Water ecological index data processing method and system combining Internet of Things and artificial intelligence

Abstract

The invention discloses a water ecological index data processing method and system combining the Internet of Things and artificial intelligence, and belongs to the technical field of water body monitoring and analysis, and the method comprises the steps: grouping sensors, synchronizing a clock, and forming a space grouping data flow; comparing the grouped data with the steady-state baseline in real time, triggering high-density event acquisition, and outputting context data pulses; obtaining and standardizing mapping environment data, and generating a multi-dimensional spatio-temporal feature cube; dynamically distributing weights to calculate a water ecological index, quantifying credibility, and outputting a dynamic water ecological index; and when the credibility is lower than a threshold value, analyzing parameter contributions and tracing abnormities, generating and issuing a calibration strategy, and realizing grouping self-calibration and sampling adjustment. According to the method, multi-source collaborative fusion, dynamic weight distribution and an intelligent self-correction mechanism are adopted, the real-time performance, intelligence and evaluation accuracy of water ecology monitoring can be improved, and efficient and reliable data support and early warning capacity are provided for complex water environment changes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water body monitoring and analysis, in particular to a water ecological index data processing method and system combining Internet of Things and artificial intelligence. BACKGROUND

[0002] In the field of water environment protection and ecological monitoring, automatic collection and data analysis of physicochemical and biological parameters in water samples are important foundations for evaluating water ecological conditions, pollution trends and resource management. Currently, common water ecological monitoring methods mostly use fixed-point sampling, laboratory analysis or distributed sensor online detection, and related data is obtained through measuring instruments and then summarized and statistically analyzed.

[0003] However, the existing technology generally has limited sampling density, insufficient real-time performance, weak data integration and comprehensive evaluation capability when facing large-scale, multi-parameter and multi-temporal and spatial changes in water environment, especially it is difficult to achieve efficient, dynamic and intelligent water ecological data processing, which limits the accurate monitoring and timely warning of complex water environment changes. SUMMARY

[0004] To solve the above problems, the present application provides a water ecological index data processing method and system combining Internet of Things and artificial intelligence, which adopts multi-source collaborative fusion, dynamic weight distribution and intelligent self-correction mechanism, and can improve the real-time performance, intelligence and evaluation accuracy of water ecological monitoring, providing efficient and reliable data support and warning capability for complex water environment changes.

[0005] The above object can be achieved by the following scheme:

[0006] A water ecological index data processing method combining the Internet of Things and artificial intelligence, comprising: grouping water quality sensors and synchronizing clocks by using water body function zoning of a target water area, setting sensor group periods and collecting basic parameters to form a spatially grouped sensor data stream; comparing the spatially grouped sensor data stream in real time with a preset water body steady-state fluctuation baseline, collecting high-density data before and after an event when a certain deviation in the spatially grouped sensor data stream is detected, and generating a context data pulse; obtaining external environment data matching the context data pulse in terms of time and geographical location, aligning and standardizing the context data pulse and the external environment data, and generating a multi-dimensional spatio-temporal feature cube; calculating a water ecological index value based on the multi-dimensional spatio-temporal feature cube through dynamic weight distribution, synchronously quantitatively analyzing a reliability score, and outputting a dynamic water ecological index in combination with the water ecological index and the reliability score; when the reliability score is lower than a preset reliability threshold, analyzing the contribution of each parameter in the dynamic water ecological index, locating abnormal parameters, and tracing back to corresponding sensor groups to generate a monitoring and calibration strategy; and issuing the monitoring and calibration strategy to the sensor groups through the Internet of Things, and executing the monitoring and calibration strategy by the sensor groups to complete self-calibration and adjust subsequent sampling behavior.

[0007] Optionally, the forming of the spatially grouped sensor data stream comprises: configuring multiple types of water quality sensors by using the water body function zoning, collecting basic parameter data of each zoning to generate an original sensor data group; setting different sampling periods and sampling parameters for the original sensor data group based on different geographical zoning to obtain an ecological zoning sampling parameter set; performing clock synchronization and spatial identification coding on the original sensor data group and the ecological zoning sampling parameter set to generate a zoning synchronous sensor data set; and performing edge fusion preprocessing and real-time missing compensation on the zoning synchronous sensor data set to generate the spatially grouped sensor data stream.

[0008] Optionally, the generating of the context data pulse comprises: performing sliding window analysis on grouped water quality parameter changes in real time based on the spatially grouped sensor data stream and the water body steady-state fluctuation baseline, detecting dynamic fluctuation amplitudes of basic parameters, and generating an abnormal fluctuation marker; when an indicated parameter of the abnormal fluctuation marker fluctuates to a certain extent, automatically collecting continuous high-density monitoring data of time periods before and after an event, obtaining a time window, a spatial grouping identifier, and an event number corresponding to the high-density monitoring data, and packaging and outputting a context data pulse.

[0009] Optionally, the generating the multi-dimensional spatio-temporal feature cube comprises: acquiring meteorological data and remote sensing satellite image data consistent with the context data pulse in time stamp and geographic coordinates, mapping the meteorological data and the remote sensing satellite image data to corresponding spatio-temporal positions to generate an external environment data set; performing feature pairing and standardization processing on the context data pulse and the external environment data set to obtain standard multi-source data; and coupling internal sensor parameters and external environmental factors based on the standard multi-source data, and integrating to output a multi-dimensional spatio-temporal feature cube.

[0010] Optionally, the method further comprises: extracting a spatio-temporal pulse feature vector based on the context data pulse to generate a spatio-temporal pulse feature group; extracting an environment coupling feature based on the multi-dimensional spatio-temporal feature cube to generate an environment coupling feature group; and performing multi-source coupling fusion optimization based on the spatio-temporal pulse feature group and the environment coupling feature group to generate a coupling dynamic calibration index.

[0011] Optionally, the outputting the dynamic water ecological index in combination with the credibility score comprises: performing multi-layer adaptive weight distribution based on the multi-dimensional spatio-temporal feature cube to dynamically adjust the weights of various types of sensor parameters and external environmental factors, and generating a preliminary water ecological index value through weighted aggregation calculation; and based on the preliminary water ecological index value, fusing the coupling dynamic calibration index to perform feedback adjustment on the influence and correlation of each parameter to generate a credibility score, and outputting the dynamic water ecological index in linkage with the preliminary water ecological index value and the credibility score.

[0012] Optionally, the generating the monitoring calibration strategy comprises: performing anomaly detection and trend analysis on a contribution degree sequence of each parameter based on the credibility score of the dynamic water ecological index and the coupling dynamic calibration index to generate a parameter anomaly contribution degree set; analyzing spatio-temporal distribution and parameter change pulse based on the parameter anomaly contribution degree set and the spatio-temporal pulse feature group to determine an abnormal sensor grouping and generate a grouping anomaly diagnosis report; and evaluating abnormal causes and influence range according to the grouping anomaly diagnosis report and the environment coupling feature group to generate a monitoring calibration strategy.

[0013] Optionally, the generating the grouping anomaly diagnosis report comprises: performing multi-dimensional clustering analysis on the parameter anomaly contribution degree set and the spatio-temporal pulse feature group to hierarchically classify and categorize abnormal sensors in space to generate a hierarchical spatial anomaly grouping matrix; and analyzing coupling effects of abnormal grouping and environmental changes in combination with the environment coupling feature group for the hierarchical spatial anomaly grouping matrix, and outputting a grouping anomaly diagnosis report.

[0014] Optionally, the monitoring calibration strategy executed by the sensor group to complete self-calibration and adjust subsequent sampling behavior comprises: issuing the monitoring calibration strategy to the control node of the sensor group through the Internet of Things control network, the control node executing the strategy instruction and increasing the sampling frequency and duration, and generating a calibration and adjustment result record; and feeding back an abnormal state in the execution process to the system based on the calibration and adjustment result record.

[0015] Based on the same inventive concept, the application further provides a water ecological index data processing system combining the Internet of Things and artificial intelligence, which comprises: a spatial grouping collection module for grouping water quality sensors and synchronizing clocks by water body function zoning of a target water area, setting sensor group periods and collecting basic parameters to form a spatial grouping sensor data stream; an abnormality monitoring trigger module for comparing a preset water body steady-state fluctuation baseline with the spatial grouping sensor data stream in real time, collecting high-density data before and after an event when a certain deviation of the spatial grouping sensor data stream is detected, and generating context data pulses; a multi-source fusion processing module for obtaining external environment data matching the context data pulses in time and geographical location, aligning and standardizing the context data pulses and the external environment data, and generating a multi-dimensional spatio-temporal feature cube; an index calculation and reliability evaluation module for calculating a water ecological index value based on the multi-dimensional spatio-temporal feature cube through dynamic weight distribution, synchronously quantitatively analyzing a reliability score, and outputting a dynamic water ecological index combining the water ecological index and the reliability score; a calibration strategy generation module for analyzing the contribution degree of each parameter in the dynamic water ecological index when the reliability score is lower than a preset reliability threshold, locating abnormal parameters and tracing back to corresponding sensor groups, and generating a monitoring calibration strategy; and a strategy issuing and executing module for issuing the monitoring calibration strategy to the sensor groups through the Internet of Things, and executing the monitoring calibration strategy by the sensor groups to complete self-calibration and adjust subsequent sampling behavior.

[0016] Compared with the prior art, the application has the following advantages:

[0017] 1. Multi-source water body data automatic collection and intelligent fusion based on spatial grouping and dynamic timing are realized, which can realize full-area, all-weather and high-precision water ecological monitoring for complex water areas, and effectively improves the coverage range of data and the real-time analysis.

[0018] 2. Artificial intelligence dynamic weight distribution and multi-dimensional spatio-temporal feature analysis are introduced, combined with event-driven adaptive sampling and sensor self-calibration mechanism, which greatly enhances the evaluation scientificity of the water ecological index, the result reliability and the intelligent level of the system, and provides a solid data basis for early warning and regulation.

[0019] Other features and advantages of the present application will be set forth in the descriptions that follow, and in part will be apparent from the description, or can be learned by practice of the application. The purposes and other advantages of the present application will be realized and attained by the structures particularly pointed out in the written description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0021] Figure 1 is a flowchart of a water ecological index data processing method combining Internet of Things and artificial intelligence according to an embodiment of the present application.

[0022] Figure 2 is a multi-dimensional spatio-temporal feature cube graph according to an embodiment of the present application.

[0023] Figure 3 is a distribution graph of attention weight of multi-source feature coupling fusion according to an embodiment of the present application.

[0024] Figure 4 is a causal attribution thermal map of abnormal causes according to an embodiment of the present application.

[0025] Figure 5 is a structural schematic diagram of a water ecological index data processing system combining Internet of Things and artificial intelligence according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0027] With reference to Figure 1 An embodiment of the present application provides a water ecological index data processing method combining Internet of Things and artificial intelligence. The method adopts multi-source collaborative fusion, dynamic weight distribution and intelligent self-correction mechanism, and can improve the real-time performance, intelligence and evaluation accuracy of water ecological monitoring, and provides efficient and reliable data support and early warning capability for complex water environment changes.

[0028] The method of the embodiment specifically comprises:

[0029] The water quality sensor is grouped and the clock is synchronized by using the water body function zoning of the target water area, the sensor group period is set and the basic parameters are collected, and a spatially grouped sensor data stream is formed;

[0030] According to the preset water body steady-state fluctuation baseline, the spatially grouped sensor data stream is compared in real time, when a certain deviation of the spatially grouped sensor data stream is detected, high-density data before and after the event is collected, and context data pulses are generated;

[0031] External environment data matching the context data pulses in time and geographical position is obtained, the context data pulses and the external environment data are aligned and standardized, and a multi-dimensional spatiotemporal feature cube is generated;

[0032] Based on the multi-dimensional spatiotemporal feature cube, a water ecological index value is calculated by dynamic weight distribution, and a reliability score is quantitatively analyzed synchronously, and a dynamic water ecological index is output in combination with the water ecological index and the reliability score;

[0033] When the reliability score is lower than a preset reliability threshold, the contribution degree of each parameter in the dynamic water ecological index is analyzed, the abnormal parameter is located and traced back to the corresponding sensor group, and a monitoring calibration strategy is generated;

[0034] The monitoring calibration strategy is issued to the sensor group through the Internet of Things, and the sensor group executes the monitoring calibration strategy to complete self-calibration and adjust subsequent sampling behavior.

[0035] The present application adopts multi-source collaborative fusion, dynamic weight distribution and intelligent self-calibration mechanism, which can improve the real-time, intelligence and evaluation accuracy of water ecological monitoring, and provides efficient and reliable data support and early warning ability for complex water environment changes.

[0036] Optionally, the forming of the spatially grouped sensor data stream comprises:

[0037] By using the water body function zoning, a plurality of types of water quality sensors are configured, basic parameter data of each zoning is collected, and an original sensor data group is generated;

[0038] Specifically, this step aims to obtain the most original water ecological information with spatial heterogeneity from the physical world. According to the water function zoning, different sensors are configured in different functional areas. For example, in the industrial discharge area, electrochemical sensors for monitoring heavy metals and specific chemical pollutants are densely deployed; in the agricultural irrigation area, optical sensors for monitoring total phosphorus, total nitrogen, and chlorophyll are configured; and in the nature reserve area, biological sensors for monitoring microbial community activity and dissolved oxygen are deployed. These sensors collectively collect basic parameter data in each subarea, and the raw sensor data set is formed.

[0039] Based on different geographical subareas, different sampling periods and sampling parameters are set for the raw sensor data set to obtain an ecological subarea sampling parameter set.

[0040] Specifically, this step aims to intelligently allocate monitoring resources in different subareas based on prior knowledge. According to the environmental sensitivity and expected change rate of each water function subarea, a dedicated sampling strategy is set for it. For example, the sampling period of the industrial discharge area may be set to the minute level to capture sudden pollution events; while the sampling period of the nature reserve area with relatively stable hydrological conditions may be set to the hour level. All these differentiated sampling strategies for different subareas collectively constitute the ecological subarea sampling parameter set.

[0041] Clock synchronization and spatial identification encoding are performed on the raw sensor data set and the ecological subarea sampling parameter set to generate a subarea synchronous sensor data set.

[0042] Specifically, this step aims to give a unified space-time reference to heterogeneous and distributed raw data. In this process, clock synchronization can be achieved through the Network Time Protocol (NTP), ensuring that the timestamps of all Internet of Things terminals are aligned with a unified time center. Spatial identification coding is a data enhancement process that converts the water function subarea information of each data point into a digital vector through One-Hot Encoding, etc., and binds it as metadata with the data point. After processing, a subarea synchronous sensor data set with accurate time and spatial identity information for each data point is generated.

[0043] For the subarea synchronous sensor data set, edge fusion preprocessing and real-time missing compensation are performed to generate a spatial grouping sensor data stream.

[0044] In particular, this step aims to perform preliminary intelligent preprocessing on the edge computing node close to the data source before uploading the data to the cloud, in order to improve data quality and reduce transmission load. In this process, edge fusion processing refers to the real-time weighted average or consistency check of the readings of multiple sensors of the same type in the same group by the edge node to smooth the noise. Real-time missing compensation refers to the use of a lightweight time series prediction model by the edge node to estimate a compensation value in real time based on the historical data of the sensor and the data of adjacent sensors to ensure the continuity of the data stream. After this step, the spatially grouped sensor data stream is finally generated.

[0045] Optionally, the generating context data pulses comprises:

[0046] Based on the spatially grouped sensor data stream and the water body steady-state fluctuation baseline, a sliding window analysis is performed on the grouped water quality parameter changes in real time to detect the dynamic fluctuation amplitude of the basic parameters and generate an abnormal fluctuation marker.

[0047] In particular, this step aims to accurately and automatically identify statistically significant real environmental events from continuous and low-frequency spatially grouped sensor data streams, rather than simply sensor noise. In this process, the sliding window analysis uses a change point detection method based on information divergence. The water body steady-state fluctuation baseline is obtained by a probability density estimation using a Gaussian mixture model on a large amount of historical normal operation data in the corresponding water body functional partition by those skilled in the art, which contains the mean and covariance matrix describing the normal fluctuation range of the water quality parameters in the partition. At each time step, the data distribution in the sliding window is compared with the historical normal distribution defined by the water body steady-state fluctuation baseline, and the Kullback-Leibler divergence (KLD) between the two is calculated, which is used to quantify the difference between the two probability distributions. The calculation formula can be:

[0048]

[0049] where P(x) is the actual probability distribution of the water quality parameter in the current sliding window, and Q(x) is the historical reference probability distribution defined by the water body steady-state fluctuation baseline. The calculated Kullback-Leibler divergence value is used as a continuous and quantitative indicator to generate an abnormal fluctuation marker.

[0050] When the indicator parameter of the abnormal fluctuation marker fluctuates, continuous high-density monitoring data of a period before and after the event occurrence is automatically collected, and a time window, a spatial grouping identifier, and an event number corresponding to the high-density monitoring data are obtained, and context data pulses are packaged and output.

[0051] Specifically, this step is the execution link of event-triggered high-frequency sampling. When the Kullback-Leibler divergence value in the abnormal fluctuation marker exceeds a statistical threshold dynamically adjusted according to historical data, it is determined that an event requiring high-density collection has occurred. High-frequency sampling instructions are immediately issued to the sensor group, and continuous high-density monitoring data including a preset time window before and after the event occurrence are automatically collected and packaged. At the same time, a unique event number is assigned to this event, and the number is bound to the corresponding time window information and the spatial grouping identifier of the sensor group triggering the event. All these information are packaged together to form a structured, complete information, and a context data pulse as the final output.

[0052] Optionally, the generating a multi-dimensional spatio-temporal feature cube comprises:

[0053] Meteorological data and remote sensing satellite image data consistent with the timestamp and geographical coordinates of the context data pulse are obtained, and the meteorological data and the remote sensing satellite image data are mapped to corresponding spatio-temporal positions to generate an external environment data set;

[0054] Specifically, this step aims to collect external environmental information that may affect the instantaneous changes of water ecology from a macro perspective. In this process, the event time window and the spatial grouping identifier contained in the context data pulse are used as an accurate spatio-temporal query index. The processing module calls external meteorological data Application Programming Interface (API) and remote sensing satellite image database services according to the index to obtain rainfall, light intensity, wind speed, and Normalized Difference Vegetation Index (NDVI) and other multi-source external data in the same time and same geographical range, and integrates these data to form an external environment data set.

[0055] The context data pulse and the external environment data set are subjected to feature pairing and standardization and normalization processing to obtain standard multi-source data.

[0056] Specifically, this step aims to unify the internal and external data with different sources, formats, and dimensions, and lay the foundation for subsequent coupling analysis. In this process, feature pairing refers to establishing a correlation analysis pair between each internal sensor parameter in the context data pulse and multiple potential related factors in the external environment data set. Standardization refers to using methods such as max-min normalization to map the values of all paired features to the interval of 0 to 1, to eliminate the influence of different physical dimensions, and finally obtain standard multi-source data.

[0057] Based on the standard multi-source data, internal sensor parameters and external environmental factors are coupled through feature correlation analysis and feature selection, and a multi-dimensional spatio-temporal feature cube is output.

[0058] Specifically, this step is a key link for constructing the final analysis input, aiming to filter out core features with strong correlation from numerous possible internal-external factor associations, and structure them. In this process, a Maximal Information Coefficient (MIC) method that can capture nonlinear relationships is used to perform feature correlation analysis. By calculating the maximal information coefficient value between each pair of internal and external features, nonlinear, lag response relationships between, for example, rainfall and turbidity can be identified. Based on the correlation analysis results, feature pairs with correlation scores higher than a certain threshold are selected for final coupling, and organized into a tensor structure with time, space, and feature type as the three dimensions. This structure is a multi-dimensional spatio-temporal feature cube, as shown in Figure 2 The multi-dimensional spatio-temporal feature cube is visualized in the form of a three-dimensional body data in the figure, where the three coordinate axes represent time, space, and feature type, and the different gray-level voxels inside the cube intuitively represent the feature intensity of the coupled internal sensor parameters and external environmental factors at a specific spatio-temporal point.

[0059] Optionally, the method further comprises:

[0060] Based on the context data pulse, a spatio-temporal pulse feature vector is extracted, and a spatio-temporal pulse feature group is generated;

[0061] Specifically, this step aims to extract deep features that can accurately describe the internal dynamic evolution law of the event from the high-density, instantaneous context data pulse. In this process, a wavelet scattering transform technique with high sensitivity to transient signals is used. This technique can capture the stable internal structural features in the context data pulse that do not change with time translation, such as the steepness of the impact, the frequency and decay rate of the oscillation, etc. These extracted multi-scale features with physical interpretability collectively constitute the spatio-temporal pulse feature group.

[0062] extracting environment coupling features based on the multi-dimensional spatio-temporal feature cube to generate an environment coupling feature group;

[0063] Specifically, this step aims to learn and extract the complex and nonlinear coupling relationship between internal sensor parameters and external environmental factors from the multi-dimensional spatio-temporal feature cube. In this process, a three-dimensional convolutional neural network (3D CNN) is used. This network can process the multi-dimensional spatio-temporal feature cube as a whole, and its three-dimensional convolution kernel can slide in time, space and feature dimensions at the same time, effectively capturing and extracting deep coupling features such as "after the upstream rainfall event, the response pattern of the downstream turbidity sensor reading at a certain time delay". The feature map compressed by the pooling layer at the end of the network is the environment coupling feature group.

[0064] performing multi-source coupling fusion optimization based on the spatio-temporal pulse feature group and the environment coupling feature group to generate a coupling dynamic calibration index.

[0065] Specifically, this step aims to intelligently fuse two groups of heterogeneous features that describe the intrinsic dynamics of events and the coupling relationship between internal and external factors, respectively, to generate a top-level index that can evaluate overall self-consistency. In this process, a fusion network with an attention mechanism is used. The fusion network contains two parallel encoder input branches for receiving the spatio-temporal pulse feature group and the environment coupling feature group and encoding them into high-dimensional vectors; then, the environment coupling features are used as the context to calculate the importance weight of each element in the spatio-temporal pulse feature; finally, the attention-weighted feature vector is spliced with the original environment coupling feature vector, and an output layer containing multiple fully connected layers and activation functions is used to map it to a single value standardized to the range of 0 to 1, which is the coupling dynamic calibration index, as shown in Figure 3 As shown in the figure, the weight distribution logic of the attention mechanism in the multi-source coupling fusion optimization is visualized in the form of a sunburst chart, where the inner circle represents different feature groups, the outer circle represents specific features within each group, and the area of the sector is proportional to the attention weight allocated to the feature in the current fusion calculation.

[0066] Optionally, the outputting the dynamic water ecological index based on the combination of the water ecological index and the credibility score comprises:

[0067] based on the multi-dimensional spatio-temporal feature cube, performing multi-layer adaptive weight distribution to dynamically adjust the weights of various types of sensor parameters and external environmental factors, and generating a preliminary water ecological index value through weighted aggregation calculation;

[0068] Specifically, this step aims to intelligently aggregate the complex, multi-level information contained in the multi-dimensional spatio-temporal feature cube into a preliminary, comprehensive index. In this process, a multi-level adaptive weight distribution is achieved using a hierarchical attention network. The attention network contains two levels of attention mechanisms: the first level is the "factor-level attention", which calculates an initial weight for each internal sensor parameter based on the external environmental factors coupled with it; the second level is the "parameter-level attention", which receives all the weighted parameters from the first level as input and calculates a higher-dimensional distribution weight again according to the contribution of different parameter combinations to the overall state of the aquatic ecosystem. The training of this network uses supervised learning, with expert evaluation indices or recognized water quality standard grades in historical data as labels, and minimizes the mean square error between the predicted index and the label as the loss function. Through this multi-level weighting aggregation from local to global, it ensures that the final calculated preliminary water ecological index value can most accurately reflect the core state of the current aquatic ecosystem.

[0069] Based on the preliminary water ecological index value, the coupling dynamic calibration index is fused to feedback adjust the influence and correlation of each parameter, generate a credibility score, and output a dynamic water ecological index in linkage with the preliminary water ecological index value and the credibility score.

[0070] Specifically, this step aims to give an objective, quantitative credibility evaluation to the final output index and achieve intelligent linkage between the two. In this process, the coupling dynamic calibration index is taken as the core input. This index is fused into a probability model to feedback adjust the preliminary water ecological index value and generate the final credibility score. The probability model can be a logistic function that nonlinearly maps the coupling dynamic calibration index with a value range in the standardized interval to a probability value between 0 and 1, which is the credibility score. A higher coupling dynamic calibration index will make the final credibility score tend to 1; conversely, a lower index will significantly lower the credibility score. Finally, the preliminary water ecological index value and the credibility score generated after feedback adjustment are linked, for example, the index value is taken as the expected value of the probability distribution, and the inverse function of the credibility score is taken as the variance of the distribution, together forming a dynamic water ecological index as the final output that contains both the optimal estimate and its uncertainty boundary.

[0071] Optionally, the generating a monitoring calibration strategy comprises:

[0072] Based on the credibility score of the dynamic water ecological index and the coupled dynamic calibration index, the contribution degree sequence of each parameter is detected for abnormality and analyzed for trend, to generate a parameter abnormal contribution degree set;

[0073] Specifically, this step aims to identify the "culprit" from the parameter level when the overall credibility decreases. The reliability threshold is not a fixed value, but a self-adaptive threshold dynamically determined by statistical analysis of historical monitoring data using the Receiver Operating Characteristic Curve (ROC) analysis method to find the optimal balance point between false positive rate and false negative rate. In this process, the credibility score of the dynamic water ecological index and the coupled dynamic calibration index are continuously monitored. When both indicate cognitive bias, an abnormal detection model analyzes the historical contribution degree time series of each basic parameter that constitutes the index. The model filters out parameters with continuously abnormal or highly volatile contribution degree values through trend analysis and outlier detection algorithms, and these identified parameters and their abnormal contribution degree values together form the parameter abnormal contribution degree set.

[0074] Based on the parameter abnormal contribution degree set and the spatio-temporal pulse feature group, analyze the spatio-temporal distribution and parameter change pulse, determine the abnormal sensor group, and generate a group abnormal diagnosis report;

[0075] Specifically, this step aims to conduct in-depth diagnosis on the initially screened abnormal parameters to distinguish whether the problem is caused by sensor failure or by its intense response to real environmental events. In this process, the parameter abnormal contribution degree set is cross-coupled with the spatio-temporal pulse feature group. If a parameter's contribution degree is abnormal and its corresponding spatio-temporal pulse feature group shows no intense, physically meaningful energy pulse in the time window, it is likely that the abnormality is caused by the sensor's own quality problem. By tracing the spatio-temporal distribution of the parameter, the abnormal sensor group to which it belongs is finally determined, and a group abnormal diagnosis report containing problem positioning, abnormal timestamp, and preliminary diagnosis opinion is generated.

[0076] According to the group abnormal diagnosis report and the environmental coupling feature group, evaluate the cause of abnormality and the scope of influence, and generate a monitoring calibration strategy.

[0077] Specifically, this step is the decision-making link for generating the final calibration strategy, aiming to "cure the disease". In this process, the grouping abnormality diagnosis report is cross-coupled with the environmental coupling feature group for secondary cross-coupling analysis. This analysis aims to further assess the deep reasons and potential impacts of the abnormality. For example, if the diagnosis report indicates that a certain sensor grouping is abnormal, and the environmental coupling feature group shows that the readings of this grouping have an abnormal strong correlation with external rainfall, it can be inferred that there may be a waterproof problem with the sensor. According to this deep diagnosis result, through a pre-trained decision tree model with diagnosis attribution labels as input and specific instruction sets as output, a differentiated and targeted monitoring and calibration strategy is automatically and deterministically generated. For example, for the above case, the generated strategy will preferentially include the "self-cleaning and re-calibration" instruction set for this sensor grouping.

[0078] Optionally, the generating of the grouping abnormality diagnosis report comprises:

[0079] Performing multi-dimensional cluster analysis on the parameter abnormality contribution degree set and the spatio-temporal pulse feature group to hierarchically classify the abnormal sensors in space and generate a hierarchical spatial abnormal grouping matrix.

[0080] Specifically, this step aims to mine abnormal event patterns with spatial aggregation from discrete and individual abnormal parameter contribution degrees. In this process, the multi-dimensional cluster analysis uses a density-based spatial clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN). This algorithm takes the geographic coordinates of each abnormal sensor as the spatial dimension and its corresponding parameter abnormality contribution degree set and spatio-temporal pulse feature group as the attribute dimension for joint clustering. Through this algorithm, sensors that are spatially adjacent and have similar abnormal patterns can be automatically classified into a class, thereby identifying, for example, "a band of turbidity abnormal sensor groups distributed along the riverbank", and these clustering results, their categories, confidence levels, and other information collectively constitute the hierarchical spatial abnormal grouping matrix.

[0081] For the hierarchical spatial abnormal grouping matrix, in combination with the environmental coupling feature group, the coupling effect of abnormal grouping and environmental changes is analyzed, and a grouping abnormality diagnosis report is output.

[0082] Specifically, this step aims to make a deep cause diagnosis on the identified abnormal sensor cluster, in order to distinguish whether it is caused by external environment driving or internal failure. In this process, a transfer entropy in information entropy theory is used to analyze the coupling effect of abnormal grouping and environmental change. The transfer entropy can quantify the information flow from one time series to another, so as to judge the causal driving relationship between them. Before calculation, first, the principal component analysis method is used to extract the time series of the first principal component of the high-dimensional environmental coupling feature group and the hierarchical space abnormal grouping matrix respectively, and the two one-dimensional time series are taken as input and substituted into the formula for calculation. The calculation formula can be:

[0083]

[0084] Where, T E→S represents the transfer entropy from the environmental coupling feature group E to the time series S of a certain sensor grouping in the hierarchical space abnormal grouping matrix; s t+1 represents the state of the sensor grouping at the next time; represents the historical state of the sensor grouping; represents the historical state of the environmental coupling feature group. A higher transfer entropy value indicates that the behavior of the abnormal grouping is largely driven by external environmental changes. Based on the quantitative analysis results of the coupling effect, the final output includes the grouping abnormal diagnosis report containing the abnormal grouping identification, spatial range, root cause attribution and confidence evaluation, as shown in Figure 4 The coupling effect of abnormal grouping and environmental change is visualized in the form of a bidirectional heat tree diagram, in which the rows of the heat map represent different abnormal sensor groupings obtained by clustering analysis, the columns represent related external environmental factors, and the color depth of the cells quantifies the transfer entropy or causal correlation strength between them; The tree diagram at the top and left of the figure respectively shows the results of hierarchical clustering of environmental factors and sensor groupings.

[0085] Optionally, after the sensor grouping performs the monitoring calibration strategy to complete self-calibration and adjust subsequent sampling behavior, the method further comprises:

[0086] The monitoring calibration strategy is issued to the control node of the sensor grouping through the Internet of Things control network, the control node executes the strategy instruction and increases the sampling frequency and duration, and generates a calibration and adjustment result record;

[0087] Specifically, this step aims to precisely and automatically execute intelligent calibration decisions at the edge. In this process, the monitoring calibration strategy is packaged into a standardized instruction data packet and issued to the target sensor group's control node through the Internet of Things control network. After receiving and verifying the instruction data packet, the control node, such as an embedded microcontroller (MCU), executes the strategy instructions contained therein. The execution process includes calling the sensor underlying driver to complete the sensor's self-cleaning or re-calibration, or dynamically modifying the configuration parameters of the data acquisition module to achieve the up-regulation of the sampling frequency and duration. All executed actions, timestamps, and parameter comparisons before and after execution are recorded to form a calibration and adjustment result record.

[0088] Based on the calibration and adjustment result record, the abnormal state during execution is fed back to the system.

[0089] Specifically, this step aims to provide a verification and feedback mechanism at the execution level for the entire closed-loop self-optimization to ensure that the instructions are successfully executed. In this process, the control node will perform a "self-check" after executing the instructions, and an execution fidelity evaluation algorithm will be used to quantify the actual execution effect of the instructions. The evaluation can calculate an execution deviation score δ exec , and the calculation formula can be:

[0090]

[0091] where f cmd and d cmd are the target sampling frequency and target duration of the instructions in the monitoring calibration strategy; f actual and d actual are the actual sampling frequency and duration achieved by the control node after execution; w f and w d are the weight coefficients of different parameters. When the execution deviation score exceeds a preset fidelity threshold, it is determined that an abnormal state at the execution level has occurred. This abnormal state, together with the calibration and adjustment result record, will be fed back as an abnormal state feedback to the system for further diagnosis or strategy re-planning.

[0092] To verify the feasibility and advancement of the present application in implementation, the present application is applied to a smart water ecological management project in the Xinwen River Basin upstream of the Xinwen Reservoir, an important drinking water source. The basin has an ecological nature reserve in the upstream, an agricultural irrigation area in the middle reaches, and an industrial discharge area near the city in the downstream, with a complex and sensitive water ecological system. Traditional periodic and fixed-point water quality monitoring methods are difficult to effectively early warn and manage sudden and cumulative ecological risks.

[0093] In this embodiment, a 12-month continuous data processing and analysis of the new river basin is carried out. The Internet of Things terminal is deployed according to the water function zoning, covering water temperature, pH value, dissolved oxygen, turbidity and other physical and chemical parameter sensors, as well as new sensors such as online microscopic imaging and underwater acoustic probe for direct observation of biological communities.

[0094] Firstly, by constructing the spatial grouping of sensor data stream, the differentiated normal monitoring of different functional areas is realized. On April 10, 2025, during the stable hydrology period, each sensor group collects data at a low frequency of 10 minutes. At 15:20 on that day, the turbidity and specific ion concentration parameters of the sensor group located in the downstream industrial discharge area deviate from the water steady-state fluctuation baseline sharply within 1 minute. The abnormal fluctuation marker is immediately detected, and a high-frequency sampling instruction is automatically issued to the group. Five minutes of high-density monitoring data before and after the event are collected, successfully capturing and packaging the context data pulse containing the complete process.

[0095] Subsequently, the rainfall data of the weather station and the high-resolution remote sensing satellite image data corresponding to the spatio-temporal context data pulse are obtained as external environmental data sets. Through feature correlation analysis, it is found that this event has no strong correlation with rainfall, but is highly coupled with the color change of the water body near a certain discharge port in the remote sensing image. Finally, a multi-dimensional spatio-temporal feature cube containing internal and external factors is integrated.

[0096] At the same time, a parallel analysis module is started, which extracts a set of spatio-temporal pulse features based on the context data pulse, and extracts a set of environmental coupling features based on the multi-dimensional spatio-temporal feature cube. Through multi-source coupling fusion optimization of these two sets of features, a coupling dynamic calibration index as high as 0.95 is generated, indicating that the internal sensor data of this event is highly consistent with the external environmental factors, and the cognition is reliable.

[0097] Based on the high-credibility multi-dimensional spatio-temporal feature cube and coupling dynamic calibration index, a dynamic water ecological index is output. The value of the index decreases sharply from the normal 0.82 to 0.45, and the credibility score is as high as 98%, clearly pointing to a sudden and high-intensity pollution event.

[0098] In another scenario, on July 22, 2025, it is monitored that the sensor group located in the middle reaches of the agricultural irrigation area has a dynamic water ecological index with a credibility score continuously below the reliability threshold of 85%. The diagnostic process is immediately started, based on the credibility score and the continuously low coupling dynamic calibration index, the contribution sequence of each parameter is analyzed, and a parameter abnormal contribution set is generated. Through the analysis of the spatio-temporal pulse feature group, it is found that the abnormal parameter has no sharp physical pulse feature, and accordingly a grouping abnormality diagnosis report is generated, determining that the abnormality is caused by the measurement drift of the biological attachment of the sensor probe, rather than the real water deterioration.

[0099] According to the diagnosis report, a differentiated monitoring calibration strategy is generated, which contains the instruction set of "ultrasonic self-cleaning and online electrochemical re-calibration" for the sensor group. The strategy is issued to the control node of the target sensor group through the Internet of Things, and is successfully executed. After execution, the calibration and adjustment result record of the group is fed back, showing that the sensor reading returns to normal, and the credibility score also rises to more than 95%, completing a complete closed-loop self-calibration.

[0100] The data show that the method of the application has significant advantages in monitoring efficiency, early warning accuracy and self-maintenance ability compared with traditional fixed frequency monitoring and manual evaluation. Through intelligent data processing method, the complex water ecological system can be evaluated comprehensively and dynamically with high precision and credibility.

[0101] Table 1 Data collection and processing of industrial discharge sudden pollution event

[0102]

[0103] Table 2 Comparison table of water ecological index calculation results and traditional methods

[0104]

[0105] Table 3 Self-calibration process verification table of sensor drift event

[0106]

[0107] From the data recorded in the above tables 1-3, it can be seen that the application performs outstandingly in the embodiments. Table 1 shows the ability of intelligent triggering high-frequency sampling in real sudden events. Table 2 clearly proves that the dynamic water ecological index of the application is not only more accurate in evaluation, but also effectively avoids misjudgment caused by sensor failure through its unique credibility score mechanism. Table 3 completely records the whole process of self-diagnosis and closed-loop calibration, verifying the great advantage of the application in ensuring long-term data quality, and providing strong technical support for realizing truly unattended and intelligent water ecological management.

[0108] Based on the same inventive concept, the application also provides a water ecological index data processing system combining Internet of Things and artificial intelligence, as shown in Figure 5 The system comprises:

[0109] The spatial grouping acquisition module is used for grouping water quality sensors and synchronizing clocks by using water body function zoning of a target water area, setting sensor group periods and acquiring basic parameters, and forming spatial grouping sensor data stream;

[0110] an anomaly monitoring triggering module configured to compare the spatially grouped sensor data stream with a preset water body steady-state fluctuation baseline in real time, and collect high-density data before and after an event when a certain deviation is detected in the spatially grouped sensor data stream, to generate a context data pulse;

[0111] a multi-source fusion processing module configured to obtain external environment data matching the context data pulse in terms of time and geographical location, and perform alignment and standardization processing on the context data pulse and the external environment data, to generate a multi-dimensional spatio-temporal feature cube;

[0112] an index calculation and credibility evaluation module configured to calculate a water ecological index value based on the multi-dimensional spatio-temporal feature cube through dynamic weight distribution, and simultaneously quantitatively analyze a credibility score, to output a dynamic water ecological index in combination with the water ecological index and the credibility score;

[0113] a calibration strategy generation module configured to analyze the contribution degree of each parameter in the dynamic water ecological index when the credibility score is lower than a preset reliability threshold, to locate an abnormal parameter and trace it back to a corresponding sensor group, and to generate a monitoring calibration strategy;

[0114] a strategy issuing and executing module configured to issue the monitoring calibration strategy to the sensor group through the Internet of Things, and to execute the monitoring calibration strategy by the sensor group to complete self-calibration and adjust subsequent sampling behavior.

[0115] It should be noted that the function division and information interaction among the above-mentioned modules are logical, and can be integrated in the same software platform or distributed in physical implementation. The connection among them represents data flow and control flow, and aims to cooperatively achieve the object of the present application. The above-mentioned only illustrates the exemplary embodiments of the present application, and cannot limit the protection scope of the present application.

Claims

1. A water ecological index data processing method combining the Internet of Things and artificial intelligence, characterized in that, The method comprises: Grouping water quality sensors and synchronizing clocks using water body function zoning of a target water area, setting sensor group periods and collecting basic parameters to form spatially grouped sensor data streams; According to a preset water body steady-state fluctuation baseline, real-time comparison is performed on the spatially grouped sensor data streams, when a certain deviation of the spatially grouped sensor data streams is detected, high-density data before and after an event is collected, and a context data pulse is generated; External environment data matching the context data pulse in time and geographical location is obtained, the context data pulse and the external environment data are aligned and standardized, and a multi-dimensional spatio-temporal feature cube is generated; Based on the multi-dimensional spatio-temporal feature cube, a water ecological index value is calculated through dynamic weight distribution, and a reliability score is quantitatively analyzed synchronously, and a dynamic water ecological index is output combined with the water ecological index and the reliability score; Using the reliability score of the dynamic water ecological index, when the reliability score is lower than a preset reliability threshold, the contribution of each parameter in the dynamic water ecological index is analyzed, the abnormal parameter is located and traced back to the corresponding sensor group, and a monitoring and calibration strategy is generated; The monitoring and calibration strategy is issued to the sensor group through the Internet of Things, and the sensor group executes the monitoring and calibration strategy to complete self-calibration and adjust subsequent sampling behavior. 2.The water ecological index data processing method combining Internet of Things and artificial intelligence according to claim 1, characterized in that, The formation of the spatially grouped sensor data stream comprises: Using the water body function zoning, configuring multiple types of water quality sensors, collecting basic parameter data of each partition, and generating an original sensor data group; Based on different geographical partitions, different sampling periods and sampling parameters are set for the original sensor data group to obtain an ecological partition sampling parameter set; Clock synchronization and spatial identification coding are performed on the original sensor data group and the ecological partition sampling parameter set to generate a partitioned synchronous sensor data set; Edge fusion preprocessing and real-time missing compensation are performed on the partitioned synchronous sensor data set to generate a spatially grouped sensor data stream. 3.The water ecological index data processing method combining Internet of Things and artificial intelligence according to claim 1, characterized in that, The generation of the context data pulse comprises: Based on the spatially grouped sensor data stream and the water body steady-state fluctuation baseline, sliding window analysis is performed on the change of grouped water quality parameters in real time, the dynamic fluctuation amplitude of the basic parameters is detected, and an abnormal fluctuation marker is generated; When the indicator parameter of the abnormal fluctuation marker fluctuates to a certain extent, continuous high-density monitoring data of the time period before and after the event is automatically collected, and the time window, spatial grouping identifier, and event number corresponding to the high-density monitoring data are obtained, and a context data pulse is packaged and output.

4. The water ecological index data processing method combining Internet of Things and artificial intelligence according to claim 1, characterized in that, The generation of the multi-dimensional spatio-temporal feature cube comprises: Meteorological data and remote sensing satellite image data consistent with the timestamp and geographical coordinates of the context data pulse are obtained, the meteorological data and the remote sensing satellite image data are mapped to the corresponding spatio-temporal location to generate an external environment data set; Feature pairing and standardization processing are performed on the context data pulse and the external environment data set to obtain standard multi-source data; Based on the standard multi-source data, through feature correlation analysis and feature screening, the internal sensor parameters are coupled with the external environmental factors, and a multi-dimensional spatio-temporal feature cube is integrated and output.

5. The water ecological index data processing method combining Internet of Things and artificial intelligence according to claim 1, characterized in that, The method further includes: Based on the context data pulse, a spatio-temporal pulse feature vector is extracted, and a spatio-temporal pulse feature group is generated; Based on the multi-dimensional spatio-temporal feature cube, an environmental coupling feature is extracted, and an environmental coupling feature group is generated; Based on the spatio-temporal pulse feature group and the environmental coupling feature group, multi-source coupling fusion optimization is performed, and a coupling dynamic calibration index is generated.

6. The water ecological index data processing method combining the Internet of Things and artificial intelligence according to claim 5, characterized in that, The output of the dynamic water ecological index combined with the reliability score includes: Based on the multi-dimensional spatio-temporal feature cube, multi-layer adaptive weight distribution is performed, the weights of each type of sensor parameter and external environmental factor are dynamically adjusted, and a preliminary water ecological index value is generated through weighted aggregation calculation; Based on the preliminary water ecological index value, the coupling dynamic calibration index is fused, the influence and correlation of each parameter are feedback adjusted, a reliability score is generated, and the preliminary water ecological index value and the reliability score are linked to output a dynamic water ecological index.

7. The water ecological index data processing method combining the Internet of Things and artificial intelligence according to claim 5, characterized in that, The generation of the monitoring calibration strategy includes: Based on the reliability score of the dynamic water ecological index and the coupling dynamic calibration index, the contribution sequence of each parameter is detected and analyzed for abnormality, and a parameter abnormal contribution degree set is generated; Based on the parameter abnormal contribution degree set and the spatio-temporal pulse feature group, the spatio-temporal distribution and parameter change pulse are analyzed, the abnormal sensor group is determined, and a grouping abnormal diagnosis report is generated; According to the grouping abnormal diagnosis report and the environmental coupling feature group, the abnormal cause and influence range are evaluated, and a monitoring calibration strategy is generated.

8. The water ecological index data processing method combining the Internet of Things and artificial intelligence according to claim 7, characterized in that, The generation of the grouping abnormal diagnosis report includes: Multi-dimensional cluster analysis is performed on the parameter abnormal contribution degree set and the spatio-temporal pulse feature group, the abnormal sensors are classified and categorized in space, and a hierarchical spatial abnormal grouping matrix is generated; For the hierarchical spatial abnormal grouping matrix, combined with the environmental coupling feature group, the coupling effect of abnormal grouping and environmental change is analyzed, and a grouping abnormal diagnosis report is output.

9. The water ecological index data processing method combining Internet of Things and artificial intelligence according to claim 1, characterized in that, After the self-calibration and adjustment of the subsequent sampling behavior of the sensor group are completed by executing the monitoring calibration strategy, the following includes: The monitoring calibration strategy is sent to the control node of the sensor group through the Internet of Things control network, the control node executes the strategy instruction and increases the sampling frequency and duration, and a calibration and adjustment result record is generated; Based on the calibration and adjustment result record, the abnormal state in the execution process is fed back to the system.

10. A water ecological index data processing system combining Internet of Things and artificial intelligence, applied to the water ecological index data processing method combining Internet of Things and artificial intelligence according to any one of claims 1-9, characterized in that, The system includes: A spatial grouping collection module is used to divide the water quality sensors into groups and synchronize the clocks by using the water body function zoning of the target water area, set the sensor group period and collect the basic parameters, and form a spatial grouping sensor data stream; An abnormal monitoring triggering module is used to compare the pre-set water body steady-state fluctuation baseline with the spatial grouping sensor data stream in real time, collect high-density data before and after the event when a certain deviation of the spatial grouping sensor data stream is detected, and generate a context data pulse; An abnormal monitoring triggering module is used to compare the pre-set water body steady-state fluctuation baseline with the spatial grouping sensor data stream in real time, collect high-density data before and after the event when a certain deviation of the spatial grouping sensor data stream is detected, and generate a context data pulse; A multi-source fusion processing module is configured to acquire external environment data matching the context data pulse in time and geographical position, align and standardize the context data pulse and the external environment data, and generate a multi-dimensional spatio-temporal feature cube; An index calculation and credibility evaluation module is configured to calculate a water ecological index value based on the multi-dimensional spatio-temporal feature cube through dynamic weight distribution, and simultaneously quantitatively analyze a credibility score, and output a dynamic water ecological index in combination with the water ecological index and the credibility score; A calibration strategy generation module is configured to analyze the contribution degree of each parameter in the dynamic water ecological index when the credibility score is lower than a preset reliability threshold, locate an abnormal parameter and trace back to a corresponding sensor group, and generate a monitoring calibration strategy; A strategy issuing and executing module is configured to issue the monitoring calibration strategy to the sensor group through the Internet of Things, and execute the monitoring calibration strategy by the sensor group to complete self-calibration and adjust subsequent sampling behavior.

Images