A method and system for intelligent processing of sewage monitoring data

Abstract

The present invention relates to the field of data processing technology, and in particular to a method and system for intelligent processing of sewage monitoring data; the method comprises the following steps: obtaining multimodal sewage real-time monitoring data; performing sliding window division on the multimodal sewage real-time monitoring data, and performing outlier filtering processing to construct filtering optimized real-time monitoring data; obtaining deep feature data of multiple modes based on filtering optimized real-time monitoring data; performing implicit correlation analysis on the deep feature data of multiple modes to generate implicit correlation data between modes; performing cross-dimensional spatial mapping on the filtering optimized real-time monitoring data based on the implicit correlation data between modes, and performing spatial correlation topological fitting to construct a modal feature point mapping space; performing multi-time point evolution simulation on the modal feature point mapping space, and performing dynamic evolution trend analysis, thereby obtaining the dynamic evolution law of feature points. The present invention realizes efficient sewage monitoring data processing and improves the decision-making ability of sewage treatment.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for intelligently processing sewage monitoring data. Background Art

[0002] With the acceleration of urbanization, the burden on sewage treatment plants has become increasingly heavy. How to improve the intelligence level of sewage treatment from the root has always been a major challenge facing the industry. At present, most sewage monitoring systems only rely on a single sensor to collect some indicators, making it difficult to fully perceive the complex changes in sewage components. At the same time, the analysis and processing of monitoring data are mostly limited to traditional statistical methods, lacking intelligent diagnosis and decision support, and often have problems such as low data processing efficiency and poor real-time responsiveness. Therefore, it is urgent to develop an innovative sewage monitoring data processing method in order to achieve accurate perception and autonomous optimization of the entire sewage treatment process. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a sewage monitoring data intelligent processing method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, the present invention provides a method for intelligently processing sewage monitoring data, comprising the following steps:

[0005] Step S1: obtaining multimodal sewage real-time monitoring data; performing sliding window division on the multimodal sewage real-time monitoring data, and performing outlier filtering processing to construct filtering and optimizing real-time monitoring data;

[0006] Step S2: obtaining deep feature data of multiple modes based on filtering and optimizing the real-time monitoring data; performing implicit correlation analysis on the deep feature data of multiple modes to generate implicit correlation data between the modes;

[0007] Step S3: Based on the implicit correlation data between modalities, the filtered and optimized real-time monitoring data is subjected to cross-dimensional spatial mapping, and spatial correlation topology fitting is performed to construct a modal feature point mapping space;

[0008] Step S4: performing multi-time point evolution simulation on the modal feature point mapping space and performing dynamic evolution trend analysis to obtain the dynamic evolution law of the feature points;

[0009] Step S5: performing multi-feature point recursive prediction calculation on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby obtaining multiple potential sewage situation prediction scenarios;

[0010] Step S6: Generate multiple potential sewage situation visualization views based on multiple potential sewage situation prediction scenarios; perform adaptive risk decision optimization on the multiple potential sewage situation visualization views to construct a sewage treatment intelligent decision model.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: acquiring multimodal real-time sewage monitoring data and historical sewage monitoring data based on sensors;

[0013] Step S12: performing historical parameterized numerical distribution calculation on the historical sewage monitoring data, thereby obtaining the historical numerical distribution range;

[0014] Step S13: performing sliding window division on the multimodal sewage real-time monitoring data, thereby obtaining a plurality of window monitoring data segments;

[0015] Step S14: performing dynamic outlier detection on multiple window monitoring data segments according to the historical value distribution range, and extracting outlier abnormal monitoring data;

[0016] Step S15: Perform outlier data filtering based on the outlier abnormal monitoring data to construct filtered and optimized real-time monitoring data.

[0017] The present invention obtains multimodal real-time sewage monitoring data and historical sewage monitoring data through sensors to better understand the evolution trend and change law of the data. The historical parameterized numerical distribution calculation can reveal the distribution law of historical data and provide a reference range of historical data for subsequent analysis. The acquisition of the historical numerical distribution range compares and analyzes the real-time data to help identify abnormal situations.

[0018] The multimodal real-time sewage monitoring data is divided into multiple window monitoring data segments through sliding windows, and the continuous data is processed in segments. The acquisition of window monitoring data segments provides more refined data units for subsequent analysis, which is conducive to in-depth mining and analysis of the data. Dynamic outlier detection is performed according to the historical numerical distribution range, which can timely discover anomalies in the monitoring data and extract problems in identifying outlier abnormal monitoring data.

[0019] Based on the outlier abnormal monitoring data, outlier data filtering and processing are carried out, and filtering and optimizing real-time monitoring data are constructed to improve data quality and accuracy. The establishment of filtering and optimizing real-time monitoring data provides a more reliable data basis for subsequent analysis and decision-making, and improves processing efficiency and accuracy.

[0020] Preferably, step S2 comprises the following steps:

[0021] Step S21: performing deep feature mining of different modes on the filtered optimized real-time monitoring data, thereby obtaining deep feature data of multiple modes;

[0022] Step S22: performing modal iterative similarity calculation on deep feature data of multiple modalities to generate a similarity matrix between modalities;

[0023] Step S23: performing implicit correlation analysis on deep feature data of multiple modalities according to the similarity matrix between modalities, and generating implicit correlation data between modalities.

[0024] The present invention mines the potential features of the data from different angles by performing deep feature mining of different modes on the filtered optimized real-time monitoring data, obtains deep feature data of multiple modes to more comprehensively understand the information content of the data, and improves the data characterization capability and mining depth. The modal iterative similarity calculation can quantify the similarity between different modes, provide quantitative indicators of the relationship between modalities for subsequent analysis, generate a similarity matrix between modalities to reveal the potential correlation between different modalities, and provide a basis for data integration and analysis.

[0025] Implicit correlation analysis based on the similarity matrix between modalities can explore the potential correlation between modalities, help understand the complex correlations between data, generate implicit correlation data between modalities, reveal the implicit rules and correlations behind the data, and provide more in-depth guidance for further analysis and decision-making.

[0026] Preferably, step S3 specifically comprises the following steps:

[0027] Step S31: performing cross-modal convolution on deep feature data of multiple modalities to generate multiple modal convolution feature maps;

[0028] Step S32: performing local perception aggregation on multiple modal convolutional feature maps according to implicit correlation data between modalities, thereby generating a local relationship feature tensor;

[0029] Step S33: performing inter-modal collaborative representation mining on the local relationship feature tensor, thereby generating an inter-modal collaborative representation link;

[0030] Step S34: performing cross-dimensional spatial mapping on multiple modal convolution feature maps to generate feature points of each modal space mapping;

[0031] Step S35: performing spatial correlation topology fitting on each modal space mapping feature point based on the inter-modal collaborative representation link to construct a modal feature point mapping space.

[0032] The present invention uses cross-modal convolution to interact and fuse information between different modal features, improves the data representation capability and the comprehensiveness of multimodal information, generates multiple modal convolution feature maps to extract higher-level features in the data, and provides richer and deeper feature representation for subsequent analysis. It captures implicit associations between modalities through local perception aggregation, improves the representation capability and correlation of features, generates local relationship feature tensors to better understand local relationships in the data, and helps to mine the deep features of the data. Inter-modal collaborative representation mining reveals the collaborative relationship between different modalities, providing more in-depth guidance for data integration and analysis, and generates inter-modal collaborative representation links to build an inter-modal association network to help identify the synergy and correlation between data.

[0033] Cross-dimensional spatial mapping can map multimodal features into a common space, improve the comparability and integration of data, generate feature points for each modal space mapping, compare and analyze the features of different modal data in a unified space, perform spatial correlation topology fitting based on the collaborative representation link between modalities, construct a modal feature point mapping space, help integrate and associate different modal data, and construct a modal feature point mapping space to better understand the relationship and interaction between data, providing more comprehensive data support for subsequent analysis and application.

[0034] Preferably, the specific steps of step S4 are:

[0035] Step S41: performing time series change analysis on the filtering optimization real-time monitoring data to generate time series change monitoring data;

[0036] Step S42: performing multi-time point evolution simulation on the modal feature point mapping space based on the time series change monitoring data, thereby generating a feature point evolution mapping space;

[0037] Step S43: Perform dynamic evolution trend analysis on the feature point evolution mapping space to generate a dynamic evolution trend trajectory of the feature points;

[0038] Step S44: performing evolution pattern clustering analysis on the dynamic evolution trend trajectory of the feature points, thereby obtaining the dynamic evolution law of the feature points.

[0039] The present invention can help understand the changing trend of data over time through time series change analysis, reveal the periodicity and regularity of data, capture the dynamic change characteristics of data, and provide a data perspective in the time dimension. Multi-time point evolution simulation is based on time series change monitoring data, and can simulate the evolution of feature points at different time points, help understand the time series evolution process of data, and show the evolution laws and trends of data features in the time dimension.

[0040] Dynamic evolution trend analysis reveals the dynamic change rules of feature points in the evolution process, helps identify data trends and regularities, visualizes the dynamic evolution process of data, and provides a more intuitive data analysis perspective. Evolutionary pattern clustering analysis can cluster the dynamic evolution trends of feature points, discover the differences in evolution patterns between different feature points, and obtain the dynamic evolution rules of feature points to identify the potential patterns and regularities of data, providing a deeper understanding of data behavior and development trends.

[0041] Preferably, the specific steps of step S42 are:

[0042] Extracting multiple modal feature points based on the modal feature point mapping space;

[0043] Conduct time-flow feature learning on time-series change monitoring data to generate time-flow features of monitoring data;

[0044] Based on multiple modal feature points as the initial evolution starting point;

[0045] The time-flow characteristics of monitoring data are used to simulate the multi-time-point evolution of the initial evolution starting point and generate a dense map of characteristic point evolution;

[0046] The spatial distribution of feature points is analyzed on the feature point evolution dense graph to generate a feature point evolution mapping space.

[0047] The present invention can integrate information between different modes by extracting multiple modal feature points, improve the dimension and richness of data features, and extract multiple modal feature points to capture the diversity and complexity of data and provide a more comprehensive data representation. Time flow feature learning helps understand how monitoring data change over time, extracts data features in the time dimension, reveals the timing characteristics of data, and provides an analysis perspective for time series data. Using the time flow characteristics of monitoring data to perform multi-time point evolution simulation on the initial evolution starting point can simulate the evolution of feature points at different time points and help understand the timing evolution process of data.

[0048] The feature point evolution dense map shows the evolution law and trend of data features in the time dimension. The feature point spatial distribution analysis of the feature point evolution dense map reveals the spatial distribution characteristics of the feature points, helps understand the spatial distribution law of the data, and generates a feature point evolution mapping space to visualize the feature points in space, providing a more intuitive data analysis perspective.

[0049] Preferably, the specific steps of step S5 are:

[0050] Step S51: performing a recursive prediction calculation of multiple feature points on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby generating prediction calculation data of multiple feature points;

[0051] Step S52: performing trend distribution prediction for future time periods on the prediction calculation data of multiple feature points to generate sewage trend distribution prediction data;

[0052] Step S53: Evolving the sewage trend distribution prediction data into multiple trend scenario features, thereby obtaining multiple potential sewage trend prediction scenarios.

[0053] The present invention uses recursive prediction calculations of multiple feature points based on the dynamic evolution laws of feature points to infer the trends and changing trends of multiple feature points in the future, predict the future development of sewage data, and provide multi-dimensional data prediction results. The trend distribution prediction of future time periods is based on the prediction calculation data of multiple feature points to predict the trend and distribution of future sewage data in different time periods, understand the development trend of sewage data in advance, and provide a reference for future sewage treatment and planning. The feature evolution of multiple trend scenarios conducts multi-faceted feature evolution analysis on the sewage trend distribution prediction data, and obtains multiple potential sewage situation prediction scenarios to explore the development of sewage under different scenarios, providing diversified prediction results for formulating response strategies and plans.

[0054] Preferably, the specific steps of step S6 are:

[0055] Step S61: Visually render multiple potential sewage situation prediction scenes to generate multiple potential sewage situation visualization views;

[0056] Step S62: performing sewage risk diffusion trajectory analysis on multiple potential sewage situation visualization views to generate multiple potential sewage diffusion trajectories;

[0057] Step S63: performing a sewage risk situation quantitative assessment on multiple potential sewage diffusion trajectories to generate a sewage risk situation quantitative value;

[0058] Step S64: Adaptively optimize the risk decision of the sewage risk situation quantification value and build a sewage treatment intelligent decision model.

[0059] The present invention presents complex data results in an intuitive manner by visually rendering multiple potential sewage situation prediction scenarios, so as to more intuitively understand the situations of different sewage situation prediction scenarios and provide visualization support for subsequent decision-making. The sewage risk diffusion trajectory analysis can explore the diffusion paths and trends of sewage in different scenarios, help understand the propagation laws and impact areas of sewage, predict the spread range of sewage and the affected areas, and provide a reference for risk assessment and response measures. The sewage risk situation quantitative assessment quantitatively evaluates multiple potential sewage diffusion trajectories and quantifies the severity of different sewage risk situations. The adaptive risk decision optimization is based on the sewage risk situation quantitative value, adjusts the decision plan according to real-time data and situation, optimizes sewage treatment decisions, and constructs a sewage treatment intelligent decision-making model to make intelligent decisions based on specific situations and risk assessment results, thereby improving the efficiency and response capabilities of sewage treatment.

[0060] Furthermore, the present invention also provides a sewage monitoring data intelligent processing system, which is used to execute the sewage monitoring data intelligent processing method as described above, comprising:

[0061] The outlier filtering module is used to obtain multi-modal real-time sewage monitoring data; the multi-modal real-time sewage monitoring data is divided into sliding windows, and outlier filtering is performed to construct filtering and optimization real-time monitoring data;

[0062] Implicit association module, used to obtain deep feature data of multiple modes based on filtering and optimizing real-time monitoring data; perform implicit association analysis on deep feature data of multiple modes to generate implicit association data between modes;

[0063] The spatial mapping module is used to perform cross-dimensional spatial mapping of the filtered and optimized real-time monitoring data based on implicit correlation data between modalities, and to perform spatial correlation topological fitting to construct a modal feature point mapping space;

[0064] The dynamic evolution module is used to simulate the multi-time evolution of the modal feature point mapping space and analyze the dynamic evolution trend, so as to obtain the dynamic evolution law of the feature points;

[0065] The sewage situation prediction module is used to perform multi-feature point recursive prediction calculations on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby obtaining multiple potential sewage situation prediction scenarios;

[0066] The intelligent decision-making module is used to generate multiple potential sewage situation visualization views based on multiple potential sewage situation prediction scenarios; adaptively optimize the risk decision of multiple potential sewage situation visualization views, and build an intelligent decision-making model for sewage treatment.

[0067] The beneficial effects of the present invention are:

[0068] 1. The present invention can effectively reduce outliers and noise in the data and improve data quality and reliability through sliding window division and outlier filtering processing. By constructing filtering and optimizing real-time monitoring data, it helps to provide more accurate and reliable sewage system status information, providing a reliable basis for subsequent analysis and decision-making. By generating implicit correlation data between modalities, the potential relationship between different modal data can be revealed, which helps to deeply understand the interaction and influence between various modalities in the sewage system. By obtaining deep feature data of multiple modalities and performing implicit correlation analysis, it helps to discover useful information hidden behind the data, providing deeper insights for subsequent analysis and decision-making.

[0069] 2. By constructing the modal feature point mapping space through spatial mapping and associated topological fitting, data of different dimensions can be integrated to form a more comprehensive data perspective. Through multi-time point evolution simulation and dynamic evolution trend analysis, we can deeply understand the changing rules of feature points at different time points and help predict the future development trend of the system.

[0070] 3. The present invention helps to discover potential problems and trends in advance based on the law of dynamic evolution, and provides a more accurate reference for system management and decision-making. Through multi-feature point recursive prediction calculations, the future state of the sewage system can be predicted based on the law of dynamic evolution, providing an important reference for system management and decision-making. Adaptive risk decision optimization through multiple potential sewage situation visualization views can help decision makers understand the system status more intuitively and make more accurate decisions. Building an intelligent decision-making model for sewage treatment can improve the scientificity and efficiency of decision-making, while reducing risks and costs, and promoting the sustainable development of sewage treatment systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic diagram of the steps of a method for intelligently processing sewage monitoring data according to the present invention;

[0072] Figure 2 Detailed implementation flow chart of step S1;

[0073] Figure 3 Detailed implementation flow chart of step S2;

[0074] Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0075] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0076] This application example provides a method and system for intelligent processing of sewage monitoring data. The execution subjects of the method and system include but are not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of this application, and the data processing platform includes but is not limited to: at least one of audio and image management system, information management system, and cloud data management system.

[0077] Example 1, a method for intelligently processing sewage monitoring data, see Figures 1 to 4 , including the following steps:

[0078] Step S1: obtaining multimodal sewage real-time monitoring data; performing sliding window division on the multimodal sewage real-time monitoring data, and performing outlier filtering processing to construct filtering and optimizing real-time monitoring data;

[0079] Step S2: obtaining deep feature data of multiple modes based on filtering and optimizing the real-time monitoring data; performing implicit correlation analysis on the deep feature data of multiple modes to generate implicit correlation data between the modes;

[0080] Step S3: Based on the implicit correlation data between modalities, the filtered and optimized real-time monitoring data is subjected to cross-dimensional spatial mapping, and spatial correlation topology fitting is performed to construct a modal feature point mapping space;

[0081] Step S4: performing multi-time point evolution simulation on the modal feature point mapping space and performing dynamic evolution trend analysis to obtain the dynamic evolution law of the feature points;

[0082] Step S5: performing multi-feature point recursive prediction calculation on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby obtaining multiple potential sewage situation prediction scenarios;

[0083] Step S6: Generate multiple potential sewage situation visualization views based on multiple potential sewage situation prediction scenarios; perform adaptive risk decision optimization on the multiple potential sewage situation visualization views to construct a sewage treatment intelligent decision model.

[0084] The present invention can provide instant sewage system status information by acquiring real-time monitoring data, which helps to timely discover abnormal situations and problems. By using sliding window division and outlier filtering processing, optimized real-time monitoring data can be constructed to reduce the impact of noise and outliers and improve data quality.

[0085] By extracting deep feature data through the above steps, the complexity and diversity of the sewage system can be better characterized, providing more representative data for subsequent analysis and modeling. Implicit association analysis can reveal the potential connections between different modal data, which helps to fully understand the operating status of the sewage system.

[0086] Secondly, through spatial mapping and associated topological fitting, the monitoring data are integrated and mapped in different dimensions to reveal the spatial relationship between the data. By constructing the modal feature point mapping space, it is helpful to form a more comprehensive and comprehensive data perspective. Through evolution simulation and dynamic evolution trend analysis, the changing patterns of feature points at different time points are observed, which helps to understand the dynamic change trend of the sewage system. The dynamic evolution pattern of the feature points is then obtained, which helps to predict the future development trend of the system.

[0087] Through recursive prediction calculations, the future state of the sewage system is predicted based on past data and dynamic laws, providing early warning and decision support for system operation. Generating multiple potential sewage situation prediction scenarios helps to fully consider the possible development of the system and prepare for different scenarios. The potential sewage situation visualization view can intuitively display the system status under different prediction scenarios, helping users to better understand and analyze the prediction results. Adaptive risk decision optimization makes intelligent decisions based on the information in the visualization view to improve the operating efficiency and safety of the sewage treatment system.

[0088] In this embodiment 1, refer to Figure 1 , is a schematic flow chart of the steps of a sewage monitoring data intelligent processing method of the present invention. In this example, the more detailed steps of the sewage monitoring data intelligent processing method include:

[0089] Step S1: obtaining multimodal sewage real-time monitoring data; performing sliding window division on the multimodal sewage real-time monitoring data, and performing outlier filtering processing to construct filtering and optimizing real-time monitoring data;

[0090] In this embodiment, various sensors, such as pH sensors, turbidity sensors, gas sensors, etc., are deployed at key monitoring points in the sewage treatment plant to build a comprehensive multimodal monitoring network. A dedicated data acquisition terminal is used to collect various physical and chemical index data output by these sensors in real time, and upload them to the cloud server through industrial Ethernet or other interfaces. In the cloud system, a centralized multimodal sewage monitoring database is established to store real-time monitoring data collected from the site.

[0091] For the time series data of each monitoring indicator, a sliding window method is used for segmented processing. The size of the sliding window can be set according to actual business needs, such as 1 hour, 6 hours, etc. A certain overlapping area can be set between adjacent windows. For the data in each sliding window, statistical features such as mean, variance, peak, etc. are extracted. These statistical features are used as representative data of the time window for subsequent outlier detection and data optimization. For the statistical feature data of each sliding window, distance-based, density-based and other outlier detection algorithms are used to identify abnormal data points. For data detected as outliers, interpolation, smoothing and other methods can be used for correction and optimization. The optimized data is reassembled into a time series to construct optimized monitoring data that has been filtered outliers.

[0092] Step S2: obtaining deep feature data of multiple modes based on filtering and optimizing the real-time monitoring data; performing implicit correlation analysis on the deep feature data of multiple modes to generate implicit correlation data between the modes;

[0093] In this embodiment, the original multimodal data (such as pH value, turbidity, gas composition, etc.) collected from the monitoring terminal in real time is cleaned, including denoising, outlier processing, etc., and feature engineering is performed on the data according to business needs to extract time series features, frequency domain features, and other features that reflect the dynamic changes of monitoring indicators. For different types of monitoring indicator data, corresponding deep neural network models are designed for feature learning. For example, LSTM is used to capture time series features for continuous indicators, and CNN is used to extract spatial features for discrete indicators. Through end-to-end deep learning training, deep feature representations of multiple modalities are extracted from the original monitoring data.

[0094] Perform principal component analysis or dimensionality reduction on the above deep features to remove redundant information and improve the discriminability of features. Use L1 / L2 regularization and other methods to sparsely process features to further highlight key features. Continuously optimize model hyperparameters through cross-validation and other methods to ensure the robustness and generalization of feature extraction. Use unsupervised representation learning methods such as autoencoders and variational autoencoders to mine potential patterns and laws from fused features. These implicit patterns reflect the complex interactions and evolution laws between different monitoring indicators. Integrate the above learned feature representations and implicit patterns into a unified knowledge base.

[0095] Step S3: Based on the implicit correlation data between modalities, the filtered and optimized real-time monitoring data is subjected to cross-dimensional spatial mapping, and spatial correlation topology fitting is performed to construct a modal feature point mapping space;

[0096] In this embodiment, real-time monitoring data collected by various sensors during the sewage treatment process are collected, including multi-dimensional indicators such as pH value, turbidity, flow rate, temperature, etc. These original monitoring data are preprocessed, including missing value filling, outlier correction, dimensionalization and other operations to ensure the consistency and reliability of the data. Deep learning techniques, such as autoencoders and contrastive learning, are used to perform nonlinear feature extraction and dimensionality reduction on the preprocessed monitoring data. Through this step, the original high-dimensional monitoring data is mapped to a potential low-dimensional feature space, the intrinsic correlation between the indicators of each dimension is retained, and representative feature points are selected from the above low-dimensional feature space.

[0097] For these feature points, their similarities and topological relationships in the feature space are calculated respectively, and the adjacency matrix between the feature points is constructed. Using graph neural networks and other technologies, these feature points and their relationships are modeled and learned to obtain a spatial correlation topology that can describe the intrinsic structure of the monitoring data. The learned topological structure is projected back to the original high-dimensional monitoring data space to construct a cross-modal feature point mapping space. According to the position and topological relationship of the feature points in the mapping space, the monitoring data is optimized and fused using interpolation, extrapolation and other methods to eliminate noise interference and improve data quality. The optimized monitoring data is fed back to subsequent data analysis and decision support modules to provide basic data support for precise perception and intelligent disposal.

[0098] Step S4: performing multi-time point evolution simulation on the modal feature point mapping space and performing dynamic evolution trend analysis to obtain the dynamic evolution law of the feature points;

[0099] In this embodiment, according to the time series characteristics of the real-time monitoring data, a time series prediction model (such as LSTM) is used to predict and simulate the positions of these feature points in the mapping space. Through a step-by-step iterative method, the dynamic changes of the feature points in the mapping space over time are simulated, and the simulated time series evolution trajectory of the feature points is intuitively displayed in the form of animation or video for easy observation and analysis. The evolution trajectory of the feature points is visualized and analyzed to observe its movement law in the mapping space. In combination with the actual sewage treatment process, the dynamic change characteristics of the monitoring data reflected by the evolution trajectory of the feature points are analyzed, and the key evolution laws such as clustering trends, mutation characteristics, and periodic fluctuations of the feature points in the mapping space are extracted. These dynamic evolution laws are compared and analyzed with the actual operating status of the sewage treatment process to explore the intrinsic connection between the two.

[0100] Step S5: performing multi-feature point recursive prediction calculation on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby obtaining multiple potential sewage situation prediction scenarios;

[0101] In this embodiment, corresponding time series prediction models are established based on the dynamic evolution trend of the feature points obtained by analysis in the mapping space. These prediction models use recurrent neural network structures such as LSTM and GRU, which can capture the temporal patterns in the evolution trajectory of feature points. For different types of feature points, such as pH value, turbidity, etc., corresponding prediction models are constructed respectively to better reflect their unique dynamic evolution laws. Based on the time series prediction model, the positions of these feature points in the mapping space are recursively predicted.

[0102] At each time step, the position of the feature point predicted at the previous moment is used as the input for the prediction at the next moment to realize the recursive update of the feature point state. Through multiple rounds of iterative prediction, the long-term series evolution trajectory of these feature points in the mapping space is obtained. The predicted evolution trajectory of each feature point is integrated into a complete mapping space model. This mapping space model reflects the linkage change process of multiple monitoring indicators in the time dimension. The mapping space model is analyzed and visualized to generate multiple sewage treatment situation prediction scenarios. These potential situation prediction scenarios provide an important basis for subsequent risk assessment and decision optimization.

[0103] Step S6: Generate multiple potential sewage situation visualization views based on multiple potential sewage situation prediction scenarios; perform adaptive risk decision optimization on the multiple potential sewage situation visualization views to construct a sewage treatment intelligent decision model.

[0104] In this embodiment, multiple potential sewage situation prediction scenarios, such as sewage component concentration, diffusion range, etc., are mapped to the visualization platform for visualization rendering. Through dynamic simulation, a series of potential sewage situation visualization views at different time nodes are generated. Combined with real-time monitoring data and historical experience, an adaptive decision optimization model based on reinforcement learning is trained. The potential situation visualization view is used as input, and the optimal emergency disposal strategy is obtained through simulation calculation.

[0105] This strategy may include a variety of measures such as adjusting the dosage, starting backup equipment, evacuating people, etc., and automatically adjust according to actual conditions. The adaptive decision-making model will be integrated into the intelligent sewage treatment system to achieve autonomous optimization of the entire emergency response process. For these predicted scenarios, the system will quickly assess its risk level and autonomously generate the optimal emergency response strategy, and feed back the optimized decision plan to the sewage treatment control execution system to achieve autonomous collaborative optimization of the entire process.

[0106] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0107] Step S11: acquiring multimodal real-time sewage monitoring data and historical sewage monitoring data based on sensors;

[0108] Step S12: performing historical parameterized numerical distribution calculation on the historical sewage monitoring data, thereby obtaining the historical numerical distribution range;

[0109] Step S13: performing sliding window division on the multimodal sewage real-time monitoring data, thereby obtaining a plurality of window monitoring data segments;

[0110] Step S14: performing dynamic outlier detection on multiple window monitoring data segments according to the historical value distribution range, and extracting outlier abnormal monitoring data;

[0111] Step S15: Perform outlier data filtering based on the outlier abnormal monitoring data to construct filtered and optimized real-time monitoring data.

[0112] In this embodiment, suitable sensors are selected, such as pH sensors, dissolved oxygen sensors, turbidity sensors, temperature sensors, etc., to obtain multimodal sewage real-time monitoring data. The data collection frequency is set according to the monitoring requirements, for example, every minute, every hour or shorter time intervals. The acquired multimodal sewage real-time monitoring data is stored in a database to facilitate subsequent analysis and processing, and historical sewage monitoring data is collected and organized, for example, historical data is obtained from databases, log files or other sources, and historical sewage monitoring data is pre-processed, for example, removing abnormal data, smoothing data, normalizing data, etc., and a suitable parameterized numerical distribution model is selected, such as normal distribution, lognormal distribution, gamma distribution, etc.

[0113] Perform parameterized numerical distribution calculation on historical sewage monitoring data, calculate the historical numerical distribution range according to the parameterized numerical distribution model, for example, calculate the mean, standard deviation, confidence interval, etc., set the size of the sliding window according to monitoring requirements, for example, set a window of 10 minutes, 30 minutes or longer, set the step size of the sliding window according to monitoring requirements, for example, set a step size of 1 minute, 5 minutes or longer, use the sliding window method to divide the multimodal sewage real-time monitoring data, and obtain multiple window monitoring data segments.

[0114] Select appropriate outlier detection methods, such as the 3σ criterion, box plot method, DBSCAN algorithm, etc., perform dynamic outlier detection on multiple window monitoring data segments according to the historical numerical distribution range, extract outlier abnormal monitoring data based on the outlier detection results, for example, mark data that exceeds the historical numerical distribution range as outlier abnormal data, set filtering strategies based on the outlier abnormal monitoring data and actual conditions, for example, directly delete outlier data, use interpolation methods to fill out outlier data, use statistical methods to smooth outlier data, etc. Filter the outlier abnormal monitoring data according to the filtering strategy, and construct filtered optimized real-time monitoring data.

[0115] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0116] Step S21: performing deep feature mining of different modes on the filtered optimized real-time monitoring data, thereby obtaining deep feature data of multiple modes;

[0117] Step S22: performing modal iterative similarity calculation on deep feature data of multiple modalities to generate a similarity matrix between modalities;

[0118] Step S23: performing implicit correlation analysis on deep feature data of multiple modalities according to the similarity matrix between modalities, and generating implicit correlation data between modalities.

[0119] In this embodiment, a suitable deep feature mining method is selected, for example, a convolutional neural network (CNN), a recurrent neural network (RNN), an autoencoder, etc., to perform deep feature mining of different modalities on the filtering optimization real-time monitoring data, and the deep feature mining method is used to extract deep features of data of different modalities, for example, extracting the trend of pH value, the fluctuation frequency of dissolved oxygen, the amplitude of change of turbidity, etc. The extracted deep feature data is vectorized, for example, using a numerical vector, a matrix or a tensor representation to calculate the similarity between deep feature data of different modalities, and an iterative method is used, for example, a k-nearest neighbor (k-NN) algorithm, a spectral clustering algorithm, etc., to perform modal iterative similarity calculation on deep feature data of multiple modalities, and the similarity calculation results between deep feature data of different modalities are stored in a matrix to form a similarity matrix between modalities.

[0120] According to the similarity matrix between modes, implicit correlation analysis is performed on the deep feature data of multiple modes. According to the results of the correlation analysis, implicit correlation relationships between data of different modes are discovered. For example, there is a negative correlation between changes in pH value and changes in dissolved oxygen, and there is a positive correlation between changes in dissolved oxygen and changes in turbidity, etc. The discovered correlation relationships are sorted out to generate implicit correlation data between modes, for example, the correlation relationships are represented as rules, models or other forms of data.

[0121] In this embodiment, reference Figure 4 The above is a schematic flow chart of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0122] Step S31: performing cross-modal convolution on deep feature data of multiple modalities to generate multiple modal convolution feature maps;

[0123] Step S32: performing local perception aggregation on multiple modal convolutional feature maps according to implicit correlation data between modalities, thereby generating a local relationship feature tensor;

[0124] Step S33: performing inter-modal collaborative representation mining on the local relationship feature tensor, thereby generating an inter-modal collaborative representation link;

[0125] Step S34: performing cross-dimensional spatial mapping on multiple modal convolution feature maps to generate feature points of each modal space mapping;

[0126] Step S35: performing spatial correlation topology fitting on each modal space mapping feature point based on the inter-modal collaborative representation link to construct a modal feature point mapping space.

[0127] In this embodiment, a suitable cross-modal convolution method is selected, for example, a multi-modal convolutional neural network (MM-CNN), a cross-modal convolution kernel (Cross-Modal Kernel), etc., and cross-modal convolution is performed on the deep feature data of multiple modalities. The cross-modal convolution method is used to perform convolution operations on the deep feature data of different modalities. For example, the deep feature data of the pH value is convolved with the deep feature data of the dissolved oxygen to generate a new feature map. The deep feature data of each modality is cross-modally convolved to generate multiple modal convolution feature maps. Each feature map contains association information between different modal data. According to the implicit association data between the modalities, each modal convolution feature map is locally perceived. For example, the negative correlation between the pH value and the dissolved oxygen is used as the basis for local perception, and the convolution feature maps of the two modalities are aggregated. Each modal convolution feature map is locally perceived and aggregated to generate a local relationship feature tensor, which contains local association information between different modal data.

[0128] Multimodal graph neural network (MM-GNN) and cross-modal self-attention mechanism (Cross-Modal Self-Attention) methods are used to mine inter-modal collaborative representations of local relationship feature tensors. The collaborative representation mining method is used to mine the collaborative representations between different modal data. For example, the collaborative change trend between pH value, dissolved oxygen and turbidity is discovered. The mined collaborative representation relationships are sorted out to generate inter-modal collaborative representation links, which represent the collaborative change relationship between different modal data. The convolution feature map of each modality is mapped to a low-dimensional space using a spatial mapping method. After mapping the convolution feature map of each modality to the low-dimensional space, the spatial mapping feature points of each modality are obtained. Each feature point represents the position of the modal data in the low-dimensional space. The appropriate spatial correlation topology fitting method is selected, such as graph embedding, multimodal graph neural network (MM-GNN), etc.

[0129] Based on the inter-modal collaborative characterization link, the spatial correlation topology fitting is performed on the spatial mapping feature points of each modality. According to the inter-modal collaborative characterization link, the spatial mapping feature points of different modalities are spatially correlated. The negative correlation between pH value and dissolved oxygen is used as the basis for spatial correlation. The feature points of these two modalities are connected, and the spatial correlation topology fitting is performed on the spatial mapping feature points of all modalities to construct a modal feature point mapping space, which contains the spatial correlation relationship between different modal data.

[0130] In this embodiment, step S4 includes the following steps:

[0131] Step S41: performing time series change analysis on the filtering optimization real-time monitoring data to generate time series change monitoring data;

[0132] Step S42: performing multi-time point evolution simulation on the modal feature point mapping space based on the time series change monitoring data, thereby generating a feature point evolution mapping space;

[0133] Step S43: Perform dynamic evolution trend analysis on the feature point evolution mapping space to generate a dynamic evolution trend trajectory of the feature points;

[0134] Step S44: performing evolution pattern clustering analysis on the dynamic evolution trend trajectory of the feature points, thereby obtaining the dynamic evolution law of the feature points.

[0135] In this embodiment, a suitable time series change analysis method is selected, for example, a moving average method, an exponential smoothing method, an autoregressive model (AR), an autoregressive moving average model (ARMA), an autoregressive integrated moving average model (ARIMA), etc., to perform time series change analysis on the filtering optimization real-time monitoring data, and the time series change analysis method is used to extract the time series characteristics of different modal data, for example, extract the trend of pH value, the fluctuation frequency of dissolved oxygen, the change amplitude of turbidity, etc. The extracted time series feature data are sorted to generate time series change monitoring data, which contains the time series change information of different modal data, and the evolution simulation method is used to simulate the evolution process of the feature points of different modal data in the time dimension, for example, simulating the change trajectory of pH value, dissolved oxygen and turbidity feature points at different time points.

[0136] The evolution results of feature points at different time points are integrated to generate a feature point evolution mapping space, which contains the evolution information of feature points of different modal data in the time dimension. Appropriate dynamic evolution trend analysis methods are selected, such as trend analysis (Trend Analysis), principal component analysis (PCA), etc., to perform dynamic evolution trend analysis on the feature point evolution mapping space. The dynamic evolution trend analysis method is used to analyze the dynamic evolution trends of feature points of different modal data, for example, to analyze the rising trend, falling trend or stable trend of pH value, dissolved oxygen and turbidity feature points. The dynamic evolution trends obtained by the analysis are sorted out to generate feature point dynamic evolution trend trajectories, which represent the dynamic evolution trends of feature points of different modal data in the time dimension.

[0137] The dynamic evolution trend trajectories of feature points of different modal data are clustered using the evolutionary pattern clustering analysis method. For example, the dynamic evolution trend trajectories of pH value, dissolved oxygen and turbidity feature points are divided into different categories. The results of the clustering analysis are sorted out to obtain the dynamic evolution law of the feature points, which represents the dynamic evolution pattern of feature points of different modal data in the time dimension.

[0138] In this embodiment, the specific steps of step S42 are:

[0139] Extracting multiple modal feature points based on the modal feature point mapping space;

[0140] Conduct time-flow feature learning on time-series change monitoring data to generate time-flow features of monitoring data;

[0141] Based on multiple modal feature points as the initial evolution starting point;

[0142] The time-flow characteristics of monitoring data are used to simulate the multi-time-point evolution of the initial evolution starting point and generate a dense map of characteristic point evolution;

[0143] The spatial distribution of feature points is analyzed on the feature point evolution dense graph to generate a feature point evolution mapping space.

[0144] In this embodiment, the original multi-dimensional time series monitoring data is preprocessed, including data cleaning, normalization and other operations, and time series feature extraction methods such as Fourier transform and wavelet transform are used to extract statistical features reflecting the law of time evolution from the original data. The extracted time flow features are modeled and learned using time series models such as recurrent neural networks (such as LSTM) to capture potential patterns in the data sequence. Representative feature points are identified from the original monitoring data, such as extreme points or turning points of key indicators such as pH value and turbidity, and multimodal feature extraction is performed on these feature points, including time features, frequency domain features, correlation features, etc.

[0145] The extracted multi-dimensional feature vector is used as the initial state of the feature point and defined as the evolution starting point. According to the feature vectors of different feature points, an initial set of evolution starting points is constructed in the high-dimensional feature space. Each feature point in the initial set of evolution starting points is traversed, and the learned time flow feature model is used to perform time series prediction on the feature point to simulate its evolution trajectory in the time dimension. For each time step, the feature vector of the feature point is updated according to the predicted value to form an evolution sequence of the feature point in the time dimension. The evolution sequences of all feature points are combined to construct an evolution dense map of the feature points in the time and space dimensions, and the density distribution of the feature points in space is estimated. For example, a kernel density estimation method is used to perform cluster analysis on the feature points to identify different evolution patterns. PCA is used to map the feature points to a low-dimensional space and analyze their spatial distribution. The results of the spatial distribution analysis are integrated to generate a feature point evolution mapping space, which contains the evolution trends and spatial distribution information of different initial evolution starting points in the time dimension.

[0146] In this embodiment, the specific steps of step S5 are:

[0147] Step S51: performing a recursive prediction calculation of multiple feature points on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby generating prediction calculation data of multiple feature points;

[0148] Step S52: performing trend distribution prediction for future time periods on the prediction calculation data of multiple feature points to generate sewage trend distribution prediction data;

[0149] Step S53: Evolving the sewage trend distribution prediction data into multiple trend scenario features, thereby obtaining multiple potential sewage trend prediction scenarios.

[0150] In this embodiment, a time series model, such as ARIMA, SARIMA, etc., is used to perform recursive prediction calculations on each feature point according to the dynamic evolution law of the feature point, and recursive prediction calculations are performed on multiple feature points in the modal feature point mapping space. For example, the changing trends of pH value, dissolved oxygen and turbidity feature points at different time points in the future are predicted, and the prediction results of each feature point at different time points in the future are sorted out to generate prediction calculation data of multiple feature points, which includes the predicted values ​​of different modal feature points at different time points in the future. The selected method is used to predict the trend distribution of multiple feature points in the future. For example, the probability distribution of pH value, dissolved oxygen and turbidity feature points in different time periods in the future is predicted, and the trend distribution prediction results of each feature point in different time periods in the future are sorted out to generate sewage trend distribution prediction data, which includes the probability distribution information of different modal feature points in different time periods in the future.

[0151] According to the sewage trend distribution prediction data, different trend scenarios are generated using rule reasoning methods. For example, if the pH value continues to rise, acidic sewage will be discharged. The selected method is used to evolve the characteristics of multiple trend scenarios of the sewage trend distribution prediction data. For example, a scenario in which the pH value rises, the dissolved oxygen decreases, and the turbidity is stable, as well as a scenario in which the pH value decreases, the dissolved oxygen increases, and the turbidity is stable are generated. The different trend scenarios obtained by evolution are sorted out to obtain multiple potential sewage trend prediction scenarios. Each scenario contains the changing trends and probability distribution information of different modal feature points in different time periods in the future.

[0152] In this embodiment, the specific steps of step S6 are:

[0153] Step S61: Visually render multiple potential sewage situation prediction scenes to generate multiple potential sewage situation visualization views;

[0154] Step S62: performing sewage risk diffusion trajectory analysis on multiple potential sewage situation visualization views to generate multiple potential sewage diffusion trajectories;

[0155] Step S63: performing a sewage risk situation quantitative assessment on multiple potential sewage diffusion trajectories to generate a sewage risk situation quantitative value;

[0156] Step S64: Adaptively optimize the risk decision of the sewage risk situation quantification value and build a sewage treatment intelligent decision model.

[0157] In this embodiment, multiple sewage situation prediction models are established based on the sewage condition monitoring data and feature analysis results obtained in the aforementioned steps, and the output results of these prediction models, such as sewage component concentration, diffusion range, etc., are mapped to a visualization platform for visualization rendering. Through dynamic simulation, a series of potential sewage situation visualization views at different time nodes are generated. For each potential sewage situation visualization view, simulation technologies such as computational fluid dynamics (CFD) are used to simulate the diffusion and propagation process of sewage in the environment.

[0158] Through simulation calculation, the diffusion trajectory of sewage in the time and space dimensions is obtained, including key parameters such as pollutant concentration gradient, diffusion rate, and impact range. The above diffusion trajectory data are superimposed on the visualization platform in the form of vector diagrams or heat maps to intuitively present the spatiotemporal evolution process of sewage diffusion. For different potential situations, multiple corresponding sewage diffusion trajectory diagrams are generated. Combined with environmental quality standards, receptor sensitivity and other factors, a targeted sewage risk assessment model is established. The diffusion trajectory data is used as input and substituted into the risk assessment model for calculation to obtain the quantitative value of the sewage risk situation under each potential situation, reflecting the potential degree of harm caused by sewage accidents to the environment and people.

[0159] These quantitative values ​​are superimposed on the diffusion trajectory diagram of the visualization platform in the form of numbers or colors to intuitively present the sewage risk situation. Combined with real-time monitoring data and historical experience, an adaptive decision-making optimization model based on reinforcement learning is trained. The various quantitative values ​​of sewage risk situations are used as input, and the optimal emergency response strategy is obtained through simulation calculation. This strategy may include adjusting the dosage, starting backup equipment, evacuating people, and other measures. The adaptive decision-making model is integrated into the intelligent sewage treatment system to achieve autonomous optimization of the entire emergency response process.

[0160] Embodiment 2: This embodiment provides a sewage monitoring data intelligent processing system, which is used to execute a sewage monitoring data intelligent processing method as described in Embodiment 1, including:

[0161] The outlier filtering module is used to obtain multi-modal real-time sewage monitoring data; the multi-modal real-time sewage monitoring data is divided into sliding windows, and outlier filtering is performed to construct filtering and optimization real-time monitoring data;

[0162] Implicit association module, used to obtain deep feature data of multiple modes based on filtering and optimizing real-time monitoring data; perform implicit association analysis on deep feature data of multiple modes to generate implicit association data between modes;

[0163] The spatial mapping module is used to perform cross-dimensional spatial mapping of the filtered and optimized real-time monitoring data based on implicit correlation data between modalities, and to perform spatial correlation topological fitting to construct a modal feature point mapping space;

[0164] The dynamic evolution module is used to simulate the multi-time evolution of the modal feature point mapping space and analyze the dynamic evolution trend, so as to obtain the dynamic evolution law of the feature points;

[0165] The sewage situation prediction module is used to perform multi-feature point recursive prediction calculations on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby obtaining multiple potential sewage situation prediction scenarios;

[0166] The intelligent decision-making module is used to generate multiple potential sewage situation visualization views based on multiple potential sewage situation prediction scenarios; adaptively optimize the risk decision of multiple potential sewage situation visualization views, and build an intelligent decision-making model for sewage treatment.

[0167] The present invention obtains multimodal real-time sewage monitoring data, and improves the quality and accuracy of monitoring data through sliding window division and outlier filtering processing, constructs filtering and optimizing real-time monitoring data to exclude outliers and noise, and provides a more reliable data basis for subsequent analysis. The present invention obtains deep feature data of multiple modes based on filtering and optimizing real-time monitoring data, reveals the potential correlation between data, generates implicit correlation data between modalities to fully understand the relationship between different monitoring data, and provides more dimensional information for subsequent analysis. The implicit correlation data between modalities is spatially mapped and correlated topologically fitted, a modal feature point mapping space is constructed, multimodal data is integrated, a modal feature point mapping space is constructed to analyze data across dimensions, and spatial correlations between data are discovered, providing a more comprehensive data perspective for subsequent analysis.

[0168] Multi-time point evolution simulation and dynamic evolution trend analysis are performed on the modal feature point mapping space to reveal the dynamic evolution law of the feature points. The dynamic evolution law of the feature points is obtained to understand the development trend of the data and provide a basis for subsequent predictions and decisions. Recursive prediction calculations of multiple feature points are performed according to the dynamic evolution law of the feature points to generate multiple potential sewage situation prediction scenarios and provide multi-dimensional prediction results. Multiple potential sewage situation prediction scenarios are provided to predict future sewage data development trends and provide diversified prediction results for decision-making. Multiple potential sewage situation visualization views are generated to allow relevant personnel to intuitively understand the sewage scenario and provide visualization support for intelligent decision-making. Adaptive risk decision-making optimization is used to build an intelligent decision-making model for sewage treatment. Intelligent decisions are made based on multiple potential scenarios to improve decision-making efficiency and accuracy.

[0169] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for intelligent processing of sewage monitoring data, characterized in that: The following steps are involved: Step S1: obtaining multimodal sewage real-time monitoring data; Divide the multimodal sewage real-time monitoring data into sliding windows, perform outlier filtering, and construct filtering and optimization real-time monitoring data; Step S2: Based on filtering and optimizing the real-time monitoring data, deep feature data of multiple modes are obtained; implicit correlation analysis is performed on the deep feature data of multiple modes to generate implicit correlation data between modes; the specific steps of step S2 are: Step S21: performing deep feature mining of different modes on the filtered optimized real-time monitoring data, thereby obtaining deep feature data of multiple modes; Step S22: performing modal iterative similarity calculation on deep feature data of multiple modalities to generate a similarity matrix between modalities; Step S23: performing implicit correlation analysis on deep feature data of multiple modes according to the similarity matrix between modes to generate implicit correlation data between modes; Step S3: Based on the implicit correlation data between modalities, the filtered and optimized real-time monitoring data is mapped across dimensions, and spatial correlation topology fitting is performed to construct a modal feature point mapping space; the specific steps of step S3 are: Step S31: performing cross-modal convolution on deep feature data of multiple modalities to generate multiple modal convolution feature maps; Step S32: performing local perception aggregation on multiple modal convolutional feature maps according to implicit correlation data between modalities, thereby generating a local relationship feature tensor; Step S33: performing inter-modal collaborative representation mining on the local relationship feature tensor, thereby generating an inter-modal collaborative representation link; Step S34: performing cross-dimensional spatial mapping on multiple modal convolution feature maps to generate feature points of each modal space mapping; Step S35: performing spatial correlation topology fitting on each modal space mapping feature point based on the inter-modal collaborative representation link to construct a modal feature point mapping space; Step S4: performing multi-time point evolution simulation on the modal feature point mapping space and performing dynamic evolution trend analysis to obtain the dynamic evolution law of the feature points; Step S5: performing multi-feature point recursive prediction calculation on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby obtaining multiple potential sewage situation prediction scenarios; Step S6: Generate multiple potential sewage situation visualization views based on multiple potential sewage situation prediction scenarios; perform adaptive risk decision optimization on the multiple potential sewage situation visualization views to construct a sewage treatment intelligent decision model.

2. The intelligent processing method for sewage monitoring data according to claim 1 is characterized in that: The specific steps of step S1 are: Step S11: acquiring multimodal real-time sewage monitoring data and historical sewage monitoring data based on sensors; Step S12: performing historical parameterized numerical distribution calculation on the historical sewage monitoring data, thereby obtaining the historical numerical distribution range; Step S13: performing sliding window division on the multimodal sewage real-time monitoring data, thereby obtaining a plurality of window monitoring data segments; Step S14: performing dynamic outlier detection on multiple window monitoring data segments according to the historical value distribution range, and extracting outlier abnormal monitoring data; Step S15: Perform outlier data filtering based on the outlier abnormal monitoring data to construct filtered and optimized real-time monitoring data.

3. The intelligent processing method for sewage monitoring data according to claim 1 is characterized in that: The specific steps of step S4 are: Step S41: performing time series change analysis on the filtering optimization real-time monitoring data to generate time series change monitoring data; Step S42: performing multi-time point evolution simulation on the modal feature point mapping space based on the time series change monitoring data, thereby generating a feature point evolution mapping space; Step S43: Perform dynamic evolution trend analysis on the feature point evolution mapping space to generate a dynamic evolution trend trajectory of the feature points; Step S44: performing evolution pattern clustering analysis on the dynamic evolution trend trajectory of the feature points, thereby obtaining the dynamic evolution law of the feature points.

4. The intelligent processing method for sewage monitoring data according to claim 3 is characterized in that: The specific steps of step S42 are: Extracting multiple modal feature points based on the modal feature point mapping space; Conduct time-flow feature learning on time-series change monitoring data to generate time-flow features of monitoring data; Based on multiple modal feature points as the initial evolution starting point; The time-flow characteristics of monitoring data are used to simulate the multi-time-point evolution of the initial evolution starting point and generate a dense map of characteristic point evolution; The spatial distribution of feature points is analyzed on the feature point evolution dense graph to generate a feature point evolution mapping space.

5. The method for intelligent processing of sewage monitoring data according to claim 1, characterized in that: The specific steps of step S5 are: Step S51: performing a recursive prediction calculation of multiple feature points on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby generating prediction calculation data of multiple feature points; Step S52: performing trend distribution prediction for future time periods on the prediction calculation data of multiple feature points to generate sewage trend distribution prediction data; Step S53: Evolving the sewage trend distribution prediction data into multiple trend scenario features, thereby obtaining multiple potential sewage trend prediction scenarios.

6. The method for intelligent processing of sewage monitoring data according to claim 1, characterized in that: The specific steps of step S6 are: Step S61: Visually render multiple potential sewage situation prediction scenes to generate multiple potential sewage situation visualization views; Step S62: performing sewage risk diffusion trajectory analysis on multiple potential sewage situation visualization views to generate multiple potential sewage diffusion trajectories; Step S63: performing a sewage risk situation quantitative assessment on multiple potential sewage diffusion trajectories to generate a sewage risk situation quantitative value; Step S64: Adaptively optimize the risk decision of the sewage risk situation quantification value and build a sewage treatment intelligent decision model.

7. A sewage monitoring data intelligent processing system, characterized in that: Used to execute the intelligent processing method for sewage monitoring data according to claim 1, comprising: The outlier filtering module is used to obtain multi-modal real-time sewage monitoring data; the multi-modal real-time sewage monitoring data is divided into sliding windows, and outlier filtering is performed to construct filtering and optimization real-time monitoring data; Implicit association module, used to obtain deep feature data of multiple modes based on filtering and optimizing real-time monitoring data; perform implicit association analysis on deep feature data of multiple modes to generate implicit association data between modes; The spatial mapping module is used to perform cross-dimensional spatial mapping of the filtered and optimized real-time monitoring data based on implicit correlation data between modalities, and to perform spatial correlation topological fitting to construct a modal feature point mapping space; The dynamic evolution module is used to simulate the multi-time evolution of the modal feature point mapping space and analyze the dynamic evolution trend, so as to obtain the dynamic evolution law of the feature points; The sewage situation prediction module is used to perform multi-feature point recursive prediction calculations on the modal feature point mapping space according to the dynamic evolution law of the feature points, thereby obtaining multiple potential sewage situation prediction scenarios; The intelligent decision-making module is used to generate multiple potential sewage situation visualization views based on multiple potential sewage situation prediction scenarios; adaptively optimize the risk decision of multiple potential sewage situation visualization views, and build an intelligent decision-making model for sewage treatment.

Images