Sewage monitoring method and system in rainstorm scene based on deep learning

Abstract

The invention relates to the technical field of environment monitoring, in particular to a sewage monitoring method and system in a rainstorm scene based on deep learning. The method comprises the following steps: acquiring sewage quality data and meteorological data based on a water quality monitoring node, and generating an environment monitoring data sequence; performing cleaning and feature extraction on the environment monitoring data sequence to generate a water quality spatial-temporal feature matrix; inputting the water quality spatial-temporal characteristic matrix into a deep learning model, predicting a sewage quality change trend and an abnormal event, and generating a prediction result sequence; based on the prediction result sequence and historical environment monitoring data, constructing a multi-level decision tree model, and generating a multi-level regulation and control scheme; and executing the regulation and control scheme and feeding back an execution effect, and optimizing the deep learning model and the multi-stage regulation and control scheme through a federated learning updating strategy. According to the invention, sewage quality change and abnormal events can be accurately monitored and predicted under extreme weather conditions such as rainstorm, and the real-time performance and accuracy of sewage treatment are improved.

Description

Technical Field

[0001] The present invention relates to the field of new generation information technology and environmental monitoring technology, and in particular to a sewage monitoring method and system in a rainstorm scenario based on deep learning. Background Art

[0002] With the increase of global climate change and extreme weather events, extreme weather such as rainstorms have brought huge challenges to urban sewage treatment systems and environmental quality. Under extreme weather conditions such as rainstorms, sewage flow and water quality change dramatically. Traditional sewage monitoring methods often cannot reflect water quality changes in a timely and accurate manner, resulting in the inability to effectively predict abnormal water quality events, thus affecting the timeliness and effectiveness of sewage treatment.

[0003] At present, the field of sewage monitoring generally relies on traditional water quality testing instruments and manual sampling methods, which have the disadvantages of long data collection cycle, poor real-time performance and great environmental interference. At the same time, meteorological factors have a great impact on sewage quality, but traditional monitoring systems often fail to effectively integrate water quality monitoring data and meteorological data, resulting in inaccurate predictions of sewage quality changes.

[0004] In order to solve the above problems, researchers have tried to use artificial intelligence technologies such as machine learning to predict sewage quality in recent years. However, due to the dynamic changes in water quality data under extreme weather conditions and the influence of multiple factors, existing technologies still find it difficult to achieve high-precision prediction and effective regulation in extreme weather conditions such as rainstorms. Therefore, how to accurately monitor and predict water quality in extreme weather scenarios through deep learning algorithms combined with multi-source water quality and meteorological data is still a technical problem that needs to be solved urgently. Summary of the invention

[0005] The present invention provides a sewage monitoring method and system in a rainstorm scenario based on deep learning, so as to solve the problem of how to accurately monitor and predict sewage water quality changes and abnormal events based on multi-source water quality, meteorological data and deep learning algorithms in extreme weather scenarios such as rainstorms, so as to achieve real-time early warning and emergency regulation.

[0006] In order to solve the above technical problems, the present invention provides a sewage monitoring method in a rainstorm scene based on deep learning, comprising: Obtain sewage water quality data and meteorological data based on water quality monitoring nodes to generate environmental monitoring data sequences; The environmental monitoring data sequence is cleaned and feature extracted to generate a water quality spatiotemporal feature matrix; the water quality spatiotemporal feature matrix extracts time series features through a sliding window method and is constructed in combination with a spatial distance matrix calculated by combining the geographic coordinates of the monitoring points; Inputting the water quality spatiotemporal feature matrix into a deep learning model, wherein the deep learning model is trained based on a long short-term memory network or / and a Transformer network; predicting sewage water quality change trends and abnormal events, and generating a prediction result sequence; Based on the prediction result sequence and historical environmental monitoring data, a multi-level decision tree model is constructed, and a multi-level control scheme including sewage discharge control, pollution control measures and equipment control instructions is generated; Execute the multi-level control scheme, and input the actual execution effect of the multi-level control scheme into the adaptive model optimization module, and use the federated learning update strategy to optimize and adjust the deep learning model and the multi-level control scheme.

[0007] Furthermore, the acquisition of sewage water quality data and meteorological data specifically includes: Acquire water quality data including water flow rate, water quality components, turbidity, and pH value from water quality monitoring nodes of multiple sewage quality monitoring stations and generate a water quality data set; Obtain meteorological data including rainfall, temperature, and wind speed from a remote meteorological data node of a meteorological monitoring station to generate a meteorological data sequence; The water quality data set is aligned with the meteorological data sequence in terms of time dimension, and data matching is performed based on the geographical location parameters of the water quality monitoring station and the meteorological monitoring station to generate an environmental monitoring data sequence.

[0008] Furthermore, the cleaning and feature extraction of the environmental monitoring data sequence specifically includes: Performing data cleaning on the environmental monitoring data sequence to remove missing values ​​and noise points, and generating a cleaned environmental monitoring data sequence; The time series and spatial geographic features of the cleaned environmental detection data sequence are extracted to generate a water quality spatiotemporal feature matrix.

[0009] Furthermore, the prediction of sewage quality change trends and abnormal events specifically includes: The water quality spatiotemporal feature matrix is ​​input into the deep learning model to predict the sewage quality change trend and generate a sewage quality change trend sequence; Performing anomaly detection on the sewage quality change trend sequence, marking possible abnormal water quality events, and generating a prediction result sequence; The prediction result sequence includes a sewage quality change trend sequence and abnormal event markers.

[0010] Furthermore, the construction of a multi-level decision tree model and the generation of a multi-level control scheme also includes: Based on the prediction result sequence and historical environmental monitoring data, a multi-level decision tree model is constructed to generate a multi-level decision path; Conduct risk assessment on key decision nodes in the multi-level decision path, match emergency response plans based on risk assessment results, and generate an environmental emergency warning report; A multi-level control plan is generated based on the environmental emergency warning report.

[0011] Furthermore, the use of the federated learning update strategy to optimize and adjust the deep learning model and the multi-level control scheme specifically includes: Execute a multi-level control scheme, obtain the actual implementation effect of the multi-level control scheme, including water quality improvement and pollution control results, and generate an actual implementation effect data set; The actual execution effect data set is compared with the prediction result sequence, the prediction error and the deep learning model performance evaluation parameters are calculated, and a performance evaluation parameter set is generated.

[0012] Furthermore, the use of a federated learning update strategy to optimize and adjust the deep learning model and the multi-level control scheme also includes: Based on the performance evaluation parameter set, the deep learning model is optimized by using a federated learning update strategy and an adaptive optimization module to obtain an optimized deep learning model; Adjust the multi-level control plan based on the optimized deep learning model.

[0013] The present invention provides a sewage monitoring system in a rainstorm scene based on deep learning, comprising: The data acquisition module acquires sewage quality data and meteorological data from multiple sensors and data sources to generate environmental monitoring data sequences; A data preprocessing and feature extraction module cleans and extracts features from the environmental monitoring data sequence to generate a water quality spatiotemporal feature matrix; A deep learning model prediction module inputs the water quality spatiotemporal feature matrix into a deep learning model to predict sewage water quality change trends and abnormal events; The multi-level control scheme generation and execution module generates multi-level control schemes based on the prediction result sequence and historical environmental monitoring data; executes the multi-level control schemes and records the actual execution results; The user interaction and visualization module provides an intuitive user interface to display the system's prediction result sequence and multi-level control scheme to help users understand and use the system; The system monitoring and adaptation module continuously monitors the operating status of the system, adaptively adjusts the system settings based on the real-time environmental monitoring data set and actual execution results, and optimizes the deep learning model and multi-level control scheme.

[0014] The present invention has the following beneficial effects: The present invention provides a sewage monitoring method and system in a rainstorm scenario based on deep learning, which can realize accurate monitoring and prediction of sewage water quality under extreme weather conditions (such as rainstorm). Compared with traditional sewage monitoring methods, the deep learning model of the present invention can more accurately process complex information from multiple data sources (such as water quality data and meteorological data), thereby generating real-time and accurate water quality change trends and abnormal event predictions. This method effectively solves the problems of slow data processing and untimely response in traditional monitoring methods, and can significantly improve the efficiency of water quality monitoring and emergency control.

[0015] By introducing a multi-level control scheme and a federated learning mechanism, the present invention can not only generate a multi-level control scheme based on historical environmental monitoring data and real-time data, but also adaptively optimize the deep learning model and the multi-level control scheme according to the execution effect of the multi-level control scheme, continuously improve the prediction accuracy and execution effect of the system, and ensure efficient operation in various complex environments. This model based on adaptive optimization continuously enhances the robustness of the system, especially the emergency response capability for sewage treatment in extreme weather such as heavy rain. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a flow chart of a sewage monitoring method in a rainstorm scenario based on deep learning provided in an embodiment of the present application; Figure 2 A structural block diagram of a sewage monitoring system for rainstorm scenarios based on deep learning provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0018] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0019] Example 1: Reference Figure 1 , is a flow chart of a sewage monitoring method in a rainstorm scenario based on deep learning provided by an embodiment of the present invention. The flow chart may at least include steps S100-S500: S100. Acquire water quality data and meteorological data based on water quality monitoring nodes to generate an environmental monitoring data sequence.

[0020] S200, cleaning and feature extraction of the environmental monitoring data sequence to generate a water quality spatiotemporal feature matrix.

[0021] S300: Input the water quality spatiotemporal feature matrix into a deep learning model, predict sewage quality change trends and abnormal events, and generate a prediction result sequence.

[0022] S400: Based on the prediction result sequence and historical environmental monitoring data, a multi-level decision tree model is constructed, and a multi-level control plan is generated.

[0023] S500, executing the multi-level control scheme, and inputting the actual execution effect of the multi-level control scheme into the adaptive model optimization module, and using the federated learning update strategy to optimize and adjust the deep learning model and the multi-level control scheme.

[0024] Step S100 at least includes steps S110-S130: S110: Acquire water quality data including water flow rate, water quality components, turbidity and pH value from water quality monitoring nodes of multiple sewage monitoring stations and generate a water quality data set.

[0025] Specifically, sewage monitoring nodes are deployed at key sewage discharge points, and multifunctional water quality sensors are installed to collect the following water quality data in real time: Water flow rate : It indicates the amount of water passing through the cross section per unit time, in units of .

[0026] Water composition : It represents the concentration of characteristic pollutant i at time t, in mg / L, including suspended solids (SS), chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), etc.

[0027] Turbidity : Indicates the concentration of suspended particles in water, the unit is NTU.

[0028] pH : Indicates the pH value of water, and the value range is usually 0-14.

[0029] Furthermore, based on the above collected data, a water quality data set is generated: ; in, Represents the water quality dataset at time t. This set will be used as input for subsequent data processing.

[0030] S120: Acquire meteorological data including real-time rainfall, temperature and wind speed of the corresponding monitoring area from the remote meteorological data node of the meteorological monitoring station, and generate a meteorological data sequence.

[0031] Obtain meteorological data corresponding to the monitoring area from the remote meteorological data node of the meteorological monitoring station, including: Rainfall : The total amount of rainfall during time period t, in units of .

[0032] temperature : Ambient temperature at time t, in °C.

[0033] Wind speed : Wind speed at time t, in m / s.

[0034] Furthermore, the meteorological data sequence is constructed according to the time sequence: ; in, Represents the meteorological data sequence at time t, which is used as the input for subsequent data matching and analysis.

[0035] S130: Align the water quality data set and the meteorological data sequence according to the time dimension, perform data matching based on the geographical location parameters of the water quality monitoring station and the meteorological monitoring station, and generate an environmental monitoring data set.

[0036] The water quality dataset generated in step S110 The meteorological data sequence in step S120 Perform alignment and matching in the time dimension. The specific steps are as follows: Use timestamp t to synchronize the two data sets and align them by time dimension to form a temporary matching data set: ; Based on the geographic coordinates (X, Y) of the water quality monitoring station and the geographic coordinates (X', Y') of the meteorological monitoring station, calculate the spatial distance: ; in, , it is considered that the data between the meteorological monitoring station and the water quality monitoring station are related, and data fusion is performed.

[0037] Finally, the environmental monitoring data sequence is generated , represents the environmental monitoring data sequence at time t, and the formula is: ; Right now: .

[0038] Step S200 at least includes steps S210-S230: S210: Perform data cleaning on the environmental monitoring data sequence to remove missing values ​​and noise points, and generate a cleaned environmental monitoring data sequence.

[0039] According to the environmental monitoring data sequence generated in step S130 , perform the following operations on the data: Missing value filling: For missing values, use linear interpolation to fill in discontinuous missing data points: ; in, Indicates that there are missing data points at time t, where t represents the time index of the missing value. Indicates that there is a missing data point at time t-1. Indicates missing data points at time t+1.

[0040] Outlier detection and removal: Use triple standard deviation method to detect noise points: ,in: : The mean of variable X; : standard deviation of variable X.

[0041] Convert data in different formats into a unified format to ensure consistency of data types.

[0042] Cleaning results: Generate a sequence of environmental monitoring data after cleaning: ; S220: performing normalization and standardization processing on the post-cleaning environmental monitoring data sequence to obtain a standardized environmental monitoring data sequence.

[0043] Based on the environmental monitoring data sequence after cleaning in step S210 , follow the steps below to normalize and standardize the data: Normalization: For numerical variables , using the Min-Max normalization formula: ; in, : Normalized variable; Respectively represent the minimum and maximum values ​​of the variable X'.

[0044] Standardization: Use the Z-score standardization formula: ; in, : The mean of variable X'; : Standard deviation of variable X'.

[0045] Standardization results: Get standardized environmental monitoring data series: ; S230: Extracting time series features and spatial geographic features from the standardized environmental monitoring data sequence to generate a water quality spatiotemporal feature matrix.

[0046] According to the standardized environmental monitoring data sequence generated in step S220 , extract time series and spatial geographic characteristics, and construct the water quality spatiotemporal characteristic matrix: Time series feature extraction: Use sliding window method to extract time series features: ;

[0047] in, : time series characteristics; : sliding window weight; is a standardized time series dataset A data point in , indicating that at time step The standardized water quality and meteorological data at the moment. n is the size of the sliding window, which means the length of the data window when extracting time series features.

[0048] Spatial geographic feature extraction: Based on the geographic coordinates (X, Y) of the water quality monitoring nodes, calculate the spatial distance matrix between the monitoring points: ; in, : Spatial geographic feature, indicating the spatial geographic distance between point i and point j.

[0049] Construction of water quality spatiotemporal characteristic matrix: Combine time series characteristics with spatial geographic characteristics to generate a water quality spatiotemporal characteristic matrix: ; in, : water quality spatiotemporal characteristic matrix; : Matrix concatenation operation.

[0050] Step S300 at least includes steps S310-S330: S310: Construct an input tensor based on the spatiotemporal feature matrix of water quality, input it into the spatiotemporal feature extraction layer of the deep learning model, and generate a spatiotemporal embedding feature matrix.

[0051] According to the water quality spatiotemporal characteristic matrix generated in step S230 , construct an input tensor suitable for a deep learning model: Input tensor construction: ; Where N: the number of water quality monitoring points; M: the number of features, including water quality data and meteorological data; T: the time step window.

[0052] Input layer processing: the input tensor Input to the spatiotemporal feature extraction layer of the deep learning model, and use convolutional neural network (CNN) to extract the spatiotemporal embedding feature matrix: ; in, Represents the extracted spatiotemporal embedding feature matrix.

[0053] S320: Based on the spatiotemporal embedding feature matrix, a deep learning model (LSTM or Transformer) is used to predict the sewage quality change trend and generate a sewage quality change trend sequence. Based on the spatiotemporal embedding feature matrix extracted in step S310 , use the deep learning model network to predict it: Deep learning model input: embedding space and time into feature matrix Input deep learning model (such as LSTM or Transformer): ; in, Represents the temporal hidden state at time step t.

[0054] Prediction of sewage quality change trend: Through the fully connected layer and linear activation function, the time series hidden state is predicted: ; in, : The predicted water quality change trend at time t; are the weight and bias parameters of the fully connected layer respectively.

[0055] Water quality trend sequence generation: ;

[0056] S330: Perform anomaly detection on the sewage water quality change trend sequence, mark possible abnormal water quality events, and generate a prediction result sequence.

[0057] According to the water quality change trend sequence generated in step S320 , perform anomaly detection on it and identify potential abnormal events: Assume the water quality prediction threshold is θ; like , then determine is an abnormal value and is determined as an abnormal event; in, : Actual historical water quality observation value; θ: Water quality prediction threshold (set through the deep learning model training process).

[0058] Use anomaly detection models based on probability distribution, such as Gaussian mixture models (GMM) or autoencoders, to further mark abnormal events: ; Generate prediction result sequence: Generate a prediction result sequence containing prediction results and abnormal event labels: ; Step S400 at least includes steps S410-S430: S410: Based on the prediction result sequence and historical environmental monitoring data, a multi-level decision tree model is constructed to generate a multi-level decision path.

[0059] According to the prediction result sequence generated in step S330, combined with historical environmental monitoring data , construct a multi-level decision path, the specific steps are as follows: Construction of decision feature matrix: The prediction result sequence is spliced ​​with the historical environmental monitoring data to generate a decision feature matrix: ; in, : Decision feature matrix, including prediction result sequence and historical environmental monitoring data; ⊕: Matrix concatenation operation.

[0060] Decision tree model construction: Use a multi-level decision tree model based on ID3 or CART algorithm to build a multi-level decision path: ; Where, θ: decision feature partitioning threshold; S: decision feature set; : Among all possible θ partition points, find the partition point that maximizes the information gain, so as to select the optimal decision feature and its corresponding partition threshold; Information Gain (S, θ): Information gain calculation function, used to measure the impact of feature S on decision under a given θ value; Information gain is used to select the best split feature.

[0061] Multi-level decision path sequence generation: The constructed multi-level decision tree model outputs different decision paths to form a multi-level decision path sequence: ; in, : Multi-level decision path sequence; : Number of decision paths.

[0062] S420: Perform risk assessment on key decision nodes in the multi-level decision path sequence, match the best emergency response plan according to the risk assessment results, and generate an environmental emergency warning report.

[0063] Based on the multi-level decision path sequence generated in step S410 , conduct risk assessment and match it with the emergency response plan. The specific steps are as follows: Risk Assessment Model Calculation: Using Risk Assessment Functions Calculate the risk level for each decision path: ; Where R is a scalar value used to quantify the path; The corresponding risk level; : Decision path i; : kth risk factor eigenvalue; : The weight of the risk factor, adjusted according to environmental sensitivity.

[0064] According to the risk assessment level , matching the best emergency response plan: ; in, : The jth emergency plan; : The minimum difference between the emergency response plan and the risk assessment level value, Indicates that in all emergency response plans , find an expression that makes the subsequent expression Smallest That is, choose the path with the same risk assessment value as the current path among all possible options. The closest emergency plan.

[0065] Generate environmental emergency warning report: Generate environmental emergency warning report based on comprehensive risk assessment level and emergency response plan: .

[0066] S430: Generate a multi-level control plan based on the environmental emergency warning report.

[0067] Based on the environmental emergency warning report generated in step S420 , generate a multi-level control plan, the specific steps are as follows: Construction of control measures instructions: Generate corresponding control measures instructions according to the emergency response plan matched in the environmental emergency warning report: ; in, : A collection of directives on sewage control measures.

[0068] Multi-level control scheme generation: Each decision path The control measures and instructions are combined to form a multi-level control plan: .

[0069] Step S500 at least includes steps S510-S530: S510: Execute a multi-level control scheme, obtain the actual execution effect of the multi-level control scheme, including pollution control results and water quality improvement, and generate an actual execution effect data set.

[0070] The multi-level control scheme generated in step S430 is executed, and an execution effect data set is generated based on the water quality data and pollution control conditions actually measured after the execution.

[0071] Execute multi-level control schemes and record the actual execution results for subsequent model optimization: ; Water quality improvement data and pollution control data collection: In the time window t' after the implementation of the multi-level control plan, the real-time water quality data collected from the water quality monitoring nodes is used to generate a data sequence of the actual implementation effect of the multi-level control plan: ; in, : Water flow rate after execution; : Water quality components after execution (such as COD, TP, TN, etc.); : Turbidity after execution; : pH value after execution.

[0072] Water quality improvement parameter recording: Based on the actual execution effect data sequence, record the water quality improvement parameters: ; in, : The prediction result data of the deep learning model at time t'.

[0073] S520: Compare the actual execution effect sequence with the prediction result sequence, and calculate the prediction error and deep learning model performance evaluation parameters.

[0074] The performance evaluation parameters are calculated according to the execution effect data set and the prediction result sequence generated in step S510.

[0075] Compare the actual execution effect data at time t' And the prediction result data , calculate the prediction error: ; in, : The prediction error vector at time t'.

[0076] Performance evaluation parameter calculation: Use mean square error (MSE) and mean absolute error (MAE) for performance evaluation: ; Where n: the number of time steps in the evaluation time window.

[0077] Generation of performance evaluation parameter set: ; Among them, "multi-level control scheme execution deviation" refers to the deviation between the actual execution effect and the expected result when implementing the multi-level control scheme generated by the multi-level decision path. This deviation reflects the inconsistency between the control scheme generated by the model and the actual operation results. It is a set of performance evaluation parameters used to measure the performance and effect of the model (deep learning model and multi-level decision tree model) during the execution process.

[0078] S530: Based on the performance evaluation parameter set, a federated learning model update strategy is constructed to adaptively optimize and retrain the deep learning model and the multi-level control scheme.

[0079] According to the performance evaluation parameter set in step S520 , the federated learning update strategy and adaptive optimization module are used to optimize the deep learning model to obtain the optimized deep learning model: Federated learning update strategy construction: Based on the performance evaluation parameter set, define the deep learning model update target: ; in, They are the new and old parameters of the deep learning model respectively; : learning rate; : Loss function based on performance evaluation parameters.

[0080] Combine the performance evaluation parameters of multiple sewage monitoring stations and execute the federated learning aggregation algorithm: ; in, : global model aggregation parameter; N: the number of sites participating in federated learning.

[0081] Deep learning model optimization: based on global model parameters , update the deep learning model and obtain the optimized deep learning model: ; Adjust the multi-level control plan based on the optimized deep learning model.

[0082] Embodiment 2: Figure 2 FIG. 2 shows a structural block diagram of a sewage monitoring system in a rainstorm scenario based on deep learning according to an embodiment of the present invention. Figure 2 As shown, the system includes: The data acquisition module 10 acquires sewage water quality data and meteorological data from multiple sensors and data sources to generate an environmental monitoring data sequence.

[0083] The data preprocessing and feature extraction module 20 cleans and extracts features from the environmental monitoring data sequence to generate a water quality spatiotemporal feature matrix.

[0084] The deep learning model prediction module 30 inputs the water quality spatiotemporal feature matrix into the deep learning model, predicts the sewage water quality change trend and abnormal events, and generates a prediction result sequence; A multi-level control scheme generation and execution module 40 generates a multi-level control scheme based on the prediction result sequence and historical environmental monitoring data, executes the multi-level control scheme, and records the actual execution effect; The user interaction and visualization module 50 provides an intuitive user interface to display the system's prediction result sequence and multi-level control scheme to help users understand and use the system; The system monitoring and adaptive module 60 continuously monitors the operating status of the system, and adaptively adjusts the system settings according to the real-time environmental monitoring data set and the actual execution effect, and optimizes the deep learning model and the multi-level control scheme.

[0085] The sewage monitoring method and system based on deep learning of the present invention can accurately monitor and predict sewage quality under extreme weather conditions such as rainstorms. The system generates water quality change trends and abnormal event predictions in real time through the fusion of multi-source data and the training of deep learning models, and provides accurate multi-level control solutions. Through the federated learning mechanism and adaptive optimization, the system can continuously optimize the prediction accuracy and decision-making effect, ensuring stable and efficient operation under different environmental conditions.

[0086] In addition, this system also has powerful user interaction functions, supports multi-scenario simulation and real-time feedback, provides reliable technical support for environmental management decisions, and has broad application prospects and social benefits.

[0087] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A sewage monitoring method in a rainstorm scene based on deep learning, characterized in that: The method comprises: Obtain sewage water quality data and meteorological data based on water quality monitoring nodes to generate environmental monitoring data sequences; The environmental monitoring data sequence is cleaned and feature extracted to generate a water quality spatiotemporal feature matrix; the water quality spatiotemporal feature matrix extracts time series features through a sliding window method and is constructed in combination with a spatial distance matrix calculated by combining the geographic coordinates of the monitoring points; Inputting the water quality spatiotemporal feature matrix into a deep learning model, wherein the deep learning model is trained based on a long short-term memory network or / and a Transformer network; predicting sewage water quality change trends and abnormal events, and generating a prediction result sequence; Based on the prediction result sequence and historical environmental monitoring data, a multi-level decision tree model is constructed, and a multi-level control scheme including sewage discharge control, pollution control measures and equipment control instructions is generated; Execute the multi-level control scheme, and input the actual execution effect of the multi-level control scheme into the adaptive model optimization module, and use the federated learning update strategy to optimize and adjust the deep learning model and the multi-level control scheme.

2. The sewage monitoring method in a rainstorm scene based on deep learning according to claim 1 is characterized in that: The generating of the environmental monitoring data sequence specifically comprises: Acquire water quality data including water flow rate, water quality components, turbidity, and pH value from water quality monitoring nodes of multiple sewage monitoring stations and generate a water quality data set; Obtain meteorological data including rainfall, temperature, and wind speed from remote meteorological data nodes of meteorological monitoring stations and generate meteorological data sequences; The water quality data set is aligned with the meteorological data sequence in terms of time dimension, and data matching is performed based on the geographical location parameters of the water quality monitoring station and the meteorological monitoring station to generate an environmental monitoring data sequence.

3. The sewage monitoring method in a rainstorm scene based on deep learning according to claim 1 is characterized in that: The cleaning and feature extraction of the environmental monitoring data sequence specifically includes: Performing data cleaning on the environmental monitoring data sequence to remove missing values ​​and noise points, and generating a cleaned environmental monitoring data sequence; The time series and spatial geographic features of the cleaned environmental detection data sequence are extracted to generate a water quality spatiotemporal feature matrix.

4. The sewage monitoring method in a rainstorm scene based on deep learning according to claim 1 is characterized in that: The prediction of sewage quality change trends and abnormal events specifically includes: The water quality spatiotemporal feature matrix is ​​input into the deep learning model to predict the sewage quality change trend and generate a sewage quality change trend sequence; Performing anomaly detection on the sewage quality change trend sequence, marking possible abnormal water quality events, and generating a prediction result sequence; The prediction result sequence includes a sewage quality change trend sequence and abnormal event markers.

5. The sewage monitoring method in a rainstorm scene based on deep learning according to claim 1 is characterized in that: The construction of a multi-level decision tree model and the generation of a multi-level control scheme also includes: Based on the prediction result sequence and historical environmental monitoring data, a multi-level decision tree model is constructed to generate a multi-level decision path; Conduct risk assessment on key decision nodes in the multi-level decision path, match emergency response plans based on risk assessment results, and generate an environmental emergency warning report; A multi-level control plan is generated based on the environmental emergency warning report.

6. The sewage monitoring method in a rainstorm scene based on deep learning according to claim 5 is characterized in that: The use of the federated learning update strategy to optimize and adjust the deep learning model and the multi-level control scheme specifically includes: Execute a multi-level control scheme, obtain the actual implementation effect of the multi-level control scheme, including water quality improvement and pollution control results, and generate an actual implementation effect data set; The actual execution effect data set is compared with the prediction result sequence, the prediction error and the deep learning model performance evaluation parameters are calculated, and a performance evaluation parameter set is generated.

7. The sewage monitoring method in a rainstorm scene based on deep learning according to claim 6 is characterized in that: The use of the federated learning update strategy to optimize and adjust the deep learning model and the multi-level control scheme also includes: Based on the performance evaluation parameter set, the deep learning model is optimized by using a federated learning update strategy and an adaptive optimization module to obtain an optimized deep learning model; Adjust the multi-level control plan based on the optimized deep learning model.

8. A sewage monitoring system in a rainstorm scenario based on deep learning, characterized in that: The system comprises: The data acquisition module acquires sewage quality data and meteorological data from multiple sensors and data sources to generate environmental monitoring data sequences; A data preprocessing and feature extraction module cleans and extracts features from the environmental monitoring data sequence to generate a water quality spatiotemporal feature matrix; A deep learning model prediction module inputs the water quality spatiotemporal feature matrix into a deep learning model, predicts sewage water quality change trends and abnormal events, and generates a prediction result sequence; The multi-level control scheme generation and execution module generates multi-level control schemes based on the prediction result sequence and historical environmental monitoring data; executes the multi-level control schemes and records the actual execution results; The user interaction and visualization module provides an intuitive user interface to display the system's prediction result sequence and multi-level control scheme to help users understand and use the system; System monitoring and adaptive module, which continuously monitors the operating status of the system, and adaptively adjusts system settings based on the real-time environmental monitoring data set and actual execution results, and optimizes deep learning models and multi-level control schemes; To realize a sewage monitoring method in a rainstorm scene based on deep learning as described in any one of claims 1-7.

Images