysis window, and using the adjusted backward hidden state as the backward feature within each local analysis window.

[0049] In implementation, this invention aggregates the forward and backward hidden states in a local bidirectional model. At that time, an attention mechanism based on key event detection is used for dynamic aggregation. First, based on historical data, a normal value range is set for hydrological element parameters (water level, flow velocity, evaporation, rainfall, etc.) using statistical methods. This establishes a reference for judging data anomalies. The normal value range here refers to the aforementioned baseline range. Next, weighting coefficients are calculated by combining parameter deviation and time distance; specifically, this can be done for each time step. The extracted subsequence is The algorithm determines the Euclidean distance between all factors in a subsequence and their corresponding normal value intervals, and assigns a certain weight ratio to the hidden state based on the Euclidean distance and the time step. This approach not only identifies critical abnormal data (such as sudden increases in rainfall or abnormal fluctuations in water levels) through bias but also highlights the impact of recent data through time distance, aligning with the temporal correlation patterns of hydrological processes. This allows the attention mechanism to assign higher weight to important information, preventing the model from over-focusing on irrelevant or redundant data. Consequently, it extracts more accurate and targeted local features, providing more reliable detailed support for subsequent global decision-making and further improving overall decision-making accuracy.

[0050] Furthermore, in specific implementation, in the above steps, the deviations of various parameters at each time step within the local analysis window from the corresponding baseline interval are calculated. Combined with the distance between each time step and the current time step, coefficients are determined to adjust the weights of the forward and backward hidden states corresponding to each time step. Specifically, this may include: for the forward hidden state corresponding to each time step, calculating the deviations of water level, flow velocity, and evaporation at each time step from the corresponding baseline interval, and combining the distance between each time step and the current time step to obtain the forward weighting factor; for the backward hidden state corresponding to each time step, calculating the deviation of rainfall at each time step from the corresponding baseline interval, and combining the distance between each time step and the current time step to obtain the backward weighting factor; and adjusting the aggregation ratio of the forward and backward hidden states corresponding to each time step based on the forward and backward weighting factors.

[0051] In implementation, this invention calculates forward weighting factors for the forward hidden state by combining deviations in elements such as water level and flow velocity with time distance; for the backward hidden state, it focuses on calculating backward weighting factors by considering precipitation deviations with time distance, and then adjusts the aggregation ratio of the two types of factors. This fully considers the temporal differences in the impact of different hydrological elements (e.g., current and past states such as water level are more closely related to forward features, while precipitation and other predictive features of subsequent changes are more related to backward features). By using differentiated weights, the model accurately captures the core value of various elements, avoiding feature confusion caused by indiscriminate aggregation, and making the final fused local features more consistent with the inherent correlation patterns of hydrological data, providing more accurate detailed support for subsequent global decision-making.

[0052] Furthermore, in practical implementation, the coefficients used to adjust the weights of the forward or backward hidden states corresponding to each time step can be determined using the following formula in the above steps:

[0053] ;

[0054] ;

[0055] ;

[0056] in, For time step The value of the water level parameter can be the value of one of the parameters such as water level, flow velocity, evaporation, and rainfall. This is the lower limit of the baseline interval. This represents the upper limit of the baseline interval; To measure time steps The difference between the values ​​of water level parameters and the corresponding benchmark interval. For the current time step, This is the length parameter of the local analysis window. For time step The distance from the current time step, i.e. the time step difference, has a greater impact on the weight when the step difference is smaller. To comprehensively analyze the data at each time step within the local analysis window and The weighting coefficients, These are coefficients used to adjust the weights of the forward or backward hidden states at each time step.

[0057] In implementation, this invention uses the quantization logic of calculating the weight adjustment coefficient through the above formula, firstly... Determine whether hydrological parameters exceed the baseline upper limit and quantify the deviation, then... The overall weight within the window is calculated by combining the deviation percentage and the time step (the smaller the step, the greater the impact), and finally... The range of normalization coefficients. This is achieved through... Accurately capture abnormal data exceeding the baseline (such as a sudden rise in water level), and then through Highlighting the reference value of recent data, making the calculated It can adjust the weights of hidden states, allowing the model to focus on high-value information, filter out redundant interference, and provide objective and accurate quantitative basis for feature extraction.

[0058] Furthermore, in specific implementation, in the hydrological station monitoring method provided in the embodiments of the present invention, when there is no situation exceeding the benchmark interval or all situations exceeding the benchmark interval, that is, when there are no extreme situations or all types of parameters are in extreme situations, the ratio of the two remains equal; when the water level, flow velocity, and evaporation exceed the benchmark interval, the forward state weight increases and the backward state weight decreases, and the closer to the current time, the closer the forward weight is to 1; when the rainfall exceeds the benchmark interval, the backward state weight increases and the forward state weight decreases, and the closer to the current time, the closer the backward weight is to 1.

[0059] In implementation, when no value exceeds the normal range or all values ​​exceed the normal range, the ratio of the forward and backward hidden states can be 0.5 and 0.5 respectively. When the water level, flow rate, and evaporation exceed the normal range, the ratio of the forward and backward hidden states ranges from [0.5, 1] ​​to [0, 0.5]. If the distance from the current time step... The closer the state is, the closer the weight of the forward hidden state is to 1; the farther away it is, the closer the weight is to 0.5. When the rainfall exceeds the normal range, the ratio of the forward and backward hidden states ranges from [0, 0.5] to [0.5, 1]. If the distance from the current time step... The closer the data is, the closer the weight assigned to the backward hidden state is to 1; the farther away it is, the closer it is to 0.5. This assigns high weights to extreme situations, such as sudden increases in water levels or future extreme rainfall, which strengthens the model's focus on these extreme situations. It highlights the impact of data from critical moments such as extreme rainfall and sudden increases in water levels on decision-making, avoids treating extreme situations and regular data indiscriminately, improves the model's sensitivity to sudden floods, and reduces the risk of misjudgment.

[0060] Furthermore, in specific implementation, in the hydrological station monitoring method provided in the embodiments of the present invention, step S105 outputs the decision result of the corresponding hydrological station gate opening and closing after processing by the global one-way model, and feeds back the decision result to the corresponding hydrological station. Specifically, it may include: using the global one-way model to traverse the local feature sequence along the positive time axis to obtain the global feature sequence of each time step; based on the global feature sequence of each time step, obtaining the decision result of whether each hydrological station needs to perform gate opening and closing operations, and feeding back the decision result to the corresponding hydrological station.

[0061] In practice, this invention can utilize a global unidirectional model to traverse local feature sequences along the time axis in a forward direction. Get the global hidden state Its essence lies in integrating the micro-characteristics of various hydrological stations to capture the overall temporal evolution pattern of watershed hydrological data (such as the chain reaction of increasing rainfall across the entire watershed on water levels at various hydrological stations), avoiding the limitation of focusing only on local aspects while ignoring the overall correlation. Based on this, for each time step... Through the final global state Determining whether to open or close the gates allows for a comprehensive approach that considers both the specific data characteristics of individual hydrological stations and the overall hydrological trends of the basin. This ensures a holistic view of decision-making. Furthermore, the rapid feedback of decision-making results establishes a closed loop from analysis to execution, significantly improving the response speed of gate scheduling and effectively avoiding the lag and bias inherent in traditional experience-based decision-making.

[0062] Furthermore, in specific implementation, in the above steps, based on the global feature sequence of each time step, the decision result of whether each hydrological station needs to perform gate opening and closing operations is obtained, and the decision result is fed back to the corresponding hydrological station. Specifically, this may include: analyzing the global feature sequence of each time step, combining it with a preset decision threshold, to obtain a preliminary decision result of whether each hydrological station needs to perform gate opening and closing operations; obtaining gate opening and closing decision information after reviewing the preliminary decision result, and feeding back the gate opening and closing decision information to the corresponding hydrological station so that the corresponding hydrological station can review the gate opening and closing decision information again.

[0063] In implementation, this invention first generates preliminary decision results by analyzing global feature sequences and combining them with thresholds. These preliminary results are then reviewed by data center staff. After review, the corresponding gate opening / closing decision information is fed back to the corresponding hydrological station, where supervisors conduct a second review. Based on the review results, the decision to open or close the gate is made. This approach achieves both high efficiency and foresight in decision-making through global time-series analysis of the model, effectively avoids the risk of misjudgment that may exist with a single model through multiple rounds of review, and involves frontline hydrological stations in the final verification. This ensures that the decision aligns with both data patterns and actual field conditions, significantly improving the accuracy and reliability of gate scheduling decisions.

[0064] Furthermore, in specific implementation, in the above-mentioned hydrological station monitoring method provided in the embodiments of the present invention, while performing step S102 to obtain the future rainfall information of each hydrological station, it may also include: feeding back the future rainfall information of each hydrological station to the corresponding hydrological station, so that the corresponding hydrological station updates the future rainfall information to the corresponding hydrological element data.

[0065] In implementation, the data center transmits future rainfall information from each hydrological station back to the corresponding station, which then incorporates this information into its hydrological data system for updates. Since future rainfall is a core predictive basis for changes in key elements such as water level and flow velocity, timely updates to the hydrological station data allow the standalone platform to possess more comprehensive hydrological forecast information. This provides a more forward-looking data foundation for the next round of data reporting and model analysis, avoiding analytical delays caused by a lack of future precipitation information, thereby improving the predictability and accuracy of subsequent decision-making.

[0066] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0067] Embodiments of the present invention also provide a hydrological station monitoring device. Figure 2 This is a schematic diagram of the hydrological station monitoring device provided in an embodiment of the present invention. This embodiment is based on functional modules, such as… Figure 2 As shown, the device is used in a data center and includes:

[0068] Data receiving module 10 is used to receive hydrological element data reported by each hydrological station;

[0069] The information acquisition module 11 is used to acquire future rainfall information for each hydrological station based on hydrological element data;

[0070] The dataset generation module 12 is used to integrate hydrological element data from various hydrological stations and future rainfall information to generate the input dataset for the hydrological decision network model; the hydrological decision network model includes a local bidirectional model and a global unidirectional model.

[0071] The local feature generation module 13 is used to input the input dataset into the local bidirectional model for computation and generate a local feature sequence.

[0072] The decision result acquisition module 14 is used to take the local feature sequence as the input parameter of the global one-way model, process it through the global one-way model, output the decision result of the corresponding hydrological station gate opening and closing, and feed the decision result back to the corresponding hydrological station.

[0073] In the hydrological station monitoring device provided in this embodiment of the invention, the interaction of the above five modules can first ensure the comprehensiveness and timeliness of the model input data, laying a data foundation for accurate decision-making; then, the local bidirectional model can be used to deeply mine the bidirectional correlation characteristics of the surrounding hydrological data at each time step, and the global unidirectional model can be used to grasp the overall temporal change pattern. The combination of the two greatly improves the accuracy of gate opening and closing decisions, effectively avoiding the subjectivity and lag of relying on experience judgment in traditional monitoring; at the same time, the decision results can be quickly fed back to the corresponding hydrological station, realizing efficient connection of the entire process from data collection and analysis to dispatching instructions, which not only improves the response speed of hydrological station gate dispatching, but also enhances the initiative and foresight of watershed hydrological monitoring, providing strong technical support for flood control and disaster reduction, and optimal allocation of water resources.

[0074] Since the embodiments of the hydrological station monitoring device and the hydrological station monitoring method correspond to each other, the descriptions of the features in the embodiments corresponding to the hydrological station monitoring device can be found in the relevant descriptions of the embodiments corresponding to the hydrological station monitoring method, and will not be repeated here. Furthermore, it has the same beneficial effects as the hydrological station monitoring method mentioned above.

[0075] Furthermore, in specific implementation, in the hydrological station monitoring device provided in the embodiments of the present invention, the data center is deployed with a server including a clustered version of the hydrological management platform, and each hydrological station is deployed with an edge server including a wireless communication component and a standalone version of the hydrological management platform; each hydrological station establishes a network communication link with the data center through the wireless communication component; the clustered version of the hydrological management platform is used to manage each standalone version of the hydrological management platform.

[0076] Accordingly, the data receiving module 10 can be used to receive hydrological element data reported by each hydrological station through the stand-alone hydrological management platform at set time intervals; the hydrological element data includes time-series data of water level, flow velocity, evaporation and rainfall collected by sensors, water level data calculated based on water gauge images monitored by cameras and the corresponding water gauge images.

[0077] Furthermore, in a specific implementation, in the hydrological station monitoring device provided in the embodiments of the present invention, the local feature generation module 13 can be used to align the relevant data in the input dataset according to a set time step to form a multi-dimensional original sequence with a unified time axis; and input the multi-dimensional original sequence into the local bidirectional model for calculation to generate a local feature sequence.

[0078] The process involves inputting the multi-dimensional original sequence into a local bidirectional model for computation to generate a local feature sequence. Specifically, this includes: extracting data segments from the original multi-dimensional sequence before and after each time step as local analysis windows; extracting forward and backward features within the local analysis windows using the local bidirectional model, merging them to obtain the local features corresponding to each time step; and integrating the local features corresponding to each time step to generate a local feature sequence. Specifically, the data corresponding to the local analysis window can be input into the local bidirectional model, and computation can be performed along the forward and backward time axes to extract the forward and backward hidden states corresponding to each time step, thus obtaining the forward and backward features within the local analysis window; the forward and backward features within the local analysis window can then be merged to obtain the local features corresponding to each time step.

[0079] Furthermore, in specific implementation, the hydrological station monitoring device provided in the embodiments of the present invention may further include: an attention mechanism module, used to activate an attention mechanism based on key event detection; according to historical data of each water station, to set corresponding benchmark intervals for hydrological element parameters through statistical methods; to calculate the deviations of each parameter of each time step within the local analysis window from the corresponding benchmark interval, and, in combination with the distance between each time step and the current time step, to determine the coefficients used to adjust the weights of the forward hidden state and the backward hidden state corresponding to each time step; and to use the adjusted forward hidden state as the forward feature within each local analysis window, and the adjusted backward hidden state as the backward feature within each local analysis window.

[0080] Specifically, for the forward hidden state corresponding to each time step, the deviations of water level, flow rate, and evaporation at each time step from the corresponding baseline interval are calculated, and the forward weighting factor is obtained by combining the distance between each time step and the current time step; for the backward hidden state corresponding to each time step, the deviations of rainfall at each time step from the corresponding baseline interval are calculated, and the backward weighting factor is obtained by combining the distance between each time step and the current time step; based on the forward weighting factor and the backward weighting factor, the aggregation ratio of the forward hidden state and the backward hidden state corresponding to each time step is adjusted.

[0081] Furthermore, in a specific implementation, in the hydrological station monitoring device provided in the embodiments of the present invention, the decision result acquisition module 14 can be specifically used to perform traversal calculations on the local feature sequence along the positive time axis using a global unidirectional model to obtain the global feature sequence of each time step; based on the global feature sequence of each time step, obtain the decision result of whether each hydrological station needs to perform gate opening and closing operations, and feed the decision result back to the corresponding hydrological station.

[0082] Specifically, the system can analyze the global feature sequence at each time step and, in conjunction with a preset decision threshold, obtain a preliminary decision on whether each hydrological station needs to perform gate opening and closing operations. It can also obtain gate opening and closing decision information after reviewing the preliminary decision results and feed the gate opening and closing decision information back to the corresponding hydrological station so that the corresponding hydrological station can review the gate opening and closing decision information again.

[0083] Furthermore, in specific implementation, the hydrological station monitoring device provided in the embodiments of the present invention may further include: an accuracy judgment module, used to judge the accuracy of water level data in hydrological element data and then feed the judgment result back to the corresponding hydrological station so that the corresponding hydrological station updates the water level data according to the judgment result; after the hydrological station updates the local water level data, the water level data reported by the corresponding hydrological station is updated.

[0084] Embodiments of the present invention also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above embodiments of the hydrological station monitoring method.

[0085] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program configured to execute the steps in any of the above embodiments of the hydrological station monitoring method when running.

[0086] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0087] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the hydrological station monitoring method.

[0088] Embodiments of the present invention also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above embodiments of the hydrological station monitoring method.

[0089] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0090] The above provides a detailed description of the hydrological station monitoring method and equipment provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only intended to help understand the method and core ideas of this invention. It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from its principles, and these improvements and modifications also fall within the protection scope of this invention.

Claims

1. A method for monitoring hydrological stations, characterized in that, For use in data centers, including: Receive hydrological data reported by various hydrological stations; Based on the aforementioned hydrological data, future rainfall information for each hydrological station is obtained; The hydrological element data from each hydrological station and the future rainfall information are integrated to generate the input dataset for the hydrological decision network model; the hydrological decision network model includes a local bidirectional model and a global unidirectional model. The relevant data in the input dataset are aligned according to the set time step to form a multi-dimensional original sequence with a unified time axis; Data segments within a set time interval before and after each time step in the multi-dimensional original sequence are extracted as local analysis windows. The data corresponding to the local analysis window is input into the local bidirectional model, and calculations are performed along the forward and reverse time axes to extract the forward and backward hidden states corresponding to each time step, thereby obtaining the forward and backward features within the local analysis window. The forward and backward features within the local analysis window are merged to obtain the local features corresponding to each time step. The local features corresponding to each time step are integrated to generate a local feature sequence. Simultaneously, an attention mechanism based on key event detection is activated; based on historical data from each water station, corresponding benchmark intervals are set for hydrological element parameters using statistical methods; the deviations of each parameter at each time step within the local analysis window from the corresponding benchmark intervals are calculated, and combined with the distance between each time step and the current time step, coefficients for adjusting the weights of the forward and backward hidden states corresponding to each time step are determined; the coefficients for adjusting the weights of the forward or backward hidden states corresponding to each time step are determined using the following formula: ； ； ； in, For time step The values ​​of water level parameters, This is the lower limit of the baseline interval. The upper limit of the baseline interval. To measure time steps The difference between the values ​​of water level parameters and the corresponding benchmark interval. For the current time step, The length parameter of the local analysis window. For time step Distance from the current time step To synthesize the data from each time step within the local analysis window and The weighting coefficients, These are coefficients used to adjust the weights of the forward or backward hidden states at each time step. The adjusted forward hidden state is used as the forward feature in each of the local analysis windows, and the adjusted backward hidden state is used as the backward feature in each of the local analysis windows. The local feature sequence is used as the input parameter of the global one-way model. After processing by the global one-way model, the decision result of the corresponding hydrological station gate opening and closing is output, and the decision result is fed back to the corresponding hydrological station.

2. The hydrological station monitoring method according to claim 1, characterized in that, The data center is equipped with servers containing a clustered version of the hydrological management platform, and each hydrological station is equipped with an edge server containing wireless communication components and a standalone version of the hydrological management platform; each hydrological station establishes a network communication link with the data center through the wireless communication components. The clustered version of the hydrology system is used to manage each of the standalone hydrology management platforms. Receive hydrological data reported by various hydrological stations, including: Receive hydrological element data reported by each hydrological station through the stand-alone hydrological management platform at set time intervals; The hydrological data includes time-series data collected by sensors, including water level, flow velocity, evaporation, and rainfall; water level data calculated based on water level gauge images monitored by cameras; and the corresponding water level gauge images.

3. The hydrological station monitoring method according to claim 1, characterized in that, Calculate the deviations of various parameters at each time step within the local analysis window from the corresponding baseline interval, and, in conjunction with the distance between each time step and the current time step, determine the coefficients used to adjust the weights of the forward and backward hidden states corresponding to each time step, including: For the forward hidden state corresponding to each time step, calculate the deviation of the water level, flow rate, evaporation rate and the corresponding baseline interval for each time step, and combine the distance between each time step and the current time step to obtain the forward weighting factor. For the backward hidden state corresponding to each time step, calculate the deviation between the rainfall at each time step and the corresponding baseline interval, and combine the distance between each time step and the current time step to obtain the backward weighting factor. Based on the forward weighting factor and the backward weighting factor, adjust the aggregation ratio of the forward hidden state and the backward hidden state corresponding to each time step.

4. The hydrological station monitoring method according to claim 1, characterized in that, After processing by the global one-way model, the decision result for the gate opening and closing of the corresponding hydrological station is output, and the decision result is fed back to the corresponding hydrological station, including: The global unidirectional model is used to traverse the local feature sequence along the positive time axis to obtain the global feature sequence at each time step; By analyzing the global feature sequences at each time step and combining them with preset decision thresholds, preliminary decision results are obtained on whether each hydrological station needs to perform gate opening and closing operations. Obtain the gate opening / closing decision information after reviewing the preliminary decision results, and feed the gate opening / closing decision information back to the corresponding hydrological station so that the corresponding hydrological station can review the gate opening / closing decision information again.

5. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the hydrological station monitoring method as described in any one of claims 1 to 4 when executing the computer program.

Images