A water conservancy project monitoring and early warning method based on big data analysis

Abstract

The present application relates to the field of big data analysis, and more particularly to a water conservancy project monitoring and early warning method based on big data analysis. The content includes: real-time acquisition of sensor data of monitoring points, and standardization processing, generating standardized sensor data vector; based on the standardized sensor data vector, a directed weighted dynamic graph is constructed through a double-layer directed dynamic coupling correlation graph construction algorithm; based on the directed weighted dynamic graph, a comprehensive early warning index is calculated through a risk perception weighted aggregation algorithm, and it is judged whether the early warning is triggered. The technical problems of different sampling frequencies of various sensors, non-uniform dimensions, and large differences in signal feature distribution, which lead to lack of unified scale in subsequent modeling analysis; traditional methods ignore the potential physical connection or information transmission path between monitoring points, leading to lagging risk prediction or lack of interpretability; and relying on static or experience set alarm threshold, false alarm and missed alarm phenomenon exist.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, and in particular to a method for monitoring and early warning of water conservancy projects based on big data analytics. Background Technology

[0002] Against the backdrop of ever-increasing demands for water conservancy project operation and disaster prevention and mitigation, traditional static monitoring and rule-based early warning methods are no longer sufficient to meet the practical needs of complex hydrogeological conditions, nonlinear disaster evolution processes, and the fusion analysis of multi-source monitoring data. With the rapid development of IoT technology, sensor networks, remote sensing monitoring, big data processing, and intelligent analysis algorithms, data-driven intelligent monitoring and early warning technologies have gradually become a research and practice hotspot, and a key technological path to improve engineering safety and build a smart water conservancy system. Both its theoretical research and practical deployment have significant value and broad prospects, not only promoting the digital transformation of water conservancy projects but also providing strong technical support for the construction of disaster prevention and mitigation systems.

[0003] However, traditional data-driven intelligent monitoring and early warning technologies still suffer from problems such as difficulty in integrating multi-source heterogeneous monitoring data, insufficient ability to model risk propagation relationships, insufficient robustness of risk early warning methods, and failure of fixed thresholds. Summary of the Invention

[0004] This invention provides a water conservancy project monitoring and early warning method based on big data analysis. It addresses the problems of inconsistent sampling frequencies, dimensions, and signal characteristic distributions among various sensors deployed in current water conservancy projects. These issues lead to a lack of unified standards in subsequent modeling and analysis, hindering fair integration and comparison, and affecting the reliability of data mining. Traditional data-driven intelligent monitoring and early warning methods treat each monitoring point as an independent information source, ignoring potential physical connections or information propagation paths between monitoring points, resulting in often delayed or uninterpretable risk predictions. Furthermore, relying on statically or empirically set alarm thresholds cannot adapt to dynamic and complex hydrological conditions, leading to false alarms and missed alarms, severely restricting the practicality of early warning systems.

[0005] The present invention provides a water conservancy project monitoring and early warning method based on big data analysis, which specifically includes the following technical solutions: A method for monitoring and early warning of water conservancy projects based on big data analysis includes the following steps: S1. Collect sensor data from monitoring points in real time, perform standardization processing to obtain standardized sensor data, and construct a standardized sensor data vector; based on the standardized sensor data vector, construct a directed weighted dynamic graph through a two-layer directed dynamic coupling association graph construction algorithm. S2. Based on the directed weighted dynamic graph, a comprehensive early warning index is calculated by integrating the anomaly degree of the monitoring point itself with the risk impact of the external monitoring point through a risk perception weighted aggregation algorithm. Based on the comprehensive early warning index, it is determined whether an early warning is triggered.

[0006] Preferably, S1 specifically includes: In the implementation of the two-layer directed dynamic coupling association graph construction algorithm, the monitoring point is used as the node, and a directed weighted dynamic graph is constructed by combining the directed edge weights; the directed edge weights are obtained by weighting the spatial structure term and the temporal causal term.

[0007] Preferably, S1 specifically includes: By combining the spatial distance and elevation difference between any two monitoring points, and using a power-law form to reflect the attenuation of propagation intensity with distance, a spatial structure term is constructed.

[0008] Preferably, S1 specifically includes: A historical time window is introduced to extract standardized sensor data vectors for any pair of monitoring points. The dimensions of the extracted standardized sensor data vectors are combined pairwise to form dimension pairs, and the maximum mutual information coefficient of the dimension pairs is calculated. Combined with a lag traversal strategy, a time causal term is constructed.

[0009] Preferably, S2 specifically includes: In the implementation of the risk perception weighted aggregation algorithm, the importance weight of the project is introduced to integrate the anomaly degree of the monitoring point itself with the risk impact of the external monitoring point to obtain a comprehensive early warning index.

[0010] Preferably, S2 specifically includes: In the implementation of the risk perception weighted aggregation algorithm, the directed in-degree centrality of the monitoring points is calculated based on the directed edge weights in the directed weighted dynamic graph, and the importance weight of the project is obtained.

[0011] Preferably, S2 specifically includes: In the implementation of the risk perception weighted aggregation algorithm, the anomaly degree of the monitoring point is calculated based on the standardized sensor data and the mean of the standardized sensor data within the historical time window. Based on the anomaly degree of the monitoring point and the directed edge weights in the directed weighted dynamic graph, the impact of the incoming risk from external monitoring points is quantified.

[0012] Preferably, S2 specifically includes: When the comprehensive early warning index is greater than or equal to the early warning threshold, an early warning is triggered; otherwise, no early warning is issued and monitoring continues. The early warning threshold is adaptively updated by performing sliding window statistical analysis on the historical time series of the comprehensive early warning index.

[0013] The beneficial effects of the technical solution of the present invention are: 1. Based on the three-dimensional spatial coordinates and elevation information of monitoring points in the engineering design drawings, a spatial structure item is constructed to accurately reflect the spatial topological path of risk propagation between monitoring points; a time causal item is constructed by adopting the maximum mutual information coefficient and the lag traversal strategy to explore the lag causal strength, breaking through the traditional modeling method based solely on distance, and realizing the dual coupling modeling of spatial physical logic and time dynamic causality, thereby enhancing the physical rationality and time foresight of risk perception.

[0014] 2. The time causal term is calculated by performing maximum mutual information coefficient (MIC) analysis on historical data between any two monitoring points. This can complete the calculation of causal strength without relying on the same type of sensor, breaking the dependence of traditional monitoring on homogeneous data. It realizes unsupervised causal chain modeling between heterogeneous monitoring points, effectively expanding the early warning coverage and generalization capabilities.

[0015] 3. By constructing a fully connected directed weighted dynamic graph, the spatial propagation path and temporal dependency of risk among different monitoring points are dynamically modeled. It supports real-time updates of directed edge weights, reflects the evolution characteristics of risk propagation paths over time, and has good risk chain tracking and link propagation modeling capabilities.

[0016] 4. A risk perception weighted aggregation algorithm is introduced, which integrates the anomaly degree of the monitoring point itself with risk input information from other monitoring points, and uses the importance weight of the project as an adjustment factor to effectively avoid single-point false alarms and signal interference, and realizes dynamic and robust early warning indicator output under multi-source information fusion.

[0017] 5. By introducing directed in-degree centrality, the degree of influence of each monitoring point in the risk propagation network is dynamically calculated, and the weight of the monitoring point in the comprehensive early warning index is determined accordingly. This makes the early warning more structurally sensitive, able to focus on the truly critical paths and key nodes, and improves the overall early warning efficiency. Attached Figure Description

[0018] Figure 1 This is a flowchart of a water conservancy project monitoring and early warning method based on big data analysis as described in this invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details a specific scheme for a water conservancy project monitoring and early warning method based on big data analysis provided by the present invention.

[0022] See attached document Figure 1 The diagram illustrates a flowchart of a water conservancy project monitoring and early warning method based on big data analysis, provided by an embodiment of the present invention. The method includes the following steps: S1. Collect sensor data from monitoring points in real time, perform standardization processing to obtain standardized sensor data, and construct a standardized sensor data vector; based on the standardized sensor data vector, construct a directed weighted dynamic graph through a two-layer directed dynamic coupling association graph construction algorithm. In water conservancy projects, monitoring points are set up, and sensor data is collected in real time at each monitoring point through sensors. For example, pore water pressure in the dam body, dam foundation and seepage field around the dam is monitored by piezometers, tensile and compressive stress state inside the concrete or steel reinforcement of the dam body is monitored by strain gauges, and water level is monitored by water level gauges, etc. Definition of the first Each monitoring point at time The sensor data vector is: , Indicates at time No. The first monitoring point Sensor-like data, Indicates the first The number of sensors at each monitoring point varies from point to point. To completely eliminate the differences in dimensions and statistical distributions between different monitoring points and sensors, and to achieve fair comparability of heterogeneous data, the collected sensor data is standardized using the existing Z-Score normalization preprocessing algorithm to obtain standardized sensor data, which then constitutes the first... Each monitoring point at time Standardized sensor data vector .

[0023] To accurately characterize the physical propagation path and dynamic causal relationship of internal risks in water conservancy projects, a fully connected directed weighted dynamic graph is constructed in real time using a two-layer directed dynamic coupling association graph construction algorithm. The nodes in the directed weighted dynamic graph are directly determined by the various monitoring points deployed in the water conservancy project, that is, each monitoring point is treated as an independent node in the directed weighted dynamic graph; The directed edge weights in a directed weighted dynamic graph are determined by introducing two weight coefficients. and The spatial structure term and the temporal causality term are determined by a weighted combination, and it is recommended that the weight coefficients be equal in weight. This is to balance the two dimensions of information: spatial layout and temporal transmission.

[0024] The formula for calculating the weight of a directed edge is: ; in, Indicates at time No. The monitoring point points to the first The directed edge weights of each monitoring point; Indicates at time No. The monitoring point points to the first Spatial structure items of each monitoring point; Indicates at time No. The monitoring point points to the first The time causal terms of each monitoring point; and These represent the weighting coefficients of the spatial structure term and the temporal causality term, respectively. Recommended value ; The spatial structure item is calculated based on the three-dimensional spatial coordinates and elevation information of the monitoring points in the engineering design drawings. Specifically, for any two monitoring points, the spatial structure item is constructed using the spatial distance and elevation difference between the two monitoring points. The influence of spatial distance is quantified using a power-law method, and a power-law factor is introduced to reflect the attenuation of propagation intensity with distance. The elevation difference is calculated by taking the difference in altitude between the two monitoring points to reflect their relative positional relationship in the vertical direction. The difference in altitude between the two monitoring points is then normalized by dividing the difference in altitude between the two monitoring points by the maximum elevation difference. The normalized altitude difference is further constrained to above zero using a maximum value function to effectively avoid negative value interference. The calculation formula is as follows: ; in, Indicates the first The monitoring point to the first The spatial distance between the monitoring points is calculated using Euclidean distance. The calculation method is a well-known technique in the art and will not be elaborated here. The spatial reference distance is defined as the average of the nearest neighbor distances among all monitoring points or the average of the first few nearest neighbor distances. The method for calculating the nearest neighbor distance of monitoring points is a well-known technique in the art and will not be elaborated here. This represents the power factor, used to control the decay rate. Based on historical monitoring data in the historical monitoring database (i.e., standardized historical sensor data), it is optimized using a data-driven minimum residual fitting method, with a value range of [value missing]. The minimum residual fitting method is a well-known technique in the art and will not be elaborated here. This represents the maximum elevation difference, used for normalization, and is obtained from hydraulic engineering design documents. and They represent the first The monitoring point and the first The altitude of each monitoring point is obtained through real-time monitoring. Indicates the first The monitoring point and the first Elevation difference between monitoring points; This represents the difference in altitude after normalization. This means that by using the maximum value function, the normalized altitude difference is limited to above zero, so as to effectively avoid negative value interference. The temporal causal term uses the maximum mutual information coefficient (MIC) within the historical time window combined with a lag traversal strategy to achieve unsupervised causal strength mining between monitoring points with heterogeneous and multidimensional data and no common sensors. The specific process is as follows: Set a length of... The historical time window is used to limit the time range of MIC analysis, considering only the data within the historical time window, while also setting the maximum lag time. , used to indicate the upper limit of the causal delay considered; Within the aforementioned historical time window, for any monitoring point... Extract the standardized sensor data vectors from each past moment; combine the dimensions of the two extracted standardized sensor data vectors pairwise to generate a total of... take There are 3 dimension pairs, each representing a possible causal path, and the calculation is performed as follows: First, the lag time is fixed. (From 0 to maximum lag time) (all integer values) and moments within the historical time window For each dimension pair between the two monitoring points, the maximum mutual information coefficient (MIC value) is calculated, and the maximum value is taken to obtain the result at the lag time. and time Optimal dimensional channel strength under given conditions; then, for a fixed lag time... Iterate through all moments within the historical time window. The lag time is obtained by taking the maximum value of all the optimal dimensional channel strengths. The strongest association strength within the window; finally, iterate through all lag times. Take the maximum value of the strongest correlation strength within all windows obtained, and use it as the correlation strength from the current time point to the first window. The monitoring point to the first The formula for calculating the time causality term for each monitoring point is: ; in, Indicates the lag time; This represents the maximum lag time. The initial value was determined using expert experience and then optimized using cross-validation. Indicates the first Each monitoring point at time The Standardized sensor data; Indicates the first Each monitoring point at time The Standardized sensor data; Indicates traversing the first... All sensor dimensions at the monitoring point and the first Dimension pairs of all sensor dimensions at each monitoring point; Represents the maximum mutual information coefficient; This indicates the length of the historical time window, which is determined based on actual needs. Indicates the first Number of sensors at each monitoring point; By using directed weighted dynamic graphs to achieve true cross-modal risk transmission modeling, causal chains can be accurately captured even when the sensor types at monitoring points are completely different, significantly improving the lead time for early warning.

[0025] S2. Based on the directed weighted dynamic graph, a comprehensive early warning index is calculated by integrating the anomaly degree of the monitoring point itself with the risk impact of the external monitoring point through a risk perception weighted aggregation algorithm. Based on the comprehensive early warning index, it is determined whether an early warning is triggered.

[0026] Based on the directed weighted dynamic graph, a comprehensive early warning indicator is generated by integrating the anomaly degree of each monitoring point with the risk impact of external monitoring points through a risk perception weighted aggregation algorithm. The risk perception weighted aggregation algorithm adopts a weighted linear combination framework, which takes into account both the anomaly degree of each monitoring point and the risk impact from external monitoring points, and integrates them through the engineering importance weight of the monitoring points; The anomaly degree of the monitoring point itself is obtained by calculating the average deviation between the standardized sensor data and the average of the standardized sensor data within the historical time window, which can eliminate the randomness of single-point fluctuations. The risk impact from external monitoring points is defined as the potential risk value propagated from all non-self monitoring points to the current monitoring point. It is calculated based on the directed coupling strength between monitoring points and the anomaly degree of the external monitoring points themselves. The directed coupling strength between monitoring points is the directed edge weight between monitoring points. During the weighted fusion process, the engineering importance weight of the monitoring points controls whether the monitoring points are more inclined to trust their own monitoring or externally transmitted anomalies in the overall risk assessment. The formula for calculating the comprehensive early warning index is: ; in, Indicates at time Comprehensive early warning indicators; Indicates all Each monitoring point is traversed and weighted to accumulate the risk contribution of each monitoring point; Indicates the first The engineering importance weight of each monitoring point is used to determine the relative proportion of the anomaly of the monitoring point itself and the anomalies of external monitoring points in the calculation of the comprehensive early warning index. The calculation formula is: ; in, Indicates the first The directed in-degree centrality of a given number of monitoring points is expressed as follows: , Indicates from the first The monitoring point to the first The directional coupling strength of each monitoring point , Indicates at time No. The monitoring point points to the first The directed edge weights of each monitoring point; The coupling strength threshold is obtained by statistically analyzing the empirical distribution of directed coupling strength in historical monitoring data, and the 90th percentile value is taken as the significance dividing point, i.e., the coupling strength threshold. This indicates an indicator function that determines the starting point from the first... The monitoring point to the first Directed coupling strength at each monitoring point Does it exceed the coupling strength threshold? If the value exceeds the limit, the indicator function outputs 1, indicating that there is a significant risk transmission path; otherwise, the indicator function is recorded as 0. This represents the sum of the directed in-degree centralities of all monitoring points, used for normalization; Indicates the first The anomaly degree of each monitoring point, that is, at time t. No. The first monitoring point The average deviation between the standardized sensor data and the mean of the standardized sensor data within the historical time window. , Indicates the first Number of sensors at each monitoring point Indicates at time No. The first monitoring point Standardized sensor data Indicates a historical time window Inner The first monitoring point The mean of the sensor data after class standardization; This indicates the risk impact from external monitoring points; Indicates the first The anomaly degree of each monitoring point; By combining real-time monitoring anomalies at the monitoring points themselves with coupled risk information from external monitoring points, potential risk sources and risk propagation chains can be dynamically captured. When the comprehensive early warning indicator is greater than or equal to the early warning threshold, an early warning is triggered immediately; otherwise, no early warning is triggered and monitoring continues. The early warning threshold is updated in real time by performing sliding window statistical analysis on the historical time series of comprehensive early warning indicators, thus avoiding the false alarm and missed alarm problems of traditional fixed threshold methods under complex working conditions.

[0027] In summary, a method for monitoring and early warning of water conservancy projects based on big data analysis has been developed.

[0028] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0029] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0030] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for monitoring and early warning of water conservancy projects based on big data analysis, characterized in that, Includes the following steps: S1. Collect sensor data from monitoring points in real time, perform standardization processing to obtain standardized sensor data, and construct a standardized sensor data vector; based on the standardized sensor data vector, construct a directed weighted dynamic graph through a two-layer directed dynamic coupling association graph construction algorithm. S2. Based on the directed weighted dynamic graph, a comprehensive early warning index is calculated by integrating the anomaly degree of the monitoring point itself with the risk impact of the external monitoring point through a risk perception weighted aggregation algorithm. Based on the comprehensive early warning index, it is determined whether an early warning is triggered.

2. The water conservancy project monitoring and early warning method based on big data analysis according to claim 1, characterized in that, S1 specifically includes: In the implementation of the two-layer directed dynamic coupling association graph construction algorithm, the monitoring point is used as the node, and a directed weighted dynamic graph is constructed by combining the directed edge weights; the directed edge weights are obtained by weighting the spatial structure term and the temporal causal term.

3. The water conservancy project monitoring and early warning method based on big data analysis according to claim 2, characterized in that, S1 specifically includes: By combining the spatial distance and elevation difference between any two monitoring points, and using a power-law form to reflect the attenuation of propagation intensity with distance, a spatial structure term is constructed.

4. The water conservancy project monitoring and early warning method based on big data analysis according to claim 2, characterized in that, S1 specifically includes: A historical time window is introduced to extract standardized sensor data vectors for any pair of monitoring points. The dimensions of the extracted standardized sensor data vectors are combined pairwise to form dimension pairs, and the maximum mutual information coefficient of the dimension pairs is calculated. Combined with a lag traversal strategy, a time causal term is constructed.

5. The water conservancy project monitoring and early warning method based on big data analysis according to claim 1, characterized in that, S2 specifically includes: In the implementation of the risk perception weighted aggregation algorithm, the importance weight of the project is introduced to integrate the anomaly degree of the monitoring point itself with the risk impact of the external monitoring point to obtain a comprehensive early warning index.

6. The water conservancy project monitoring and early warning method based on big data analysis according to claim 5, characterized in that, S2 specifically includes: In the implementation of the risk perception weighted aggregation algorithm, the directed in-degree centrality of the monitoring points is calculated based on the directed edge weights in the directed weighted dynamic graph, and the importance weight of the project is obtained.

7. The water conservancy project monitoring and early warning method based on big data analysis according to claim 5, characterized in that, S2 specifically includes: In the implementation of the risk perception weighted aggregation algorithm, the anomaly degree of the monitoring point is calculated based on the standardized sensor data and the mean of the standardized sensor data within the historical time window. Based on the anomaly degree of the monitoring point and the directed edge weights in the directed weighted dynamic graph, the impact of the incoming risk from external monitoring points is quantified.

8. The water conservancy project monitoring and early warning method based on big data analysis according to claim 1, characterized in that, S2 specifically includes: When the comprehensive early warning index is greater than or equal to the early warning threshold, an early warning is triggered; otherwise, no early warning is issued and monitoring continues. The early warning threshold is adaptively updated by performing sliding window statistical analysis on the historical time series of the comprehensive early warning index.

Images