Digital twinning-based multi-dimensional water supply dynamic scheduling decision-making system

Abstract

The invention discloses a multi-dimensional water supply dynamic scheduling decision-making system based on digital twinning. The system comprises a multi-source data fusion module, a digital twinning body construction module, a water consumption prediction module, an optimization scheduling module, a water supply path optimization module and a feedback module. The multi-source data fusion module is used for collecting physical sensing data flow and external data flow of the water supply pipe network, performing fusion processing on the physical sensing data flow and the external data flow and static data, and constructing a water demand feature library; the digital twin building module can build a three-dimensional geometric model based on static data; the water consumption prediction module can calculate a final water consumption demand prediction value based on a pre-trained hybrid prediction model. According to the method, the water demand feature library is constructed, the hybrid prediction model is adopted, and multi-dimensional external data streams such as historical water consumption, weather forecast and public calendar are fused, so that high-precision and multi-scale prediction can be carried out, and peak water shortage and valley energy waste are avoided.

Description

Technical Field

[0001] This invention relates to the field of water supply scheduling technology, specifically to a multi-dimensional dynamic water supply scheduling decision system based on digital twins. Background Technology

[0002] With the acceleration of urbanization and the continuous expansion of the population, the complexity and dynamism of water supply systems are becoming increasingly prominent. On the one hand, the continuous expansion of urban boundaries has led to an exponential increase in the scale of water supply networks, with the proportion of pipelines operating under overload increasing year by year, and frequent leaks and bursts in aging pipelines. On the other hand, extreme weather events caused by climate change (such as short-term heavy rainfall and sustained high temperatures) and the emergence of new water use scenarios such as artificial lakes and commercial complexes have resulted in non-linear fluctuations in water consumption. Therefore, a dynamic water supply scheduling system is needed.

[0003] However, traditional water supply dispatching systems mainly rely on the experience of dispatchers, historical operational data, and limited monitoring instruments (such as pressure and flow meters) distributed at key nodes of the pipeline network. With the acceleration of urbanization and the diversification and dynamic nature of user water demands, traditional water supply dispatching systems are gradually revealing the following problems:

[0004] 1. Insufficient accuracy in water demand forecasting: Water demand is influenced by a combination of factors, including weather, temperature, holidays, social activities (such as sporting events and celebrations), and public emergencies. Traditional forecasting methods based on historical averages cannot capture these complex nonlinear relationships, leading to large forecasting errors. This often results in insufficient water pressure during peak hours and excessive water supply and idle pumping stations during off-peak hours at night—a "supply-demand mismatch" that negatively impacts user experience and wastes significant amounts of electricity.

[0005] 2. Delayed Dispatch Response: Faced with dynamic changes in water demand or sudden anomalies in the pipeline network (such as pipe bursts or water pollution), existing systems often rely on manual judgment and operation. From discovering the anomaly, locating the problem, formulating a solution, to executing the dispatch, the process is time-consuming and the impact can easily spread. Summary of the Invention

[0006] To address these issues, the present invention provides a multi-dimensional dynamic water supply scheduling decision system based on digital twins.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A multi-dimensional dynamic scheduling and decision-making system for water supply based on digital twins includes a multi-source data fusion module, a digital twin construction module, a water use prediction module, an optimized scheduling module, a water supply path optimization module, and a feedback module;

[0009] The multi-source data fusion module is used to collect physical sensor data streams and external data streams from the water supply network, and to fuse them with static data to build a water demand feature database.

[0010] The digital twin construction module can build a three-dimensional geometric model based on static data, use the physical sensor data stream in the water demand feature library as boundary conditions, and synchronize it to the corresponding node of the digital twin according to the mapping rule table.

[0011] The water demand prediction module can calculate the final water demand prediction value based on a pre-trained hybrid prediction model.

[0012] The optimized scheduling module can generate the optimal scheduling instruction sequence for each water supply device in the future cycle based on the final water demand forecast, combined with the water demand feature library and static data constructed by the multi-source data fusion module, in the virtual environment created by the digital twin construction module, through hydraulic simulation and optimization calculation.

[0013] The water supply path optimization module can dynamically generate multiple alternative physical paths based on real-time monitoring of pipeline node pressure values, pipeline GIS topology data, and graph search algorithms, and select the optimal path.

[0014] The feedback module is used to dynamically monitor the execution effect of the instructions generated by the optimization scheduling module. When the actual water supply pressure, flow rate or water quality deviates from the expected target, the deviation data is sent back to the water use prediction module and the optimization scheduling module.

[0015] Furthermore, the multi-source data fusion module includes a sensor array submodule and a fusion submodule;

[0016] The sensor array submodule includes a pressure transmitter, an electromagnetic flow meter, a turbidity sensor, and a pH sensor;

[0017] The fusion submodule is used to establish a unified spatiotemporal reference and data association mapping, which can assign standardized timestamps to physical sensing data streams and external data streams.

[0018] The multi-source data fusion module can also obtain public calendar data through an API interface.

[0019] Furthermore, the water usage prediction module includes a construction submodule and a calculation submodule;

[0020] The construction submodule can build and pre-train a hybrid prediction model based on the Prophet time series decomposition algorithm;

[0021] The calculation submodule can integrate historical water use data, meteorological data, and public calendar data based on a hybrid prediction model, and calculate the final water demand forecast.

[0022] Furthermore, the specific contents of the calculation submodule are as follows:

[0023] 1) Hybrid prediction models can integrate historical water usage, meteorological data, and public calendar data into multidimensional feature vectors. Then, the trend term is fitted using the Bayesian algorithm. Periodic terms and the impact of the event Determine historical water use time series data Historical water use time series data The calculation formula is as follows:

[0024] in, These are the preset coefficient terms in the hybrid prediction model;

[0025] 2) Then, using the trained hybrid prediction model, perform "in-sample" predictions on the historical data interval to obtain the baseline sequence data. Then, calculate each time point in the entire historical sequence. residual ;

[0026] 3) When the current residual fluctuation range is within the preset threshold range, the residual will be... With real-time multidimensional feature vectors Alignment and concatenation are performed to construct enhanced feature sequences. Ten similar samples from historical data with a Euclidean distance of less than 5 from the current time period are selected, and the mean of the target values ​​of these ten similar samples is taken as the final water demand forecast. .

[0027] Furthermore, the final water demand forecast The calculation formula is as follows:

[0028]

[0029] in, For the multidimensional feature vectors of 10 samples, For distance weighting function, For the first The historical residual value corresponding to each similar sample, where 10 represents the number of similar samples.

[0030] Furthermore, the specific content of the optimized scheduling module is as follows:

[0031] 1) Extract parameter data of water supply equipment within the scheduling cycle from the static data of the multi-source data fusion module to form known conditions and constraint boundaries;

[0032] 2) Calculate the optimal value with the goal of minimizing energy consumption based on the objective function and constraint boundaries. Optimization value The calculation formula is as follows:

[0033]

[0034] in, The total number of time periods in the scheduling cycle. The total number of water pumps participating in the scheduling. For time period index, For water pump index, This is the rated energy consumption characteristic coefficient of the water pump. For water pumps Relative speed, The duration of each scheduling cycle;

[0035] 3) Generate a large number of candidate scheduling schemes based on genetic algorithms, and submit each scheme to the hydraulic simulation engine of the digital twin for full-cycle dynamic simulation, calculate and output the global pressure distribution of each scheme. .

[0036] Furthermore, when performing full-cycle dynamic simulation, the digital twin uses pipeline network GIS topology data and pipeline attribute data as its framework, and the final water demand forecast value as its basis. and optimized value For driving input.

[0037] Furthermore, the global pressure distribution ,in, For the first Each node in the time period Pressure The total number of time periods in the scheduling cycle, and the pressure. The calculation formula is as follows:

[0038]

[0039] in, For the pressure of the reference node, water pump During the period The actual head, The density of water, It is the acceleration due to gravity. For water to flow through the pipe section Energy loss due to friction For the pipe section, For the first Each node.

[0040] The present invention has the following advantages: By constructing a water demand feature database and adopting a hybrid prediction model, the present invention integrates multi-dimensional external data streams such as historical water use, weather forecasts and public calendars, and realizes high-precision, multi-scale prediction of water demand under the influence of complex factors such as weather, holidays and social activities. This reduces the "supply and demand mismatch" caused by prediction deviations from the source and avoids peak water shortages and off-peak energy waste.

[0041] Meanwhile, this invention fundamentally solves the problem of delayed scheduling response by leveraging the virtual environment constructed through a digital twin and the rapid decision-making capabilities of the optimized scheduling module and water supply path optimization module. Faced with changing demand or network anomalies, the system can automatically complete the entire process from simulation and generation of the optimal scheduling instruction sequence to deduction and verification, replacing the slow process relying on manual judgment and operation. This shortens the anomaly response and handling time, minimizes the impact of faults, and improves the resilience and operational efficiency of the water supply system.

[0042] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. Attached Figure Description

[0043] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0044] Figure 1 This is a module architecture diagram of a multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to the present invention. Detailed Implementation

[0045] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these embodiments are merely for further explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Technical engineers in the field can make some non-essential improvements and adjustments to the present invention based on the above-described content. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Please see Figure 1A multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins includes a multi-source data fusion module, a digital twin construction module, a water use prediction module, an optimized scheduling module, a water supply path optimization module, and a feedback module.

[0047] The multi-source data fusion module is used to collect physical sensor data streams and external data streams from the water supply network, and to fuse them with static data to build a water demand feature database. The physical sensor data stream consists of water pressure, flow rate, and water quality data collected in real time by the sensor array at key nodes of the water supply network.

[0048] External data streams include meteorological data, which is collected by the timed task scheduler built into the multi-source data fusion module. The timed task scheduler can actively call the API interface of the external meteorological forecast data source at a preset frequency (such as every 1 hour) and extract structured key information from it through the rule engine.

[0049] For example, the weather forecast might indicate that "in the next 3 hours, the probability of rainfall in area XX is 70%, with an estimated precipitation of 10-20 mm."

[0050] Static data includes GIS topology data of water supply network and pipeline attribute data (such as point nodes, end nodes, material, diameter, length and laying year).

[0051] The multi-source data fusion module includes a sensor array submodule and a fusion submodule. The sensor array submodule includes a pressure transmitter, an electromagnetic flow meter, a turbidity sensor, and a pH sensor. The pressure transmitter monitors the water pressure at key nodes in the pipeline network in real time based on the piezoelectric sensing principle. The electromagnetic flow meter collects the instantaneous flow rate and cumulative water volume in the pipeline in real time based on Faraday's law of electromagnetic induction. The turbidity sensor uses the principle of light scattering to monitor the content of suspended particulate matter in the water, i.e., the turbidity of the water. The pH sensor measures the hydrogen ion concentration based on the glass electrode method and is used to measure the acidity and alkalinity of the water in real time, facilitating the detection of water quality in the water supply network through the turbidity sensor and pH sensor.

[0052] The fusion submodule is used to establish a unified spatiotemporal reference and data association mapping. It assigns standardized timestamps to physical sensor data streams and external data streams, unifying them to Coordinated Universal Time (UTC), and then maps them to the pipeline topology nodes in the static data, constructing a spatiotemporal index. Through this index, the fusion submodule can automatically associate and align physical sensor data streams, external data, and static data, forming a time-series consistent water demand feature library organized by geographical region, providing structured foundational data for subsequent analysis.

[0053] The multi-source data fusion module can also obtain public calendar data containing national statutory holidays, local events, and schedules of large-scale commercial and cultural sports events through API interfaces, and transform it into structured feature data through the rule engine for subsequent fusion.

[0054] For example: "Event type: concert, affected area: stadium area, intensity coefficient: 0.8".

[0055] The digital twin construction module can build a 3D geometric model based on static data, then use physical sensor data streams from the water demand feature database as boundary conditions, and synchronize them to the corresponding nodes of the digital twin according to the mapping rule table, realizing a virtual mirror of the water supply network, providing accurate simulation and prediction support for subsequent dynamic scheduling decisions. The specific working principle is as follows:

[0056] 1) Construct a three-dimensional geometric model based on static data (pipeline network GIS topology, pipeline attributes). Then, use the aligned physical sensor data stream from the water demand feature database as part of the boundary conditions. At the same time, use the structured external data of weather forecasts and calendar events from the water demand feature database as input to the water demand prediction model, so as to facilitate the subsequent calculation of the water supply demand load of each region in the future period by combining historical water use patterns.

[0057] 2) Receive physical sensor data streams in real time and map them to the corresponding nodes of the digital twin through a predefined mapping rule table to achieve state synchronization, so that the state values ​​of the three-dimensional model in the digital twin change in real time with the actual situation.

[0058] The water demand prediction module can calculate the final water demand prediction based on a pre-trained hybrid prediction model.

[0059] The water usage prediction module includes a construction submodule and a computation submodule. The construction submodule can build and pre-train a hybrid prediction model based on the Prophet time series decomposition algorithm;

[0060] First, the Prophet time series decomposition algorithm is used to train the baseline model, taking historical water consumption sequences and their corresponding dates, holidays, and event markers as input. Then, by fitting the trend, seasonal, and event impact parameters of the additive model, a hybrid forecasting model capable of capturing deterministic patterns in water consumption is generated. This hybrid forecasting model can accurately trace patterns in historical data, facilitating the calculation of baseline water consumption forecasts for any future time point based on time variables and future event schedules.

[0061] The calculation submodule integrates historical water usage data, meteorological data, and public calendar data based on a hybrid forecasting model to calculate the final water demand forecast. Details are as follows:

[0062] 1) Hybrid prediction models can integrate historical water usage, meteorological data, and public calendar data into multidimensional feature vectors. Then, the trend term is fitted using the Bayesian algorithm. Periodic terms (Including daily, weekly, and yearly cycles) and event impact items Determine historical water use time series data Historical water use time series data The calculation formula is as follows:

[0063] in, These are the preset coefficient terms in the hybrid prediction model.

[0064] 2) Then, using the trained hybrid prediction model, perform "in-sample" predictions on the historical data interval to obtain the baseline sequence data. This value is the fitting result obtained by performing "in-sample" predictions on each time point t within the historical data interval using a trained Prophet model. It represents the deterministic patterns captured by the model from the historical data. Subsequently, the values ​​for each time point in the entire historical sequence are calculated. residual residual The residual serves as the primary learning objective in the LSTM neural network within the hybrid prediction model. The calculation formula is as follows:

[0065]

[0066] 3) When the current residual fluctuation range is within the preset threshold range (e.g., -0.5 to 0.5), the residual will be... With real-time multidimensional feature vectors Alignment and concatenation are performed to construct enhanced feature sequences. Ten similar samples from historical data with a Euclidean distance of less than 5 from the current time period are selected, and the mean of the target values ​​of these ten similar samples is taken as the final water demand forecast. Final water demand forecast The calculation formula is as follows:

[0067]

[0068] in, For the multidimensional feature vectors of 10 samples, For distance weighting function, For the first The historical residual value corresponding to each similar sample, where 10 represents the number of similar samples.

[0069] The optimized scheduling module, based on the final water demand forecast and combined with the water demand feature library and static data constructed by the multi-source data fusion module, generates the optimal scheduling instruction sequence for each water supply device within a future cycle through hydraulic simulation and optimization calculations in a virtual environment created by the digital twin construction module. This facilitates the resolution of "supply-demand mismatch" and energy waste caused by forecasting errors. Details are as follows:

[0070] 1) Extract parameter data of water supply equipment within the scheduling cycle from the static data of the multi-source data fusion module to form known conditions and constraint boundaries. The parameter data includes the rated energy consumption characteristic coefficient, efficiency curve, and allowable start-stop frequency of each water pump.

[0071] 2) Calculate the optimal value with the goal of minimizing energy consumption based on the objective function and constraint boundaries. Optimization value The calculation formula is as follows:

[0072]

[0073] in, The total number of time periods in the scheduling cycle. The total number of water pumps participating in the scheduling. For time period index, For water pump index, This is the rated energy consumption characteristic coefficient of the water pump. For water pumps Relative speed is typically normalized to the range [0,1], where 0 represents shutdown and 1 represents operation at rated speed. The duration of each scheduling cycle.

[0074] 3) The optimization scheduling module generates a large number of candidate scheduling schemes based on a genetic algorithm, and submits each scheme to the hydraulic simulation engine of the digital twin for full-cycle dynamic simulation. The digital twin uses the network GIS topology data and pipeline attribute data as its framework, and the final water demand forecast value as its basis. and optimized value To drive the input, calculate and output the global pressure distribution for each scenario. This facilitates obtaining the global pressure distribution of each candidate scheduling scheme throughout the entire cycle T; .in, For the first Each node in the time period The pressure is calculated using the following formula:

[0075]

[0076] in, For the pressure of the reference node, water pump During the period The actual head, The density of water, It is the acceleration due to gravity. For water to flow through the pipe section Energy loss due to friction For the pipe section, For the first Each node.

[0077] Actual head The calculation formula is as follows:

[0078]

[0079] in, This refers to the design head of the water pump at its rated speed. This is the efficiency curve of the water pump at this speed.

[0080] Energy loss The calculation formula is as follows:

[0081]

[0082] in, For pipe section Length, For pipe section The pipe wall roughness system, For pipe section Pipe diameter.

[0083] The aforementioned genetic algorithm refers to a class of heuristic optimization algorithms that simulate the natural evolutionary process, primarily used for efficiently searching for near-optimal solutions in complex constraint spaces. In this system, the genetic algorithm initializes a set of random scheduling schemes as a "population," with each scheme encoding the start / stop status and speed settings of each pump at different times. The algorithm then iteratively performs selection, crossover, and mutation operations, using the objective function J combined with a penalty term for pressure constraint violations to evaluate and select superior schemes. Particle swarm optimization or simulated annealing algorithms can also be used, but the core purpose is to replace exhaustive search, finding the optimal balance between energy consumption and pressure stability from a massive number of possible schemes with the assistance of a digital twin simulation environment.

[0084] 4) After obtaining the global pressure distribution B from the digital twin simulation output of all candidate scheduling schemes, the optimization scheduling module initiates the screening process. First, based on the pipeline node pressure safety thresholds set in the static data, each scheme is verified one by one within the entire cycle T, covering all nodes. pressure If all boundary conditions are met, the solution is marked as feasible. If any node exceeds the pressure limit at any time, the solution is marked as infeasible.

[0085] Subsequently, the corresponding optimized values ​​are extracted from all feasible candidate solutions. The algorithm performs sorting and comparison, automatically selecting and outputting feasible optimized values. The lowest-cost option is selected as the optimal scheduling instruction sequence. This sequence specifies the start / stop instructions, target speed, and expected running time of each pump for each time period within the future scheduling cycle. It can be directly sent to the on-site control system of the water supply equipment for execution, thereby achieving precise and energy-saving dynamic scheduling.

[0086] The water supply path optimization module can dynamically generate multiple alternative physical paths that balance physical connectivity and flow capacity based on real-time monitoring of pipeline node pressure values, pipeline GIS topology data, and graph search algorithms. Finally, each alternative path undergoes hydraulic simulation verification in a digital twin. By calculating the global pressure distribution and optimized values, the optimal emergency path is automatically selected, and control commands are generated. This facilitates the rapid and automatic calculation and verification of a safe and feasible emergency water supply path in the event of pipeline anomalies, ensuring the continuity and reliability of the water supply system. The specific content of the water supply path optimization module is as follows:

[0087] 1) The water supply path optimization module can operate based on the pipeline GIS topology data, pipeline attribute data, and real-time physical sensor data stream provided by the multi-source data fusion module. When the pressure value of a certain area... An abnormal drop in water level is identified as a pipeline network anomaly.

[0088] 2) Using the anomaly point and water source as the starting and ending points, run a graph search algorithm on the pipeline topology map to quickly generate multiple alternative physical paths to bypass the anomaly area.

[0089] The graph search algorithm constructs a graph model based on the topology data of the pipeline network GIS, where pipeline nodes are vertices and pipes are edges. Each edge is dynamically weighted according to pipe attribute data (such as length, pipe diameter, and roughness coefficient) and real-time physical sensor data stream (such as current flow rate and pressure). The algorithm starts from anomalies and ends at available water sources. It traverses the weighted graph through dynamic programming or heuristic search (such as the A* algorithm) to quickly calculate multiple alternative physical paths that balance physical distance, flow capacity, and real-time load, while satisfying basic hydraulic connectivity.

[0090] 3) After each alternative path is generated, a hydraulic simulation is performed in the virtual environment created by the digital twin construction module. The simulation uses the final predicted water demand value. Using the required parameters as input, and the pipe diameter, length, roughness coefficient, and actual pump head of the selected path as boundary conditions, the global pressure distribution B of the entire network after path switching is calculated using the hydraulic equations. It is then determined whether the pressure at all nodes remains within a safe threshold, and the optimal value is calculated. Finally, the path with the smallest optimal value is selected as the optimal path.

[0091] The feedback module collects real-time physical sensor data streams from the water supply network (such as water pressure, flow rate, and water quality), and combines this with the synchronization status of the water demand feature database and the digital twin construction module to dynamically monitor the execution effect of commands generated by the optimization scheduling module. When the actual water supply pressure, flow rate, or water quality deviates from the expected target, the feedback module sends the deviation data back to the water use prediction module and the optimization scheduling module, which in turn coordinates with the water supply path optimization module to reassess the path schemes for abnormal areas of the network, thus forming a closed-loop control mechanism to ensure the dynamic balance and stability of the water supply system.

[0092] This invention effectively solves the problem of insufficient prediction accuracy of traditional methods by constructing a water demand feature database and adopting a hybrid prediction model. The model deeply integrates multi-dimensional external data streams such as historical water use, weather forecasts, and public calendars, enabling high-precision, multi-scale prediction of water demand under the influence of complex factors such as weather, holidays, and social activities. This reduces "supply-demand mismatch" caused by prediction bias at the source and avoids peak water shortages and off-peak energy waste.

[0093] Meanwhile, this invention fundamentally solves the problem of delayed scheduling response by leveraging the virtual environment constructed through a digital twin and the rapid decision-making capabilities of the optimized scheduling module and water supply path optimization module. Faced with changing demand or network anomalies, the system can automatically complete the entire process from simulation and generation of the optimal scheduling instruction sequence to deduction and verification, replacing the slow process relying on manual judgment and operation. This shortens the anomaly response and handling time, minimizes the impact of faults, and improves the resilience and operational efficiency of the water supply system.

[0094] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins, characterized in that, It includes a multi-source data fusion module, a digital twin construction module, a water use prediction module, an optimization scheduling module, a water supply path optimization module, and a feedback module; The multi-source data fusion module is used to collect physical sensor data streams and external data streams from the water supply network, and to fuse them with static data to build a water demand feature database. The digital twin construction module can build a three-dimensional geometric model based on static data, use the physical sensor data stream in the water demand feature library as boundary conditions, and synchronize it to the corresponding node of the digital twin according to the mapping rule table. The water demand prediction module can calculate the final water demand prediction value based on a pre-trained hybrid prediction model. The optimized scheduling module can generate the optimal scheduling instruction sequence for each water supply device in the future cycle based on the final water demand forecast, combined with the water demand feature library and static data constructed by the multi-source data fusion module, in the virtual environment created by the digital twin construction module, through hydraulic simulation and optimization calculation. The water supply path optimization module can dynamically generate multiple alternative physical paths based on real-time monitoring of pipeline node pressure values, pipeline GIS topology data, and graph search algorithms, and select the optimal path. The feedback module is used to dynamically monitor the execution effect of the instructions generated by the optimization scheduling module. When the actual water supply pressure, flow rate or water quality deviates from the expected target, the deviation data is sent back to the water use prediction module and the optimization scheduling module.

2. The multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to claim 1, characterized in that, The multi-source data fusion module includes a sensor array submodule and a fusion submodule; The sensor array submodule includes a pressure transmitter, an electromagnetic flow meter, a turbidity sensor, and a pH sensor; The fusion submodule is used to establish a unified spatiotemporal reference and data association mapping, which can assign standardized timestamps to physical sensing data streams and external data streams. The multi-source data fusion module can also obtain public calendar data through an API interface.

3. The multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to claim 1, characterized in that, The water usage prediction module includes a construction submodule and a calculation submodule; The construction submodule can build and pre-train a hybrid prediction model based on the Prophet time series decomposition algorithm; The calculation submodule can integrate historical water use data, meteorological data, and public calendar data based on a hybrid prediction model, and calculate the final water demand forecast.

4. The multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to claim 3, characterized in that, The specific contents of the calculation submodule are as follows: 1) Hybrid prediction models can integrate historical water usage, meteorological data, and public calendar data into multidimensional feature vectors. Then, the trend term is fitted using the Bayesian algorithm. Periodic terms and the impact of the event Determine historical water use time series data Historical water use time series data The calculation formula is as follows: ; in, These are the preset coefficient terms in the hybrid prediction model; 2) Then, using the trained hybrid prediction model, perform "in-sample" predictions on the historical data interval to obtain the baseline sequence data. Then, calculate each time point in the entire historical sequence. residual ; 3) When the current residual fluctuation range is within the preset threshold range, the residual will be... With real-time multidimensional feature vectors Alignment and concatenation are performed to construct enhanced feature sequences. Ten similar samples from historical data with a Euclidean distance of less than 5 from the current time period are selected, and the mean of the target values ​​of these ten similar samples is taken as the final water demand forecast. .

5. A multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to claim 4, characterized in that, The final water demand forecast The calculation formula is as follows: ； in, For the multidimensional feature vectors of 10 samples, For distance weighting function, For the first The historical residual value corresponding to each similar sample, where 10 represents the number of similar samples.

6. The multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to claim 1, characterized in that, The specific contents of the optimized scheduling module are as follows: 1) Extract parameter data of water supply equipment within the scheduling cycle from the static data of the multi-source data fusion module to form known conditions and constraint boundaries; 2) Calculate the optimal value with the goal of minimizing energy consumption based on the objective function and constraint boundaries. Optimization value The calculation formula is as follows: ； in, The total number of time periods in the scheduling cycle. The total number of water pumps participating in the scheduling. For time period index, For water pump index, This is the rated energy consumption characteristic coefficient of the water pump. For water pumps Relative speed, The duration of each scheduling cycle; 3) Generate a large number of candidate scheduling schemes based on genetic algorithms, and submit each scheme to the hydraulic simulation engine of the digital twin for full-cycle dynamic simulation, calculate and output the global pressure distribution of each scheme. .

7. A multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to claim 6, characterized in that, When performing full-cycle dynamic simulation, the digital twin uses pipeline network GIS topology data and pipeline attribute data as its framework, and the final water demand forecast as its base. and optimized value For driving input.

8. A multi-dimensional dynamic water supply scheduling and decision-making system based on digital twins according to claim 6, characterized in that, The global pressure distribution ,in, For the first Each node in the time period Pressure The total number of time periods in the scheduling cycle, and the pressure. The calculation formula is as follows: ； in, For the pressure of the reference node, water pump During the period The actual head, The density of water, It is the acceleration due to gravity. For water to flow through the pipe section Energy loss due to friction For the pipe section, For the first Each node.

Images