Comprehensive environment monitoring data collaborative management method based on cloud platform

Abstract

The invention discloses an integrated environmental monitoring data collaborative management method based on a cloud platform, and relates to the technical field of environmental monitoring, and the method comprises the following steps: S1, data collection and preprocessing; s2, iterative optimization is carried out through an improved WarpAdm algorithm, and a trained LSTNet model is obtained; s3, analyzing the real-time time sequence data, and generating a predicted value of the environmental data of each time step; s4, generating an environment monitoring data analysis report; s5, realizing multi-user data access control on the cloud platform; s6, supporting data interaction and sharing between users, and supporting collaborative management; and S7, dynamically adjusting CPU and memory resources of the cloud platform. The method overcomes the limitations of dependence on manual analysis, insufficient prediction precision and early warning delay in traditional environmental monitoring data management, provides an efficient, accurate and real-time solution for environmental monitoring and management decision, and has high application value and wide popularization prospects.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, and in particular to a collaborative management method for integrated environmental monitoring data based on a cloud platform. Background Technology

[0002] With the increasing demands for environmental protection and sustainable development, traditional environmental monitoring data management methods are facing growing challenges. In modern environmental management, real-time monitoring and intelligent analysis of comprehensive environmental monitoring data are not only crucial for ensuring environmental quality and safety, but also key to improving the accuracy of environmental protection decisions and management efficiency. However, current environmental monitoring data management methods mostly rely on manual judgment, periodic inspections, or analysis based on simple rules, which are cumbersome and slow to respond. While these traditional methods can provide basic environmental information, their lack of real-time data acquisition and intelligent analysis capabilities makes it difficult to meet the complex environmental monitoring needs and the real-time data processing requirements of cross-domain collaborative management.

[0003] The main limitations of traditional environmental monitoring data management methods lie in their slow data processing and insufficient accuracy. Existing methods typically rely on fixed thresholds, expert experience, or rule-based judgments to analyze environmental anomalies, which is time-consuming and struggles to accurately and promptly capture changes in environmental monitoring data. When environmental monitoring data exhibits characteristics such as multi-source, time-varying, and nonlinearity, the adaptability and generalization ability of traditional methods are severely limited. Especially when dealing with multiple types of environmental monitoring data, such as air pollution, meteorological changes, and water quality monitoring, traditional single-data analysis methods cannot efficiently and accurately identify environmental risks and potential problems, leading to difficulties in the timely detection and handling of environmental issues, thus impacting environmental safety and sustainable development.

[0004] Furthermore, traditional methods in environmental monitoring data management often overlook the spatial correlations and multidimensional dependencies between various data types, making it difficult to fully utilize the potential relationships between data. For example, in complex environmental monitoring systems, traditional single-monitoring-point data analysis methods cannot effectively integrate the correlation information between multi-regional, multi-dimensional monitoring data and various environmental factors, resulting in inaccurate monitoring results, high false alarm rates, and delayed responses. Even when some methods employ machine learning or deep learning techniques, they still fail to fully explore the deep feature relationships between different data sources, making it difficult to achieve efficient, accurate, and real-time environmental risk prediction and management decisions.

[0005] Therefore, how to provide a comprehensive environmental monitoring data collaborative management method based on a cloud platform is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] This invention proposes a cloud-based integrated environmental monitoring data collaborative management method, combining an improved WarpAdam optimization algorithm with the LSTNet model, enabling more accurate real-time monitoring and risk warning of environmental monitoring data. This method automatically extracts the temporal characteristics of environmental monitoring data such as air quality, water quality, and meteorology, significantly improving the accuracy and real-time performance of environmental monitoring data analysis. By introducing the LSTNet model, the predictive ability for long-term dependencies and short-term fluctuations is further enhanced, improving the accuracy of environmental monitoring and early warning. Furthermore, the improved WarpAdam algorithm optimizes the learning process by introducing adaptive learning rate and gradient pruning mechanisms, avoiding the gradient explosion and overfitting problems common in traditional optimization algorithms, ensuring efficient training and stability of the model on complex environmental monitoring data. This method captures subtle changes in environmental monitoring data in real time, achieving accurate prediction and timely early warning of environmental risks, and is particularly suitable for addressing various environmental monitoring tasks such as environmental pollution and climate change. This method overcomes the limitations of traditional environmental monitoring data management, such as reliance on manual analysis, insufficient prediction accuracy, and delayed early warning, providing an efficient, accurate, and real-time solution for environmental monitoring and management decision-making, with high application value and broad prospects for promotion.

[0007] According to an embodiment of the present invention, a collaborative management method for integrated environmental monitoring data based on a cloud platform includes the following steps: S1. Collect environmental monitoring data from environmental monitoring equipment and preprocess it on the cloud platform to obtain an environmental monitoring dataset; S2. Based on the environmental monitoring dataset, input it into the LSTNet model for training, and iteratively optimize it through the improved WarpAdam algorithm to obtain the trained LSTNet model; S3. Based on the environmental monitoring dataset, use the trained LSTNet model to perform real-time time series data analysis and generate predicted values ​​of environmental monitoring data for each time step. S4. Based on the predicted values ​​of environmental monitoring data at each time step, determine data anomalies and trigger anomaly alarms, continuously monitor data fluctuation trends, generate environmental monitoring data analysis reports, and store them on the cloud platform. S5. Implement multi-user data access control on the cloud platform and provide permission-based user management functions; S6. Cloud platform-based data sharing and collaborative management functions support data interaction and sharing between users, and support collaborative management; S7. Dynamically adjust the CPU and memory resources of the cloud platform according to real-time data processing needs.

[0008] Optionally, S1 specifically includes: S11. Collect environmental monitoring data through environmental monitoring equipment, including temperature, humidity, air pressure, wind speed and direction, precipitation, carbon dioxide concentration, nitrogen dioxide concentration, sulfur dioxide concentration, ozone concentration, PM2.5 concentration, PM10 concentration, water temperature, pH value, water turbidity and environmental noise data. S12. Perform data denoising processing on environmental monitoring data using the Kalman filter algorithm; S13. Perform Z-score standardization and Min-Max normalization on environmental monitoring data; S14. Convert the environmental monitoring data into a unified JSON format; S15. Add metadata attributes to the environmental monitoring data, including sensor ID, collection time, and geographical location; S16. Store and output environmental monitoring data.

[0009] Optionally, the improved WarpAdam algorithm specifically includes: Initialize parameters, including the weight matrix and bias term; initialize hyperparameters, including first-order moment estimate, second-order moment estimate, bias correction term, and learning rate; The loss function is set as the mean squared error function. The training samples are input into the model, and the predicted values ​​are obtained through forward propagation. Based on the predicted values ​​and the true values, the loss function value is calculated using the mean squared error function. Through backpropagation, the gradient of the loss function with respect to each parameter is calculated layer by layer. The first-order moment estimate is updated by taking a weighted average of the current gradient and the previous first-order moment estimate; the second-order moment estimate is updated by squaring the current gradient; and the first-order moment estimate and the second-order moment estimate are updated by performing a bias correction based on the bias correction term. Set a gradient change magnitude threshold and a gradient clipping threshold. Calculate the gradient change magnitude based on the root mean square of the current gradient and the previous gradient. If the gradient change magnitude exceeds the gradient change magnitude threshold, sum the squares of all elements in the gradient and take the square root to obtain the gradient L2 norm. Divide the clipping threshold by the gradient L2 norm to obtain the clipping ratio. Multiply the clipping ratio by the gradient to obtain the updated gradient. An adaptive learning rate mechanism is introduced, which divides the previous learning rate by the square root of the second moment estimate to obtain the updated learning rate; The parameter update factor is obtained by dividing the first moment estimate by the second moment estimate; the update step size is obtained by multiplying the parameter update factor by the learning rate; and the updated parameters are obtained by subtracting the update step size from the parameter values. Continue iterative optimization until the loss function converges.

[0010] Optionally, S3 specifically includes: S31. Sort the environmental monitoring dataset in ascending order according to timestamps; set the time window length and divide the environmental monitoring dataset into several time windows according to the time window length, with each time window serving as an independent time-series data sample; S32. Input the time series data samples into the convolutional layer, perform convolution operation through the convolution kernel, extract local features and generate feature maps, perform nonlinear transformation on the feature maps through the ReLU activation function, and perform dimensionality reduction processing using average pooling operation to obtain local feature vectors. S33. Input the local feature vectors into the LSTM layer; multiply the local feature vectors of each time step by the magnitude of the local feature vectors of other time steps to obtain the magnitude product; divide the dot product of the local feature vectors of each time step by the magnitude product to obtain the cosine similarity of the local feature vectors of each time step. S34. Based on the cosine similarity of the local feature vectors at each time step, the attention weights of the local feature vectors at each time step are obtained by processing them through the Softmax function; the local feature vectors at each time step are multiplied by the corresponding attention weights, and the results are summed to obtain the attention-weighted feature vectors at each time step. S35. Based on the attention-weighted feature vector of the current time step and the hidden state vector of the previous time step, the hidden state vector is selectively updated, forgotten, and output through the memory unit and gating mechanism of the LSTM layer to generate the hidden state vector of the current time step. S36. Input the hidden state vector into the fully connected layer; multiply the hidden state vector with the weight matrix of the fully connected layer, and add the bias term of the fully connected layer to generate the predicted value of the environmental monitoring data at each time step.

[0011] Optionally, S4 specifically includes: S41. Compare the predicted and actual values ​​of environmental monitoring data at each time step, and calculate the prediction deviation. S42. Set a prediction deviation threshold. If the prediction deviation exceeds the prediction deviation threshold, the environmental monitoring data is determined to be abnormal, an abnormal warning message is generated, an abnormal alarm is triggered, and a notification is sent to the user through the cloud platform. The abnormal warning message includes the abnormal time point, the abnormal type, and the abnormal data. S43. Continuously monitor the changes in the predicted and actual values ​​of environmental monitoring data, and generate an environmental monitoring data fluctuation trend chart. S44. Based on the abnormal early warning information and the environmental monitoring data fluctuation trend chart, generate an environmental monitoring data analysis report, which includes trend change analysis of environmental monitoring data and analysis of abnormal environmental monitoring data. Store the environmental monitoring data analysis report on the cloud platform.

[0012] Optionally, S5 specifically includes: S51. Build a multi-user data access control module on the cloud platform to support user identity verification and create different user roles according to business needs, including ordinary users, data analysts, and administrators. S52. Assign different access permissions to each user role, including read, modify, and delete permissions for environmental monitoring data and environmental monitoring data analysis reports, and administrator permissions; administrator permissions refer to the ability for administrator users to modify the roles and access permissions of other users; S53. Monitor user actions in real time and generate access logs, including user access time, accessed data, type of operation, result of operation, and remarks.

[0013] Optionally, S6 specifically includes: S61. Provides data sharing functionality through API interface, supporting users to share environmental monitoring data and environmental monitoring data analysis reports with other designated users; S62. When users modify environmental monitoring data and environmental monitoring data analysis reports, a real-time messaging system is used to synchronize and share the data with all users in real time. S63. During real-time synchronization and sharing, retain copies of historical versions of environmental monitoring data and environmental monitoring data analysis reports, generate updated new versions, and support viewing and restoring copies of historical versions. S64. Supports the creation of projects on the cloud platform and the definition of different collaborative tasks and goals for each project; each project can contain multiple sub-tasks, and each sub-task can be assigned specific work content, time nodes and responsible persons; after each sub-task is assigned, the cloud platform tracks the execution progress of the task and reminds the responsible person to complete it on time.

[0014] Optionally, S7 specifically includes: S71. Monitor the CPU and memory usage of each container in real time; S72. Set a CPU utilization threshold for a single container. When the real-time CPU utilization exceeds the CPU utilization threshold, add a new container instance to share the CPU load. S73. Set a memory utilization threshold for a single container. When the real-time memory utilization exceeds the memory utilization threshold, add a new container instance to share the memory load. S74. Continuously monitor system performance, set recycling time thresholds, and periodically clean up resources that have been idle for longer than the recycling time thresholds to release CPU resources and storage space.

[0015] The beneficial effects of this invention are: (8) This invention significantly improves the stability and optimization efficiency of the training process of the environmental monitoring data analysis model by introducing an improved WarpAdam algorithm. Traditional WarpAdam algorithms may exhibit excessive gradient fluctuations, slow training speeds, or overfitting in practical applications. To address these issues, this invention innovatively introduces a gradient clipping mechanism and an adaptive learning rate. The gradient clipping mechanism appropriately scales the gradient by setting a clipping threshold, effectively preventing gradient explosion and making the training process more stable. The adaptive learning rate mechanism dynamically adjusts the learning rate based on the previous learning rate and gradient information, optimizing the model's learning process and ensuring more accurate parameter updates.

[0016] (9) This invention effectively solves the technical challenge of integrating the temporal characteristics of comprehensive environmental monitoring data with multi-source data fusion by employing the LSTNet model. Traditional data processing methods often neglect the complex relationships between environmental monitoring data from different sources, making it difficult to accurately capture environmental change patterns and potential environmental risks. The LSTNet model effectively integrates and processes various environmental monitoring data by adaptively learning the long-term dependencies of time-series data and combining it with a deep learning network structure. This model can automatically learn the correlation weights between different features when analyzing environmental monitoring data, thereby improving the accuracy and robustness of environmental monitoring data prediction. This method breaks through the limitations of traditional data processing methods in multi-dimensional data fusion, significantly improves the accuracy and real-time performance of environmental monitoring data management, and provides efficient and accurate intelligent support for environmental monitoring and decision-making. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0018] Figure 1 This is an overall flowchart of a cloud-based integrated environmental monitoring data collaborative management method proposed in this invention. Figure 2 This is a flowchart of the improved WarpAdam algorithm for a cloud-based integrated environmental monitoring data collaborative management method proposed in this invention. Detailed Implementation

[0019] The invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0020] refer to Figure 1-2 A collaborative management method for integrated environmental monitoring data based on a cloud platform includes the following steps: S1. Collect environmental monitoring data from environmental monitoring equipment and preprocess it on the cloud platform to obtain an environmental monitoring dataset; S2. Based on the environmental monitoring dataset, input it into the LSTNet model for training, and iteratively optimize it through the improved WarpAdam algorithm to obtain the trained LSTNet model; S3. Based on the environmental monitoring dataset, use the trained LSTNet model to perform real-time time series data analysis and generate predicted values ​​of environmental monitoring data for each time step. S4. Based on the predicted values ​​of environmental monitoring data at each time step, determine data anomalies and trigger anomaly alarms, continuously monitor data fluctuation trends, generate environmental monitoring data analysis reports, and store them on the cloud platform. S5. Implement multi-user data access control on the cloud platform and provide permission-based user management functions; S6. Cloud platform-based data sharing and collaborative management functions support data interaction and sharing between users, and support collaborative management; S7. Dynamically adjust the CPU and memory resources of the cloud platform according to real-time data processing needs.

[0021] In this embodiment, S1 specifically includes: S11. Collect environmental monitoring data through environmental monitoring equipment, including temperature, humidity, air pressure, wind speed and direction, precipitation, carbon dioxide concentration, nitrogen dioxide concentration, sulfur dioxide concentration, ozone concentration, PM2.5 concentration, PM10 concentration, water temperature, pH value, water turbidity and environmental noise data. S12. Perform data denoising processing on environmental monitoring data using the Kalman filter algorithm; S13. Perform Z-score standardization and Min-Max normalization on environmental monitoring data; S14. Convert the environmental monitoring data into a unified JSON format; S15. Add metadata attributes to the environmental monitoring data, including sensor ID, collection time, and geographical location; S16. Store and output environmental monitoring data.

[0022] This implementation method collects various environmental monitoring data in real time through multiple environmental monitoring devices, including temperature, humidity, air pressure, wind speed and direction, precipitation, carbon dioxide concentration, nitrogen dioxide concentration, sulfur dioxide concentration, ozone concentration, PM2.5 concentration, PM10 concentration, water temperature, pH value, and water turbidity, ensuring comprehensive data acquisition. Data is transmitted to a cloud platform via a wired network, ensuring real-time performance and efficiency. Kalman filtering is used to denoise the environmental monitoring data, effectively improving data quality. Z-score standardization ensures that various monitoring data are analyzed at the same scale, avoiding errors caused by different data scales. Min-Max normalization further standardizes the data, ensuring a uniform representation within the [0,1] range. By converting environmental monitoring data from different sources and in different formats into a unified JSON format, data format uniformity and compatibility are achieved, facilitating subsequent data processing and analysis. Metadata attributes, including sensor ID, collection time, and geographical location, are added to each data record, ensuring data traceability and integrity. Finally, the processed environmental monitoring data is stored and output, ensuring effective data management and real-time querying.

[0023] In this embodiment, the improved WarpAdam algorithm specifically includes: Initialize parameters, including the weight matrix and bias term; initialize hyperparameters, including first-order moment estimate, second-order moment estimate, bias correction term, and learning rate; The loss function is set as the mean squared error function. The training samples are input into the model, and the predicted values ​​are obtained through forward propagation. Based on the predicted values ​​and the true values, the loss function value is calculated using the mean squared error function. Through backpropagation, the gradient of the loss function with respect to each parameter is calculated layer by layer. The first-order moment estimate is updated by taking a weighted average of the current gradient and the previous first-order moment estimate; the second-order moment estimate is updated by squaring the current gradient; and the first-order moment estimate and the second-order moment estimate are updated by performing a bias correction based on the bias correction term. Set a gradient change magnitude threshold and a gradient clipping threshold. Calculate the gradient change magnitude based on the root mean square of the current gradient and the previous gradient. If the gradient change magnitude exceeds the gradient change magnitude threshold, sum the squares of all elements in the gradient and take the square root to obtain the gradient L2 norm. Divide the clipping threshold by the gradient L2 norm to obtain the clipping ratio. Multiply the clipping ratio by the gradient to obtain the updated gradient. An adaptive learning rate mechanism is introduced, which divides the previous learning rate by the square root of the second moment estimate to obtain the updated learning rate; The parameter update factor is obtained by dividing the first moment estimate by the second moment estimate; the update step size is obtained by multiplying the parameter update factor by the learning rate; and the updated parameters are obtained by subtracting the update step size from the parameter values. Continue iterative optimization until the loss function converges.

[0024] This implementation provides basic parameter settings for the model optimization process by initializing the first-order moment estimate, second-order moment estimate, bias correction term, learning rate, gradient change magnitude threshold, and gradient clipping threshold. Through backpropagation, the gradient of the loss function relative to the model parameters is calculated, providing necessary information for subsequent parameter updates. The first-order moment estimate is updated by weighted averaging of the current gradient and the previous first-order moment estimate, allowing the model to better utilize past gradient information in each iteration. The second-order moment estimate is updated by squaring the current gradient, enabling the second moment to more accurately reflect the gradient variance and further improving the accuracy of learning rate adjustment. Based on the root mean square of the current gradient and the previous gradient, the gradient change magnitude is calculated to determine whether gradient clipping is necessary. If the gradient change magnitude exceeds a set threshold, the squares of all elements in the gradient are summed, and the square root is taken to obtain the gradient L2 norm. The clipping threshold is divided by the gradient L2 norm to obtain the clipping ratio, and the clipping ratio is multiplied by the gradient to obtain the updated gradient. This effectively avoids the gradient explosion problem and ensures the stability of the training process. An adaptive learning rate mechanism is introduced. The updated learning rate is obtained by dividing the previous learning rate by the square root of the second-order moment estimate. This automatic adjustment of the learning rate ensures that the model converges quickly during training while avoiding oscillations caused by an excessively large learning rate. Each model parameter is updated through the learning rate, gradient, and first and second-order moment estimates, ensuring the efficiency and accuracy of the optimization process. Through continuous iterative optimization until the loss function converges, high-precision training and optimization of the model are achieved.

[0025] In this embodiment, S3 specifically includes: S31. Sort the environmental monitoring dataset in ascending order according to timestamps; set the time window length and divide the environmental monitoring dataset into several time windows according to the time window length, with each time window serving as an independent time-series data sample; S32. Input the time series data samples into the convolutional layer, perform convolution operation through the convolution kernel, extract local features and generate feature maps, perform nonlinear transformation on the feature maps through the ReLU activation function, and perform dimensionality reduction processing using average pooling operation to obtain local feature vectors. S33. Input the local feature vectors into the LSTM layer; multiply the local feature vectors of each time step by the magnitude of the local feature vectors of other time steps to obtain the magnitude product; divide the dot product of the local feature vectors of each time step by the magnitude product to obtain the cosine similarity of the local feature vectors of each time step. S34. Based on the cosine similarity of the local feature vectors at each time step, the attention weights of the local feature vectors at each time step are obtained by processing them through the Softmax function; the local feature vectors at each time step are multiplied by the corresponding attention weights, and the results are summed to obtain the attention-weighted feature vectors at each time step. S35. Based on the attention-weighted feature vector of the current time step and the hidden state vector of the previous time step, the hidden state vector is selectively updated, forgotten, and output through the memory unit and gating mechanism of the LSTM layer to generate the hidden state vector of the current time step. S36. Input the hidden state vector into the fully connected layer; multiply the hidden state vector with the weight matrix of the fully connected layer, and add the bias term of the fully connected layer to generate the predicted value of the environmental monitoring data at each time step.

[0026] This implementation divides the environmental monitoring dataset into several time windows by arranging them in ascending order according to timestamps and setting a time window length. Each time window becomes an independent time-series data sample, providing a clear structure for time-series data analysis. By inputting the time-series data samples into a convolutional layer, convolution operations are performed using convolution kernels to extract local features and generate feature maps. A nonlinear transformation is performed using the ReLU activation function, and dimensionality reduction is achieved using average pooling, resulting in local feature vectors and improving the data's feature extraction capability. These local feature vectors are then input into an LSTM layer, and the cosine similarity of the local feature vectors at each time step is calculated using the modulus and dot product, further enhancing the correlation between features and providing a basis for subsequent attention weighting. Based on the cosine similarity of the local feature vectors at each time step, the attention weight for each time step is calculated using the Softmax function. This weight is then multiplied by the local feature vector and summed to generate an attention-weighted feature vector, improving the model's focus on important features. By selectively updating, forgetting, and outputting the attention-weighted feature vector at the current time step and the hidden state vector from the previous time step, the LSTM layer generates the hidden state vector for the current time step, effectively capturing the dependencies and dynamic changes in time-series data. Finally, the hidden state vector is input into a fully connected layer, and the predicted environmental monitoring data value for each time step is generated through the weight matrix and bias term, completing the final prediction task and ensuring that the model can accurately predict environmental monitoring data.

[0027] In this embodiment, S4 specifically includes: S41. Compare the predicted and actual values ​​of environmental monitoring data at each time step, and calculate the prediction deviation. S42. Set a prediction deviation threshold. If the prediction deviation exceeds the prediction deviation threshold, the environmental monitoring data is determined to be abnormal, an abnormal warning message is generated, an abnormal alarm is triggered, and a notification is sent to the user through the cloud platform. The abnormal warning message includes the abnormal time point, the abnormal type, and the abnormal data. S43. Continuously monitor the changes in the predicted and actual values ​​of environmental monitoring data, and generate an environmental monitoring data fluctuation trend chart. S44. Based on the abnormal early warning information and the environmental monitoring data fluctuation trend chart, generate an environmental monitoring data analysis report, which includes trend change analysis of environmental monitoring data and analysis of abnormal environmental monitoring data. Store the environmental monitoring data analysis report on the cloud platform.

[0028] This implementation compares the predicted and actual values ​​of environmental monitoring data at each time step to calculate the prediction deviation, thereby evaluating prediction accuracy. A prediction deviation threshold is set; if the deviation exceeds the threshold, the environmental monitoring data is considered abnormal, generating an anomaly warning message, triggering an alarm, and sending a notification to the user via the cloud platform. This ensures users can promptly access information about abnormal data. The anomaly warning message includes the anomaly time point, anomaly type, and anomaly data, providing a comprehensive anomaly report. The system continuously monitors the changes in the predicted and actual values ​​of environmental monitoring data, generating an environmental monitoring data fluctuation trend chart. This chart displays data change trends, helping users intuitively understand the changes in environmental monitoring data. Based on the anomaly warning message and the environmental monitoring data fluctuation trend chart, an environmental monitoring data analysis report is generated, including trend change analysis and anomaly analysis. This allows users to analyze the causes and trends of data anomalies more deeply. The environmental monitoring data analysis report is stored on the cloud platform for easy access and data management, improving the system's monitoring and analysis capabilities.

[0029] In this embodiment, S5 specifically includes: S51. Build a multi-user data access control module on the cloud platform to support user identity verification and create different user roles according to business needs, including ordinary users, data analysts, and administrators. S52. Assign different access permissions to each user role, including read, modify, and delete permissions for environmental monitoring data and environmental monitoring data analysis reports, and administrator permissions; administrator permissions refer to the ability for administrator users to modify the roles and access permissions of other users; S53. Monitor user actions in real time and generate access logs, including user access time, accessed data, type of operation, result of operation, and remarks.

[0030] This implementation method achieves user authentication by building a multi-user data access control module on a cloud platform. Different user roles, including ordinary users, data analysts, and administrators, are created according to business needs, providing a flexible role management mechanism for the system. Different access permissions are assigned to each user role, including read, modify, and delete permissions for environmental monitoring data and environmental monitoring data analysis reports, as well as administrator permissions. This ensures that users in different roles can operate on data according to their responsibilities and permissions. Administrator permissions allow administrators to modify other users' roles and access permissions, ensuring system security and manageability. Real-time monitoring of user operations generates access logs, recording user access time, accessed data, operation type, operation result, and remarks, providing comprehensive operation auditing capabilities. This helps track user behavior and ensure data security and compliance.

[0031] In this embodiment, S6 specifically includes: S61. Provides data sharing functionality through API interface, supporting users to share environmental monitoring data and environmental monitoring data analysis reports with other designated users; S62. When users modify environmental monitoring data and environmental monitoring data analysis reports, a real-time messaging system is used to synchronize and share the data with all users in real time. S63. During real-time synchronization and sharing, retain copies of historical versions of environmental monitoring data and environmental monitoring data analysis reports, generate updated new versions, and support viewing and restoring copies of historical versions. S64. Supports the creation of projects on the cloud platform and the definition of different collaborative tasks and goals for each project; each project can contain multiple sub-tasks, and each sub-task can be assigned specific work content, time nodes and responsible persons; after each sub-task is assigned, the cloud platform tracks the execution progress of the task and reminds the responsible person to complete it on time.

[0032] This implementation provides data sharing functionality through an API interface, allowing users to share environmental monitoring data and analysis reports with designated other users, improving data sharing and collaboration efficiency. When users modify environmental monitoring data and analysis reports, a real-time messaging system ensures real-time synchronization and sharing among all users, guaranteeing data update consistency and timeliness. During real-time synchronization and sharing, historical versions of environmental monitoring data and analysis reports are retained, and updated versions are generated, allowing users to view and restore historical versions. This provides version management and data backtracking functions, ensuring data security and traceability. It supports creating projects on the cloud platform and defining different collaborative tasks and goals for each project, improving the flexibility and organization of project management. Each project can contain multiple sub-tasks, which can be assigned specific work content, time nodes, and responsible persons. The cloud platform tracks task execution progress and reminds responsible persons to complete tasks on time, ensuring efficient project progress and timely task completion.

[0033] In this embodiment, S7 specifically includes: S71. Monitor the CPU and memory usage of each container in real time; S72. Set a CPU utilization threshold for a single container. When the real-time CPU utilization exceeds the CPU utilization threshold, add a new container instance to share the CPU load. S73. Set a memory utilization threshold for a single container. When the real-time memory utilization exceeds the memory utilization threshold, add a new container instance to share the memory load. S74. Continuously monitor system performance, set recycling time thresholds, and periodically clean up resources that have been idle for longer than the recycling time thresholds to release CPU resources and storage space.

[0034] This implementation ensures real-time control and optimization of system resources by monitoring the CPU and memory utilization of each container. After setting a CPU utilization threshold for a single container, when the real-time CPU utilization exceeds the threshold, the system automatically adds new container instances to share the CPU load, avoiding performance bottlenecks for individual containers. Similarly, after setting a memory utilization threshold for a single container, when the real-time memory utilization exceeds the threshold, the system automatically adds new container instances to share the memory load, ensuring stable system operation and efficient resource utilization. Continuous monitoring of system performance and setting a recycling time threshold allow for the periodic cleanup of resources that have been idle for longer than the threshold, releasing CPU resources and storage space, effectively avoiding resource waste and improving system resource utilization and operating efficiency.

[0035] Example 1: To verify the feasibility of this invention in the collaborative management of integrated environmental monitoring data, the method of this invention was applied to the data management system of an intelligent environmental monitoring platform (hereinafter referred to as "Platform F") of an environmental protection company. In traditional environmental monitoring data management systems, rule-based threshold judgment methods or conventional machine learning models are typically used for data analysis and risk assessment. These methods not only have limited prediction accuracy but also cannot effectively handle multi-source data with strong temporal characteristics and spatial correlations, easily leading to data analysis delays or false alarms. To solve these problems, Platform F decided to adopt the cloud-based collaborative management method for integrated environmental monitoring data proposed in this invention.

[0036] During implementation, platform F first uses equipment such as temperature and humidity sensors, gas sensors, and air quality sensors installed at various monitoring points to collect environmental monitoring data in real time. After noise reduction, standardization, and data normalization, a high-quality structured data sequence is formed.

[0037] Platform F extracts temporal features from monitoring data using the LSTNet model. The LSTNet model combines CNN and LSTM, effectively extracting features from multi-scale time-series data. The CNN part captures short-term local features, while the LSTM part handles long-term dependencies, thus improving the model's ability to model time-series data. By integrating CNN and LSTM, LSTNet can capture deeper patterns and evolutionary laws in complex time-series data, accurately capturing long-term dependencies and sudden changes in environmental monitoring. By extracting efficient temporal features, LSTNet enables the model to more accurately predict future data trends.

[0038] This invention introduces an improved WarpAdam algorithm to optimize the model training process. The improved WarpAdam algorithm incorporates gradient clipping and adaptive learning rate mechanisms, effectively improving the stability and efficiency of model training. The gradient clipping mechanism prevents gradient explosion during training, ensuring the stability of the training process, especially when dealing with deep networks or large-scale datasets, avoiding training instability caused by excessive parameter updates. The adaptive learning rate mechanism dynamically adjusts the learning rate according to gradient changes, allowing the model to learn quickly in the early stages of training and gradually reduce the step size during convergence, avoiding oscillations caused by excessively large learning rates. Simultaneously, it effectively escapes local optima at lower learning rates, improving the model's global optimization capability. Through this combination, the improved WarpAdam algorithm not only accelerates the model's convergence speed but also enhances its generalization ability, reduces the risk of overfitting, and thus significantly improves prediction accuracy and training efficiency.

[0039] During implementation, the technical team of Platform F discovered that, compared with traditional threshold rule methods and conventional machine learning methods, the method of this invention significantly improves the accuracy and response speed of environmental monitoring data analysis. Through the LSTNet model and the improved WarpAdam algorithm, intelligent analysis and early warning of environmental monitoring data are effectively realized.

[0040] To further verify the actual performance of the method of the present invention, platform E conducted a detailed comparative test between the method of the present invention and the traditional method. The specific performance data is shown in Table 1: Table 1. Performance Comparison of Environmental Monitoring Data Analysis Methods on Platform E

[0041] As shown in Table 1, the performance of the environmental monitoring data analysis system was comprehensively improved after applying the method of this invention. The data analysis accuracy increased from 80.2% with the traditional method to 96.8%, while the false alarm rate and false negative rate decreased by 6.6% and 9.3% respectively, significantly improving analysis precision and avoiding resource waste and safety hazards. The processing time for a single prediction decreased from 72 seconds with the traditional method to 22 seconds, and the real-time data processing capability was greatly improved, increasing from 150 data points / second to 550 data points / second, significantly enhancing prediction efficiency and response speed. Furthermore, the prediction result feedback time decreased from 30 seconds to 8 seconds, effectively improving the real-time decision-making response speed. Operation and maintenance costs decreased from 600,000 yuan per month to 420,000 yuan, saving 30% of costs, and user service quality satisfaction increased from 83.5% to 94.7%.

[0042] Through the method of this invention, platform E successfully achieved rapid and accurate analysis of environmental monitoring data, effectively reduced potential risks, ensured safe and reliable environmental operation, significantly improved the intelligence and automation level of the environmental monitoring system, significantly reduced manual operation and maintenance costs, enhanced the stability and robustness of the system, and provided strong technical support for intelligent environmental operation and maintenance.

[0043] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A collaborative management method for integrated environmental monitoring data based on a cloud platform, characterized in that, Includes the following steps: S1. Collect environmental monitoring data from environmental monitoring equipment and preprocess it on the cloud platform to obtain an environmental monitoring dataset; S2. Based on the environmental monitoring dataset, input it into the LSTNet model for training, and iteratively optimize it through the improved WarpAdam algorithm to obtain the trained LSTNet model; S3. Based on the environmental monitoring dataset, use the trained LSTNet model to perform real-time time series data analysis and generate predicted values ​​of environmental monitoring data for each time step. S4. Based on the predicted values ​​of environmental monitoring data at each time step, determine data anomalies and trigger anomaly alarms, continuously monitor data fluctuation trends, generate environmental monitoring data analysis reports, and store them on the cloud platform. S5. Implement multi-user data access control on the cloud platform and provide permission-based user management functions; S6. Cloud platform-based data sharing and collaborative management functions support data interaction and sharing between users, and support collaborative management; S7. Dynamically adjust the CPU and memory resources of the cloud platform according to real-time data processing needs.

2. The method for collaborative management of integrated environmental monitoring data based on a cloud platform according to claim 1, characterized in that, S1 specifically includes: S11. Collect environmental monitoring data through environmental monitoring equipment, including temperature, humidity, air pressure, wind speed and direction, precipitation, carbon dioxide concentration, nitrogen dioxide concentration, sulfur dioxide concentration, ozone concentration, PM2.5 concentration, PM10 concentration, water temperature, pH value, water turbidity and environmental noise data. S12. Perform data denoising processing on environmental monitoring data using the Kalman filter algorithm; S13. Perform Z-score standardization and Min-Max normalization on environmental monitoring data; S14. Convert the environmental monitoring data into a unified JSON format; S15. Add metadata attributes to the environmental monitoring data, including sensor ID, collection time, and geographical location; S16. Store and output environmental monitoring data.

3. The method for collaborative management of integrated environmental monitoring data based on a cloud platform according to claim 1, characterized in that, The improved WarpAdam algorithm specifically includes: Initialize parameters, including the weight matrix and bias term; initialize hyperparameters, including first-order moment estimate, second-order moment estimate, bias correction term, and learning rate; The loss function is set as the mean squared error function. The training samples are input into the model, and the predicted values ​​are obtained through forward propagation. Based on the predicted values ​​and the true values, the loss function value is calculated using the mean squared error function. Through backpropagation, the gradient of the loss function with respect to each parameter is calculated layer by layer. The first-order moment estimate is updated by taking a weighted average of the current gradient and the previous first-order moment estimate; the second-order moment estimate is updated by squaring the current gradient; and the first-order moment estimate and the second-order moment estimate are updated by performing a bias correction based on the bias correction term. Set a gradient change magnitude threshold and a gradient clipping threshold. Calculate the gradient change magnitude based on the root mean square of the current gradient and the previous gradient. If the gradient change magnitude exceeds the gradient change magnitude threshold, sum the squares of all elements in the gradient and take the square root to obtain the gradient L2 norm. Divide the clipping threshold by the gradient L2 norm to obtain the clipping ratio. Multiply the clipping ratio by the gradient to obtain the updated gradient. An adaptive learning rate mechanism is introduced, which divides the previous learning rate by the square root of the second moment estimate to obtain the updated learning rate; The parameter update factor is obtained by dividing the first moment estimate by the second moment estimate; the update step size is obtained by multiplying the parameter update factor by the learning rate; and the updated parameters are obtained by subtracting the update step size from the parameter values. Continue iterative optimization until the loss function converges.

4. The method for collaborative management of integrated environmental monitoring data based on a cloud platform according to claim 1, characterized in that, S3 specifically includes: S31. Sort the environmental monitoring dataset in ascending order according to timestamps; set the time window length and divide the environmental monitoring dataset into several time windows according to the time window length, with each time window serving as an independent time-series data sample; S32. Input the time series data samples into the convolutional layer, perform convolution operation through the convolution kernel, extract local features and generate feature maps, perform nonlinear transformation on the feature maps through the ReLU activation function, and perform dimensionality reduction processing using average pooling operation to obtain local feature vectors. S33. Input the local feature vectors into the LSTM layer; multiply the local feature vectors of each time step by the magnitude of the local feature vectors of other time steps to obtain the magnitude product; divide the dot product of the local feature vectors of each time step by the magnitude product to obtain the cosine similarity of the local feature vectors of each time step. S34. Based on the cosine similarity of the local feature vectors at each time step, the attention weights of the local feature vectors at each time step are obtained by processing them through the Softmax function; the local feature vectors at each time step are multiplied by the corresponding attention weights, and the results are summed to obtain the attention-weighted feature vectors at each time step. S35. Based on the attention-weighted feature vector of the current time step and the hidden state vector of the previous time step, the hidden state vector is selectively updated, forgotten, and output through the memory unit and gating mechanism of the LSTM layer to generate the hidden state vector of the current time step. S36. Input the hidden state vector into the fully connected layer; multiply the hidden state vector with the weight matrix of the fully connected layer, and add the bias term of the fully connected layer to generate the predicted value of the environmental monitoring data at each time step.

5. The method for collaborative management of integrated environmental monitoring data based on a cloud platform according to claim 1, characterized in that, S4 specifically includes: S41. Compare the predicted and actual values ​​of environmental monitoring data at each time step, and calculate the prediction deviation. S42. Set a prediction deviation threshold. If the prediction deviation exceeds the prediction deviation threshold, the environmental monitoring data is determined to be abnormal, an abnormal warning message is generated, an abnormal alarm is triggered, and a notification is sent to the user through the cloud platform. The abnormal warning message includes the abnormal time point, the abnormal type, and the abnormal data. S43. Continuously monitor the changes in the predicted and actual values ​​of environmental monitoring data, and generate an environmental monitoring data fluctuation trend chart. S44. Based on the abnormal early warning information and the environmental monitoring data fluctuation trend chart, generate an environmental monitoring data analysis report, which includes trend change analysis of environmental monitoring data and analysis of abnormal environmental monitoring data. Store the environmental monitoring data analysis report on the cloud platform.

6. The method for collaborative management of integrated environmental monitoring data based on a cloud platform according to claim 1, characterized in that, S5 specifically includes: S51. Build a multi-user data access control module on the cloud platform to support user identity verification and create different user roles according to business needs, including ordinary users, data analysts, and administrators. S52. Assign different access permissions to each user role, including read, modify, and delete permissions for environmental monitoring data and environmental monitoring data analysis reports, and administrator permissions; administrator permissions refer to the ability for administrator users to modify the roles and access permissions of other users; S53. Monitor user actions in real time and generate access logs, including user access time, accessed data, type of operation, result of operation, and remarks.

7. The method for collaborative management of integrated environmental monitoring data based on a cloud platform according to claim 1, characterized in that, S6 specifically includes: S61. Provides data sharing functionality through API interface, supporting users to share environmental monitoring data and environmental monitoring data analysis reports with other designated users; S62. When users modify environmental monitoring data and environmental monitoring data analysis reports, a real-time messaging system is used to synchronize and share the data with all users in real time. S63. During real-time synchronization and sharing, retain copies of historical versions of environmental monitoring data and environmental monitoring data analysis reports, generate updated new versions, and support viewing and restoring copies of historical versions. S64. Supports the creation of projects on the cloud platform and the definition of different collaborative tasks and goals for each project; each project can contain multiple sub-tasks, and each sub-task can be assigned specific work content, time nodes and responsible persons; after each sub-task is assigned, the cloud platform tracks the execution progress of the task and reminds the responsible person to complete it on time.

8. The method for collaborative management of integrated environmental monitoring data based on a cloud platform according to claim 1, characterized in that, S7 includes the following steps: S71. Monitor the CPU and memory usage of each container in real time; S72. Set a CPU utilization threshold for a single container. When the real-time CPU utilization exceeds the CPU utilization threshold, add a new container instance to share the CPU load. S73. Set a memory utilization threshold for a single container. When the real-time memory utilization exceeds the memory utilization threshold, add a new container instance to share the memory load. S74. Continuously monitor system performance, set recycling time thresholds, and periodically clean up resources that have been idle for longer than the recycling time thresholds to release CPU resources and storage space.

Images