Water supply digital scheduling method and system based on water volume prediction and water storage regulation

Abstract

The invention discloses a water supply digital scheduling method and system based on water volume prediction and water storage regulation, and relates to the technical field of water plant water supply scheduling, and the method comprises the steps: creating a scheduling scene model, and obtaining the historical data of a target water plant; determining the water demand in a period of time in the future through a water consumption prediction model; determining each pipe network end unfavorable point and a pressure control range thereof; constructing a training sample set, and establishing and training a voltage drop fitting model; on the basis of the water consumption prediction model and the pressure drop fitting model, according to historical data, peak and valley electricity prices are matched to the maximum degree, and a water plant delivery pressure scheduling scheme, a water production scheme and a water pump scheduling scheme in a period of time in the future are formulated; and real-time factory instant flow data of the target water plant is obtained, dynamic self-correction is carried out in combination with the scheme and historical data, and real-time water plant water supply scheduling schemes including a real-time factory pressure scheduling scheme and a real-time water pump scheduling scheme are formulated. The accuracy and efficiency of water supply digital scheduling of the water plant can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of water supply scheduling in water plants, and in particular to a water supply digital scheduling method and system based on water volume prediction and water storage regulation. Background Art

[0002] The digital dispatching of water supply in water plants is to reasonably organize and coordinate the operation and management of various components of the water supply system in water plants, and realize the digital and precise dispatching between water supply and water use, so as to ensure the safety of water supply, improve service quality and reduce operating costs. At present, among the existing digital dispatching methods for water supply in water plants, there is a water consumption prediction method based on the combined model of GNN (Graph Neural Network) and LSTM (Long Short-Term Memory). This method makes full use of the combined model of GNN and LSTM to predict water consumption. Although it can accurately predict water consumption based on neural network technology, it only divides the relevant influencing factors of historical water consumption into time marks and meteorological marks, and does not make a more refined distinction. Ultimately, the accuracy of the digital dispatching scheme for water supply in water plants is still low, and the precise dispatching of water cannot be achieved. In addition, the existing water supply digital scheduling method of water plants also includes a pipe network optimization scheduling method based on hydraulic models and neural network algorithms. This method combines hydraulic models and neural network algorithms for water supply scheduling, but it relies on the modeling process of hydraulic models. At the stage when the digitization of most water supply pipe networks is not yet fully mature, the pipe network hydraulic model constructed based on the collected incomplete water supply pipe network structure data is not accurate enough, which ultimately leads to the low accuracy of the water supply digital scheduling plan of the water plant. In addition, the modeling cycle of the hydraulic model is generally relatively long, which requires a lot of manpower and material resources, making the efficiency of water supply scheduling of water plants low.

[0003] In summary, the current water supply digital scheduling method for water plants generally has the problems of low accuracy and low efficiency. Therefore, how to provide a water supply digital scheduling method for water plants with higher accuracy and efficiency has become a technical problem to be solved urgently in this field. Summary of the invention

[0004] The purpose of this application is to provide a water supply digital scheduling method and system based on water volume prediction and water storage regulation, which can improve the accuracy and efficiency of water supply digital scheduling of water plants.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a water supply digital scheduling method based on water volume prediction and water storage regulation, the water supply digital scheduling method based on water volume prediction and water storage regulation comprises the following steps:

[0007] Create a scheduling scenario model and obtain historical data of the target water plant; the historical data includes historical water consumption data, historical factory pressure data, historical water pump operation data, and historical pipe network pressure data and historical pressure drop data of each pipe network pressure measurement point in the water supply area of ​​the target water plant; the historical water consumption data includes historical factory instantaneous flow data counted by minutes, hours, weeks, holidays, months, quarters, temperature and weather dimensions; the historical pressure drop data is the difference between the historical factory pressure data at the same time and the historical pipe network pressure data of each pipe network pressure measurement point.

[0008] According to the historical water consumption data and the historical pressure drop data, the terminal unfavorable points in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable points are determined.

[0009] A training sample set is constructed based on the historical water usage data and the historical pressure drop data of the terminal unfavorable points.

[0010] A water consumption prediction model and a pressure drop fitting model are established respectively, and the water consumption prediction model and the pressure drop fitting model are trained respectively using the training sample set to obtain a trained water consumption prediction model and a trained pressure drop fitting model; wherein the water consumption prediction model refers to a model established based on the historical water use data and using a sliding window mean filtering algorithm to predict the water demand in a future period of time, the input of the water consumption prediction model is the historical water use data, minutes, hours, weeks, holidays, months, quarters, temperature and weather, and the output of the water consumption prediction model is the water demand in the future period of time; the pressure drop fitting model is a model established based on the historical water use data and the historical pressure drop data of each terminal unfavorable point to fit the relationship between water consumption and the pressure drop of each terminal unfavorable point, the input of the pressure drop fitting model is the historical water use data and the historical pressure drop data of each terminal unfavorable point, and the output of the pressure drop fitting model is the fitting relationship between water consumption and the pressure drop of each terminal unfavorable point.

[0011] Based on the trained water consumption prediction model and the trained pressure drop fitting model, and according to the historical data, a water supply scheduling plan for the water plant in the future period is formulated; the water supply scheduling plan for the water plant in the future period includes a water plant outlet pressure scheduling plan for the future period, a water production plan for the future period, and a water pump scheduling plan for the water plant in the future period.

[0012] The real-time instantaneous outflow flow data of the target water plant is obtained, and a real-time water supply scheduling plan for the water plant is formulated according to the real-time instantaneous outflow flow data, the historical data and the water supply scheduling plan for the water plant in the future period; the real-time water supply scheduling plan for the water plant includes a real-time water plant outflow pressure scheduling plan and a real-time water plant water pump scheduling plan.

[0013] In the second aspect, the present application provides a digital water supply scheduling system based on water quantity prediction and water storage regulation, and the digital water supply scheduling system based on water quantity prediction and water storage regulation supports local deployment and cloud deployment; the digital water supply scheduling system based on water quantity prediction and water storage regulation executes the digital water supply scheduling method based on water quantity prediction and water storage regulation described in the first aspect.

[0014] The water supply digital scheduling system based on water volume prediction and water storage regulation specifically includes the following functional modules.

[0015] A scenario generation module is used to create a scheduling scenario model and obtain historical data of the target water plant; the historical data includes historical water consumption data, historical factory pressure data, historical water pump operation data, and historical pipe network pressure data and historical pressure drop data of each pipe network pressure measurement point within the water supply area of ​​the target water plant; the historical water consumption data includes historical factory instantaneous flow data counted according to minutes, hours, weeks, holidays, months, quarters, temperature and weather dimensions; the historical pressure drop data is the difference between the historical factory pressure data at the same time and the historical pipe network pressure data of each pipe network pressure measurement point.

[0016] The data analysis module is used to determine the terminal unfavorable points in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable points according to the historical water use data and the historical pressure drop data; and to construct a training sample set according to the historical water use data and the historical pressure drop data of the terminal unfavorable points. The data analysis module also includes the regularity analysis of water consumption and the adaptability analysis of the factory pressure and the water consumption law. The regularity analysis of water consumption refers to the analysis of the historical water consumption data according to the minute, hour, week, month, quarter, holiday, temperature and weather analysis dimensions, and provides the analysis results of the overall water use regularity, water use peaks and valleys, and the ranking of the influence of each analysis dimension on the water use law. The adaptability analysis of the factory pressure and the water consumption law refers to the analysis of the matching degree of the factory pressure and the water use law, as well as the statistics of the maximum, minimum and average pressures of the period at the terminal points of the pipe network.

[0017] A modeling and evaluation module is used to establish a water consumption prediction model and a pressure drop fitting model respectively, and use the training sample set to train the water consumption prediction model and the pressure drop fitting model respectively to obtain a trained water consumption prediction model and a trained pressure drop fitting model; wherein the water consumption prediction model refers to a model established based on the historical water use data and using a sliding window mean filtering algorithm to predict the water demand in a future period of time, the input of the water consumption prediction model is the historical water use data, minutes, hours, weeks, holidays, months, quarters, temperature and weather, and the output of the water consumption prediction model is the water demand in the future period of time; the pressure drop fitting model is a model established based on the historical water use data and the historical pressure drop data of each terminal unfavorable point for fitting the relationship between water consumption and the pressure drop of each terminal unfavorable point, the input of the pressure drop fitting model is the historical water use data and the historical pressure drop data of each terminal unfavorable point, and the output of the pressure drop fitting model is the fitting relationship between water consumption and the pressure drop of each terminal unfavorable point.

[0018] The model publishing module is used to deploy the model to a local server or a cloud server based on the trained water consumption prediction model and the trained pressure drop fitting model, and formulate a water supply scheduling plan for the water plant in the future period according to the historical data; the water supply scheduling plan for the water plant in the future period includes a water plant ex-factory pressure scheduling plan for the future period, a water plant water production plan for the future period, and a water plant water pump scheduling plan for the future period; obtain the real-time ex-factory instantaneous flow data of the target water plant, and formulate a real-time water supply scheduling plan for the water plant according to the real-time ex-factory instantaneous flow data, the historical data, and the water supply scheduling plan for the water plant in the future period; the real-time water supply scheduling plan for the water plant includes a real-time water plant ex-factory pressure scheduling plan and a real-time water plant water pump scheduling plan.

[0019] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0020] The present application provides a digital scheduling method and system for water supply based on water quantity prediction and water storage regulation. The method collects historical water consumption data according to multiple dimensions such as minutes, hours, weeks, holidays, months, quarters, temperature and weather, and then uses a sliding window mean filtering algorithm to establish a water consumption prediction model. The water consumption prediction model is trained based on the historical water consumption data, and the changes in water consumption are studied from multiple relevant influencing factors such as time periods, weeks, months, holidays, etc., so as to improve the prediction accuracy of the water consumption prediction model and more accurately predict the water demand in the future. By combining the water consumption prediction model with the pressure drop fitting model, the historical water consumption data and the historical pressure drop data of each terminal unfavorable point are input into the pressure drop fitting model, and the pressure drop relationship between the water consumption and each terminal unfavorable point is fitted by the pressure drop fitting model. Therefore, according to the water demand in the future period predicted by the water consumption prediction model and the pressure drop relationship between the water consumption and each terminal unfavorable point fitted by the pressure drop fitting model, a water supply scheduling plan for the water plant in the future period can be formulated, including the water plant factory pressure scheduling plan in the future period, the water plant water production plan in the future period and the water plant water pump scheduling plan in the future period. Furthermore, under the premise of real-time factory instantaneous flow data, combined with historical data and the water supply scheduling plan for the water plant in the future period, a real-time water supply scheduling plan for the water plant can be further formulated, including the real-time water plant factory pressure scheduling plan and the real-time water plant water pump scheduling plan. Therefore, an accurate and reliable water supply scheduling plan can be formulated more quickly and specifically, which can improve the accuracy and efficiency of digital scheduling of water supply of water plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 A flow chart of a method for digitally scheduling water supply based on water volume prediction and water storage regulation is provided in accordance with an embodiment of the present application.

[0023] Figure 2 A schematic diagram of the structure of a water supply digital scheduling system based on water volume prediction and water storage regulation provided in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0025] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0026] like Figure 1 As shown, this embodiment proposes a water supply digital scheduling method based on water quantity prediction and water storage regulation. The water supply digital scheduling method based on water quantity prediction and water storage regulation is mainly used to explore water demand prediction and water storage regulation, that is, in order to achieve a balance between the predicted water demand and the water storage capacity of the water plant reservoir, the water supply scheduling and water storage adjustment of the water plant are carried out according to the predicted water demand, so as to help determine the scheduling schemes such as the water plant's factory pressure scheduling scheme, water production scheme and water pump scheduling scheme. The water supply digital scheduling method based on water quantity prediction and water storage regulation specifically includes the following steps.

[0027] Step S1: Create a scheduling scenario model and obtain historical data of the target water plant.

[0028] In this embodiment, a digital water supply scheduling scenario of the target water plant is constructed by creating a scheduling scenario model. Historical data includes historical water consumption data, historical factory pressure data, historical water pump operation data, and historical pipe network pressure data and historical pressure drop data of each pipe network pressure measurement point in the water supply area of ​​the target water plant. Among them, the historical water consumption data includes historical factory instantaneous flow data counted according to minute, hour, week, holiday, month, quarter, temperature and weather dimensions. The historical pressure drop data refers to the difference between the historical factory pressure data at the same time and the historical pipe network pressure data of each pipe network pressure measurement point.

[0029] Step S2: determining the terminal unfavorable points in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable points according to the historical water consumption data and the historical pressure drop data.

[0030] In this embodiment, step S2 determines the terminal unfavorable points in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable points according to the historical water consumption data and the historical pressure drop data, and specifically includes the following steps.

[0031] Step S21, based on the historical water use data and the historical pressure drop data, analyze the sensitivity of each pipe network pressure measurement point to the factory pressure, and select some pressure measurement points from each pipe network pressure measurement point as terminal unfavorable points in combination with the geographical coordinate position of each pipe network pressure measurement point and the location of users whose number of complaints about water pressure is greater than the threshold. Among them, the location of users whose number of complaints about water pressure is greater than the threshold refers to the location of users whose number of complaints due to low water pressure (greater than a preset threshold) is determined by water pressure survey. Usually, the number of complaints about low water pressure in these areas is much greater than the number of complaints about low water pressure in other areas. For example, the preset threshold for the number of complaints in a single month is 5 times, and the number of complaints in a single month due to low water pressure in a certain area is 20 times, then it will be considered to select terminal unfavorable points from the corresponding pipe network pressure measurement points in the area.

[0032] In this embodiment, step S21 selects some pressure measurement points from each pipe network pressure measurement point as terminal unfavorable points, which specifically includes the following steps.

[0033] Step S211: create a scheduling scenario based on the historical water consumption data, historical factory pressure data, and historical pipe network pressure data of each pipe network pressure measurement point within the water supply area of ​​the target water plant, configure the scheduling scenario, and determine the scenario configuration information corresponding to the scheduling scenario.

[0034] In this embodiment, the Panda AIoT (Artificial Intelligence Internet of Things) intelligent application platform can be used to create scheduling scenarios and scheduling scenario models, configure the scheduling scenarios, and determine the scene configuration information corresponding to the scheduling scenarios.

[0035] Step S212: According to the scenario configuration information, a terminal unfavorable point selection method is adopted to select some pressure measurement points from each pipe network pressure measurement point to obtain each terminal unfavorable point.

[0036] In this embodiment, the terminal unfavorable point selection method includes the first terminal unfavorable point selection method, the second terminal unfavorable point selection method or the third terminal unfavorable point selection method. Among them, the first terminal unfavorable point selection method is a sensitivity analysis method, including the Pearson correlation coefficient, the Spearman correlation coefficient and the Kendall correlation coefficient method. The output of the sensitivity analysis method is the sensitivity of the pressure of each pipe network pressure measurement point to the factory pressure. The selected terminal unfavorable point is a manually selected pipe network pressure measurement point with a sensitivity range of 40% to 80%. The second terminal unfavorable point selection method is to manually select the terminal point based on the geographical coordinate position and elevation information of each pipe network pressure measurement point. The third terminal unfavorable point selection method is to investigate the location of users who have complained to the water plant about water pressure in a historical period of time and the number of complaints counted by location, and select the pipe network pressure measurement point that is closer to the geographical coordinate position of the location of the user whose number of complaints about water pressure is greater than the threshold.

[0037] Step S22: adopt the confidence interval method to take the upper and lower boundaries of a certain range of confidence intervals for the historical pipe network pressure data of the selected terminal unfavorable points as the pressure control range of the terminal unfavorable points; the pressure control range of the terminal unfavorable points supports manual adjustment.

[0038] Step S3: construct a training sample set based on the historical water usage data and the historical pressure drop data of the terminal unfavorable points.

[0039] In this embodiment, step S3 constructs a training sample set based on the historical water usage data and the historical pressure drop data of the terminal unfavorable point. Constructing the training sample set specifically includes the following steps.

[0040] Step S31, calculating the difference between the historical factory pressure data and the historical pipe network pressure data of the terminal unfavorable point, and obtaining the historical pressure drop data of the terminal unfavorable point.

[0041] Step S32: performing data analysis and data cleaning on the historical water usage data and the historical pressure drop data of the terminal unfavorable points, generating training samples, and obtaining a training sample set.

[0042] Step S33: While analyzing and cleaning the data, conduct regularity analysis of water consumption and adaptability analysis of factory pressure and water consumption pattern. The regularity analysis of water consumption mainly analyzes historical data of water consumption based on analysis dimensions such as minutes, hours, weeks, months, quarters, holidays, temperature and weather, and provides analysis results of overall water consumption regularity, water consumption peaks and valleys, and the ranking of the influence of each analysis dimension on water consumption pattern. The adaptability analysis of factory pressure and water consumption pattern mainly analyzes the matching degree of factory pressure and water use pattern, as well as the statistics of the maximum, minimum and average pressures of the time period at the terminal points of the pipe network.

[0043] In this embodiment, data cleaning includes data screening processing and interpolation value-added processing, etc., wherein data screening processing includes self-increment determination, outlier determination, jitter determination, null value determination, time determination, repeated value determination and / or fixed range determination, etc. Interpolation value-added processing includes linear interpolation, average value interpolation, specific value interpolation and / or time correction, etc. Through the above data cleaning operation, the accuracy and quality of the training samples can be guaranteed, thereby improving the efficiency and effect of model training and the accuracy of model prediction.

[0044] Step S4: establish a water consumption prediction model and a pressure drop fitting model respectively, and use the training sample set to perform model training on the water consumption prediction model and the pressure drop fitting model respectively to obtain a trained water consumption prediction model and a trained pressure drop fitting model.

[0045] In this embodiment, the water consumption prediction model refers to a model established based on the historical water consumption data and using a moving average filtering (MAF) algorithm to predict the water demand in a future period of time. The input of the water consumption prediction model is the historical water consumption data, minutes, hours, weeks, holidays, months, quarters, temperature and weather, etc. The output of the water consumption prediction model is the water demand in the future period of time.

[0046] In this embodiment, the pressure drop fitting model is a model established based on the historical water use data and the historical pressure drop data of each terminal unfavorable point for fitting the relationship between water consumption and the pressure drop of each terminal unfavorable point. The input of the pressure drop fitting model is the historical water use data and the historical pressure drop data of each terminal unfavorable point, and the output of the pressure drop fitting model is the fitting relationship between water consumption and the pressure drop of each terminal unfavorable point. The pressure drop fitting model can select a BP neural network model.

[0047] In this embodiment, in the model training stage, for the trained water consumption prediction model and pressure drop fitting model, there is also a model evaluation step for evaluating the performance of the trained water consumption prediction model and pressure drop fitting model, which is conducive to obtaining the optimal water consumption prediction model and pressure drop fitting model, thereby further improving the performance and accuracy of model prediction. The model evaluation specifically includes the following steps.

[0048] (1) Using MSE (mean square error), MAE (mean absolute error), RMSE (root mean square error) and R 2 (determination coefficient) and other evaluation indicators are used to respectively evaluate the trained water consumption prediction model and the trained pressure drop fitting model to obtain model evaluation results.

[0049] (2) According to the model evaluation results, the optimal water consumption prediction model and the optimal pressure drop fitting model are determined respectively. The optimal water consumption prediction model is used as the trained water consumption prediction model to predict the water demand in the future period. The optimal pressure drop fitting model is used as the trained pressure drop fitting model to fit the fitting relationship between the water demand in the future period output by the optimal water consumption prediction model and the pressure drop of each terminal unfavorable point.

[0050] In this embodiment, the optimal water consumption prediction model and the optimal pressure drop fitting model can be determined by model evaluation, and are used to replace the trained water consumption prediction model and pressure drop fitting model for data prediction and scheduling scheme generation. For example, based on the water demand in the future period predicted by the optimal water consumption prediction model, the fitting relationship between the water demand in the future period of the target water plant output by the optimal water consumption prediction model and the pressure drop of each terminal unfavorable point, combined with the pressure control range of the terminal unfavorable point, the water plant delivery pressure scheduling scheme in the future period is generated.

[0051] Step S5: Based on the trained water consumption prediction model and the trained pressure drop fitting model, and according to the historical data, formulate a water supply scheduling plan for the water plant in the future.

[0052] In this embodiment, the water supply scheduling plan of the water plant in the future period includes the water plant delivery pressure scheduling plan in the future period, the water production plan of the water plant in the future period and the water pump scheduling plan of the water plant in the future period.

[0053] In this embodiment, step S5 formulates a water supply scheduling plan for the water plant in the future based on the trained water consumption prediction model and the trained pressure drop fitting model and according to the historical data, which specifically includes the following steps.

[0054] Step S51, based on the water demand in the future predicted by the trained water consumption prediction model, the fitting relationship between the water consumption and the pressure drop of each terminal unfavorable point fitted by the trained pressure drop fitting model, combined with the pressure control range of the terminal unfavorable point, generate a water plant delivery pressure scheduling plan for the future.

[0055] In this embodiment, the trained water consumption prediction model and the trained pressure drop fitting model can be published to the local server or cloud server of the target water plant. Through the above model publishing operation, it is convenient for the target water plant to directly apply the above trained water consumption prediction model and the trained pressure drop fitting model to respectively perform water demand prediction and pressure drop relationship fitting in the future period, so as to quickly, efficiently and accurately formulate a water supply scheduling plan for the water plant in the future period.

[0056] Step S52, based on the water demand in the future period and the target liquid level setting of the water reservoir of the target water plant, an optimization algorithm is used to generate a water production plan for the water plant in the future period; the water production plan for the water plant in the future period refers to making full use of the electricity price advantage in the low electricity rate period according to the characteristics of peak and valley electricity prices, maximizing the matching of peak and valley electricity prices, and pre-filling the water reservoir according to the water demand of the target water plant in the future period during the low electricity rate period, reducing the water storage in the water reservoir during the high electricity rate period, alleviating the daytime peak water supply pressure while reducing the water production electricity fee.

[0057] As a preferred embodiment, the future period of time may refer to the next 24 hours, or may refer to another time period in the future, for example, the next 12 hours, the next 36 hours, the next 48 hours, etc. The present embodiment preferably uses the next 24 hours. The water production plan of the water plant in the future period of time refers to the plan in which the water production is reduced between 6-24 o'clock in the next 24 hours, the water production is increased between 0-6 o'clock, and the clear water tank level of the target water plant does not have a bottoming low alarm and a high overflow alarm, so that the preferential electricity price at night (such as 0-6 o'clock) can be fully utilized to increase the water production, and the water production is reduced during non-idle hours (such as 6-24 o'clock), while ensuring that the clear water tank level will not bottom out the low alarm and the high overflow alarm. Through the accurate prediction of the water demand in the next 24 hours, it is convenient for the water plant to dynamically adjust the reservoir, thereby maximizing the advantage of the low electricity fee period, providing a water production plan, and achieving the purpose of reducing the water production cost of the water plant.

[0058] Step S53, based on the water demand in the future period, the water plant's factory pressure in the future period, historical water use data, historical factory pressure data and historical water pump operation data, a decision tree algorithm is used to generate a water plant pump scheduling plan for the future period; the water plant pump scheduling plan for the future period refers to a water supply pump start-stop combination plan that is generated based on historical water pump operation data and water supply pump operation parameters and can meet the water demand in the future period and the factory pressure in the future period.

[0059] Step S6, obtaining the real-time instantaneous outflow flow data of the target water plant, and formulating a real-time water supply scheduling plan for the water plant according to the real-time instantaneous outflow flow data, the historical data and the water supply scheduling plan for the water plant in the future period; the real-time water supply scheduling plan for the water plant includes a real-time water plant outflow pressure scheduling plan and a real-time water plant water pump scheduling plan.

[0060] In this embodiment, step S6 obtains the real-time instantaneous flow data of the target water plant, and formulates a real-time water supply scheduling plan for the water plant based on the real-time instantaneous flow data, the historical data and the water supply scheduling plan for the water plant in the future period of time, which specifically includes the following steps.

[0061] Step S61, obtaining the real-time instantaneous outflow data of the target water plant.

[0062] In this embodiment, relative to the historical factory instantaneous flow data in the historical water usage data, the real-time factory instantaneous flow data refers to the current real-time factory instantaneous flow data calculated based on dimensions such as minutes, hours, weeks, holidays, months, quarters, temperature and weather.

[0063] Step S62, the real-time instantaneous factory flow data and the water demand in the future predicted by the trained water consumption prediction model are used as water consumption data according to weights, and the water demand data fitted by the trained pressure drop fitting model are combined with the fitting relationship between the pressure drop of each terminal unfavorable point and the pressure control range of the terminal unfavorable point, and the water demand forecast for the future is dynamically corrected to adjust the water supply, so as to adjust the water supply scheduling plan of the water plant in the future in real time, optimize the operation strategy of the target water plant, and generate a real-time water plant factory pressure scheduling plan.

[0064] This embodiment dynamically corrects the water demand forecast for a period of time in the future based on real-time data such as real-time factory instantaneous flow data, so as to adjust the water supply, adjust the working status of the water pump, and optimize the water supply scheduling plan, thereby helping to improve the accuracy of the model's prediction of water demand and enhance the water plant's control and stable operation capabilities.

[0065] Step S63, using the real-time instantaneous factory flow data and the real-time water plant factory pressure scheduling plan as water demand data and factory pressure data, combined with the historical water usage data, the historical factory pressure data and the historical water pump operation data, using a decision tree algorithm to generate a real-time water plant water pump scheduling plan.

[0066] This embodiment combines artificial intelligence, machine learning, deep learning and other technologies to analyze and model the outlet water pressure regulation scenario in the water plant, and uses the model to intelligently predict future water demand and outlet water pressure. The water demand forecasting model, a time series forecasting model, intelligently predicts the water demand in the next 24 hours. According to the water demand in the next 24 hours, it can provide the water plant with a recommended outlet water pressure, and provide a time-segmented pressure regulation plan for the next 24 hours, so as to achieve efficient, intelligent and energy-saving water plant scheduling. Through the accurate prediction of future water volume and the dynamic adjustment of the reservoir, the advantages of low electricity fee periods can be maximized, providing economic and environmental benefits for water plant operations. At the same time, based on the water use forecast and real-time data for the next 24 hours, the system will automatically correct and adjust the water supply, adjust the working status of the water pump, and optimize the operation strategy of the water plant.

[0067] In an exemplary embodiment, Figure 2 As shown, a water supply digital scheduling system based on water quantity prediction and water storage regulation is provided. The water supply digital scheduling system based on water quantity prediction and water storage regulation is an intelligent scheduling system corresponding to a water supply digital scheduling method based on water quantity prediction and water storage regulation. It performs standardized analysis and modeling on the outlet water pressure regulation scenario in the water plant, and uses the model to intelligently predict the future outlet water pressure, guide the actual production scheduling of the water plant, so as to achieve efficient, intelligent and energy-saving scheduling of the water plant. The water supply digital scheduling system based on water quantity prediction and water storage regulation mainly includes modules such as scenario generation module, data analysis module, modeling evaluation module and model publishing module. The functions of each module are as follows:

[0068] The scenario generation module is mainly used to create scheduling scenario models and generate scheduling scenarios and data acquisition, including obtaining historical data of the target water plant. The historical data includes historical water consumption data, historical factory pressure data, historical water pump operation data, and historical pipe network pressure data and historical pressure drop data of each pipe network pressure measurement point in the water supply area of ​​the target water plant. The historical water consumption data includes historical factory instantaneous flow data counted by minute, hour, week, holiday, month, quarter, temperature and weather dimensions. The historical pressure drop data is the difference between the historical factory pressure data at the same time and the historical pipe network pressure data of each pipe network pressure measurement point.

[0069] The data analysis module is mainly used for data analysis and processing, including determining the terminal unfavorable points in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable points according to the historical water use data and the historical pressure drop data, and constructing a training sample set according to the historical water use data and the historical pressure drop data of the terminal unfavorable points. The data analysis module also includes the regularity analysis of water consumption and the adaptability analysis of the factory pressure and the water consumption regularity. The regularity analysis of water consumption mainly analyzes the historical water consumption data according to the analysis dimensions of minutes, hours, weeks, months, quarters, holidays, temperature and weather, and provides the analysis results of the overall water consumption regularity, water consumption peaks and valleys, and the ranking of the influence of each analysis dimension on the water consumption regularity. The adaptability analysis of the factory pressure and the water consumption regularity mainly analyzes the matching degree of the factory pressure and the water use regularity, as well as the statistics of the maximum, minimum and average values ​​of the period pressure at the terminal point of the pipe network.

[0070] The modeling and evaluation module is mainly used to establish models and model evaluation, including respectively establishing a water consumption prediction model and a pressure drop fitting model, and using the training sample set to respectively train the water consumption prediction model and the pressure drop fitting model to obtain a trained water consumption prediction model and a trained pressure drop fitting model; wherein, the water consumption prediction model refers to a model for predicting water demand in the future period established by using a sliding window mean filtering algorithm based on the historical water consumption data, and the input of the water consumption prediction model is historical water consumption data, minutes, hours, weeks, holidays, months, quarters, temperature and weather, and the output of the water consumption prediction model is water demand in the future period. The pressure drop fitting model is a model for fitting the relationship between water consumption and the pressure drop of each terminal unfavorable point established based on the historical water consumption data and the historical pressure drop data of each terminal unfavorable point, and the input of the pressure drop fitting model is the historical water consumption data and the historical pressure drop data of each terminal unfavorable point, and the output of the pressure drop fitting model is the fitting relationship between water consumption and the pressure drop of each terminal unfavorable point.

[0071] The model publishing module is mainly used for model publishing and model application as well as the formulation and generation of scheduling plans, including publishing the model to the local server or cloud server based on the trained water consumption prediction model and the trained pressure drop fitting model, and formulating the water supply scheduling plan for the water plant in the future period according to historical data. Among them, the water supply scheduling plan for the water plant in the future period includes the water plant discharge pressure scheduling plan for the future period, the water production plan for the future period, and the water pump scheduling plan for the future period; and obtaining the real-time discharge instantaneous flow data of the target water plant, and formulating the real-time water supply scheduling plan for the water plant according to the real-time discharge instantaneous flow data, historical data and the water supply scheduling plan for the water plant in the future period. The real-time water supply scheduling plan for the water plant includes the real-time discharge pressure scheduling plan for the water plant and the real-time water pump scheduling plan for the water plant.

[0072] In actual application, the water supply digital scheduling system based on water volume prediction and water storage regulation involves data access, data collection, data preprocessing, scenario generation, scenario configuration, data sampling, data cleaning, data analysis, data modeling, model evaluation and model publishing. The working process between each module and the specific steps of prediction and scheduling are as follows:

[0073] Step 1: On the basis of establishing the IoT data of the water plant, use the scenario generation module to first create a new water supply digital scheduling scenario according to the creation navigation prompt; then bind key indicators, select the water outlet pressure equipment, water outlet flow equipment, terminal pressure equipment type and its point and statistical type, and click OK to save the binding information; then select the time range and secondary sampling frequency, and click Start Verification to verify whether the data of the bound device meets the modeling requirements; then select the device as the terminal point based on the data quality of the terminal device and the sensitivity of each data to the water outlet pressure, water outlet flow and other information.

[0074] This embodiment adopts the Panda AIoT intelligent application platform, which pre-sets the necessary input and output indicators of the model used in the scheduling scenario. The water consumption prediction in this embodiment uses a sliding window mean filtering algorithm, and the pressure drop hydraulic relationship and pressure prediction use a fully connected neural network algorithm. It can create a digital water supply scheduling scenario with one click, and fill in the necessary configuration information of the scheduling scenario. These configurations are mainly used to obtain training samples from the Panda AIoT intelligent application platform.

[0075] After acquiring samples based on the configuration information of the Panda AIoT intelligent application platform, this embodiment analyzes the acquired samples. The data analysis module mainly has functions such as data sampling, data cleaning and data analysis. In the data cleaning process, a variety of data screening methods are used, including self-increment judgment, outlier judgment, jitter judgment, null value judgment, time judgment, repeated value judgment, fixed range judgment and other data screening methods, so as to eliminate some noise data. At the same time, a variety of interpolation value-added methods are also used, such as linear interpolation, average value interpolation, specific value interpolation, time correction and other means to fill in the missing data, and finally obtain a stable sample set that can participate in modeling.

[0076] This embodiment uses the data analysis module to automatically generate training samples according to the scene configuration information, and can automatically perform data cleaning operations on the platform background. After the data cleaning is completed, the sample set can be manually screened, added and modified. The data analysis function of the data analysis module can be used to display the water consumption and influencing factors analysis and the pressure-water volume change law analysis. The historical data can be viewed and whether the terminal point selection is reasonable.

[0077] This embodiment also uses a clustering algorithm to analyze water consumption data, and divides the water consumption pattern into five water consumption periods: high peak, peak, flat valley, low peak, and low low peak. With the help of BP neural network, various relevant influencing factors such as time period, week, month, quarter, holiday, temperature, weather, etc. are added to the input and output of the water consumption prediction model. The water consumption period is combined with pressure to analyze the matching degree of water pressure and water consumption pattern, as well as the pressure distribution of the selected pipe network terminal point during the period, to determine whether the pressure and flow have a direct impact, and whether there is room for pressure reduction.

[0078] Step 2: Using the modeling evaluation module, first call the built-in water consumption prediction model for water consumption prediction, and check the difference between the actual water consumption and the predicted water consumption. Then, after configuring the training parameters, click on training to train a pressure drop fitting model for pressure prediction and pressure regulation, and check the actual water outlet pressure, predicted water outlet pressure, and terminal pressure. Both water consumption prediction and pressure prediction will call the model evaluation function to automatically calculate the model performance, model stability, and model accuracy, as well as the model's MSE, MAE, RMSE, and R 2 And other evaluation indicators.

[0079] After determining the input, output and training samples of the model, this embodiment trains the water consumption prediction model and the pressure drop fitting model respectively, and performs model evaluation after the training is completed. The MSE, MAE, RMSE and R 2 And other evaluation indicators can be used to achieve accurate evaluation of the water consumption prediction model and the pressure drop fitting model, which is conducive to improving the prediction accuracy of the water consumption prediction model and the pressure drop fitting model, and ensuring more accurate and reliable prediction results.

[0080] In this embodiment, the water consumption prediction model uses a time series prediction model. First, it is necessary to explore the relevant factors that affect water consumption. Generally speaking, water consumption is closely related to the water use period. Peak water use periods may occur in the morning, afternoon and evening. The overall water use pattern shows a "peak-valley-peak-valley-peak" trend. In addition, the number of household water users is also an important factor affecting water consumption, but generally speaking, the number of water users is proportional to water consumption, and water consumption increases with the increase in the number of water users, so it is not included in the analysis. In addition, among the external environmental factors, factors such as holidays and weather have a direct impact on people, and thus will also affect people's water consumption and water use patterns.

[0081] This embodiment designs a sliding window mean filter algorithm based on the basic principle of water supply prediction and combines mathematical models, and optimizes the sliding window mean filter algorithm. 1 , Q 2 , …, Q n, let i be the corresponding sequence of the same Sunday (week) / same holiday / same hour / same weather / ..., i = 1, 2, 3, ..., k; α i As the time constant corresponding to each sequence, the water supply corresponding to time t is expressed as follows:

[0082]

[0083] Among them, n i The size of the window used to calculate the average for factors such as the same day of the week, the same holiday, the same hour, and the same weather, also known as the size of the sliding window; To calculate α backward from time t i j The weights corresponding to factors such as the same Sunday, the same holiday, the same hour, and the same weather when the duration is long; P i The weight ratio of factors such as the same Sunday, the same holiday, the same hour and the same weather; Represents the backward calculation of α from time t i The water supply data corresponding to the j time period, is with The related median filter function, the relationship between the two is as follows:

[0084]

[0085] Among them, m is the filter window length, m is an even number, Q lmin and Q lmax They are the minimum and maximum values ​​of the normal range of water supply respectively.

[0086] This embodiment uses a sliding window mean filtering algorithm, and compares the sliding window mean filtering algorithm with open source packaged time series prediction algorithms such as N-HITS (neural network hierarchical interpolation algorithm), N-BEATS (neural network bottom layer extension analysis algorithm), TimesNet (time series network algorithm), CRNN (convolutional recurrent neural network algorithm), etc. A water plant is selected as the experimental object. The historical data of the water plant from January to December 2023 is used as sample training, and the historical data of the first few days of January 2024 is used as the test set. In order to summarize the evaluation into the same dimension, MSE and MAE are used as evaluation indicators to evaluate the model accuracy. The prediction result data is shown in Table 1.

[0087] Table 1 Basic situation of water consumption in the area under the jurisdiction of a water plant

[0088]

[0089]

[0090] It can be seen that the MSE and MAE calculated using the sliding window mean filtering algorithm are the smallest, which means that this model has the best prediction effect. Therefore, the system gives priority to using this model as the algorithm model for water consumption prediction during model evaluation.

[0091] In this embodiment, the pressure drop fitting model uses a BP neural network model. The pressure drop fitting model is mainly used to predict the factory water pressure and fit the pressure drop relationship between water consumption and each terminal unfavorable point. When predicting the factory water pressure, it is first necessary to clarify the relationship between the factory water pressure, the factory water flow rate and the water pressure of each node and terminal monitoring point of the pipe network. In the selection of terminal points, by finding the lowest pressure point of each section of the pipe network and the pressure monitoring point near the pipe section with more complaints from residential users, 8 terminal unfavorable points are determined as sampling points, and then the corresponding factory water pressure data are collected, and the pressure drop of each terminal unfavorable point is calculated. Combined with the time variable, factory water flow rate and node flow rate data, the sample data information is merged and aligned (because the water pressure fluctuation propagates faster in the pipeline, the data delay can be ignored). The data is the average value collected by the Panda Unified Access Platform for 5 minutes, and horizontal alignment is performed by time.

[0092] It should be noted that the Panda AIoT intelligent application platform and Panda unified access platform used in this embodiment are existing platforms developed by Panda Smart Water Co., Ltd. The above platforms have been widely used and will not be repeated here.

[0093] This embodiment establishes a single-layer multi-neuron BP neural network. Based on the data of time variables, outgoing water flow and node flow, historical water consumption data and historical pressure drop data of each terminal unfavorable point are used as input, and the fitting relationship between water consumption and pressure drop of each terminal unfavorable point is used as output to obtain a black box model of flow and pressure drop. The fitting of the relationship between the outgoing water flow of the water plant and the pressure drop from the water plant to the terminal pressure measuring point is shown in Table 2.

[0094] Table 2 Fitting of the relationship between the water flow rate of a water plant and the pressure drop from the water plant to the terminal pressure measuring point

[0095]

[0096] According to Table 2, it can be seen that the prediction error is roughly within 0.01MPa, which is within the acceptable range for water plant scheduling.

[0097] This embodiment combines the water consumption prediction model to calculate the water consumption in the next 24 hours with hourly accuracy, and uses it as the input of the outgoing water flow parameter in the pressure drop fitting model to calculate the pressure drop change in the next 24 hours. Add the minimum service head requirement at each point, take the maximum value as the predicted value of the outgoing water pressure, and the outgoing water pressure in each period is as follows:

[0098]

[0099] Where P represents the outlet water pressure, is the voltage drop at the ith monitoring point, miny i is the minimum service head requirement for the ith monitoring point.

[0100] Step 3: Using the model publishing module, you can browse and select historical training models. Click Switch to switch the published model. Click Deploy to deploy the qualified pressure drop fitting model to the dispatching system in a single water plant to achieve real-time pressure prediction. When predicting real-time pressure, you can use the following formula: real-time outflow flow*0.9+water consumption in the next hour*0.1=input water volume of the current pressure drop fitting model to calculate the input water volume of the current pressure drop fitting model. In addition, you can also configure self-learning settings, configure self-learning trigger conditions, self-learning end conditions, and self-learning iterative replacement methods to achieve autonomous learning iterative updates. After the model is successfully published, the model can be used to predict future water consumption and outlet water pressure, provide time-based pressure regulation plans, and guide future water supply scheduling of the water plant.

[0101] Step 4: The water consumption prediction model predicts the water consumption in the next 24 hours and generates a water production plan. When producing water during the idle time of 0-6 o'clock at night, ensure that the clear water tank is at the lowest liquid level and the water storage tank reaches the maximum water storage capacity, thereby reducing the water production during the day and increasing the water production at night, and achieving the purpose of energy saving, water saving and optimizing water resource utilization. This embodiment uses the water consumption prediction model to predict the water consumption in the next 24 hours. Through the optimization algorithm, it can obtain a water production plan that reduces water production between 6 o'clock and 24 o'clock, increases water production between 0 o'clock and 6 o'clock, and does not have a low-level alarm or a high-level overflow alarm when the liquid level in the clear water tank touches the bottom, thereby improving the efficiency and effect of water supply scheduling in the water plant, improving water production efficiency, reducing water production costs, and ensuring low-cost and stable operation of water plant equipment.

[0102] Step 5: Continuously predict the water consumption and water outlet pressure in the next 24 hours. Based on the water consumption forecast and real-time water consumption data in the next 24 hours, the water plant's water production system will automatically adjust the water supply, adjust the working status of the water pump, optimize the water plant's operating strategy, and realize real-time updating and optimization of the water plant's water supply scheduling and water production plan.

[0103] This embodiment aims to achieve full coverage of the standard scenario modeling process, covering the complete chain from data access, data collection, data preprocessing, data sampling, data cleaning, data modeling, model evaluation, model release, and finally to the data model to generate practical application value, ensuring the smooth transmission and mining of data, and realizing the standardization, agility, and scenario modeling process of data. This embodiment aims to realize the water consumption prediction method of time series multi-model, analyze the water consumption situation from multiple related influencing factors such as time period, week, month, quarter, holiday, temperature, weather, etc., and use multiple neural network time series prediction algorithm models to comprehensively predict water consumption. The system evaluates the model capability according to the prediction results and selects one or more models with good prediction effect to use. In addition, this embodiment abandons the traditional cumbersome hydraulic model construction process and adopts advanced BP neural network algorithm to efficiently process complex pipe network hydraulic relations. This method fully incorporates key parameters such as time variables, factory water flow, node flow, and pressure data of each node and terminal into the modeling process, and accurately fits the dynamic relationship model between factory water flow and terminal pressure drop through deep learning mechanism. Compared with traditional hydraulic modeling methods, it significantly improves modeling efficiency, shortens the model development cycle, and better implements optimized scheduling and intelligent decision-making.

[0104] Compared with the prior art, this embodiment shows significant technical, economic, social and environmental benefits, which are specifically reflected in the following:

[0105] (1) Energy saving and consumption reduction and improvement of economic benefits.

[0106] Reducing the unit electricity consumption of water plants: This embodiment uses digital water supply scheduling technology to reduce the unit electricity consumption by 5% to 10%, significantly reducing operating costs, improving energy efficiency, and optimizing resource use.

[0107] Reducing pipe network leakage: This embodiment intelligently provides real-time water supply pressure based on users' real-time water usage data and water usage trend forecasts to better meet users' actual water needs. It reduces pipe network leakage and pipe burst accidents by optimizing factory water pressure.

[0108] Water use self-correction function: This embodiment realizes the water use self-correction function for the next 24 hours. For example, the electricity fee is about 1 yuan from 6:00 to 11:00 in normal time, and the value is about 0.3 yuan from 0:00 to 6:00 at night. By calculating the water tank at a low level and the night valley (according to the actual production capacity of the water plant, the normal water plant design is greater than the actual usage), the water tank is stored at a high level to achieve energy saving. The entire production capacity changes and the water supply scheduling plan of the water plant is automatically adjusted according to the calculation capacity ratio.

[0109] (2) Improvement of water supply safety and quality.

[0110] Security assurance: This embodiment uses the Internet of Things technology to comprehensively monitor the operating data of the water system, conduct intelligent analysis and early warning, reduce safety risks, and ensure safe and stable water supply.

[0111] Water pressure guarantee: This embodiment uses artificial intelligence technology to provide a digital water supply scheduling solution based on users' real-time water usage data and trend forecasts to accurately meet users' water needs and ensure stable water pressure.

[0112] (3) Both production efficiency and management efficiency are improved.

[0113] Intelligent modeling and management platform: This embodiment generates a highly available prediction model through a fast and agile modeling method, reduces interference from human factors, improves production stability and reliability, and reduces manual operation, analysis, and management costs.

[0114] Optimize the water system structure: This embodiment transforms from relying on personal experience and decentralized management to automation, centralization, and intelligence, improving management efficiency and reducing operating costs.

[0115] (4) Promote the coordinated development of upstream and downstream of the industrial chain.

[0116] Promote the development of the water industry chain: The realization of digital water supply scheduling in this embodiment relies on intelligent hardware, automation level and infrastructure construction. It can promote close cooperation between upstream and downstream of the water industry chain through intelligent transformation, optimize supply and demand matching, and improve overall production efficiency.

[0117] Reduce operating costs and improve resource utilization efficiency: This embodiment uses intelligent means to achieve efficient resource utilization, promote maximum economic benefits, and create more cooperation opportunities and value for upstream and downstream companies.

[0118] (5) Significant social and environmental benefits.

[0119] Improve service quality: Digital water supply scheduling directly improves residents’ water use experience and service quality by optimizing water supply pressure.

[0120] Promoting sustainable development: The application of digital water supply scheduling in this embodiment not only improves the level of intelligence in the water industry, but also promotes the conservation and recycling of water resources, which is in line with the sustainable development strategy.

[0121] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0122] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A water supply digital scheduling method based on water volume prediction and water storage regulation, characterized in that: The water supply digital scheduling method based on water quantity prediction and water storage regulation comprises the following steps: Create a scheduling scenario model and obtain historical data of the target water plant; the historical data includes historical water consumption data, historical factory pressure data, historical water pump operation data, and historical pipe network pressure data and historical pressure drop data of each pipe network pressure measurement point in the water supply area of ​​the target water plant; the historical water consumption data includes historical factory instantaneous flow data counted by minute, hour, week, holiday, month, quarter, temperature and weather dimensions; the historical pressure drop data is the difference between the historical factory pressure data at the same time and the historical pipe network pressure data of each pipe network pressure measurement point; Determine the terminal unfavorable point in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable point according to the historical water use data and the historical pressure drop data; Constructing a training sample set according to the historical water use data and the historical pressure drop data of the terminal unfavorable point; A water consumption prediction model and a pressure drop fitting model are established respectively, and the water consumption prediction model and the pressure drop fitting model are trained respectively using the training sample set to obtain a trained water consumption prediction model and a trained pressure drop fitting model; wherein the water consumption prediction model refers to a model established based on the historical water consumption data and using a sliding window mean filtering algorithm to predict the water demand in a future period of time, the input of the water consumption prediction model is the historical water consumption data, minutes, hours, weeks, holidays, months, quarters, temperature and weather, and the output of the water consumption prediction model is the water demand in the future period of time; the pressure drop fitting model is a model established based on the historical water consumption data and the historical pressure drop data of each terminal unfavorable point for fitting the relationship between water consumption and the pressure drop of each terminal unfavorable point, the input of the pressure drop fitting model is the historical water consumption data and the historical pressure drop data of each terminal unfavorable point, and the output of the pressure drop fitting model is the fitting relationship between water consumption and the pressure drop of each terminal unfavorable point; Based on the trained water consumption prediction model and the trained pressure drop fitting model, and according to the historical data, a water supply scheduling plan for the water plant in the future period is formulated; the water supply scheduling plan for the water plant in the future period includes a water plant outlet pressure scheduling plan for the future period, a water plant water production plan for the future period, and a water plant water pump scheduling plan for the future period; The real-time instantaneous outflow flow data of the target water plant is obtained, and a real-time water supply scheduling plan for the water plant is formulated according to the real-time instantaneous outflow flow data, the historical data and the water supply scheduling plan for the water plant in the future period; the real-time water supply scheduling plan for the water plant includes a real-time water plant outflow pressure scheduling plan and a real-time water plant water pump scheduling plan.

2. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 1 is characterized in that: Determining the terminal unfavorable points in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable points according to the historical water consumption data and the historical pressure drop data, specifically includes: According to the historical water use data and the historical pressure drop data, the sensitivity of each pipe network pressure measurement point to the factory pressure is analyzed, and combined with the geographical coordinates of each pipe network pressure measurement point and the location of users whose number of complaints about water pressure is greater than a threshold, some pressure measurement points are selected from each pipe network pressure measurement point as terminal unfavorable points; By adopting the confidence interval method, the upper and lower boundaries of a certain range of confidence intervals are taken for the historical pipe network pressure data of the selected terminal unfavorable points as the pressure control range of the terminal unfavorable points; the pressure control range of the terminal unfavorable points supports manual adjustment.

3. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 2 is characterized in that: Select some pressure measurement points from each pipe network pressure measurement point as terminal unfavorable points, including: Creating a scheduling scenario based on the historical water consumption data, the historical factory pressure data, and the historical pipe network pressure data of each pipe network pressure measurement point within the water supply area of ​​the target water plant, and performing scenario configuration on the scheduling scenario to determine the scenario configuration information corresponding to the scheduling scenario; According to the scenario configuration information, a terminal unfavorable point selection method is adopted to select some pressure measurement points from each pipe network pressure measurement point to obtain each terminal unfavorable point; The terminal unfavorable point selection method includes a first terminal unfavorable point selection method, a second terminal unfavorable point selection method or a third terminal unfavorable point selection method; wherein, the first terminal unfavorable point selection method is a sensitivity analysis method, including Pearson correlation coefficient, Spearman correlation coefficient and Kendall correlation coefficient method, the output of the sensitivity analysis method is the sensitivity of the pressure of each pipe network pressure measurement point to the factory pressure, and the selected terminal unfavorable point is a manually selected pipe network pressure measurement point with a sensitivity range of 40% to 80%; the second terminal unfavorable point selection method is to manually select terminal points according to the geographical coordinate position and elevation information of each pipe network pressure measurement point; the third terminal unfavorable point selection method is to investigate the location of users who have complained to the water plant about water pressure in a period of history and the number of complaints counted by location, and select pipe network pressure measurement points that are closer to the geographical coordinate position of the location of users whose number of water pressure complaints is greater than the threshold; According to the historical water use data and the historical pressure drop data of the terminal unfavorable point, a training sample set is constructed. The construction of the training sample set specifically includes: Calculate the difference between the historical factory pressure data and the historical pipe network pressure data of the terminal unfavorable point to obtain the historical pressure drop data of the terminal unfavorable point; Data analysis and data cleaning are performed on the historical water usage data and the historical pressure drop data of the terminal unfavorable points to generate training samples and obtain a training sample set.

4. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 1 is characterized in that: Based on the trained water consumption prediction model and the trained pressure drop fitting model, and according to the historical data, a water supply scheduling plan for the water plant in the future is formulated, specifically including: Generate a water plant delivery pressure scheduling plan for a period of time in the future based on the water demand predicted by the trained water demand prediction model, the fitting relationship between the water demand and the pressure drop of each terminal unfavorable point fitted by the trained pressure drop fitting model, and the pressure control range of the terminal unfavorable point; According to the water demand in the future period and the target liquid level setting of the water reservoir of the target water plant, an optimization algorithm is used to generate a water production plan for the water plant in the future period; the water production plan for the water plant in the future period refers to taking full advantage of the electricity price during the low electricity fee period according to the characteristics of peak and valley electricity prices, pre-filling the water reservoir according to the water demand of the target water plant in the future period during the low electricity fee period, reducing the water storage in the water reservoir during the high electricity fee period, alleviating the daytime peak water supply pressure and reducing the water production electricity fee; Based on the water demand in the future period, the water plant's factory pressure in the future period, historical water use data, historical factory pressure data and historical water pump operation data, a decision tree algorithm is used to generate a water plant pump scheduling plan for the future period; the water plant pump scheduling plan for the future period refers to a water supply pump start-stop combination plan that is generated based on historical water pump operation data and water supply pump operation parameters and can meet the water demand in the future period and the factory pressure in the future period.

5. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 4 is characterized in that: The water production plan for the water plant in the future period refers to a plan in which the water production volume is reduced between 6-24 o'clock and the water production volume is increased between 0-6 o'clock in the next 24 hours. At the same time, the liquid level in the clear water tank of the target water plant does not bottom out and trigger a low-level alarm or a high-level overflow alarm.

6. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 1 is characterized in that: The sliding window mean filtering algorithm is used for a water supply time series data Q1, Q2, ..., Q n , let i be the corresponding sequence of the same Sunday, the same holiday, the same hour and the same weather, i = 1, 2, 3, ..., k; α i For each sequence, the time constant is the corresponding time constant, and the water supply corresponding to time t is expressed by the following formula: Among them, n i The window size used to calculate the average for each day of the week, holiday, hour, and weather factor; To calculate α backward from time t i j is the weight corresponding to the factors of same Sunday, same holiday, same hour and same weather; P i The weight ratio of each factor on the same Sunday, holiday, hour and weather; Represents the backward calculation of α from time t i The water supply data corresponding to the j time period, is with The related median filter function, and The relationship is expressed as: Among them, m is the filter window length, m is an even number, Q lmin and Q lmax They are the minimum and maximum values ​​of the normal range of water supply respectively.

7. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 1 is characterized in that: After the steps of respectively establishing a water consumption prediction model and a pressure drop fitting model, and respectively training the water consumption prediction model and the pressure drop fitting model using the training sample set to obtain a trained water consumption prediction model and a trained pressure drop fitting model, the water supply digital scheduling method based on water quantity prediction and water storage regulation further includes: Using MSE, MAE, RMSE and R 2 Evaluation indicators, respectively performing model evaluation on the trained water consumption prediction model and the trained pressure drop fitting model to obtain model evaluation results; According to the model evaluation results, the optimal water consumption prediction model and the optimal pressure drop fitting model are respectively determined; the optimal water consumption prediction model is used as the trained water consumption prediction model to predict the water demand in the future period, and the optimal pressure drop fitting model is used as the trained pressure drop fitting model to fit the fitting relationship between the water demand in the future period output by the optimal water consumption prediction model and the pressure drop of each terminal unfavorable point; Based on the water demand in the future predicted by the optimal water consumption prediction model, the fitting relationship between the water demand of the target water plant in the future output by the optimal water consumption prediction model and the pressure drop of each terminal unfavorable point fitted by the optimal pressure drop fitting model, and combined with the pressure control range of the terminal unfavorable point, a water plant outlet pressure scheduling plan for the future is generated.

8. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 1 is characterized in that: Acquire the real-time instantaneous outflow flow data of the target water plant, and formulate a real-time water supply scheduling plan for the water plant according to the real-time instantaneous outflow flow data, the historical data and the water supply scheduling plan for the water plant in the future period of time, specifically including: Obtaining real-time instantaneous outflow flow data of the target water plant; The real-time instantaneous outflow data and the water demand in the future predicted by the trained water consumption prediction model are used as water consumption data according to weights, and the water demand forecast in the future is dynamically corrected and the water supply is adjusted in combination with the fitting relationship between the water demand data fitted by the trained pressure drop fitting model and the pressure drop of each terminal unfavorable point and the pressure control range of the terminal unfavorable point, so as to adjust the water supply scheduling plan of the water plant in the future in real time, optimize the operation strategy of the target water plant, and generate a real-time water plant outflow pressure scheduling plan; The real-time instantaneous flow data out of the factory and the real-time water plant pressure scheduling plan are used as water demand data and factory pressure data, combined with the historical water use data, the historical factory pressure data and the historical water pump operation data, and a decision tree algorithm is used to generate a real-time water plant pump scheduling plan.

9. The water supply digital scheduling method based on water quantity prediction and water storage regulation according to claim 1 is characterized in that: The pressure drop fitting model is a BP neural network model.

10. A water supply digital scheduling system based on water volume prediction and water storage regulation, characterized in that: The water supply digital scheduling system based on water volume prediction and water storage regulation supports local deployment and cloud deployment; the water supply digital scheduling system based on water volume prediction and water storage regulation implements the water supply digital scheduling method based on water volume prediction and water storage regulation according to any one of claims 1 to 9; The water supply digital scheduling system based on water volume prediction and water storage regulation specifically includes: A scenario generation module is used to create a scheduling scenario model and obtain historical data of the target water plant; the historical data includes historical water consumption data, historical factory pressure data, historical water pump operation data, and historical pipe network pressure data and historical pressure drop data of each pipe network pressure measurement point in the water supply area of ​​the target water plant; the historical water consumption data includes historical factory instantaneous flow data counted according to minute, hour, week, holiday, month, quarter, temperature and weather dimensions; the historical pressure drop data is the difference between the historical factory pressure data at the same time and the historical pipe network pressure data of each pipe network pressure measurement point; A data analysis module is used to determine the terminal unfavorable points in each pipe network pressure measurement point and the pressure control range of the terminal unfavorable points according to the historical water use data and the historical pressure drop data; a training sample set is constructed according to the historical water use data and the historical pressure drop data of the terminal unfavorable points; the data analysis module also includes a regularity analysis of water consumption and an adaptability analysis of the factory pressure and the water consumption regularity. The regularity analysis of water consumption refers to analyzing the historical data of water consumption according to the analysis dimensions of minutes, hours, weeks, months, quarters, holidays, temperature and weather, and providing the analysis results of the overall water consumption regularity, water consumption peaks and valleys, and the ranking of the influence of each analysis dimension on the water consumption regularity. The adaptability analysis of the factory pressure and the water consumption regularity refers to analyzing the matching degree of the factory pressure and the water use regularity, as well as the statistics of the maximum, minimum and average values ​​of the time period pressure at the terminal points of the pipe network; A modeling and evaluation module, for respectively establishing a water consumption prediction model and a pressure drop fitting model, and using the training sample set to respectively train the water consumption prediction model and the pressure drop fitting model to obtain a trained water consumption prediction model and a trained pressure drop fitting model; wherein the water consumption prediction model refers to a model established based on the historical water consumption data and using a sliding window mean filtering algorithm to predict the water demand in a future period of time, the input of the water consumption prediction model is the historical water consumption data, minutes, hours, weeks, holidays, months, quarters, temperature and weather, and the output of the water consumption prediction model is the water demand in the future period of time; the pressure drop fitting model is a model established based on the historical water consumption data and the historical pressure drop data of each terminal unfavorable point for fitting the relationship between water consumption and the pressure drop of each terminal unfavorable point, the input of the pressure drop fitting model is the historical water consumption data and the historical pressure drop data of each terminal unfavorable point, and the output of the pressure drop fitting model is the fitting relationship between water consumption and the pressure drop of each terminal unfavorable point; The model publishing module is used to deploy the model to a local server or a cloud server based on the trained water consumption prediction model and the trained pressure drop fitting model, and formulate a water supply scheduling plan for the water plant in the future period according to the historical data; the water supply scheduling plan for the water plant in the future period includes a water plant ex-factory pressure scheduling plan for the future period, a water plant water production plan for the future period, and a water plant water pump scheduling plan for the future period; obtain the real-time ex-factory instantaneous flow data of the target water plant, and formulate a real-time water supply scheduling plan for the water plant according to the real-time ex-factory instantaneous flow data, the historical data, and the water supply scheduling plan for the water plant in the future period; the real-time water supply scheduling plan for the water plant includes a real-time water plant ex-factory pressure scheduling plan and a real-time water plant water pump scheduling plan.

Images