A water supply system scheduling optimization method and device, electronic equipment and storage medium

Abstract

The application discloses a water supply system scheduling optimization method and device, electronic equipment and storage medium, and relates to the field of intelligent scheduling of water supply systems. The method comprises the following steps: collecting and preprocessing multidimensional data to obtain a data set; analyzing pressure data from the time and space dimensions to accurately identify unfavorable points and their pressure requirements; constructing a water quantity prediction model using a time series model, and constructing a total water head difference prediction model using a LightGBM gradient boosting tree combined with a MultiOutputRegressor multi-output regression framework; constructing a minimum total water production cost objective function based on the prediction results, iteratively optimizing in combination with water quantity and pressure constraints, and outputting an optimal scheduling scheme; and establishing a model updating mechanism to ensure dynamic adaptation of the strategy. The application solves the problems of insufficient pressure guarantee, extensive cost control, and weak model practicality and generalization ability, overcomes the inherent contradiction between the high complexity of traditional mechanism models and the poor generalization and lack of physical meaning of pure data-driven pressure prediction models, and provides safe, stable and efficient intelligent technical support for water supply systems.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology for water supply systems, and in particular to a method, apparatus, electronic device, and storage medium for optimizing water supply system scheduling. Background Technology

[0002] With the acceleration of urbanization and the increasing demand for water supply, urban water supply systems are facing unprecedented challenges. In the existing water supply dispatching system, insufficient pressure guarantee and high energy consumption are the key factors restricting the stable and efficient operation of water supply. Specifically, traditional water supply dispatching methods rely heavily on experience or simple pressure threshold control, resulting in significant head loss along the pipeline during peak hours and difficulty in guaranteeing pressure, or excessive pressure due to blind pressure maintenance, causing ineffective energy consumption; during off-peak hours, pipeline overpressure increases both ineffective energy consumption and the risk of pipe bursts.

[0003] The crude control of water production costs in the system is also a major problem in current water supply scheduling. Existing multi-water plant water supply scheduling often sets the upper and lower limits of the outlet pressure over time based on experience, lacking dynamic consideration of the differences in water production costs among multiple water plants (such as electricity costs and chemical consumption costs). This makes it impossible to effectively optimize water allocation and operation strategies, and makes it difficult to achieve optimal system-level energy efficiency.

[0004] Secondly, existing prediction models lack practicality and generalization ability, making it difficult to support real-time optimization scheduling. On the one hand, traditional mechanistic models are highly dependent on pipeline parameters, have high modeling costs, and are difficult to adapt to changes in the pipeline network; on the other hand, the performance of pure data-driven pressure prediction models is heavily dependent on the range of training data, has poor generalization ability in optimization exploration, and is prone to generating unrealistic scheduling schemes.

[0005] More importantly, the existing water supply scheduling system lacks an intelligent collaborative closed loop. Water plant, pipeline network, and environmental data are isolated from each other and have not formed a closed loop. Scheduling decisions rely heavily on human experience, and the response lags behind changes in actual demand.

[0006] Therefore, there is an urgent need for a water supply system scheduling optimization method that can accurately identify and meet pressure demands, optimize system-level scheduling costs, improve the practicality and generalization ability of prediction models, and form an intelligent collaborative closed loop. Summary of the Invention

[0007] This invention provides a water supply system scheduling optimization method to address the problems of insufficient pressure guarantee, crude water production cost control, insufficient practicality and generalization ability of prediction models, isolated water plant network environmental data, and lack of intelligent collaborative closed-loop systems in existing technologies. The technical solution is as follows:

[0008] According to one aspect of the present invention, a water supply system scheduling optimization method includes: acquiring multidimensional data of the water supply system and preprocessing it to obtain a dataset; analyzing the pressure data in the dataset from a spatiotemporal dimension to obtain a set of unfavorable points and corresponding pressure demands to construct a demand set; the multidimensional data includes data from the water plant side, the pipeline network side, and the environment side; constructing a water volume prediction model based on water volume data, meteorological information, and time characteristics in the dataset using a time series model; and optimizing the model by generating samples and evaluating the prediction accuracy using a sliding window method and the mean absolute percentage error; the time characteristics include holiday attributes and seasonal attributes; and employing LightGBM gradient boosting tree. Combining the MultiOutputRegressor framework, a total head difference prediction model is constructed based on hydraulic mechanisms, the outflow of each water plant in the dataset, and the total flow of the area. The model accuracy is evaluated and optimized by using the mean absolute error over multiple time periods. An objective function is constructed based on the outflow data and cost of each water plant. Water volume constraints are obtained based on the output of the water volume prediction model, the water supply capacity of each water plant, equipment and process requirements, and pressure constraints are obtained based on the demand set. An optimization algorithm is used to iteratively optimize the water volume and pressure constraints and minimize the water production cost until the set conditions are met, and the optimal scheduling scheme of the water supply system is output.

[0009] In one embodiment, acquiring and preprocessing multidimensional data from the water supply system to obtain a dataset is achieved through the following steps: Data from the water plant side is collected via a SCADA system; data from the pipeline network side is collected via a remote pressure monitoring terminal; and environmental data is obtained via a meteorological department interface, resulting in a multidimensional raw dataset. The water plant side data includes the outflow rate, pressure, and meter elevation of each water plant; the pipeline network side data includes pressure data and meter elevation at pipeline monitoring points; and the environmental data includes temperature, humidity, rainfall data, and time characteristics. A quantization threshold is set based on the equipment range and actual operating patterns of the water supply system. Data exceeding the quantization threshold and missing data segments in the raw dataset are marked as outliers and missing values, respectively. Monitoring points with outliers and missing values ​​exceeding a set proportion are removed, and their related data is deleted to obtain the dataset.

[0010] In one embodiment, analyzing the pressure data in the dataset from a spatiotemporal perspective to obtain a set of unfavorable points and corresponding pressure demands is achieved through the following steps: Based on the water supply system's network topology, identify the network endpoints with distances greater than a set value from the water plant and the high points with elevation differences greater than a set value as unfavorable point sets to construct low-pressure areas; extract the spatial features of the low-pressure areas from the dataset; the spatial features include transport distance and elevation difference; analyze the data of the low-pressure areas from a time perspective based on the dataset to obtain low-pressure periods; and based on the spatial features of the low-pressure areas and low-pressure periods, combined with daily pressure complaint data, clarify the pressure demands of each unfavorable point to construct a demand set; the pressure demands include a single value and a curve that changes over time.

[0011] In one embodiment, a total head difference prediction model is constructed based on hydraulic mechanisms, the outflow rates of each water plant in the dataset, and the total flow rate of the area, using a LightGBM gradient boosting tree combined with a MultiOutputRegressor framework. This is achieved through the following steps: A two-dimensional matrix is ​​constructed using the outflow rates of each water plant and the total flow rate of the area as input features. After ensuring spatiotemporal alignment with the two-dimensional matrix, the total head difference of the water supply system is calculated to construct a total head difference matrix. The two-dimensional matrix and the total head difference matrix are aligned, filtered, and normalized to obtain a training set. The LightGBM gradient boosting tree combined with the MultiOutputRegressor framework is used, with mean squared error as the loss function. L1 regularization is introduced to suppress overfitting, and the hyperparameters are tuned using Bayesian optimization. The total head difference prediction model is then constructed and trained based on the training set.

[0012] In one embodiment, the model accuracy is evaluated and optimized by using the average absolute error over multiple time periods through the following steps: the prediction accuracy of the total head difference prediction model is evaluated by using the average absolute error over multiple time periods; the characteristic contribution of each water plant's flow rate to each total head difference is analyzed by using the SHAP value to ensure the consistency between the model and the hydraulic mechanism; if the prediction accuracy does not meet the standard, training data is supplemented or the model structure is adjusted to optimize the model.

[0013] In one embodiment, an objective function is constructed based on the effluent data and cost of each water plant. Water constraints are obtained based on the output of the water volume prediction model, the water supply capacity of each water plant, equipment and process requirements, and pressure constraints are obtained based on the demand set. An optimization algorithm is used to iteratively optimize the water supply system by satisfying the water volume and pressure constraints and minimizing the water production cost until the set conditions are met, outputting the optimal scheduling scheme for the water supply system. This is achieved through the following steps: obtaining the effluent flow rate of each water plant and deriving a flow matrix based on water balance; constructing an objective function based on the flow matrix, the effluent pressure of each water plant, and various cost coefficients; obtaining water constraints based on the output of the water volume prediction model, the equipment and process requirements of each water plant, and water supply capacity; and then iteratively optimizing the scheme based on the demand set to satisfy the water volume and pressure constraints and minimize the water production cost until the set conditions are met. The demand set is used to obtain pressure constraints; the flow matrix is ​​input into the total head difference prediction model to obtain the total head difference prediction matrix; based on the prediction matrix, the outflow pressure is calculated based on the principle of prioritizing the demand set; and the objective function is substituted to calculate the current solution under the premise of satisfying the water volume constraint and pressure constraint; if the sequential least squares programming algorithm is selected as the optimization algorithm, the current solution is iteratively solved through sequential quadratic programming, gradient calculation and solution verification until the set conditions are met, and the optimal scheduling scheme of the water supply system is output; if the genetic algorithm or particle swarm optimization algorithm is used as the optimization algorithm, the current solution is iteratively solved through heuristic search until the set conditions are met, and the optimal scheduling scheme of the water supply system is output.

[0014] In one embodiment, the method further includes the following steps: if the pipeline network data changes, the dataset is re-collected, the demand set is reconstructed, and the total head difference prediction model is retrained and its accuracy is evaluated until the core indicators meet the standards; if the water plant data changes, corresponding new data is added and the relevant parameters of the objective function are updated, and the validity of the parameters is verified through historical operating conditions; when environmental data is abnormal, the model is updated, the water volume prediction model is updated first, and the water volume prediction accuracy is evaluated until the core indicators meet the standards; the environmental data anomalies include extreme weather and sudden events; when the prediction accuracy of the water volume prediction model or the total head difference prediction model decreases, drift is diagnosed through data distribution testing, monitoring data is added, and the model is updated using incremental training; if the model still does not meet the standards after training, the sensor status is checked or the model structure is adjusted; when the set update time is reached, all models are retrained and their accuracy is evaluated until the core indicators meet the standards; in scenarios where all models are changed, the new model that meets the standards is deployed, and the reason for triggering the model update, the changed parameters, and the verification results are recorded; the core indicators include water volume prediction accuracy, total head difference model accuracy, and water production cost.

[0015] According to one aspect of the present invention, a water supply system scheduling optimization device includes: a data acquisition and processing module, configured to acquire multidimensional data of the water supply system and preprocess it to obtain a dataset, analyze the pressure data in the dataset from a spatiotemporal dimension, obtain a set of unfavorable points and corresponding pressure demands to construct a demand set; the multidimensional data includes data from the water plant side, the pipeline network side, and the environment side; a water volume prediction module, configured to construct a water volume prediction model based on the water volume data, meteorological information, and time characteristics in the dataset using a time series model, and optimize the model by generating samples and average absolute percentage error using a sliding window method; the time characteristics include holiday attributes and seasonal attributes; and a total head difference prediction module, configured to predict the total head difference using a Lig... The htGBM gradient boosting tree, combined with the MultiOutputRegressor framework, constructs a total head difference prediction model based on hydraulic mechanisms, the outflow from each water plant in the dataset, and the total flow in the area. The model accuracy is evaluated and optimized by using the mean absolute error over multiple time periods. The scheduling scheme generation module constructs an objective function based on the outflow data and cost of each water plant. Based on the output of the water volume prediction model, the water supply capacity of each water plant, equipment and process requirements, water volume constraints are obtained. Pressure constraints are obtained based on the demand set. An optimization algorithm is used to iteratively optimize the water volume and pressure constraints and minimize the water production cost until the set conditions are met, and outputs the optimal scheduling scheme for the water supply system.

[0016] According to one aspect of the present invention, an electronic device includes at least one processor and at least one memory, wherein computer-readable instructions are stored in the memory; the computer-readable instructions are executed by one or more of the processors to cause the electronic device to implement the water supply system scheduling optimization method as described above.

[0017] According to one aspect of the present invention, a storage medium stores computer-readable instructions thereon, which are executed by one or more processors to implement the water supply system scheduling optimization method as described above.

[0018] The beneficial effects of the technical solution provided by this invention are:

[0019] In the above technical solution, this invention first collects and preprocesses multi-dimensional data from the water plant, pipeline network, and environment to construct a high-quality dataset. Then, it analyzes pressure data from a spatiotemporal perspective to accurately identify unfavorable points and their pressure demands, forming a demand set. Subsequently, it uses a time series model combined with the LightGBM gradient boosting tree and the MultiOutputRegressor framework to construct a water volume prediction model and a total head difference prediction model, ensuring prediction accuracy and practicality. Based on the prediction results and cost data, it constructs a system with the objective function of minimizing water production costs, combining water volume and pressure constraints, and iteratively optimizes using algorithms such as sequential least squares programming to output the optimal scheduling scheme. Furthermore, the method includes a dynamic model update mechanism. For changes in pipeline network, water plant, and environmental data, or a decrease in prediction accuracy, it differentiates and promotes model retraining or parameter adjustment to ensure that the scheduling strategy adapts to real-time changes in system dynamics. This method effectively improves the accuracy and energy efficiency of water supply system scheduling, solves problems such as insufficient pressure guarantee, crude cost control and weak model generalization ability in existing technologies, overcomes the inherent technical contradiction between "high complexity of traditional mechanism models" and "poor generalization and lack of physical meaning of pure data-driven pressure prediction models", and provides reliable technical support for the safe, stable and efficient intelligent operation of water supply systems. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a water supply system scheduling optimization method according to an exemplary embodiment;

[0022] Figure 2 This is a flowchart illustrating a water supply system scheduling optimization method in an exemplary embodiment;

[0023] Figure 3 This is a block diagram of a water supply system scheduling optimization device according to an exemplary embodiment;

[0024] Figure 4 This is a hardware structure diagram of an electronic device according to an exemplary embodiment;

[0025] Figure 5 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0026] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0027] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this disclosure means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0028] This invention provides a water supply system scheduling optimization method. By constructing a closed-loop scheduling system encompassing "data fusion, demand analysis, intelligent prediction, and dynamic optimization," it achieves stable, efficient, and intelligent coordination of the water supply system. This solves the problems of insufficient pressure guarantee, crude water production cost control, and poor practicality of prediction models in existing technologies. This water supply system scheduling optimization method is applicable to water supply system scheduling optimization devices, which can be electronic devices. The water supply system scheduling optimization method in this invention can be applied to various scenarios, such as the scheduling optimization of urban water supply systems.

[0029] Please see Figure 1 This invention provides a method for optimizing the scheduling of a water supply system, which is applicable to electronic devices.

[0030] In the following method embodiments, for ease of description, the execution subject of each step of the method is an electronic device, but this does not constitute a specific limitation.

[0031] like Figure 1 As shown, the method may include the following steps:

[0032] Step 110: Obtain multidimensional data of the water supply system and preprocess it to obtain a dataset. Analyze the pressure data in the dataset from the spatiotemporal dimensions to obtain a set of unfavorable points and corresponding pressure requirements to construct a demand set.

[0033] In one possible implementation, data from the water plant is collected through a SCADA system, data from the pipeline network is collected through a remote pressure monitoring terminal, and environmental data is obtained through an interface with the meteorological department to obtain a multidimensional raw dataset. A quantization threshold is set based on the equipment range and actual operating patterns of the water supply system. Data exceeding the quantization threshold and missing data segments in the raw dataset are marked as outliers and missing values, respectively. Monitoring points with outliers and missing values ​​exceeding a set proportion are removed and their related data is deleted to obtain the dataset.

[0034] The multidimensional data includes data from the water plant, the pipeline network, and the environment. The water plant data includes the outflow rate, pressure, and meter elevation of each water plant, etc.; the pipeline network data includes the pressure data and meter elevation of the pipeline monitoring points, etc.; and the environmental data includes temperature, humidity, rainfall data, and time characteristics, etc., without any restrictions.

[0035] In one possible implementation, based on the topology of the water supply system, the network end points that are more than a set value away from the water plant and the high points with a terrain elevation difference greater than a set value are identified as unfavorable points to construct a low-pressure area, and the spatial features of the low-pressure area are extracted from the dataset.

[0036] It should be noted that the set of unfavorable points is not a traditional, single, fixed most unfavorable point, but rather a pre-selected set of key monitoring points, which are then dynamically determined based on real-time hydraulic data.

[0037] Furthermore, by analyzing the data of low-pressure areas from a time perspective based on the dataset, low-pressure periods are obtained. Based on the spatial characteristics of low-pressure areas and low-pressure periods, combined with daily stress complaint data, stress demands at each disadvantage point are identified to construct a demand set.

[0038] Spatial characteristics include conveying distance and elevation difference, while pressure requirements include single values ​​and curves that change over time, etc., none of which are specified here.

[0039] Specifically, the SCADA system collects outflow, pressure, and meter elevation data for each water plant. This data reflects the real-time operating status of the water plants and forms the basis for scheduling optimization. Remote pressure monitoring terminals collect pressure data and meter elevation from monitoring points in the pipeline network. This data is used to analyze the pressure distribution of the pipeline network and identify low-pressure areas. Daily temperature, humidity, and rainfall data are obtained through an interface with the meteorological department, combined with time characteristics (such as holiday attributes and seasonal attributes). This data reflects the impact of the external environment on water supply demand.

[0040] Furthermore, by combining the equipment range and actual operating patterns of the water supply system, quantitative thresholds are set for each characteristic variable. Data exceeding the threshold range is marked as outliers. Missing data segments appearing during the data collection process are marked, and monitoring points with outliers and missing values ​​exceeding the set proportion are removed. Related data samples are also deleted to ensure the integrity and reliability of the dataset.

[0041] Specifically, based on the pipeline topology of the water supply system, the network endpoints with distances greater than a set value from the water plant and the high points with elevation differences greater than a set value are identified as unfavorable points. Spatial features of these unfavorable points, including transport distance and elevation difference, are extracted from the dataset to construct a low-pressure area model.

[0042] Furthermore, the data on low-pressure areas is analyzed from a temporal perspective based on the dataset to identify low-pressure periods, particularly peak and off-peak water usage times. By combining daily pressure complaint data with spatiotemporal analysis results, the pressure demand at each disadvantaged point is clarified. Pressure demand can be represented by a single value or a curve varying over time.

[0043] In the above process, this embodiment of the invention ensures the integrity and reliability of the dataset through multi-source data acquisition and a rigorous data preprocessing procedure, providing a high-quality data foundation for subsequent analysis. The purpose of this step is to eliminate noise interference and improve the accuracy of the model input data, thereby providing reliable data support for subsequent pressure demand analysis, predictive model construction, and optimized scheduling. By combining spatial characteristics and temporal patterns in pressure demand analysis methods, unfavorable points in the water supply system and their pressure demands are accurately identified, constructing a detailed demand set. The purpose of this step is to provide accurate pressure demand basis for subsequent predictive model construction and optimized scheduling, ensuring that the scheduling scheme can meet the water supply needs of each region and improve water supply reliability.

[0044] Step 120: A water volume prediction model is constructed based on water volume data, meteorological information and time characteristics in the dataset using a time series model. The model is then optimized by generating samples and evaluating the mean absolute percentage error using the sliding window method.

[0045] Specifically, an input feature set is constructed based on historical hourly water volume data, combined with meteorological information (daily temperature, humidity, and rainfall) and temporal characteristics (holiday attributes and seasonal attributes). A time series model (such as LSTM or Prophet) is used for training, and samples are generated using a sliding window method. The prediction accuracy is evaluated using the mean absolute percentage error (MAPE). If the prediction accuracy does not meet the standard, historical data is added for retraining.

[0046] In the above process, the embodiments of the present invention construct a water volume prediction model by selecting appropriate input features and model algorithms. The purpose of this step is to provide accurate prediction data and provide a reliable basis for subsequent optimization scheduling. The water volume prediction model can capture the periodicity and trend of water use patterns.

[0047] Step 130: Using LightGBM gradient boosting tree combined with MultiOutputRegressor framework, a total head difference prediction model is constructed based on hydraulic mechanism, outflow from each water plant in the dataset and total flow in the area. The model accuracy is evaluated and optimized by the mean absolute error over multiple time periods.

[0048] In one possible implementation, the outflow from each water plant and the total flow in the area are used as input features to construct a two-dimensional matrix. After ensuring spatiotemporal alignment with the two-dimensional matrix, the total head difference of the water supply system is calculated to construct a total head difference matrix. The training set is obtained by aligning, filtering and normalizing the two-dimensional matrix and the total head difference matrix.

[0049] Furthermore, a LightGBM gradient boosting tree combined with a MultiOutputRegressor framework is adopted, with mean squared error as the loss function. L1 regularization is introduced to suppress overfitting, and hyperparameters are tuned through Bayesian optimization. Based on the training set, a total head difference prediction model is constructed and trained.

[0050] Furthermore, the prediction accuracy of the total head difference prediction model is evaluated by using the mean absolute error over multiple time periods. The characteristic contribution of each water plant's flow rate to each total head difference is analyzed by using the SHAP value to ensure the consistency between the model and the hydraulic mechanism. If the prediction accuracy does not meet the standard, training data is supplemented or the model structure is adjusted to optimize the model.

[0051] Specifically, the outflow rates of each water plant and the total flow rate of the area in the dataset are used as input features to construct a two-dimensional matrix, and the total head difference of the water supply system is calculated to construct the total head difference matrix. A LightGBM gradient boosting tree combined with a MultiOutputRegressor framework is adopted, with mean squared error as the loss function, and L1 regularization is introduced to suppress overfitting. The hyperparameters are tuned through Bayesian optimization to construct and train the total head difference prediction model.

[0052] In the above process, this embodiment of the invention constructs a total head loss prediction model by selecting appropriate input features and model algorithms. The purpose of this step is to provide accurate prediction data, providing a reliable basis for subsequent optimized scheduling. The total head loss prediction model ensures the accuracy and reliability of the prediction results through the hydraulic mechanism that "flow rate determines head loss".

[0053] Step 140: Construct an objective function based on the water output data and cost of each water plant. Obtain water volume constraints based on the output of the water volume prediction model, the water supply capacity of each water plant, equipment and process requirements, and obtain pressure constraints based on the demand set. Use optimization algorithms to iteratively optimize the water volume and pressure constraints and minimize water production costs until the set conditions are met, and output the optimal scheduling scheme for the water supply system.

[0054] In one possible implementation, the outflow rate of each water plant is obtained and the flow rate matrix is ​​derived by combining the water balance. Based on the flow rate matrix, the outflow pressure of each water plant and the cost coefficients, an objective function is constructed. Based on the output of the water prediction model, the equipment, process requirements and water supply capacity of each water plant, the water quantity constraint is obtained. Based on the demand set, the pressure constraint is obtained.

[0055] Furthermore, the flow matrix is ​​input into the total head difference prediction model to obtain the prediction matrix of the total head difference. Based on the prediction matrix, the pressure demand at the most unfavorable point in the current period is used as the core constraint to back-calculate the outlet pressure. The objective function is then substituted to calculate the current solution under the premise of satisfying the water volume constraint and pressure constraint.

[0056] Furthermore, if the sequential least squares programming algorithm is selected as the optimization algorithm, then the optimal scheduling scheme of the water supply system is output by iteratively solving the current solution through sequential quadratic programming, gradient calculation and solution verification until the set conditions are met. If the genetic algorithm or particle swarm optimization algorithm is used as the optimization algorithm, then the optimal scheduling scheme of the water supply system is output by iteratively solving the current solution through heuristic search until the set conditions are met.

[0057] The optimization algorithms include, but are not limited to, SLSQP (Sequential Least Squares Programming), genetic algorithms, and particle swarm optimization. The selection is based on the characteristics of water supply system optimization (continuous decision variables, multiple constraints, and real-time requirements). This embodiment of the invention uses the SLSQP algorithm to find the optimal objective function. The SLSQP algorithm has advantages such as fast convergence speed, strong constraint handling capability, and high solution accuracy, making it particularly suitable for this type of engineering optimization problem.

[0058] Specifically, an objective function is constructed based on the effluent and cost data of each water plant, with minimizing the total water production cost as the core objective. Water quantity constraints are obtained based on the output of the water quantity prediction model, the equipment and process requirements of each water plant, and the water supply capacity. Pressure constraints are obtained based on the demand set. The Sequential Least Squares Programming (SLSQP) algorithm is used for iterative optimization until the set conditions are met, outputting the optimal scheduling scheme for the water supply system.

[0059] In one possible implementation, if the pipeline network data changes, the dataset is re-collected, the demand set is reconstructed, and the total head difference prediction model is retrained and its accuracy is evaluated until the core indicators meet the standards. If the water plant data changes, the corresponding new data is supplemented and the relevant parameters of the objective function are updated and their validity is verified through historical operating conditions. When the environmental data is abnormal, the model is updated, the water volume prediction model is updated first, and the accuracy of the water volume prediction is evaluated until the core indicators meet the standards.

[0060] Anomalies in environmental data include extreme weather, emergencies, etc., which are not specified here.

[0061] Furthermore, when the prediction accuracy of the water volume prediction model or the total head difference prediction model decreases, the drift is diagnosed by data distribution test, monitoring data is supplemented and the model is updated by incremental training. If the model still fails to meet the standard after training, the sensor status is checked or the model structure is adjusted.

[0062] Furthermore, when the set update time is reached, all models are retrained and their accuracy is evaluated until the core metrics are met.

[0063] Furthermore, in all scenarios where the model is modified, the new model that meets the criteria is deployed, and the reasons for triggering the model update, the changed parameters, and the verification results are recorded; the core indicators include water volume prediction accuracy, total head difference model accuracy, and water production cost.

[0064] Among them, environmental data anomalies include situations that deviate significantly from historical patterns, such as extreme weather and emergencies. The core indicators include the accuracy of water volume prediction, the accuracy of the total head difference model, and the cost of water production.

[0065] In the above process, this embodiment of the invention constructs an objective function with minimizing the total water production cost as its core, and iteratively optimizes it by combining water quantity constraints and pressure constraints, outputting the optimal scheduling scheme for the water supply system. Simultaneously, a model update mechanism is established to adapt to dynamic changes in the system, ensuring the long-term effectiveness and accuracy of the scheduling strategy. The purpose of this step is to achieve stable and efficient operation of the water supply system, reduce water production costs, and improve user satisfaction.

[0066] Through the above process, this invention achieves stable and efficient operation of the water supply system by data acquisition and preprocessing, pressure demand analysis, predictive model construction, and optimized scheduling and modeling. It constructs an accurate demand set by combining spatial characteristics and temporal patterns in pressure demand analysis; it builds water volume prediction models and total head difference prediction models by selecting appropriate input features and model algorithms; it iteratively optimizes the system with the core objective of minimizing total water production costs and constraints, outputting the optimal scheduling scheme; and it establishes a model update mechanism to adapt to dynamic changes in the system. This method improves water supply stability, reduces system water production costs, enhances the practicality and generalization ability of the predictive model, and achieves intelligent collaboration across the entire chain, resulting in significant economic and social benefits.

[0067] In an exemplary embodiment, the process of optimizing water supply system scheduling by the water supply system scheduling method of the present invention is illustrated.

[0068] like Figure 2 As shown, the following steps may be included:

[0069] Step S1: Screening of mechanism constraint features.

[0070] Step S1.1, Analysis of Influence Factors.

[0071] Step S1, Data Acquisition and Fusion Module.

[0072] Step S1.1, Data Acquisition.

[0073] Water Plant Side Data Acquisition: Key data such as outflow rate, pressure, and meter elevation are collected in real-time and accurately from each water plant using a SCADA system. These data reflect the production and operation status of the water plant. For example, the outflow rate reflects the water plant's supply capacity, while pressure data is closely related to the stability and coverage of the water supply. It should be noted that the "meter elevation" mentioned in this invention refers to the absolute elevation of the pressure monitoring point, which is the basis for hydraulic calculations. It is used to convert measured pressure values ​​into head values, thereby unifying the data from measuring points in different geographical locations to the same reference plane for analysis.

[0074] Pipeline-side data acquisition: Remote pressure monitoring terminals are widely distributed at various monitoring points in the pipeline network to acquire pressure data and meter elevations in real time. Pipeline pressure data is an important indicator for understanding the operational status of the pipeline network and can promptly detect abnormalities such as leaks and blockages; meter elevation information helps in analyzing the dynamic characteristics of water flow.

[0075] Environmental data collection: Daily temperature, humidity, and rainfall data are obtained through the meteorological department interface, combined with time characteristics (such as holiday attributes and seasonal attributes). Temperature and humidity affect users' water usage habits; for example, residential water consumption usually increases during hot weather. Rainfall affects urban outdoor water consumption; for example, water consumption for greening and road cleaning usually decreases during rainy days. Holiday and seasonal attributes can also reflect the changing patterns of water demand; for example, water consumption may decrease in some areas during the Spring Festival, while demand increases significantly during the summer peak water consumption period.

[0076] Step S1.2, data processing.

[0077] Anomaly Data Labeling: Based on the equipment range and actual operating patterns of the water supply system, reasonable quantification thresholds are set for each data point. When collected data exceeds the threshold range, it is labeled as an anomaly. For example, the outlet pressure of a water plant typically fluctuates within a certain range; exceeding this range may indicate equipment malfunction or abnormal operating conditions.

[0078] Missing value imputation: Missing data segments encountered during data collection are handled meticulously. For data segments with short consecutive missing periods, an arithmetic sequence generated from the preceding and following valid values ​​is used to imput them; this method maintains the continuity and trend of the data. Missing values ​​for lag features or sliding window statistics are marked.

[0079] Invalid monitoring points and sample removal: Calculate the percentage of outliers and missing values. If the percentage exceeds a set threshold (e.g., 10%), remove the monitoring point and delete the related data samples. This ensures the quality of the dataset and avoids invalid data interfering with subsequent analysis.

[0080] Step S2, water supply network pressure demand analysis.

[0081] Step S2.1, spatial analysis of water supply pressure.

[0082] Unfavorable Point Identification: Based on the water supply system's pipeline topology, and using graph theory and other relevant knowledge, we identify pipeline endpoints that are farther from the water plant than a set value, as well as high points with elevation differences greater than a set value, and define them as unfavorable points. These points often face significant pressure challenges during water supply and are key areas for ensuring water supply reliability.

[0083] Spatial Feature Extraction: Spatial features of low-pressure areas, including transport distance and elevation difference, are extracted from the dataset. Transport distance affects water flow resistance, while elevation difference is closely related to head loss. These spatial features provide important physical basis for analyzing pressure demand.

[0084] Step S2.2, Water supply pressure and time analysis.

[0085] Temporal Pattern Capture: In-depth analysis of low-pressure area data is conducted over time, employing time series analysis methods to capture the patterns of pressure changes over time. For example, by analyzing historical data, pressure variation patterns can be identified at different times of day, on different days of week, and in different seasons, revealing pressure characteristics during peak and off-peak water usage periods.

[0086] Step S2.3, quantitative analysis of adverse pressure demand.

[0087] Clear Demand: Combining the spatial characteristics of low-pressure areas, temporal analysis results, and daily pressure complaint data, the impact of various factors on pressure demand is comprehensively considered. For example, during peak water usage periods, unfavorable locations may require higher pressure to meet water demand; while daily pressure complaint data reflects users' actual feelings and needs regarding pressure. Through comprehensive analysis, the pressure demand of each unfavorable location is clearly defined, constructing a demand set that includes both single values ​​and time-varying curves, providing a precise basis for subsequent scheduling decisions.

[0088] Step S3: Construction of water supply prediction model.

[0089] Step S3.1, Water volume prediction model construction.

[0090] Input feature construction: Based on historical hourly water volume data, an input feature set is constructed by combining meteorological information (daily temperature, humidity, rainfall) and temporal characteristics (holiday attributes, seasonal attributes). These features can comprehensively reflect various factors affecting water consumption. For example, rising temperatures may lead to increased residential water consumption, and holidays may change industrial and commercial water consumption patterns.

[0091] Model Selection and Training: Time series models (such as LSTM and Prophet) are used for training. LSTM (Long Short-Term Memory) can handle time series data with long-term dependencies, while Prophet has good predictive performance for data with seasonality and trends. Samples are generated using the sliding window method, and the prediction accuracy is evaluated using the mean absolute percentage error (MAPE). If the prediction accuracy does not meet the standard, historical data is added and the model is retrained until the accuracy requirements are met, ensuring the accuracy of water volume prediction.

[0092] Step S3.2, Construction of the total head difference prediction model.

[0093] The total head difference, representing the total head loss between the water plant and the monitoring points in the pipeline network, is a key physical quantity for calculating the energy consumption of the pipeline network. Its calculation depends on the pressure and meter elevation data of the water plant and the monitoring points.

[0094] Data Preparation: A two-dimensional matrix is ​​constructed using the outflow rates of each water plant and the total flow rate of the area as input features. While ensuring spatiotemporal alignment with the two-dimensional matrix, the total head difference of the water supply system is accurately calculated to construct a total head difference matrix. Spatiotemporal alignment is crucial for ensuring data consistency and model accuracy; for example, it must be ensured that flow rate data at the same point in time corresponds to the corresponding total head difference data.

[0095] Data preprocessing: The two-dimensional matrix and the total head difference matrix are aligned, filtered, and normalized. Alignment filtering removes invalid or erroneous data points, while normalization scales the data to a specific range, facilitating model training and improving model convergence speed.

[0096] Model Construction and Training: A LightGBM gradient boosting tree combined with a MultiOutputRegressor framework was adopted. LightGBM offers advantages such as fast training speed, low memory consumption, and high prediction accuracy, while MultiOutputRegressor can handle multiple outputs. Mean squared error was used as the loss function, and L1 regularization was introduced to suppress overfitting. Hyperparameters were tuned using Bayesian optimization. A total head difference prediction model was constructed and trained based on the training set. Simultaneously, the mean absolute error over multiple time periods was used to evaluate model accuracy, and the SHAP value was used to analyze the characteristic contribution of each water plant's flow rate to each total head difference, ensuring the consistency between the model and hydraulic mechanisms. If the prediction accuracy did not meet the target, additional training data was added or the model structure was adjusted to optimize the model, ensuring that the model could accurately predict the total head difference.

[0097] Step S4: Construction of water supply optimization scheduling model.

[0098] Step S4.1, Construction of the objective function system.

[0099] Flow matrix derivation: The flow rate of each water plant is obtained and the flow matrix is ​​derived by combining it with the principle of water balance. Water balance is a fundamental principle of water supply systems, which allows for accurate calculation of the flow distribution relationship between water plants.

[0100] Objective function construction: Based on the flow matrix, the outlet pressure of each water plant, and various cost coefficients, an objective function is constructed with minimizing the total water production cost as the core objective. Cost coefficients include water production costs (excluding distribution costs) and distribution costs. By constructing a reasonable objective function, costs can be minimized while meeting water supply demands.

[0101] Step S4.2, calculate the objective function.

[0102] Constraints are determined as follows: Water volume constraints are derived based on the output of the water volume prediction model, the equipment and process requirements of each water plant, and its supply capacity. The water volume prediction model provides forecasts of future water consumption, while the equipment and process requirements of each water plant determine the upper limit of its supply capacity, thus defining the range of water volume constraints. Pressure constraints are obtained based on the demand set to ensure that the water supply pressure meets the demand at unfavorable points.

[0103] Objective function calculation: Under the premise of satisfying water quantity and pressure constraints, calculate the value of the objective function. This step needs to consider the interaction between various constraints to ensure the feasibility of the scheduling scheme.

[0104] Step S4.3: Optimize the objective function.

[0105] Algorithm selection: Sequential Least Squares Programming (SLSQP) is adopted, which has good performance in handling constrained optimization problems. Iterative solutions are performed to satisfy water quantity and pressure constraints.

[0106] Iterative solution: Through steps such as sequential quadratic programming, gradient calculation, and solution verification, the scheduling scheme is continuously adjusted until the set conditions are met, outputting the optimal scheduling scheme that satisfies the objective function. For example, the descent direction of the objective function is determined by calculating the gradient, and solution verification ensures that the scheme satisfies all constraints.

[0107] Step S5: Model update and iteration.

[0108] Step S5.1, Model update trigger condition.

[0109] Pipeline-side data changes: When significant changes occur in the pipeline network, such as the addition of new pipelines or pipeline expansion and reconstruction, the network topology and flow characteristics will change. In this case, it is necessary to re-collect the dataset, reconstruct the demand set, retrain the total head difference prediction model, and evaluate the accuracy of the total head difference model until the core indicators meet the requirements to adapt to the new pipeline network structure.

[0110] Changes in water plant data: If water plant data changes, such as adjustments to water plant processes or changes in electricity pricing policies, it will affect the water plant's effluent data and cost coefficients. In this case, it is necessary to supplement the corresponding new data, update the relevant parameters of the objective function, and verify the validity of the parameters through historical operating conditions to ensure the accuracy of the objective function.

[0111] Anomalies in environmental data: When environmental data becomes abnormal, such as extreme weather (7 consecutive days of high temperatures, or heavy rain with a daily rainfall of ≥50mm) or sudden public events (such as large-scale sporting events), it can lead to significant changes in water usage patterns. In this case, a model update is triggered, prioritizing the updating of the water volume prediction model and evaluating the accuracy of water volume prediction until the core indicators meet the standards, in order to adapt to the new water demand.

[0112] Decreased Model Prediction Accuracy: When the prediction accuracy of the water volume or total head difference model decreases, drift is diagnosed through data distribution analysis. For example, analyze the difference in distribution between predicted and actual values ​​to determine whether the data distribution has changed or if there is a problem with the model itself. Supplement monitoring data and update the model using incremental training. If the results are still unsatisfactory after training, check the sensor status or adjust the model structure to ensure the model's prediction accuracy.

[0113] Regular updates: When the set update time is reached, all core model update processes are triggered. Regular updates ensure that the model adapts to the gradual changes in the system over long-term operation, guaranteeing the model's timeliness.

[0114] Step S5.2, Model update mechanism.

[0115] Differentiated update process: The model update process is advanced in a differentiated manner based on different triggering reasons. For example, for changes in pipeline network data, it is necessary to re-collect data, construct the demand set, and retrain the model; for changes in water plant data, the main focus is on updating the relevant parameters of the objective function.

[0116] Model Evaluation and Deployment: After retraining the model, evaluate its accuracy until core indicators (such as water volume prediction accuracy, total head difference model accuracy, and water production cost) meet the standards. Deploy the new model after it meets the standards, and record the reasons for triggering the model update, the changed parameters, and the verification results, so as to track and analyze the model update process in the future and ensure that the model always adapts to the dynamic changes of the water supply system.

[0117] Through the above process, the water supply system scheduling optimization method provided in this embodiment achieves intelligent management of the entire water supply system process. In the data acquisition and fusion stage, the comprehensiveness, accuracy, and reliability of the data are ensured, providing a solid foundation for subsequent analysis. Pressure demand analysis accurately identifies unfavorable points and their pressure demands, providing key basis for scheduling decisions. The construction of the water supply prediction model improves the prediction accuracy of water volume and total head difference, providing reliable data support for optimized scheduling. The construction of the water supply optimization scheduling model achieves precise scheduling with the goal of minimizing water production costs, improving the economic efficiency of the water supply system. The model update and iteration mechanism ensures that the model can adapt to the dynamic changes of the water supply system in real time, ensuring the long-term effectiveness of the scheduling strategy. This method significantly improves the stability, reliability, and intelligence level of the water supply system, and has broad application prospects and promotional value.

[0118] In an application scenario, the actual operation of a certain water supply area is described. This area includes two main water plants, WTP1 and WTP2, and three unfavorable points in the pipeline network: PE1, PH1, and PH2. The water supply system scheduling optimization method proposed in this invention is used for water supply system scheduling optimization, specifically including the following steps.

[0119] The first step is data collection and fusion.

[0120] Specifically, the SCADA system continuously collected minute-level outflow and pressure data, hourly outflow and pressure meter elevation data for water plants WTP1 and WTP2 for 15 consecutive days without any anomalies. Utilizing 82 remote pressure monitoring terminals deployed at key nodes of the pipeline network, minute-level network pressure and meter elevation data for the area were collected. Temperature, humidity, and rainfall data were acquired daily from a publicly available interface with meteorological departments, and weekday / holiday identification was collected daily, with seasonal attributes updated monthly.

[0121] Furthermore, based on the measurement range of the water supply system equipment and actual operating patterns, quantitative thresholds were set. For example, the pressure on the water plant side was set to 0.10-0.6 MPa, and the flow rate to 100-50000 m³ / h; the pressure on the pipeline side was set to 0.05-0.6 MPa. Data exceeding the threshold range was marked as outliers. Missing data segments encountered during the data collection process were uniformly marked as missing values. Using a statistical period of 15 consecutive days of data collection, monitoring points with a "outlier + missing value" ratio exceeding 10% were removed. The original number of time periods was 15 × 24 × 60 = 21600, and after removal, the final sample consisted of 20662 data points.

[0122] The second step is to analyze the pressure demand of the water supply network.

[0123] Specifically, based on the analysis of the water supply pressure in the area, PE1 was identified as the end point of the pipeline network, while PH1 and PH2 were identified as high-altitude points. These points, due to their distance from the water plant or their high altitude, were considered unfavorable points in the water supply system.

[0124] Furthermore, the spatial characteristics of PE1, PH1, and PH2 were extracted, including their distances from the water plant (PE1 is 7.46 km from WTP1 / WTP2, and PH1 / PH2 is 1.27 km from WTP1 / WTP2, and PH1 / PH2 is 1.43 km from WTP1 / WTP2, and PH1 / PH2 is 2.04 km from WTP1 / WTP2, and PH1 / PH2 is 16.81 km from WTP1 / WTP2, and PH1 / PH2 is 19.59 km from WTP1 / WTP2, and PH1 / PH2 is 2.04 m from WTP1 / PH1, and PH1 / PH2, respectively) and their elevation differences (PE1 / PH1 / PH2 is -2.04 m from WTP1 / PH1, and PH1 / PH2 is 19.59 m from WTP1, and PH1 / PH2 is 4.59 m from WTP1 / PH1, and PH1 / PH2, respectively).

[0125] Further analysis of the pressure variation distribution of PE1, PH1, and PH2 during peak water supply months revealed that the most unfavorable pressure for PE1 occurred during the evening peak water consumption period (21:00-23:00), while the most unfavorable pressure for PH1 and PH2 occurred during the early morning off-peak water consumption period (00:00-06:00).

[0126] Finally, combining spatial and temporal analysis results with daily pressure complaints, it was determined that the minimum pressure demand threshold for PE1 is peak pressure, while the minimum pressure demand thresholds for PH1 and PH2 are early morning low-peak pressure. A dataset of unfavorable locations and pressure demands was constructed to provide a basis for subsequent scheduling strategies.

[0127] The third step is to build a water supply prediction model.

[0128] Specifically, based on the 30-day historical hourly total water demand data collected in the first step, an input feature set is constructed by combining meteorological information (daily temperature, humidity, and rainfall) and temporal characteristics (weekday / holiday identifiers, seasonal attributes). The Prophet time series model is used, taking historical water consumption, meteorological information, and temporal characteristics as input. By learning the periodicity and trend of water consumption patterns, it predicts the hourly water demand for the next 24 hours. The model training uses a sliding window method to generate samples, predicting water demand for the next M=24 time periods based on historical water consumption over N=72 time periods. The accuracy is evaluated using MAPE (Modal Analysis for Predicting Water Demand). The measured MAPE of the model is 2.6%, meeting the scheduling accuracy requirements.

[0129] Furthermore, the outflow rates of water plants WTP1 and WTP2, along with the total flow rate of the area, were used as input features to construct a T×(2+1) column two-dimensional matrix (T=20662). Simultaneously, a total head difference matrix was constructed to record the total head difference between each time period, each water plant, and the unfavorable point. The two-dimensional matrix and the total head difference matrix were aligned, filtered, and normalized to eliminate the influence of dimensions. A LightGBM gradient boosting tree combined with a MultiOutputRegressor framework was adopted, using mean squared error as the loss function. L1 regularization was introduced to suppress overfitting, and hyperparameters were optimized using Bayesian optimization. Model evaluation used the mean absolute error across multiple time periods, with an overall error of 0.0026 MPa, meeting the accuracy requirements.

[0130] The fourth step is to construct a water supply scheduling optimization model.

[0131] Specifically, a 24×1 dimensional flow matrix is ​​constructed using the hourly outflow rate of WTP1 over the next 24 hours as an independent decision variable. The hourly outflow rate of WTP2 is derived from the water balance relationship. With the core objective of "minimizing the total water production cost of the two water plants over 24 hours", an objective function is constructed by combining the water production cost coefficients and water distribution power consumption cost coefficients of WTP1 and WTP2, excluding water distribution costs.

[0132] The cost coefficients include: WTP1 water production cost coefficient excluding water distribution cost: 0.05742 yuan / m³ (electricity cost 0.01122 yuan / m³ + chemical cost 0.04619 yuan / m³); WTP2 water production cost coefficient excluding water distribution cost: 0.05688 yuan / m³ (electricity cost 0.00939 yuan / m³ + chemical cost 0.04750 yuan / m³); WTP1 water distribution electricity consumption cost coefficient: 0.2914 yuan / (m³·MPa); WTP2 water distribution electricity consumption cost coefficient: 0.2829 yuan / (m³·MPa).

[0133] Specifically, water volume constraints are determined based on the output of the water volume prediction model; pressure constraints are determined based on the demand set. This ensures that each water plant meets its water supply capacity requirements, with hourly and daily flow rates within reasonable ranges. The total head difference prediction model is invoked, inputting the hourly outflow matrices of the two water plants and the total flow matrix of the region in the current optimization iteration, and outputting the total head difference matrix. Based on the predicted total head difference, the outflow pressure is calculated using the "unfavorable point priority" principle, ensuring that the pressure at all unfavorable points meets the constraints. The calculated outflow pressure is then substituted into the objective function to calculate the total 24-hour water production cost.

[0134] Furthermore, the SLSQP algorithm is employed for optimization to minimize the total water production cost while satisfying water balance, flow threshold, and pressure constraints. The scheduling scheme is iteratively optimized through steps such as sequential quadratic programming, gradient calculation, and solution verification. The final output is the optimal flow scheduling scheme, including the outflow rate and corresponding outflow pressure of each water plant for each time period within the next 24 hours. In this case, the optimized total water production cost decreased from 76,103 yuan to 72,749 yuan, a reduction of 4.4%.

[0135] The fifth step is to update and iterate the model.

[0136] Pipeline-side data changes: If there are significant changes to the pipeline topology, such as the addition of new pipelines or pipeline reconstruction and expansion, pressure data of the new nodes in the pipeline after the change and flow-pressure correlation data of the water plant are collected again. The set of unfavorable points is re-identified and the pressure threshold is adjusted. The total head difference prediction model is retrained and the accuracy of the total head difference model is evaluated until the core indicators meet the standards.

[0137] Water plant side data changes: If the water plant process is adjusted, or the electricity price policy changes, etc., new operating condition data such as cost coefficients and peak and valley periods are added, the corresponding parameters in the objective function are updated and optimized, and the validity of the parameters is verified through historical operating conditions.

[0138] Environmental data anomalies: If extreme weather or sudden public events occur, causing water use patterns to deviate significantly from historical patterns, data should be collected again, the model updated, and the accuracy of water volume prediction assessed until the core indicators meet the standards.

[0139] Decreased model prediction accuracy: If the accuracy of the water volume or total head difference prediction model decreases, diagnose drift by checking the data distribution, supplement monitoring data, and update the model using incremental training. If the accuracy still does not meet the requirements, check the sensor status or adjust the model structure.

[0140] Regular updates: Force a model update periodically (e.g., every 15 days) to adapt to the gradual changes that occur during long-term system operation.

[0141] Specifically, the model update process is differentiated based on different triggering reasons. For example, for changes in pipeline network data, data is re-collected and the total head difference model is trained; for changes in water plant data, the objective function parameters are updated. After retraining the model, the accuracy is evaluated until the core indicators meet the standards. The new model that meets the standards is deployed, and the triggering reasons, parameter changes, and verification results are recorded to ensure that the model always adapts to the dynamic changes of the water supply system.

[0142] Through the above process, the water supply system scheduling optimization method of this invention was implemented in this area, effectively solving the problems of insufficient terminal pressure during peak hours and pipeline overpressure during off-peak hours, significantly improving water supply reliability and user satisfaction. With the goal of minimizing the total water production cost, and considering the differences in unit water production costs among multiple water plants, an intelligent optimization algorithm was used to optimize the 24-hour flow scheme, resulting in a measured reduction of 4.4% in the total water production cost. Extending this method to a multi-water plant collaborative scheduling system can achieve a 2%-5% reduction in water production costs.

[0143] The following are embodiments of the apparatus of the present invention, which can be used to execute the water supply system scheduling optimization method involved in the present invention. For details not disclosed in the embodiments of the apparatus of the present invention, please refer to the method embodiments of the water supply system scheduling optimization method involved in the present invention.

[0144] Please see Figure 3 This invention provides a water supply system scheduling optimization device 800.

[0145] The water supply system scheduling optimization device 800 includes, but is not limited to: a data acquisition and processing module 810, a water volume prediction module 830, a total head difference prediction module 850, and a scheduling scheme generation module 870.

[0146] Among them, the data acquisition and processing module 810 is used to acquire multidimensional data of the water supply system and preprocess it to obtain a dataset. It analyzes the pressure data in the dataset from the spatiotemporal dimension to obtain a set of unfavorable points and corresponding pressure requirements to construct a demand set. The multidimensional data includes data from the water plant side, the pipeline network side and the environment side.

[0147] The water volume prediction module 830 is used to construct a water volume prediction model based on water volume data, meteorological information and time characteristics in the dataset using a time series model, and to optimize the model by generating samples and evaluating the prediction accuracy using the sliding window method and the mean absolute percentage error; the time characteristics include holiday attributes and seasonal attributes.

[0148] The Total Head Difference Prediction Module 850 uses the LightGBM gradient boosting tree combined with the MultiOutputRegressor framework to build a total head difference prediction model based on hydraulic mechanisms, the outflow of each water plant in the dataset and the total flow of the area. The model accuracy is evaluated and optimized by the mean absolute error over multiple time periods.

[0149] The scheduling scheme generation module 870 is used to construct an objective function based on the water output data and cost of each water plant, obtain water volume constraints based on the output of the water volume prediction model, the equipment, process requirements and water supply capacity of each water plant, obtain pressure constraints based on the demand set, and iterate to optimize the water supply system with the goal of satisfying the water volume constraints and pressure constraints and minimizing the water production cost until the set conditions are met, and output the optimal scheduling scheme of the water supply system.

[0150] It should be noted that the water supply system scheduling optimization provided in the above embodiments is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed. That is, the internal structure of the water supply system scheduling optimization device will be divided into different functional modules to complete all or part of the functions described above.

[0151] Furthermore, the embodiments of the water supply system scheduling optimization device and the water supply system scheduling optimization method provided in the above embodiments belong to the same concept. The specific way in which each module performs its operation has been described in detail in the method embodiments, and will not be repeated here.

[0152] Figure 4 A schematic diagram of the structure of an electronic device according to an exemplary embodiment is shown.

[0153] It should be noted that this electronic device is merely an example adapted to the present invention and should not be construed as providing any limitation on the scope of use of the present invention. Furthermore, this electronic device should not be interpreted as requiring or depending on having... Figure 4 One or more components of the exemplary electronic device 2000 shown.

[0154] The hardware structure of electronic devices 2000 can vary significantly due to differences in configuration or performance, such as... Figure 4As shown, the electronic device 2000 includes: a power supply 210, an interface 230, at least one memory 250, and at least one central processing unit (CPU) 270.

[0155] Specifically, power supply 210 is used to provide operating voltage for various hardware devices on electronic device 2000.

[0156] Interface 230 includes at least one wired or wireless network interface 231 for interacting with external devices. Of course, in other examples adapted to this invention, interface 230 may further include at least one serial-to-parallel conversion interface 233, at least one input / output interface 235, and at least one USB interface 237, etc. Figure 4 As shown, this does not constitute a specific limitation.

[0157] The memory 250 serves as a carrier for resource storage and can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored on it include the operating system 251, application programs 253, and data 255, etc., and the storage method can be temporary storage or permanent storage.

[0158] The operating system 251 is used to manage and control the various hardware devices and application programs 253 on the electronic device 2000, so as to enable the central processing unit 270 to perform calculations and processing on the massive data 255 in the memory 250. It can be Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0159] Application 253 is a computer-readable instruction based on operating system 251 that performs at least one specific task, and may include at least one module ( Figure 4 (Not shown), each module may contain computer-readable instructions for the electronic device 2000. For example, the water supply system scheduling optimization device can be considered as an application program 253 deployed on the electronic device 2000.

[0160] Data 255 may be signal information, etc., and is stored in memory 250.

[0161] The central processing unit 270 may include one or more processors and is configured to communicate with the memory 250 via at least one communication bus to read computer-readable instructions stored in the memory 250, thereby enabling the computation and processing of massive amounts of data 255 in the memory 250. For example, a water supply system scheduling optimization method can be implemented by the central processing unit 270 reading a series of computer-readable instructions stored in the memory 250.

[0162] Furthermore, the present invention can also be implemented through hardware circuits or a combination of hardware circuits and software. Therefore, the implementation of the present invention is not limited to any specific hardware circuit, software, or combination thereof.

[0163] Please see Figure 5 This invention provides an electronic device 4000, which may include: a desktop computer, a laptop computer, a server, etc., with sensor recognition capabilities.

[0164] exist Figure 5 In this context, the electronic device 4000 includes at least one processor 4001 and at least one memory 4003.

[0165] The data interaction between the processor 4001 and the memory 4003 can be achieved through at least one communication bus 4002. This communication bus 4002 may include a path for transmitting data between the processor 4001 and the memory 4003. The communication bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0166] Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0167] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0168] The memory 4003 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program instructions or code in the form of instructions or data structures and accessible by the electronic device 4000, but not limited thereto.

[0169] The memory 4003 stores computer-readable instructions, and the processor 4001 can read the computer-readable instructions stored in the memory 4003 through the communication bus 4002.

[0170] The computer-readable instructions are executed by one or more processors 4001 to implement the water supply system scheduling optimization method in the above embodiments.

[0171] Furthermore, this embodiment of the invention provides a storage medium storing computer-readable instructions, which are executed by one or more processors to implement the water supply system scheduling optimization method described above.

[0172] This invention provides a computer program product, which includes computer-readable instructions stored in a storage medium. One or more processors of an electronic device read the computer-readable instructions from the storage medium, load and execute the computer-readable instructions, thereby enabling the electronic device to implement the water supply system scheduling optimization method as described above.

[0173] Compared with related technologies, the beneficial effects of the present invention are:

[0174] 1. This invention can improve water supply stability; by integrating spatial-temporal pressure demand analysis and multi-model prediction, it accurately identifies unfavorable points and their pressure demands, effectively solving the problems of insufficient pressure at the end of the pipeline during peak hours and overpressure in the pipeline during off-peak hours, thereby significantly improving water supply reliability and user satisfaction.

[0175] 2. This invention effectively reduces the system's water production cost. By constructing an optimization objective function system centered on minimizing the total water production cost of each water plant, and considering the differences in unit electricity costs and chemical consumption costs among multiple water plants, an intelligent optimization algorithm is used to dynamically adjust the operating strategies of each water plant, achieving optimal system-level energy efficiency. Compared to traditional scheduling modes, this solution reduces the measured total water production cost by 4.4%. When extended to multi-water plant collaborative scheduling systems, it can achieve a 2%-5% reduction in water production costs. Simultaneously, precise pressure control avoids pipeline overpressure, reducing pipe burst rates and pipeline maintenance costs.

[0176] 3. This invention improves the practicality and generalization ability of the prediction model. By constructing a total head loss prediction model that relies solely on the water plant's outlet flow rate, it avoids the problems of strong dependence on pipeline parameters and high modeling costs associated with traditional mechanistic models. Simultaneously, it overcomes the prediction distortion problem caused by insufficient generalization ability in traditional mathematical models. By focusing on the hydraulic essence of "flow rate determining head loss," this model ensures prediction reliability across the entire flow rate exploration range, significantly improving the model's practicality and application scope.

[0177] 4. This invention achieves end-to-end intelligent collaboration; by integrating data from water plants, pipelines, and environmental systems, it constructs a closed-loop water supply scheduling system encompassing "data fusion, demand analysis, intelligent prediction, and dynamic optimization." This system can respond to changes in water demand in real time, dynamically adapt to system operating status, reduce manual intervention, and significantly improve the intelligent management level of the water supply system. Simultaneously, by constructing a model-differentiated update mechanism, it ensures that the scheduling strategy can adapt to dynamic changes in the system over the long term, maintaining the continuous effectiveness and adaptability of the scheduling system.

[0178] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0179] The above description is only a partial embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing the scheduling of a water supply system, characterized in that, The method includes: Multidimensional data of the water supply system is acquired and preprocessed to obtain a dataset. Pressure data in the dataset is analyzed from a spatiotemporal perspective to obtain a set of unfavorable points and corresponding pressure demands to construct a demand set. The multidimensional data includes data from the water plant side, the pipeline network side, and the environment side. A water volume prediction model is constructed based on the water volume data, meteorological information, and time characteristics in the dataset using a time series model. The prediction accuracy is evaluated and the model is optimized by generating samples and the mean absolute percentage error using the sliding window method. The time characteristics include holiday attributes and seasonal attributes. The outflow from each water plant and the total flow in the area in the dataset are used as input features to construct a two-dimensional matrix. After ensuring that the two-dimensional matrix is ​​spatiotemporally aligned with the two-dimensional matrix, the total head difference of the water supply system is calculated to construct a total head difference matrix. The two-dimensional matrix and the total head difference matrix are aligned, filtered and normalized to obtain the training set. A LightGBM gradient boosting tree combined with a MultiOutputRegressor framework is adopted, with mean squared error as the loss function. L1 regularization is introduced to suppress overfitting, and hyperparameters are tuned by Bayesian optimization. A total head difference prediction model is constructed and trained based on the training set, and the model accuracy is evaluated and optimized by the mean absolute error of multiple time periods. The outflow rate of each water plant is obtained and the flow rate matrix is ​​derived by combining the water balance. An objective function is constructed based on the flow rate matrix, the outflow pressure of each water plant and each cost coefficient. Water constraints are obtained based on the output of the water prediction model, the equipment, process requirements and water supply capacity of each water plant. Pressure constraints are obtained based on the demand set. The flow rate matrix is ​​input into the total head difference prediction model to obtain the total head difference prediction matrix. Based on the prediction matrix, the outlet pressure is calculated in reverse according to the principle of prioritizing the demand set. The current solution is calculated by substituting the objective function with the premise of satisfying the water volume constraint and pressure constraint. If the sequential least squares programming algorithm is selected as the optimization algorithm, then the optimal scheduling scheme of the water supply system is output by iteratively solving the current solution through sequential quadratic programming, gradient calculation and solution verification until the set conditions are met. If the genetic algorithm or particle swarm optimization algorithm is used as the optimization algorithm, then the optimal scheduling scheme of the water supply system is output by iteratively solving the current solution through heuristic search until the set conditions are met.

2. The water supply system scheduling optimization method as described in claim 1, characterized in that, The process of acquiring and preprocessing multidimensional data from the water supply system to obtain a dataset includes: Data is collected from the water plant side via the SCADA system, from the pipeline network side via the remote pressure monitoring terminal, and from the environmental side via the meteorological department interface to obtain a multidimensional raw dataset. The water plant side data includes the outflow rate, pressure, and meter elevation of each water plant. The pipeline network side data includes the pressure data and meter elevation of the pipeline network monitoring points. The environmental side data includes temperature, humidity, rainfall data, and time characteristics. A quantization threshold is set based on the equipment range and actual operating patterns of the water supply system. Data exceeding the quantization threshold and missing data segments in the original dataset are marked as outliers and missing values, respectively. Monitoring points with outliers and missing values ​​exceeding a set proportion are removed and their related data is deleted to obtain the dataset.

3. The water supply system scheduling optimization method as described in claim 1, characterized in that, The process of analyzing the stress data in the dataset from a spatiotemporal perspective to obtain a set of adverse points and corresponding stress requirements to construct a demand set includes: Based on the topology of the water supply system, identify the pipe network end points and the high points with elevation differences greater than the set value that are more than the set value away from the water plant as unfavorable points to construct a low-pressure area. Extract the spatial features of the low-pressure area from the dataset. The spatial features include the transport distance and elevation difference. Based on the dataset, the data of the low-pressure area is analyzed from the time dimension to obtain the low-pressure period. Based on the spatial characteristics of the low-pressure area and the low-pressure period, combined with daily stress complaint data, the stress demand of each of the unfavorable points is identified to construct a demand set; the stress demand includes a single value and a curve that changes over time.

4. The water supply system scheduling optimization method as described in claim 1, characterized in that, The model accuracy optimization model, which evaluates the mean absolute error over multiple time periods, includes: The prediction accuracy of the total head difference prediction model is evaluated by using the mean absolute error over multiple time periods. The characteristic contribution of each water plant's flow rate to each total head difference is analyzed by using the SHAP value to ensure the consistency between the model and the hydraulic mechanism. If the prediction accuracy does not meet the standard, training data is supplemented or the model structure is adjusted to optimize the model.

5. The water supply system scheduling optimization method as described in claim 1, characterized in that, The method includes: If the pipeline data changes, the dataset is re-collected, the demand set is reconstructed, the total head difference prediction model is retrained, and the accuracy of the total head difference prediction model is evaluated until the core indicators meet the standards. If the data from the water plant changes, the corresponding new data is added and the relevant parameters of the objective function are updated. The validity of the parameters is then verified by checking the historical operating conditions. When environmental data is abnormal, the model is updated. The water volume prediction model is updated first, and the accuracy of water volume prediction is evaluated until the core indicators meet the standards. The abnormal environmental data includes extreme weather and sudden events. When the prediction accuracy of the water volume prediction model or the total head difference prediction model decreases, the drift is diagnosed by data distribution test, monitoring data is supplemented and the model is updated by incremental training. If the accuracy is still not met after training, the sensor status is checked or the model structure is adjusted. When the set update time is reached, all models are retrained and their accuracy is evaluated until the core metrics are met. In all scenarios where the model is modified, deploy the new model that meets the criteria, and record the reason for triggering the model update, the changed parameters, and the verification results; the core indicators include water volume prediction accuracy, total head difference model accuracy, and water production cost.

6. A water supply system scheduling optimization device, characterized in that, The apparatus is applied to the water supply system scheduling optimization method according to claim 1, and the apparatus comprises: The data acquisition and processing module is used to acquire multidimensional data of the water supply system and preprocess it to obtain a dataset. It analyzes the pressure data in the dataset from the spatiotemporal dimension to obtain a set of unfavorable points and corresponding pressure demands to construct a demand set. The multidimensional data includes data from the water plant side, the pipeline network side, and the environment side. The water volume prediction module is used to construct a water volume prediction model based on the water volume data, meteorological information and time characteristics in the dataset using a time series model, and to evaluate the prediction accuracy and optimize the model by generating samples and average absolute percentage error using the sliding window method; the time characteristics include holiday attributes and seasonal attributes. The total head difference prediction module is used to combine the LightGBM gradient boosting tree with the MultiOutputRegressor multi-output regression framework to build a total head difference prediction model based on hydraulic mechanisms, the outflow of each water plant in the dataset and the total flow of the area. The model accuracy is evaluated and the model is optimized by evaluating the mean absolute error of multiple time periods. The scheduling scheme generation module is used to construct an objective function based on the water output data and cost of each water plant, obtain water volume constraints based on the output of the water volume prediction model, the water supply capacity of each water plant, equipment and process requirements, obtain pressure constraints based on the demand set, and use optimization algorithms to iteratively optimize the water volume constraints and pressure constraints and minimize water production costs until the set conditions are met, and output the optimal scheduling scheme of the water supply system.

7. An electronic device, characterized in that, include: At least one processor and at least one memory, wherein, The memory stores computer-readable instructions; The computer-readable instructions are executed by one or more of the processors, causing the electronic device to implement the water supply system scheduling optimization method as described in any one of claims 1 to 5.

8. A storage medium having computer-readable instructions stored thereon, characterized in that, The computer-readable instructions are executed by one or more processors to implement the water supply system scheduling optimization method as described in any one of claims 1 to 5.

Images