A reservoir water conservancy monitoring system based on multi-sensor fusion technology

Abstract

The present invention relates to the field of open-air water source survey technology, and specifically to a reservoir water conservancy monitoring system based on multi-sensor fusion technology. The present invention first collects and processes reservoir water conservancy survey data of the reservoir through a variety of sensors, including water source parameter measurement data and reservoir structure survey data; then comprehensively analyzes the hydrological changes of the reservoir water conservancy, the stress changes, pressure changes and vibration strain changes of the reservoir structure from three perspectives: statistical characteristics, time series characteristics and spatial characteristics, and obtains flood assessment indicators of different dimensions; finally, flood warning information is output according to the analysis results. The present invention improves the safety monitoring accuracy of reservoir water conservancy during the flood season by surveying the hydrological and structural changes of the reservoir structure at multiple points, and provides real-time and intelligent risk warning and decision support for open-air water source safety survey management.

Description

Technical Field

[0001] The invention relates to the technical field of open-air water source survey, and in particular to a reservoir water conservancy monitoring system based on multi-sensor fusion technology. Background Art

[0002] Reservoir water conservancy monitoring refers to the systematic monitoring and evaluation of the water conservancy environment of the reservoir and its surroundings to ensure the safety and efficient use of the reservoir. Reservoir water conservancy monitoring includes internal monitoring and external monitoring; internal monitoring mainly focuses on the internal structure of the reservoir and the state of the water body, while external monitoring focuses more on the environment and meteorological conditions around the reservoir.

[0003] The safety of reservoirs is directly related to the safety of life and property of downstream residents. Especially during the flood season, reservoirs are not only a barrier to flood control, but also an important basis for ecological protection to ensure water supply. Reservoirs can not only effectively regulate floods, relieve water level pressure downstream and reduce the risk of flood disasters, but also store abundant precipitation and provide valuable water resources in the dry season.

[0004] Due to the particularity of water conservancy conditions during the flood season, reservoir water conservancy monitoring during this period requires higher requirements than reservoir water conservancy monitoring during normal periods. In the prior art, reservoir water conservancy monitoring during the flood season mainly focuses on early warning monitoring of water conditions. Through multi-type and multi-point distributed sensors, neural network models and other means are used to compare historical data and real-time data to identify abnormal conditions and risks of reservoirs during the flood season. Although the above method has achieved early warning and assessment of water conditions during the flood season, it also has some shortcomings. For example, before the flood season arrives, the geological structure of the reservoir dam will be subject to more complex impact conditions than usual, and flow rate and flow parameters will also change in a relatively short period of time, which may cause sudden cracks or leakage. The prior art has not conducted in-depth research on the impact of reservoir structure impact conditions under such extreme conditions as the flood season on flood warnings.

[0005] In order to solve the above problems, the present invention proposes a reservoir water conservancy monitoring system based on multi-sensor fusion technology. Summary of the invention

[0006] The purpose of the present invention is to provide a reservoir water conservancy monitoring system based on multi-sensor fusion technology, which improves the accuracy of safety monitoring of reservoir water conservancy during flood season by monitoring the hydrological and structural changes of reservoir structures at multiple points, and provides real-time and intelligent risk warning and decision support for water conservancy project safety management. The present invention first collects and processes the sensor data of the reservoir, and then comprehensively analyzes the hydrological changes of the reservoir water conservancy, the structural characteristic data changes of the reservoir structure, and the vibration strain changes from the three perspectives of statistical characteristics, time series characteristics, and spatial characteristics, and obtains flood assessment indicators of different dimensions; finally, flood warning information is output according to the analysis results.

[0007] To achieve the above object, the present invention provides a reservoir water conservancy monitoring system based on multi-sensor fusion technology, a data acquisition and processing unit: used to collect reservoir water conservancy survey data and pre-process it, including a data acquisition module and a pre-processing module; the reservoir water conservancy survey data includes water source survey data and reservoir structure data; a preliminary feature extraction unit: used to extract time features and spatial features from the reservoir water conservancy survey data;

[0008] Data fusion and analysis unit: used to fuse and analyze the reservoir water conservancy survey data from different sensors; use statistical analysis methods to perform time series analysis on the water source survey data to obtain the first flood assessment index; extract the time series features in the reservoir structure data to construct a linear regression equation for the reservoir structure, predict the stress change and pressure change of the reservoir structure, and obtain the second flood assessment index; construct a reservoir structure safety assessment model to monitor the leakage of the reservoir structure in real time, and obtain the third flood assessment index;

[0009] Decision-making and response unit: used to adjust the risk warning threshold and output the corresponding warning information according to the results of the fusion analysis; feedback maintenance unit: record all the reservoir water conservancy survey data and decision response measures, build an open water source survey database to adjust the monitoring threshold and optimize the monitoring model.

[0010] Furthermore, the steps of the data collection and processing unit include: monitoring the key water source parameters of the reservoir water area, triggering the collection and transmission of the reservoir water conservancy survey data when the key water source parameters reach a set threshold; dynamically adjusting the data collection frequency in combination with changes in the reservoir environment;

[0011] The collected reservoir water conservancy survey data is converted into a time series format, outliers are removed and missing parts are completed; the completed survey data is smoothed and processed into an independent data set according to data characteristics.

[0012] Furthermore, the step of extracting the temporal and spatial features includes: dividing the pre-processed reservoir water conservancy survey data into time series data in time windows of fixed length; calculating the rate of change and acceleration of the time series data in each time window; applying Fourier transform to convert the time series data into frequency threshold signals, extracting the main frequency and energy spectrum, and obtaining the temporal features of the reservoir structure state;

[0013] For the preprocessed reservoir water conservancy survey data, a spatial data set is constructed according to different monitoring points in the reservoir; the correlation coefficients between the monitoring points are calculated to evaluate the correlation between the survey data at different locations, and the spatial features are extracted.

[0014] Furthermore, the process of obtaining the first flood assessment index includes: for the water source survey data, calculating the data statistics in each time window, and applying the autoregressive moving average method to model the time series data; the autoregressive moving average model is expressed as:

[0015] ;

[0016] in, for The water source survey data value at the moment, is the intercept, and are the parameters of the autoregressive part and the moving average part, respectively. and are the orders of the autoregressive and moving average parts, respectively; For the past The water source survey data value at a moment; For historical moments The error term, is the white noise error term;

[0017] Collect the real-time water source survey data values ​​and predict them through the autoregressive moving average model Hydrological data at the moment , and obtain the first flood situation assessment index.

[0018] Furthermore, the process of obtaining the second flood situation assessment index includes:

[0019] The time characteristics of the reservoir structure data are constructed as a time characteristic vector, and the correlation analysis is used to select a stress-related temporal characteristic, and pressure-related time features, forming a feature set ;in, For feature concentration The moment The stress characteristics, For feature concentration The moment Pressure characteristics;

[0020] For feature set Standardized into structural characteristic data; the stress changes of the reservoir structure With pressure changes A linear regression equation is constructed between each of the structural feature data, and the mean square error is used to perform fitting optimization on the linear regression equation; the linear regression equation is:

[0021] ;

[0022] in, and They are The first The stress characteristics and Pressure characteristics, and is the intercept term, and Respectively The stress characteristics and The regression coefficient of each pressure feature, and is the error term;

[0023] Collect the real-time reservoir structure data and predict the Reservoir structure stress data at the moment and pressure data , and obtain the second flood situation assessment index.

[0024] Furthermore, the process of obtaining the third flood situation assessment index includes:

[0025] For the reservoir structure data, The temporal and spatial features of each moment are processed into a spatiotemporal feature matrix ,in, is the number of time steps, is the number of monitoring points, The data dimension for each monitoring point;

[0026] The spatiotemporal feature matrix is ​​input into the convolution kernel with a size of Convolution operation is performed in the CNN to extract the time-space relationship between different monitoring points in the reservoir structure ;in, is the convolution window size in the time dimension, and is the convolution window size in spatial dimension;

[0027] Will The time series characteristics of the vibration and strain of the reservoir structure are captured by inputting into the bidirectional gated loop control unit, and the predicted leakage of the reservoir structure is output through the fully connected layer and the activation function to obtain the third flood situation assessment index.

[0028] Furthermore, the decision-making and response unit includes: continuously collecting and preprocessing the reservoir water conservancy survey data, and continuously outputting the first flood situation assessment index, the second flood situation assessment index and the third flood situation assessment index through the data fusion and analysis unit, which are respectively recorded as , and ;

[0029] Set up the first warning mechanism when , and When the set threshold is exceeded, an early warning is triggered and the response early warning information is output; , and When the set threshold is not exceeded, the second early warning mechanism is adopted.

[0030] Furthermore, the second early warning mechanism is: when the first flood assessment index The second flood assessment indicator and the third flood assessment indicator When the set threshold is not exceeded, the specified time step is continuously calculated. The rate of change of indicators within ; express Moment The rate of change of each flood assessment indicator;

[0031] ;

[0032] When the indicator change rate exceeds the set safety value, an early warning is triggered and a response early warning message is output.

[0033] Furthermore, the step of outputting the response warning information includes:

[0034] According to the characteristics of reservoir water body and structure, the first flood assessment index is defined The second flood assessment indicator and the third flood assessment indicator Develop a risk level matrix based on the threshold range;

[0035] When an alert is triggered, , and The respective threshold ranges are adjusted dynamically , and The weight coefficient is used to calculate the matrix score and output the corresponding warning information.

[0036] Furthermore, the feedback maintenance unit includes: recording all the collected reservoir water conservancy survey data, evaluation indicators and warning information into the system database, and optimizing the evaluation indicator acquisition model in the data fusion and analysis unit through the error between the actual feedback and the warning information.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. The present invention divides the pre-processed reservoir water conservancy survey data into time series data, extracts the time features in the sensor data; then, according to the deployment position of the sensor, calculates the correlation between different monitoring data to obtain spatial features; the time features and spatial features not only reflect the dynamic changes of the reservoir structure status data over time and the correlation of the spatial layout, but also provide streamlined data input for the subsequent more complex comprehensive evaluation of flood conditions. By extracting the time series features in the reservoir structure data, a linear regression equation is constructed to predict the stress changes and pressure changes of the reservoir structure, thereby achieving accurate monitoring of the safety status of the reservoir structure; the second flood assessment index obtained can reflect the potential risks of the reservoir structure in terms of structural deformation and stress concentration. By real-time monitoring of these key changes, the relevant personnel of the reservoir management can take necessary reinforcement and repair measures for the reservoir in the early stage of the flood. The stress change and pressure change prediction made by this method for the reservoir structure provides scientific data and theoretical basis for the survey and management decision-making of open water sources.

[0039] 2. The present invention extracts the time series characteristics of reservoir structure vibration and strain through convolutional neural network, and then combines the bidirectional gated cyclic unit to analyze the time-space relationship of different monitoring points in the reservoir structure, conducts dynamic monitoring of vibration and strain, and evaluates the leakage risk of the reservoir in real time to generate the third flood assessment index. In the calculation situation during the flood season, the impact changes on the reservoir structure are more complex than usual. This method identifies potential leakage hazards in the structure through a deep understanding of the vibration and strain of the reservoir structure, and predicts possible leakage in the future. Through leakage prediction, reservoir management personnel can know whether there are potential risks and specific risk situations in the reservoir before the flood, and reduce the impact of leakage on the safety of the reservoir and the water body. This method collects and analyzes multi-source data in real time, monitors the change pattern of the state of open-air water source water conservancy structure, and provides a scientific basis for more accurate flood warning assessment.

[0040] 3. The present invention continuously collects and pre-processes sensor data, and continuously outputs the first, second and third flood assessment indicators through the data fusion and analysis unit; under the first warning mechanism, when the assessment indicator exceeds the set threshold, the warning is triggered and the response warning information is output; under the second warning mechanism, the assessment indicators do not exceed the set threshold, and the fluctuation of the assessment indicators is calculated; when the fluctuation is large, the warning is also triggered. The present invention formulates a risk level matrix according to the threshold range of the three assessment indicators. When the warning is triggered, the weights are dynamically adjusted according to the threshold range of the three assessment indicators, the matrix score is calculated, and the corresponding response warning information is output. The present invention improves the sensitivity and real-time performance of the flood warning response by real-time monitoring and continuous calculation of the indicator change rate; the set risk level matrix combined with the assessment indicator threshold can provide clear warning information and scientific and reasonable decision-making basis for the management of reservoirs, water conservancy and other open-air water source structures, and realize more accurate flood warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a reservoir water conservancy monitoring system based on multi-sensor fusion technology;

[0042] Figure 2 It is a structural diagram of a reservoir water conservancy monitoring system based on multi-sensor fusion technology;

[0043] Figure 3 This is a schematic diagram of the structure of the early warning response mechanism. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] In order to achieve more accurate flood warning assessment by monitoring the changes in the reservoir structure status before the flood season, the present invention proposes a reservoir water conservancy monitoring system based on multi-sensor fusion technology. Figure 1 and Figure 2 This application will describe the specific implementation methods of the flood situation assessment method and early warning mechanism in the reservoir water conservancy monitoring system proposed by the present invention through Example 1 and Example 2.

[0046] Embodiment 1

[0047] As an embodiment of the present invention, refer to Figure 1S10 in the embodiment is applied to a data acquisition and preprocessing module of a reservoir water conservancy monitoring system based on multi-sensor fusion technology. This embodiment collects sensor data of a medium-sized reservoir A, including water level, pressure (including water pressure and pressure on the reservoir structure), flow, temperature, vibration, strain and other data that can reflect the operating status of the reservoir.

[0048] Furthermore, a water level threshold a and a flow rate threshold b are set. When the water level and flow rate monitored by the sensor exceed the threshold, the data acquisition module is triggered to collect reservoir water conservancy survey data, including water source survey data and reservoir structure data, and the collected data is transmitted to the back-end processing unit of the system for further analysis and processing. The water level threshold a and flow rate threshold b are adjusted according to the actual situation, and the frequency of sensor collection is dynamically adjusted according to changes in the reservoir environment.

[0049] The collected reservoir water conservancy survey data is converted from sensor format to time series format. Each record contains the sensor type, timestamp and corresponding value. For example, water level data can be represented as , is the timestamp, For the corresponding The water level value at a certain moment.

[0050] The Z-Score statistical method is used to detect outliers in the data, and the missing data is completed by linear interpolation. After the data is completed, the sensor data is smoothed using a sliding average method, and the data is divided into independent data sets according to the sensor type.

[0051] This embodiment triggers the collection and transmission of reservoir water conservancy survey data by setting the threshold of water source parameters, and pre-processes the collected sensor data, including outlier detection and missing value completion, which can eliminate the accuracy caused by noise data and missing data, and ensure the accuracy of subsequent analysis and modeling. In addition, the volatility of reservoir detection data is reduced by smoothing, providing consistent and reliable data input for subsequent feature extraction.

[0052] Further, refer to Figure 1 S20 in the figure is used for a preliminary feature extraction unit of a reservoir water conservancy monitoring system based on multi-sensor fusion technology. Specifically, the pre-processed time series data is divided into time windows of fixed length. The length of the time window is determined by the sampling frequency of the sensor.

[0053] For the reservoir water conservancy survey data in each time window, Calculate the data change rate; where, For the The data change rate within a time window, is the data value at the current moment, is the time interval between two time points;

[0054] Based on the data change rate, Calculate the acceleration of data change; where, is the acceleration, and is the data change rate between two adjacent time points.

[0055] For the reservoir water conservancy survey data in each time window, Fourier transform is applied to convert the time domain signal into the frequency domain signal, and the main frequency and energy spectrum are extracted to obtain the time characteristics of the reservoir structure state;

[0056] In this embodiment, the reservoir water conservancy survey data of each monitoring point can be represented as a multi-dimensional vector containing multiple types of sensor data; is the number of monitoring points, is the data dimension for each monitoring point; each monitoring point The sensor data is represented as ,The Pearson correlation coefficient between the sensor data at different monitoring points was calculated to evaluate the correlation between the monitoring data.

[0057] The sensor data of different monitoring points and their correlations are used to form a spatial feature matrix, which describes the spatial relationship between different monitoring points. The spatial feature matrix is ​​reduced in dimension by principal component analysis to extract spatial features.

[0058] By extracting temporal and spatial features, this embodiment enables the system to refine key data and use it as input for subsequent analysis. Temporal features help capture trends and anomalies in reservoir hydrological and structural changes, while spatial features explain the correlation between different monitoring points and the dynamic behavior of local areas, providing high-quality input data for subsequent data analysis and model building.

[0059] Further, refer to Figure 1 S30, S40 and S50 in the above, S30, S40 and S50 are applied to a data fusion and analysis unit of a reservoir water conservancy monitoring system based on multi-sensor fusion technology, and the reservoir water conservancy survey data from different sensors are fused and analyzed;

[0060] Using a statistical analysis method to perform time series analysis on the water source survey data to obtain a first flood assessment index; extracting time series features from the reservoir structure data to construct a multivariate autoregressive model of the reservoir structure, predicting stress and pressure changes of the reservoir structure, and obtaining a second flood assessment index;

[0061] A convolutional neural network is used to extract the time-space relationship of different monitoring points in the reservoir structure data, and a bidirectional gated cyclic unit is used to process the vibration and strain characteristics, and a reservoir structure safety assessment model is constructed to monitor the leakage of the reservoir structure in real time, and obtain the third flood assessment index. Specifically:

[0062] Basic statistics are calculated for the water source survey data in each time window, including mean, variance, maximum and minimum values. The autoregressive moving average method is used to model the time series data; the autoregressive moving average model is expressed as:

[0063] ;

[0064] in, for The water source survey data value at the moment, is the intercept, and are the parameters of the autoregressive part and the moving average part, respectively. and are the orders of the autoregressive and moving average parts, respectively; For the past The water source survey data value at a moment; For historical moments The error term, is the white noise error term;

[0065] In this example, the intercept term in the autoregressive moving average model is The value is the mean of the time window, and the parameters of the autoregressive part are adjusted by the mean and variance. and order ; The standard deviation is used to adjust the parameters of the moving average part and coefficient .

[0066] Collect real-time water source survey data and predict it through the autoregressive moving average model Hydrological data at the moment , obtain the first flood assessment index. This embodiment collects water source survey data of reservoir A simulating flood season water conditions, with a collection frequency of 1 time / second and a time window size of 10; the autoregressive moving average model is used to predict the future time Hydrological data at all times, obtain future time The first flood situation assessment indicator at the moment. Some examples of collected data and predicted data are given in Table 1.

[0067] Table 1 Some examples of obtaining the first flood assessment index

[0068]

[0069] Table 1 gives the hydrological data forecasts for six time points. For time point t11, the predicted water level and predicted flow are normalized and linearly weighted to obtain the first flood situation assessment index of 0.15.

[0070] This embodiment calculates basic statistics for water source survey data in each time window and combines it with the autoregressive moving average algorithm to build a model, which can accurately predict future hydrological changes and obtain the first flood assessment index to provide basic hydrological data support for the overall flood assessment. The basic statistics provide a basis for adjusting the parameters of the autoregressive moving average model, improve the accuracy and adaptability of the prediction, and provide a reliable flood assessment for reservoir management.

[0071] The process of obtaining the second flood situation assessment index includes: constructing the time characteristics of the reservoir structure data into a time characteristic vector, selecting a stress-related temporal characteristic, and pressure-related time features, forming a feature set ;in, For feature concentration The moment The stress characteristics, For feature concentration The moment Pressure characteristics;

[0072] For feature set Standardized into structural characteristic data; the stress changes of the reservoir structure With pressure changes A linear regression equation is constructed between each of the structural feature data, and the mean square error is used to perform fitting optimization on the linear regression equation; the linear regression equation is:

[0073] ;

[0074] in, and They are The first The stress characteristics and Pressure characteristics, and is the intercept term, and Respectively The stress characteristics and The regression coefficient of each pressure feature, and is the error term;

[0075] Collect the real-time reservoir structure data and predict the Reservoir structure stress data at the moment and pressure data , and obtain the second flood situation assessment index.

[0076] This example collects sensor data of reservoir A's simulated flood season water conditions, with a collection frequency of 1 time / second and a time window size of 10; the linear regression equation is used to predict the future time Stress and pressure data at the moment, obtain future time Table 2 gives some examples of the second flood situation assessment indicators at the time.

[0077] Table 2 Some examples of obtaining the second flood assessment indicators

[0078]

[0079] Table 2 gives the stress and pressure data predictions for six time points. For time point t11, the predicted pressure and predicted stress are normalized and weighted to obtain the second flood situation assessment index of 0.18.

[0080] This embodiment establishes a linear regression model between the stress change and pressure change of the reservoir structure and the time characteristic data, and uses mean square error optimization to improve the accuracy of stress and pressure prediction and ensure the reliability of real-time monitoring and prediction results; the linear regression model has low computational complexity and is suitable for real-time monitoring and emergency response of reservoir water conservancy, thereby improving the response speed and accuracy of the system. It provides a scientific basis for flood risk warning and reservoir safety management.

[0081] The process of obtaining the third flood situation assessment index includes: for the reservoir structure data, The temporal and spatial features of each moment are processed into a spatiotemporal feature matrix ,in, is the number of time steps, is the number of monitoring points, The data dimension for each monitoring point;

[0082] The spatiotemporal feature matrix is ​​input into the convolution kernel with a size of Convolution operation is performed in the CNN to extract the time-space relationship between different monitoring points in the reservoir structure ;in, is the convolution window size in the time dimension, and is the convolution window size in spatial dimension;

[0083] Will The time series characteristics of the vibration and strain of the reservoir structure are captured by inputting into the bidirectional gated loop control unit, and the predicted leakage of the reservoir structure is output through the fully connected layer and the activation function to obtain the third flood situation assessment index.

[0084] This embodiment collects reservoir structure data of reservoir A simulating flood season water conditions, with a collection frequency of 10 times / hour and a time window size of 5;

[0085] Predicting the future time through CNN-BiGRU The leakage of the reservoir structure at the moment, and obtain the future time The third flood situation assessment indicator at the moment.

[0086] Some examples are given in Table 3.

[0087] Table 3 Some examples of obtaining the third flood assessment index

[0088]

[0089] Table 1 gives the leakage predictions at five time points. For monitoring point 1, the predicted leakage rate is normalized to obtain the first flood situation assessment index of 0.08.

[0090] This embodiment combines the temporal and spatial characteristics of reservoir structure data, uses convolutional neural networks and bidirectional gated recurrent unit models to extract the temporal-spatial relationship between different monitoring points, and detects and evaluates the leakage of reservoir structures. Through the fusion of multi-dimensional features, the accuracy of reservoir leakage risk assessment is improved, and reservoir structure survey data can be processed in real time and future leakage risks can be predicted; the third flood assessment index obtained provides decision-making data support for reservoir survey management during the flood season.

[0091] Further, refer to Figure 1 S60 in the example is applied to a decision-making and response unit of a reservoir water conservancy monitoring system based on multi-sensor fusion technology, continuously collects and pre-processes reservoir water conservancy survey data, and continuously outputs the first flood situation assessment index, the second flood situation assessment index, and the third flood situation assessment index through a data fusion and analysis unit, which are respectively denoted as , and ;

[0092] Set up the first warning mechanism when , and When the set threshold is exceeded, an early warning is triggered and the response early warning information is output; , and When the set threshold is not exceeded, the second early warning mechanism is adopted.

[0093] The second early warning mechanism is: when the first flood assessment indicator The second flood assessment indicator and the third flood assessment indicator When the set thresholds of 0.2, 0.1 and 0.2 are not exceeded, the specified time step is continuously calculated. The rate of change of indicators within ; express Moment The rate of change of each flood assessment indicator;

[0094] ;

[0095] When the indicator change rate exceeds the set safety value, an early warning is triggered and a response early warning message is output.

[0096] In this embodiment, reservoir water survey data of reservoir A is continuously collected and preprocessed, and the first flood situation assessment index, the second flood situation assessment index and the third flood situation assessment index are continuously output through the data fusion and analysis unit; Table 4 gives an example of the calculation of the change rate of the assessment index of monitoring point 3.

[0097] Table 4 Some examples of change rates of flood assessment indicators

[0098]

[0099] According to the index change rate calculation in Table 4, the first, second and third flood situation assessment index change rate thresholds are set to 20.0%, 50% and 25% respectively, then at time point 1 and time point 3, monitoring point 3 triggers an early warning and outputs a response early warning information. In this embodiment, various thresholds are set by those skilled in the art based on the specific situation of reservoir A, combined with historical data and expert experience fitting.

[0100] This embodiment can realize dynamic monitoring of reservoirs and accurate flood prediction by continuously collecting and preprocessing reservoir water survey data and combining the data fusion unit to continuously calculate and output the first, second and third flood assessment indicators. The dual early warning mechanism established ensures timely response in different situations, improving the real-time and effectiveness of water assessment and early warning during the flood season.

[0101] This embodiment sets up a second early warning mechanism to continuously monitor the change rate of each indicator when the first, second and third flood assessment indicators do not exceed the set thresholds, ensuring that potential risks can be discovered in a timely manner. By dynamically calculating the change rate and setting a safety value, the system can trigger an early warning when the indicator does not exceed the set threshold but changes rapidly, further enhancing the sensitivity and response capability of the reservoir in extreme situations such as flood season.

[0102] Furthermore, the feedback maintenance unit includes: recording all collected data, evaluation indicators and warning information into a system database, and optimizing the evaluation indicator acquisition model in the data fusion and analysis unit through the error between the actual feedback and the warning information.

[0103] This embodiment introduces a feedback maintenance mechanism into the reservoir water conservancy monitoring system, which can continuously optimize the evaluation index acquisition model and improve the flood forecasting accuracy of the system. By recording the collected reservoir water conservancy survey data, flood assessment indicators and warning information into the system data, and comparing and analyzing the errors between the actual feedback and warning information, the system can achieve dynamic adjustment and self-optimization, and enhance the prediction ability of future flood conditions.

[0104] Embodiment 2

[0105] As an embodiment of the present invention, refer to Figure 3 , Figure 3 It is used to explain the specific implementation method of the second early warning mechanism in a reservoir water conservancy monitoring system based on multi-sensor fusion technology proposed by the present invention.

[0106] For medium-sized reservoir A, the first flood assessment index is defined based on the characteristics of its water body and structure. The second flood assessment indicator and the third flood assessment indicator Table 5 shows the risk level matrix specified for the water body and structural characteristics of medium-sized reservoir A.

[0107] Table 5 Some examples of change rates of flood assessment indicators

[0108]

[0109] When the warning of monitoring point 4 of reservoir A is triggered, , and , dynamically adjust , and The weight coefficient is used to calculate the matrix score of monitoring point 4 to be 0.0868, and the corresponding warning information is output as "medium risk, attention". In this embodiment, the thresholds and weights are set by technicians in this field according to the specific situation of reservoir A, combined with historical data and expert experience.

[0110] This embodiment improves the accuracy and flexibility of flood risk assessment by formulating a dedicated risk level matrix for the reservoir and dynamically adjusting the weight of the assessment indicators. When the assessment indicators of the monitoring point trigger an early warning, the system can quickly calculate the score based on real-time data and output clear early warning information to help managers take effective measures in a timely manner to ensure the safe operation of the reservoir during the flood season.

[0111] In order to achieve more accurate flood warning assessment by monitoring the changes in the reservoir structure status before the flood season, this embodiment proposes a reservoir water conservancy monitoring system based on multi-sensor fusion technology for medium-sized reservoir A. By integrating multiple sensor technologies and data analysis methods, the problem of insufficient accuracy and timeliness of reservoir monitoring in the extreme case of flood season in the existing technology is solved. This embodiment obtains reservoir water conservancy survey data through the data acquisition and processing unit, and extracts key time features and spatial features in combination with the preliminary feature extraction unit, captures the dynamic changes of the water body and structure of the reservoir, and lays the foundation for subsequent analysis.

[0112] In this embodiment, the data fusion and analysis unit uses a statistical analysis method to perform time series analysis on the water source survey data, thereby obtaining the first evaluation index; by constructing a linear regression equation for the reservoir structure, the stress and pressure of the reservoir structure are predicted to obtain the second flood assessment index; a convolutional neural network and a bidirectional gated loop unit are used to deeply analyze the spatiotemporal relationship in the reservoir structure data, monitor the leakage of the reservoir structure, and obtain the third flood assessment index. The multi-angle evaluation improves the monitoring accuracy and timeliness of reservoir safety during the flood season. This embodiment dynamically adjusts the risk warning threshold according to the fusion analysis results, and outputs the corresponding warning information in a timely manner; in addition, the feedback maintenance unit records all reservoir water conservancy survey data and corresponding decision-making measures, builds a database to optimize the monitoring model and adjust the threshold, and ensures that the system always maintains the best monitoring state.

[0113] This embodiment uses multi-sensor fusion technology to provide a comprehensive and accurate flood season water conservancy monitoring and evaluation tool for the reservoir, which improves the accuracy of monitoring the changes in the reservoir structure status before the flood season. It not only provides reliable data support and decision-making basis for timely flood warning and reservoir safety management, but also provides theoretical and technical references for the safety monitoring and operation management of similar open-air water source structures.

[0114] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A reservoir water conservancy monitoring system based on multi-sensor fusion technology, characterized in that: include: Data acquisition and processing unit: used to collect reservoir water conservancy survey data and perform preprocessing, including data acquisition module and preprocessing module; The reservoir water conservancy survey data includes water source survey data and reservoir structure data; a preliminary feature extraction unit: used to extract time features and space features from the reservoir water conservancy survey data; Data fusion and analysis unit: used to fuse and analyze the reservoir water conservancy survey data from different sensors; use statistical analysis methods to perform time series analysis on the water source survey data to obtain the first flood assessment index; extract the time series features in the reservoir structure data to construct a linear regression equation for the reservoir structure, predict the stress and pressure changes of the reservoir structure, and obtain the second flood assessment index; Constructing a reservoir structure safety assessment model to monitor the leakage of the reservoir structure in real time and obtain the third flood assessment index; The process of obtaining the first flood situation assessment indicator includes: For the water source survey data, the data statistics in each time window are calculated, and the time series data are modeled using the autoregressive moving average method; the autoregressive moving average model is expressed as: ； in, for The water source survey data value at the moment, is the intercept, and are the parameters of the autoregressive part and the moving average part, respectively. and are the orders of the autoregressive and moving average parts, respectively; For the past The water source survey data value at a moment; For historical moments The error term, is a white noise error term; collect the real-time water source survey data value, and predict it through the autoregressive moving average model Hydrological data at the moment , obtain the first flood situation assessment index; The process of obtaining the second flood situation assessment index includes: The time characteristics of the reservoir structure data are constructed as a time characteristic vector, and the correlation analysis is used to select a stress-related temporal characteristic, and pressure-related time features, forming a feature set ;in, For feature concentration The moment The stress characteristics, For feature concentration The moment pressure features; for feature set Standardized into structural characteristic data; the stress changes of the reservoir structure With pressure changes A linear regression equation is constructed between each of the structural feature data, and the mean square error is used to perform fitting optimization on the linear regression equation; the linear regression equation is: ； in, and They are The first The stress characteristics and Pressure characteristics, and is the intercept term, and Respectively The stress characteristics and The regression coefficient of each pressure feature, and is the error term; collect the real-time reservoir structure data and predict Reservoir structure stress data at the moment and pressure data , obtain the second flood situation assessment index; The process of constructing the reservoir structure safety assessment model to obtain the third flood situation assessment index includes: For the reservoir structure data, The temporal and spatial features of each moment are processed into a spatiotemporal feature matrix ,in, is the number of time steps, is the number of monitoring points, The data dimension for each monitoring point; The spatiotemporal feature matrix is ​​input into the convolution kernel with a size of Convolution operation is performed in the CNN to extract the time-space relationship between different monitoring points in the reservoir structure ;in, is the convolution window size in the time dimension, and is the convolution window size in spatial dimension; Will The time series characteristics of the vibration and strain of the reservoir structure are captured by inputting into the bidirectional gated loop control unit, and the predicted leakage of the reservoir structure is output through the fully connected layer and the activation function to obtain the third flood situation assessment index; Decision-making and response unit: used to adjust the risk warning threshold and output the corresponding warning information according to the results of the fusion analysis; Feedback maintenance unit: record all the reservoir water conservancy survey data and decision response measures, build an open water source survey database to adjust the monitoring threshold and optimize the monitoring model; The decision and response unit comprises: Continuously collect and preprocess the reservoir water conservancy survey data, and continuously output the first flood situation assessment index, the second flood situation assessment index and the third flood situation assessment index through the data fusion and analysis unit, which are respectively recorded as , and ; Set up the first warning mechanism, when , and When the set threshold is exceeded, an early warning is triggered and the response early warning information is output; , and When the set threshold is not exceeded, a second early warning mechanism is adopted; The second early warning mechanism is: When the first flood assessment indicator The second flood assessment indicator and the third flood assessment indicator When the set threshold is not exceeded, the specified time step is continuously calculated. The rate of change of indicators within ; express Moment The change rate of each flood assessment indicator; ； When the indicator change rate exceeds the set safety value, an early warning is triggered and a response early warning message is output; The step of outputting the response warning information comprises: According to the characteristics of reservoir water body and structure, the first flood assessment index is defined The second flood assessment indicator and the third flood assessment indicator The threshold range is set to develop a risk level matrix; when the warning is triggered, , and The respective threshold ranges are adjusted dynamically , and The weight coefficient is used to calculate the matrix score and output the corresponding warning information.

2. A reservoir water conservancy monitoring system based on multi-sensor fusion technology according to claim 1, characterized in that: The steps of the data acquisition and processing unit include: Monitor the key water source parameters of the reservoir water area, and trigger the collection and transmission of the reservoir water conservancy survey data when the key water source parameters reach the set threshold; dynamically adjust the data collection frequency in combination with changes in the reservoir environment; The collected reservoir water conservancy survey data is converted into a time series format, outliers are removed and missing parts are completed; the completed survey data is smoothed and processed into an independent data set according to data characteristics.

3. A reservoir water conservancy monitoring system based on multi-sensor fusion technology according to claim 1, characterized in that: The step of extracting the temporal features and the spatial features comprises: Divide the pre-processed reservoir water conservancy survey data into time series data in time windows of fixed length; calculate the rate of change and acceleration of the time series data in each time window; apply Fourier transform to convert the time series data into frequency threshold signals, extract the main frequency and energy spectrum, and obtain the time characteristics of the reservoir structure state; For the preprocessed reservoir water conservancy survey data, a spatial data set is constructed according to different monitoring points in the reservoir; the correlation coefficients between the monitoring points are calculated to evaluate the correlation between the survey data at different locations, and the spatial features are extracted.

4. A reservoir water conservancy monitoring system based on multi-sensor fusion technology according to claim 1, characterized in that: The feedback maintenance unit includes recording all the collected reservoir water conservancy survey data, evaluation indicators and early warning information into the system database, and optimizing the evaluation indicator acquisition model in the data fusion and analysis unit through the error between the actual feedback and the early warning information.

Images