A Method and System for Monitoring and Controlling Urban Sewage Pipeline Networks Based on Multi-Source Data

Abstract

This invention discloses a method and system for monitoring and controlling urban sewage pipe networks based on multi-source data, belonging to the field of urban sewage management technology. The method includes: preprocessing and adaptively weighting and fusing raw data, external auxiliary data, and standard data from the pipe network sensing layer; assessing the risk indices of blockage, overflow, and water quality exceeding standards based on the fused data; constructing a Gaussian process regression prediction model with a spatiotemporal composite kernel; dynamically setting control thresholds based on design and operation and maintenance cost parameters; generating real-time control commands for the pipe network; adjusting control parameters based on equipment feedback; and constructing a digital twin to achieve pipe network visualization and synchronous updates. This invention improves monitoring reliability through precise fusion of multi-source data; achieves higher prediction accuracy through spatiotemporal coupling; balances pipe network safety and energy conservation through dynamic thresholds and multi-objective optimization; and effectively improves the real-time performance, traceability, and visualized management level of control through closed-loop feedback and digital twins.

Description

Technical Field

[0001] This invention relates to the field of urban wastewater management technology, and in particular to a method and system for monitoring and controlling urban wastewater pipe networks based on multi-source data. Background Technology

[0002] Urban sewage pipe networks are core municipal infrastructure for the collection, transportation, and discharge of urban sewage. Their operational status directly determines the effectiveness of urban flood control, water pollution prevention, and municipal drainage safety. Monitoring and control of urban sewage pipe networks involves real-time sensing, data analysis, risk assessment, and enforcement mechanism regulation of key parameters such as pipe network flow, liquid level, water quality, and equipment operating status. This allows for the timely identification of operational hazards such as blockages, overflows, and water quality exceeding standards, and the optimization of operating strategies for pump stations, valves, and other equipment. It is a core technological means to achieve smart water management, ensure the resilience of urban drainage systems, and reduce operation and maintenance costs and environmental risks.

[0003] Existing urban sewage pipe network monitoring and control technologies are mostly based on single-point liquid level and flow sensor data acquisition, and use fixed thresholds for abnormal alarms; some technologies introduce traditional machine learning models for simple time series prediction, and rely on manual experience or conventional proportional algorithms to adjust the operating parameters of pump stations and valves; a few solutions combine GIS systems to realize the display of basic pipe network information. Overall, the main technical approach is "single-point monitoring, fixed thresholds, passive control, and one-way display".

[0004] This results in existing technical solutions having several problems: a lack of adaptive fusion mechanisms for multi-source heterogeneous monitoring data; unreasonable integration of temporal, spatial, environmental, and equipment status data; fixed weights; low data utilization; large monitoring errors; insufficient accuracy in risk prediction under rainfall disturbances due to the reliance of status prediction models on single-dimensional time-series data; inability to dynamically adjust control thresholds based on prediction results and maintenance costs, making it difficult to balance the multi-objective requirements of overflow prevention, blockage prevention, and energy saving; and the lack of a closed-loop feedback correction mechanism after control command execution, leading to lag in regulation, poor visualization, and poor traceability. Summary of the Invention

[0005] This invention provides a method and system for monitoring and controlling urban sewage pipe networks based on multi-source data. It addresses the following issues in existing technologies: lack of adaptive fusion mechanisms for multi-source heterogeneous monitoring data; unreasonable integration of temporal, spatial, environmental, and equipment status data; fixed weights; low data utilization; large monitoring errors; insufficient accuracy in risk prediction under rainfall disturbances due to the reliance on single-dimensional time-series data for status prediction models; inability to dynamically adjust control thresholds based on prediction results and maintenance costs, making it difficult to balance multiple objectives such as overflow prevention, blockage prevention, and energy conservation; and the lack of a closed-loop feedback correction mechanism after control command execution, resulting in lag in regulation, poor visualization, and poor traceability.

[0006] On the one hand, embodiments of the present invention provide a method for monitoring and controlling urban sewage pipe networks based on multi-source data, including: The raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network are preprocessed to obtain a standardized multi-source dataset; the standardized multi-source dataset is then classified and weighted to obtain a multi-source fusion dataset with a unified format. Based on the pipeline design parameters and the multi-source fusion dataset, a pipeline status assessment is performed to obtain pipeline status assessment results including blockage risk index, overflow risk index and water quality exceedance risk index; A pipeline network status prediction model is established based on historical fusion datasets, rainfall forecast data, and pipeline network status assessment results. The pipeline network status prediction model is a Gaussian process regression model with a spatiotemporal composite kernel function. The dynamic control threshold of the pipeline network is set based on the pipeline network design parameters, operation and maintenance cost parameters, and the prediction results of the pipeline network status prediction model. A real-time control instruction set for the pipeline network is generated based on the operating parameters of the pump station valves, the prediction results, and the dynamic control threshold of the pipeline network. The adjusted control parameters and deviation dataset are obtained based on the feedback data from the execution equipment and the real-time control instruction set of the pipeline network. Based on the control parameters and deviation dataset, a twin model is constructed to enable visualization and synchronous updating of the urban pipeline network.

[0007] In one possible implementation, the raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network are preprocessed to obtain a standardized multi-source dataset, including: The flow data of the urban pipeline network is collected by an insertion ultrasonic flow meter in the sensing layer; COD, ammonia nitrogen and SS data of the pipeline network are collected by a multi-parameter water quality monitor; pressure and level data of the pipeline network are collected by a pressure and level sensor; and gas composition data of the pipeline network are collected by a gas sensor. The original data is obtained by packaging the flow rate data, COD data, ammonia nitrogen data, SS data, pressure data, liquid level data, and gas composition data. The external auxiliary data for the urban pipeline network includes rainfall data from meteorological stations, GIS pipeline network topology data, historical fault data, and operating parameters of pumping station equipment.

[0008] In one possible implementation, classifying and weighting the standardized multi-source dataset to obtain a uniformly formatted multi-source fused dataset includes: The data from the standardized multi-source dataset is classified into time-series data, spatial data, state-related data, and environmental data. Based on the classification methods of time-series data, spatial data, state-type data, and environment-type data, the basic weights of the weighted fusion algorithm are reset to obtain an adaptive weighted fusion algorithm; The standardized multi-source dataset is obtained by weighted fusion of the data using the adaptive weighted fusion algorithm.

[0009] In one possible implementation, a pipeline network condition prediction model is established based on historical fusion datasets, rainfall forecast data, and the pipeline network condition assessment results. This pipeline network condition prediction model is a Gaussian process regression model with a spatiotemporal composite kernel function, comprising: An improved Gaussian process regression model with a spatiotemporal composite kernel function is obtained by superimposing a time kernel function and a spatial kernel function on the standard Gaussian process regression model. The historical fusion dataset and the corresponding pipeline status evaluation results are divided into training set and validation set according to a preset ratio; The prediction error RMSE on the validation set is optimized using a Bayesian optimization algorithm. The optimal hyperparameter combination that fits the characteristics of the pipeline network data is obtained by iteratively optimizing the kernel variance and length scale hyperparameter of the spatiotemporal composite kernel function using the Bayesian optimization algorithm. The improved Gaussian process regression model is trained by optimizing and minimizing the RMSE (Reference Error of Validation Set) and the optimal combination of hyperparameters.

[0010] In one possible implementation, setting the pipeline dynamic control threshold based on pipeline design parameters, operation and maintenance cost parameters, and the prediction results of the pipeline state prediction model includes: By inputting rainfall forecast data and current pipeline status assessment results into the trained improved Gaussian process regression model, prediction results for the blockage risk index, overflow risk index, and water quality exceedance risk index for the future preset period are obtained.

[0011] In one possible implementation, generating a real-time control instruction set for the pipeline network based on the pump station valve operating parameters, the prediction results, and the pipeline network dynamic control threshold includes: The optimization objectives are to minimize the energy consumption of pump station operation, the risk of pipeline blockage, the risk of overflow, and the risk of water quality exceeding standards. A multi-objective optimization objective function for the pipeline network is constructed using the blockage risk index, overflow risk index, and water quality exceedance risk index from the prediction results. Constraints on liquid level, flow rate, and equipment operating status are established based on the pipeline network dynamic control threshold and the pump group power characteristics, valve opening and flow rate correspondence, and maximum number of pump groups in the pump station valve operating parameters. An improved multi-objective particle swarm optimization algorithm incorporating an adaptive clustering mechanism is used to iteratively solve the multi-objective optimization objective function within the constraints to obtain a set of Pareto optimal solutions; Based on the pipeline network operation and maintenance priority, the optimal control scheme is selected from the Pareto optimal solution set to generate a real-time pipeline network control instruction set containing the number of pump units in operation, valve opening degree, and running time.

[0012] In one possible implementation, constructing a twin model based on the control parameters and deviation dataset to achieve visualization and synchronous updating of the urban pipeline network includes: Based on the GIS topology data and design parameters of the urban pipeline network, a digital twin basic geometric model and a hydraulic simulation sub-model matching the hydraulic characteristics of the pipeline network are constructed. A digital twin model of the urban pipe network is constructed based on the aforementioned digital twin basic geometric model and a hydraulic simulation sub-model that matches the hydraulic characteristics of the pipe network. The adjusted control parameters and deviation datasets are combined with the full-process data of pipeline monitoring, condition assessment, prediction, and control commands and imported into the digital twin model for hydraulic parameter calibration and correction. The real-time operating status, control command execution results, and deviation data of the urban pipeline network are synchronously mapped to the calibrated and corrected digital twin model of hydraulic parameters to achieve a consistent matching between the urban pipeline network and the digital twin model. Based on the synchronously completed digital twin model, the network topology, equipment operating status, risk level, control parameters and deviation information are visualized, achieving real-time visualization and dynamic synchronous updates of the urban network.

[0013] On the other hand, embodiments of the present invention provide a monitoring and control system for urban sewage pipe networks based on multi-source data, including: The data acquisition module is used to preprocess the raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network to obtain a standardized multi-source dataset; The data fusion processing module is used to classify and weightedly fuse the standardized multi-source dataset to obtain a multi-source fusion dataset with a unified format. The status assessment module is used to perform a pipeline status assessment based on the pipeline design parameters and the multi-source fusion dataset to obtain a pipeline status assessment result that includes a blockage risk index, an overflow risk index, and a water quality exceedance risk index. The model building module is used to establish a pipeline network status prediction model based on historical fusion datasets, rainfall forecast data and pipeline network status assessment results. The pipeline network status prediction model is a Gaussian process regression model with a spatiotemporal composite kernel function. The threshold setting module is used to set the dynamic control threshold of the pipeline network based on the pipeline network design parameters, operation and maintenance cost parameters and the prediction results of the pipeline network status prediction model. The instruction generation module is used to generate a real-time control instruction set for the pipeline network based on the operating parameters of the pump station valves, the prediction results, and the pipeline network dynamic control threshold. The digital twin display module is used to obtain adjusted control parameters and deviation datasets based on feedback data from the execution equipment and the real-time control instruction set of the pipeline network; and to construct a twin model based on the control parameters and deviation datasets to achieve visualization and synchronous updating of the urban pipeline network.

[0014] The urban sewage pipe network monitoring and control method and system based on multi-source data in this invention have the following advantages: (1) By classifying time-series, spatial, state-based and environmental data, and dynamically adjusting the weights based on sensor reliability, the data from multiple sources can be accurately integrated, thereby significantly improving data utilization and monitoring reliability.

[0015] (2) An improved Gaussian process regression model with a spatiotemporal composite kernel function is adopted to simultaneously capture the temporal change law of pipeline network status and the topological spatial correlation characteristics. Combined with rainfall forecast data, the prediction accuracy of blockage, overflow and water quality exceedance risks is significantly improved.

[0016] (3) Based on the prediction results, pipeline design parameters and operation and maintenance costs, the liquid level and flow control thresholds are dynamically generated. Combined with the improved multi-objective particle swarm optimization algorithm, collaborative optimization is achieved, which effectively reduces the risk of overflow and blockage while significantly reducing the energy consumption of pump station operation.

[0017] (4) Closed-loop correction of control parameters is achieved by executing equipment feedback data, and a digital twin model is constructed to complete the real-time synchronous mapping between the physical pipeline network and the virtual model, thereby improving the real-time control, decision traceability and pipeline network visualization management level. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the urban sewage pipe network monitoring and control method based on multi-source data provided in this application embodiment; Figure 2 A schematic diagram of the structure of a city sewage pipe network monitoring and control system based on multi-source data provided in this application embodiment. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Figure 1 This is a flowchart illustrating the urban sewage pipe network monitoring and control method based on multi-source data provided in an embodiment of the present invention. The embodiment of the present invention provides an urban sewage pipe network monitoring and control method based on multi-source data, including: The raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network are preprocessed to obtain a standardized multi-source dataset; the standardized multi-source dataset is then classified and weighted to obtain a multi-source fusion dataset with a unified format. Based on the pipeline design parameters and the multi-source fusion dataset, a pipeline status assessment is performed to obtain pipeline status assessment results including blockage risk index, overflow risk index and water quality exceedance risk index; A pipeline network status prediction model is established based on historical fusion datasets, rainfall forecast data, and pipeline network status assessment results. The pipeline network status prediction model is a Gaussian process regression model with a spatiotemporal composite kernel function. The dynamic control threshold of the pipeline network is set based on the pipeline network design parameters, operation and maintenance cost parameters, and the prediction results of the pipeline network status prediction model. A real-time control instruction set for the pipeline network is generated based on the operating parameters of the pump station valves, the prediction results, and the dynamic control threshold of the pipeline network. The adjusted control parameters and deviation dataset are obtained based on the feedback data from the execution equipment and the real-time control instruction set of the pipeline network. Based on the control parameters and deviation dataset, a twin model is constructed to enable visualization and synchronous updating of the urban pipeline network.

[0022] The standardized multi-source dataset is obtained by preprocessing the raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network, including: The flow data of the urban pipeline network is collected by an insertion ultrasonic flow meter in the sensing layer; COD, ammonia nitrogen and SS data of the pipeline network are collected by a multi-parameter water quality monitor; pressure and level data of the pipeline network are collected by a pressure and level sensor; and gas composition data of the pipeline network are collected by a gas sensor. The original data is obtained by packaging the flow rate data, COD data, ammonia nitrogen data, SS data, pressure data, liquid level data, and gas composition data. The external auxiliary data for the urban pipeline network includes rainfall data from meteorological stations, GIS pipeline network topology data, historical fault data, and operating parameters of pumping station equipment.

[0023] The standardized multi-source dataset is classified and weighted to obtain a unified multi-source fusion dataset, including: The data from the standardized multi-source dataset is classified into time-series data, spatial data, state-related data, and environmental data. Based on the classification methods of time-series data, spatial data, state-type data, and environment-type data, the basic weights of the weighted fusion algorithm are reset to obtain an adaptive weighted fusion algorithm; The standardized multi-source dataset is obtained by weighted fusion of the data using the adaptive weighted fusion algorithm.

[0024] A pipeline network condition prediction model is established based on historical fusion datasets, rainfall forecast data, and the pipeline network condition assessment results. This model is a Gaussian process regression model with a spatiotemporal composite kernel function, comprising: An improved Gaussian process regression model with a spatiotemporal composite kernel function is obtained by superimposing a time kernel function and a spatial kernel function on the standard Gaussian process regression model. The historical fusion dataset and the corresponding pipeline status evaluation results are divided into training set and validation set according to a preset ratio; The prediction error RMSE on the validation set is optimized using a Bayesian optimization algorithm. The optimal hyperparameter combination that fits the characteristics of the pipeline network data is obtained by iteratively optimizing the kernel variance and length scale hyperparameter of the spatiotemporal composite kernel function using the Bayesian optimization algorithm. The improved Gaussian process regression model is trained by optimizing and minimizing the RMSE (Reference Error of Validation Set) and the optimal combination of hyperparameters.

[0025] Setting dynamic control thresholds for the pipeline network based on pipeline design parameters, operation and maintenance cost parameters, and the prediction results of the pipeline network status prediction model includes: By inputting rainfall forecast data and current pipeline status assessment results into the trained improved Gaussian process regression model, prediction results for the blockage risk index, overflow risk index, and water quality exceedance risk index for the future preset period are obtained.

[0026] The real-time control instruction set for the pipeline network is generated based on the pump station valve operating parameters, the prediction results, and the pipeline network dynamic control threshold, including: The optimization objectives are to minimize the energy consumption of pump station operation, the risk of pipeline blockage, the risk of overflow, and the risk of water quality exceeding standards. A multi-objective optimization objective function for the pipeline network is constructed using the blockage risk index, overflow risk index, and water quality exceedance risk index from the prediction results. Constraints on liquid level, flow rate, and equipment operating status are established based on the pipeline network dynamic control threshold and the pump group power characteristics, valve opening and flow rate correspondence, and maximum number of pump groups in the pump station valve operating parameters. An improved multi-objective particle swarm optimization algorithm incorporating an adaptive clustering mechanism is used to iteratively solve the multi-objective optimization objective function within the constraints to obtain a set of Pareto optimal solutions; Based on the pipeline network operation and maintenance priority, the optimal control scheme is selected from the Pareto optimal solution set to generate a real-time pipeline network control instruction set containing the number of pump units in operation, valve opening degree, and running time.

[0027] Constructing a twin model based on the control parameters and deviation dataset to achieve visualization and synchronous updating of the urban pipeline network includes: Based on the GIS topology data and design parameters of the urban pipeline network, a digital twin basic geometric model and a hydraulic simulation sub-model matching the hydraulic characteristics of the pipeline network are constructed. A digital twin model of the urban pipe network is constructed based on the aforementioned digital twin basic geometric model and a hydraulic simulation sub-model that matches the hydraulic characteristics of the pipe network. The adjusted control parameters and deviation datasets are combined with the full-process data of pipeline monitoring, condition assessment, prediction, and control commands and imported into the digital twin model for hydraulic parameter calibration and correction. The real-time operating status, control command execution results, and deviation data of the urban pipeline network are synchronously mapped to the calibrated and corrected digital twin model of hydraulic parameters to achieve a consistent matching between the urban pipeline network and the digital twin model. Based on the synchronously completed digital twin model, the network topology, equipment operating status, risk level, control parameters and deviation information are visualized, achieving real-time visualization and dynamic synchronous updates of the urban network.

[0028] For example, data acquisition in the urban pipeline network sensing layer is achieved by collecting the actual flow rate of the pipeline network using an insertion ultrasonic flow meter. Multi-parameter water quality monitors collect actual chemical oxygen demand. Pressure and level sensors collect pressure data and actual liquid levels in the pipeline network. Gas sensors collect gas composition data and package it to form raw data. Then, the actual rainfall was obtained through weather stations. GIS pipeline topology data, historical fault data, and pump station equipment operating parameters are packaged to form external auxiliary data for the urban pipeline network. Based on national / industry / local standards and specifications for the design, operation, and management of urban sewage pipe networks, including design flow rate, design level, design pressure, water quality discharge standard limits, and rated operating parameters of pumping stations / valve equipment, urban pipe network standard and specification data are formed and denoted as [data missing]. .

[0029] Then to , and After performing timestamp alignment, unit unification, linear interpolation of missing values, and outlier removal using the 3σ criterion, a standardized multi-source dataset is obtained, denoted as . .

[0030] The standardized multi-source dataset is divided into time-series data, spatial data, state-based data, and environmental data. Based on the characteristics of these four data types, the basic weights are reset to form an adaptive weighted fusion algorithm. This algorithm is used to perform a weighted summation of the standardized data, resulting in fused data with a unified format. The formula is as follows: ; in, For adaptively adjusted weights, Let represent the mean values ​​of time-series data, spatial data, state-based data, and environment-based data after data denoising, respectively, and satisfy the weight normalization condition. .

[0031] After denoising the four types of standardized data, their means were calculated, and then a weighted sum was obtained using an adaptive weighted fusion algorithm to obtain a multi-source fusion dataset with a uniform format. The fusion calculation formula is as follows: ; in, It is a multi-source fusion dataset, which is a one-dimensional row vector containing all the core feature indicators of pipeline operation; It is time series data D t The mean value after noise reduction is a scalar that reflects the overall level of the time series data; It is spatial data D s The mean value after noise reduction is a scalar that reflects the core characteristics of the pipeline network spatial topology; It is state-type data D sta The mean value after noise reduction is a scalar quantity that reflects the overall operating status of the pipeline network and equipment. It is environmental data D env The mean value after noise reduction is a scalar quantity that reflects the comprehensive impact of the external environment on the pipeline network.

[0032] The result of fusion calculation The data is converted into a structured data table format, containing four core fields: monitoring time, monitoring node, feature index, and fused value, thus completing the construction of the multi-source fusion dataset.

[0033] Based on pipeline design parameters and combined with real-time monitoring fusion values ​​from multi-source fusion datasets, index calculation models are constructed for three core operational risks: blockage, overflow, and water quality exceeding standards, to quantitatively assess the current operational status of the pipeline network; from the multi-source fusion datasets... Core indicators directly related to three types of risks were selected: real-time flow rate and pipeline pressure for blockage risk; real-time liquid level and rainfall for overflow risk; and COD, ammonia nitrogen, and SS concentrations for water quality exceeding standards risk. The actual combined values ​​of each risk-related indicator were compared with the baseline values ​​of the pipeline network design parameters to obtain normalized indicator values, eliminating dimensional differences. The normalization formula is as follows: ; in, It is the normalized value of the indicator, which is a scalar; From The fusion value extracted is a scalar; These are the baseline values ​​of the pipeline design parameters corresponding to the indicators, and are scalar quantities.

[0034] The blockage risk is negatively correlated with the actual flow rate of the pipeline and positively correlated with the actual pressure of the pipeline. The blockage risk index calculation model is as follows: ; in, It is a blockage risk index, dimensionless, with a value range of [0,1]. The larger the value, the higher the risk of blockage in the pipeline network. It is the weighting coefficient of the traffic indicator; It is the weighting coefficient of the pressure indicator, which satisfies The determination is based on pipeline network operation and maintenance experience and the analytic hierarchy process. It is the normalized value of real-time traffic, derived from the formula. Calculated; It is the normalized value of the pipeline pressure, derived from the formula. Calculations show that This is the normalized value.

[0035] Overflow risk is positively correlated with both the actual liquid level in the pipeline network and rainfall. Furthermore, a rainfall impact coefficient γ is introduced to reflect the amplifying effect of rainfall on overflow risk. The overflow risk index calculation model is constructed as follows: ; in, It is an overflow risk index, dimensionless, with a value range of [0,1]. The larger the value, the higher the risk of pipeline overflow. It is the weighting coefficient of the liquid level index, 0 < <1, determined by the liquid level safety factor in the pipeline design; It is the normalized value of the real-time liquid level, derived from the formula. Calculated. It is the rainfall impact coefficient. ≥0, when rainfall is 0 =0, the greater the rainfall The higher the value, the greater the degree of rainwater and sewage mixing in the regional pipe network; It is the normalized value of rainfall, derived from the formula. Calculation yields ( (Rainfall carrying capacity threshold designed for the pipeline network).

[0036] The risk of water quality exceeding standards is positively correlated with the actual concentrations of COD, ammonia nitrogen, and SS. The calculation model for the water quality exceeding standard risk index is as follows: ; in, It is a water quality exceedance risk index, which is dimensionless and ranges from [0,1]. The larger the value, the higher the risk of water quality exceeding the standard in the pipe network. , , The weighting coefficients for COD, ammonia nitrogen, and SS are respectively, satisfying the following conditions. The limits are determined by the water quality standards and the degree of hazard posed by the pollutants. These are the normalized values ​​of COD, ammonia nitrogen, and SS, respectively, all derived from the formula... Calculation yields ( (These are the standard limits for the corresponding water quality indicators).

[0037] The calculated , and The data is integrated to form a pipeline network status assessment result, using the following formula: The risk index is classified into levels (e.g., low risk: 0≤I<0.3, medium risk: 0.3≤I<0.7, high risk: 0.7≤I≤1) to complete the quantitative assessment of the current operating status of the pipeline network.

[0038] Then, a Gaussian process regression model with a spatiotemporal composite kernel function is constructed. By superimposing the time kernel function and the spatial kernel function, the spatiotemporal composite kernel function is constructed so that it can simultaneously capture the temporal variation law of the pipeline network status and the topological spatial correlation characteristics. The optimal hyperparameters of the model are then found through the Bayesian optimization algorithm.

[0039] Let the input of the Gaussian process regression model be... The output is in The feature vector of the pipeline network status; For the corresponding pipeline network status assessment results Then the standard Gaussian process regression model satisfies: ; in, Let m(x) be the mean function of a Gaussian process, and take m(x) = 0 (the mean of the pipeline network condition risk index approaches 0). It is the standard kernel function, which is the covariance function of a Gaussian process and reflects the correlation between input samples.

[0040] Time kernel function Radial basis functions (RBF) are used to construct a system that captures the continuous change in the network state over time. The formula is as follows: ; in, The kernel variance of the time kernel. It is the length-scale hyperparameter of the time kernel. For different time points; Spatial kernel function The quadratic exponent kernel (SE) is used to capture the spatial correlation characteristics between nodes in different topologies of the pipeline network. The formula is as follows: ; in, It is the kernel variance of the spatial kernel. It is the length-scale hyperparameter of the spatial kernel. These are different pipeline spatial nodes. It is the Euclidean norm.

[0041] Time kernel function and space kernel function Superposition and multiplication yield the spatiotemporal composite kernel function. As the covariance function of the improved Gaussian process regression model, the formula is: ; make (The kernel variance of the spatiotemporal composite kernel), then the formula simplifies to: ; Replace the kernel function of the standard Gaussian process regression model with the aforementioned spatiotemporal composite kernel function. An improved Gaussian process regression model with a spatiotemporal composite kernel is obtained, and the model expression is: ; Dataset partitioning and model hyperparameter optimization include fusing historical multi-source datasets. and the corresponding pipeline network status assessment results The training set is randomly divided according to a preset ratio of 7:3. and verification set ,in Used for model training. Used for model hyperparameter optimization and performance verification.

[0042] With the goal of minimizing the root mean square error (RMSE) of the validation set predictions, a Bayesian optimization algorithm is used to iteratively optimize the hyperparameters of the improved Gaussian process regression model. The optimized hyperparameter is the kernel variance of the spatiotemporal composite kernel function. Time length scale Spatial length scale The optimization range of hyperparameters is determined by the temporal granularity and spatial topological characteristics of the pipeline network data.

[0043] Iterative computation using a Bayesian optimization algorithm (100 iterations) is performed, calculating the RMSE value on the validation set in each iteration, ultimately yielding the optimal hyperparameter combination that minimizes the RMSE. The formula for calculating the RMSE of the validation set is: ; in, It is the number of samples in the validation set; It is the verification set number i The model prediction value for each sample; It is the verification set number i The actual true value of each sample.

[0044] training set Input the improved Gaussian process regression model and substitute the optimal hyperparameter combination. The model is trained using the maximum likelihood estimation method to obtain a trained pipeline network state prediction model. .

[0045] Verification set Input training completed The model's prediction accuracy, MAE (mean absolute error), RMSE, and other performance metrics are calculated. If the performance metrics meet the preset requirements (e.g., prediction accuracy ≥ 90%, RMSE ≤ 0.1), the model training is complete. If not, the hyperparameter optimization range is readjusted, and the hyperparameter optimization and model training process is repeated until the model performance meets the requirements.

[0046] Rainfall forecast data And the current pipeline network status assessment results Input the completed prediction model The predicted pipeline risk index for a future preset time period T (e.g., 1 hour, 2 hours, 6 hours, which can be adjusted according to pipeline operation and maintenance needs) is obtained and denoted as... , where t∈[0,T].

[0047] Organize pipeline network operation and maintenance cost parameters This cost is converted into operation and maintenance cost coefficients, including: pump station energy consumption cost coefficient λ1 (yuan / (kW·h)), blockage fault repair cost coefficient λ2 (yuan / unit blockage risk), overflow fault repair cost coefficient λ3 (yuan / unit overflow risk), and water quality exceeding standard penalty cost coefficient λ4 (yuan / unit water quality exceeding standard risk). These cost coefficients are determined by the actual operation and maintenance costs of the pipeline network operator and industry standards.

[0048] With the goal of minimizing the total operation and maintenance cost of the pipeline network and the constraint of pipeline network operation safety (risk index not exceeding 1), the dynamic control threshold optimization function is constructed as follows: ; ; in, This is the total operation and maintenance cost of the pipeline network within a future preset time period T, expressed in yuan; It is the estimated energy consumption of the pumping station within a preset time period T, in kW·h, and is negatively correlated with the risk index threshold (the higher the threshold, the lower the operating intensity of the pumping station and the lower the energy consumption). The dynamic control thresholds for the risk indices of blockage, overflow, and water quality exceeding standards are to be solved; the constraint condition indicates that the dynamic threshold values ​​are in the range of (0,1] to ensure the safe operation of the pipeline network.

[0049] The above optimization function is solved by gradient descent to obtain the dynamic control threshold of the pipeline risk index within a preset time period T. Based on the mathematical correlation between the risk index and liquid level and flow rate (from the risk index calculation model in step two), by... The dynamic control threshold for liquid level is obtained by reverse calculation. T H and flow dynamic control threshold T Q This forms a complete dynamic control threshold system for the pipeline network, providing constraints for the generation of subsequent control commands.

[0050] With the goal of minimizing pump station operating energy consumption and minimizing the risk of pipeline blockage / overflow / water quality exceeding standards, the Pareto optimal solution is obtained by improving the multi-objective particle swarm optimization algorithm, and the optimal control scheme is selected and control commands are generated based on the pipeline operation and maintenance priority. The optimization objectives are to minimize pump station operating energy consumption, minimize pipeline blockage risk, minimize overflow risk, and minimize water quality exceeding standards, combined with risk index prediction results. The objective function for multi-objective optimization of the pipeline network is constructed as follows:

[0051] in, The optimization variable is the control parameter to be solved. This represents the number of pump units in operation (a positive integer). Valve opening degree (%, 0≤) ≤100), Runtime (min); It is the objective function of energy consumption for pump station operation. , The rated power (kW) of a single pump unit is determined by... supply; It is the objective function for pipeline blockage risk. ,Depend on supply; It is the objective function for pipeline overflow risk. ,Depend on supply; It is the objective function for the risk of water quality exceeding standards in the pipe network. Depend on supply.

[0052] Based on the dynamic control threshold Tdyn and the operating parameters of the pump station valves Constraints are established for liquid level, flow rate, and equipment operating status to ensure that the optimized solution conforms to the actual operating capacity of the pipeline network and equipment. The constraint expressions are as follows:

[0053] in, This is the maximum number of pump units in operation, determined by... supply; The real-time liquid level in the pipeline network is determined by the control parameter X. It is the real-time flow rate of the pipeline network determined by the control parameter X; , and These are risk indices for real-time pipe network blockage, overflow, and water quality exceeding standards, determined by the control parameter X.

[0054] Based on the traditional multi-objective particle swarm optimization (MOPSO) algorithm, an adaptive clustering mechanism is introduced to adaptively cluster the non-dominated solutions of the particle swarm, eliminating redundant solutions and improving the distribution of Pareto optimal solutions and solution efficiency. The cluster centers of the adaptive clustering are determined by the particle density and distance, and the cluster radius is dynamically adjusted with the number of iterations.

[0055] The parameters for the improved MOPSO algorithm are set as follows: particle swarm size. N =50, maximum number of iterationsG max =200, learning factors c1=c2=2, and inertia weight ω linearly decreases from 0.9 to 0.4. The multi-objective optimization objective function and constraints are input into the improved MOPSO algorithm for iterative solution, ultimately yielding a set of Pareto optimal solutions satisfying all constraints. Each solution in the set is a set of optimal control parameters. Furthermore, the solutions are not mutually dominant.

[0056] The pipeline operation and maintenance priority shall be determined by the pipeline operation and maintenance unit based on the actual operation and maintenance needs. For example, during the flood season, the lowest overflow risk shall be given the highest priority, and during the non-flood season, the lowest pump station energy consumption shall be given the highest priority.

[0057] Based on the preset maintenance priorities, from the Pareto optimal solution set Selecting the unique optimal control solution The optimal control solution Convert into standardized equipment operation instructions to build a real-time control instruction set for the pipeline network. The instruction set contains seven core fields: control node, number of pump units in operation, valve opening degree, running time, execution time, and instruction validity period. The format conforms to the communication protocol of pump station / valve actuators.

[0058] Then, by comparing the actual feedback data of the execution device with the baseline data of the control command, the deviation is calculated and the control parameters are adjusted to solve the problems of no closed-loop feedback and control lag in the existing technology.

[0059] Real-time control instruction set The instructions are sent to the control system of the pipeline equipment, and the equipment executes the operation according to the instructions; at the same time, the actual operation feedback data of the equipment is collected through the equipment's sensors and control cabinet. ,in This represents the actual number of pump units in operation. This represents the actual valve opening. Actual operating time of the equipment This represents the actual energy consumption of the equipment.

[0060] Calculate feedback data Dfed and real-time control instruction set China benchmark data Estimated energy consumption The deviation values ​​are used to construct a deviation dataset. D dev The deviation calculation formula is:

[0061] in, EThese deviations are: pump unit number deviation, valve opening deviation, running time deviation, and energy consumption deviation. These deviation values ​​are then integrated with monitoring time and control nodes to form a structured deviation dataset. D d ev.

[0062] Based on the deviation dataset Ddev, a proportional-integral (PI) compensation algorithm is used to adjust the original control parameters. Perform adaptive adjustment to obtain the adjusted control parameters. The correction is as follows: ; in, It is a proportionality coefficient. It is the integral coefficient. It is the control parameter deviation vector, from .

[0063] If the deviation value If all deviations are less than the preset deviation threshold, then the adjusted control parameters will be... The parameters are sent to the equipment to complete the calibration. If the deviation exceeds the preset threshold, a deviation warning is triggered to remind maintenance personnel to check for equipment malfunctions (such as valve jamming, pump set failure, etc.) and manually intervene to adjust the control parameters.

[0064] Based on the GIS topology data and pipeline design parameters of the urban pipeline network A digital twin basic geometric model is constructed using 3D modeling software (such as AutoCAD and Revit) at a 1:1 scale to the physical pipeline network. The model includes all physical entities of the pipeline network, such as pipes, nodes, pump stations, valves, and monitoring equipment, accurately restoring the spatial topology and equipment appearance features of the pipeline network.

[0065] Based on fluid mechanics and pipeline hydraulic calculation theory, combined with pipeline design parameters (e.g., pipe diameter, pipe length, design flow rate, design liquid level) Construct a hydraulic simulation sub-model that matches the hydraulic characteristics of the pipe network. This model can simulate the flow state of sewage in the pipe network, calculate the variation law of hydraulic parameters such as liquid level, flow rate, and pressure, and associate and bind it with the basic geometric model.

[0066] By integrating the basic geometric model with the hydraulic simulation sub-model and adding information tags such as equipment attributes, pipeline parameters, and monitoring points, a digital twin basic model of the urban sewage pipeline network is formed, denoted as [model name missing]. .

[0067] With bias dataset As a basis for calibration, the flow coefficient, resistance coefficient and other parameters of the hydraulic simulation sub-model are adjusted so that the deviation between the hydraulic parameters calculated by the model and the actual monitoring parameters of the physical pipe network is less than the preset threshold. Using full-process data as training samples, the hydraulic simulation sub-model is iteratively optimized to improve the model's simulation accuracy of the actual operating state of the pipeline network; Adjusted control parameters Import the model and update the equipment's operating parameter settings to ensure that the model's control parameters are consistent with the physical equipment. After calibration and correction, a digital twin model that accurately matches the physical pipeline network is obtained. .

[0068] Establish physical pipeline network and digital twin model The real-time data communication link transmits the real-time operating status of the physical pipeline network, the execution results of control commands, and deviation data to the digital twin model in real time, enabling: Geometric state synchronization: The operating status of equipment in the physical pipeline network (such as pump start-up and shutdown, valve opening and closing) is mapped to the three-dimensional geometric entity of the model in real time, realizing the synchronous simulation of equipment actions; Hydraulic status synchronization: The real-time hydraulic parameters of the physical pipe network are updated to the hydraulic simulation sub-model of the model in real time, realizing the real-time simulation of the sewage flow state within the pipe network; Risk Status Synchronization: The real-time risk level of the physical pipeline network is labeled on the corresponding pipeline nodes in the model, enabling real-time visualization of the risk status. Ultimately, this achieves a consistent match between the physical pipeline network and the digital twin model, resulting in a digital twin model that is synchronized with the physical pipeline network in real time. .

[0069] Simultaneously, a dynamic synchronization update mechanism for the model is set up, and the data communication link collects real-time data of the physical pipeline network at a preset time granularity (e.g., 1 minute) to update the digital twin model. Dynamic updates are performed to ensure real-time synchronization between the model and the physical pipe network, enabling real-time visualization and dynamic synchronous updates of the urban sewage pipe network.

[0070] Figure 2 This is a schematic diagram of the urban sewage pipe network monitoring and control system based on multi-source data provided in an embodiment of the present invention; the embodiment of the present invention provides an urban sewage pipe network monitoring and control system based on multi-source data, including: The data acquisition module is used to preprocess the raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network to obtain a standardized multi-source dataset; The data fusion processing module is used to classify and weightedly fuse the standardized multi-source dataset to obtain a multi-source fusion dataset with a unified format. The status assessment module is used to perform a pipeline status assessment based on the pipeline design parameters and the multi-source fusion dataset to obtain a pipeline status assessment result that includes a blockage risk index, an overflow risk index, and a water quality exceedance risk index. The model building module is used to establish a pipeline network status prediction model based on historical fusion datasets, rainfall forecast data and pipeline network status assessment results. The pipeline network status prediction model is a Gaussian process regression model with a spatiotemporal composite kernel function. The threshold setting module is used to set the dynamic control threshold of the pipeline network based on the pipeline network design parameters, operation and maintenance cost parameters and the prediction results of the pipeline network status prediction model. The instruction generation module is used to generate a real-time control instruction set for the pipeline network based on the operating parameters of the pump station valves, the prediction results, and the pipeline network dynamic control threshold. The digital twin display module is used to obtain adjusted control parameters and deviation datasets based on feedback data from the execution equipment and the real-time control instruction set of the pipeline network; and to construct a twin model based on the control parameters and deviation datasets to achieve visualization and synchronous updating of the urban pipeline network.

[0071] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0072] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring and controlling urban sewage pipe networks based on multi-source data, characterized in that, include: The raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network are preprocessed to obtain a standardized multi-source dataset; The standardized multi-source dataset is classified and weighted to obtain a multi-source fused dataset with a unified format; Based on the pipeline design parameters and the multi-source fusion dataset, a pipeline status assessment is performed to obtain pipeline status assessment results including blockage risk index, overflow risk index and water quality exceedance risk index; A pipeline network status prediction model is established based on historical fusion datasets, rainfall forecast data, and pipeline network status assessment results. The pipeline network status prediction model is a Gaussian process regression model with a spatiotemporal composite kernel function. The dynamic control threshold of the pipeline network is set based on the pipeline network design parameters, operation and maintenance cost parameters, and the prediction results of the pipeline network status prediction model. A real-time control instruction set for the pipeline network is generated based on the operating parameters of the pump station valves, the prediction results, and the dynamic control threshold of the pipeline network. The adjusted control parameters and deviation dataset are obtained based on the feedback data from the execution equipment and the real-time control instruction set of the pipeline network. Based on the control parameters and deviation dataset, a twin model is constructed to enable visualization and synchronous updating of the urban pipeline network.

2. The urban sewage pipe network monitoring and control method based on multi-source data according to claim 1, characterized in that, The standardized multi-source dataset is obtained by preprocessing the raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network, including: The flow data of the urban pipeline network is collected by an insertion ultrasonic flow meter in the sensing layer; COD, ammonia nitrogen and SS data of the pipeline network are collected by a multi-parameter water quality monitor; pressure and level data of the pipeline network are collected by a pressure and level sensor; and gas composition data of the pipeline network are collected by a gas sensor. The original data is obtained by packaging the flow rate data, COD data, ammonia nitrogen data, SS data, pressure data, liquid level data, and gas composition data. The external auxiliary data for the urban pipeline network includes rainfall data from meteorological stations, GIS pipeline network topology data, historical fault data, and operating parameters of pumping station equipment.

3. The urban sewage pipe network monitoring and control method based on multi-source data according to claim 1, characterized in that, The standardized multi-source dataset is classified and weighted to obtain a unified multi-source fusion dataset, including: The data from the standardized multi-source dataset is classified into time-series data, spatial data, state-related data, and environmental data. Based on the classification methods of time-series data, spatial data, state-type data, and environment-type data, the basic weights of the weighted fusion algorithm are reset to obtain an adaptive weighted fusion algorithm; The standardized multi-source dataset is obtained by weighted fusion of the data using the adaptive weighted fusion algorithm.

4. The urban sewage pipe network monitoring and control method based on multi-source data according to claim 1, characterized in that, A pipeline network condition prediction model is established based on historical fusion datasets, rainfall forecast data, and the pipeline network condition assessment results. This model is a Gaussian process regression model with a spatiotemporal composite kernel function, comprising: An improved Gaussian process regression model with a spatiotemporal composite kernel function is obtained by superimposing a time kernel function and a spatial kernel function on the standard Gaussian process regression model. The historical fusion dataset and the corresponding pipeline status evaluation results are divided into training set and validation set according to a preset ratio; The prediction error RMSE on the validation set is optimized using a Bayesian optimization algorithm. The optimal hyperparameter combination that fits the characteristics of the pipeline network data is obtained by iteratively optimizing the kernel variance and length scale hyperparameter of the spatiotemporal composite kernel function using the Bayesian optimization algorithm. The improved Gaussian process regression model is trained by optimizing and minimizing the RMSE (Reference Error of Validation Set) and the optimal combination of hyperparameters.

5. The urban sewage pipe network monitoring and control method based on multi-source data according to claim 1, characterized in that, Setting dynamic control thresholds for the pipeline network based on pipeline design parameters, operation and maintenance cost parameters, and the prediction results of the pipeline network status prediction model includes: By inputting rainfall forecast data and current pipeline status assessment results into the trained improved Gaussian process regression model, prediction results for the blockage risk index, overflow risk index, and water quality exceedance risk index for the future preset period are obtained.

6. The urban sewage pipe network monitoring and control method based on multi-source data according to claim 1, characterized in that, The real-time control instruction set for the pipeline network is generated based on the pump station valve operating parameters, the prediction results, and the pipeline network dynamic control threshold, including: The optimization objectives are to minimize the energy consumption of pump station operation, the risk of pipeline blockage, the risk of overflow, and the risk of water quality exceeding standards. A multi-objective optimization objective function for the pipeline network is constructed using the blockage risk index, overflow risk index, and water quality exceedance risk index from the prediction results. Constraints on liquid level, flow rate, and equipment operating status are established based on the pipeline network dynamic control threshold and the pump group power characteristics, valve opening and flow rate correspondence, and maximum number of pump groups in the pump station valve operating parameters. An improved multi-objective particle swarm optimization algorithm incorporating an adaptive clustering mechanism is used to iteratively solve the multi-objective optimization objective function within the constraints to obtain a set of Pareto optimal solutions; Based on the pipeline network operation and maintenance priority, the optimal control scheme is selected from the Pareto optimal solution set to generate a real-time pipeline network control instruction set containing the number of pump units in operation, valve opening degree, and running time.

7. The urban sewage pipe network monitoring and control method based on multi-source data according to claim 1, characterized in that, Constructing a twin model based on the control parameters and deviation dataset to achieve visualization and synchronous updating of the urban pipeline network includes: Based on the GIS topology data and design parameters of the urban pipeline network, a digital twin basic geometric model and a hydraulic simulation sub-model matching the hydraulic characteristics of the pipeline network are constructed. A digital twin model of the urban pipe network is constructed based on the aforementioned digital twin basic geometric model and a hydraulic simulation sub-model that matches the hydraulic characteristics of the pipe network. The adjusted control parameters and deviation datasets are combined with the full-process data of pipeline monitoring, condition assessment, prediction, and control commands and imported into the digital twin model for hydraulic parameter calibration and correction. The real-time operating status, control command execution results, and deviation data of the urban pipeline network are synchronously mapped to the calibrated and corrected digital twin model of hydraulic parameters to achieve a consistent matching between the urban pipeline network and the digital twin model. Based on the synchronously completed digital twin model, the network topology, equipment operating status, risk level, control parameters and deviation information are visualized, achieving real-time visualization and dynamic synchronous updates of the urban network.

8. A monitoring and control system for urban sewage pipe networks based on multi-source data, characterized in that, include: The data acquisition module is used to preprocess the raw data collected by the urban pipeline network sensing layer equipment, the external auxiliary data of the urban pipeline network, and the standard specification data of the urban pipeline network to obtain a standardized multi-source dataset; The data fusion processing module is used to classify and weightedly fuse the standardized multi-source dataset to obtain a multi-source fusion dataset with a unified format. The status assessment module is used to perform a pipeline status assessment based on the pipeline design parameters and the multi-source fusion dataset to obtain a pipeline status assessment result that includes a blockage risk index, an overflow risk index, and a water quality exceedance risk index. The model building module is used to establish a pipeline network status prediction model based on historical fusion datasets, rainfall forecast data and pipeline network status assessment results. The pipeline network status prediction model is a Gaussian process regression model with a spatiotemporal composite kernel function. The threshold setting module is used to set the dynamic control threshold of the pipeline network based on the pipeline network design parameters, operation and maintenance cost parameters and the prediction results of the pipeline network status prediction model. The instruction generation module is used to generate a real-time control instruction set for the pipeline network based on the operating parameters of the pump station valves, the prediction results, and the pipeline network dynamic control threshold. The digital twin display module is used to obtain adjusted control parameters and deviation datasets based on feedback data from the execution equipment and the real-time control instruction set of the pipeline network; and to construct a twin model based on the control parameters and deviation datasets to achieve visualization and synchronous updating of the urban pipeline network.

Images