Intelligent water division control method based on time series prediction

Abstract

The application discloses a smart water distribution control method based on time series prediction, and belongs to the technical field of urban water supply control. The application aims at the problems that the existing water supply pressure and water quantity control relies on the change of pipeline diameter, the construction is difficult, and the sudden water use condition cannot be coped with. The application comprises the following steps: collecting average water consumption data and average water pressure data of each household per hour; identifying abnormal values based on quartiles, and adjusting the abnormal values into normal values; obtaining original data of daily hourly average water consumption and original data of average water pressure of each household in each season of the adjacent previous year from the adjusted data; simultaneously obtaining reference data of daily hourly average water consumption and reference data of average water pressure of each household in the current season of the current year; obtaining predicted data of daily hourly average water consumption and predicted data of average water pressure of each household in the current year the next day based on the original data and the reference data by adopting a SARIMA model; and adjusting the valve of a water distributor according to the predicted data. The application is used for urban water supply distribution control.

Description

Technical Field

[0001] The present invention relates to an intelligent water distribution control method based on time series prediction, and belongs to the technical field of urban water supply control. Background Art

[0002] In urban water supply systems, water supply pipes typically need to pass through residential areas and enter various buildings. Inside the building, the water supply pipes are first distributed to each floor through a riser. The water supply system on each floor distributes water to each household through a manifold. However, due to differences in water use habits and sanitary fixtures, different households have different water consumption and peak water use, which can lead to certain problems in this water distribution method. Specifically, when residents on the ground floor have no water demand during a certain period of time, residents on higher floors may face increased water demand, resulting in excessive water pressure on the ground floor and insufficient water pressure on higher floors. This uneven water pressure distribution not only affects the daily quality of life of residents but also places higher demands on the water distribution system of the manifold. Therefore, smart manifolds are gradually gaining attention.

[0003] Currently, numerous technologies have made significant progress in long-distance data transmission. For example, the use of multimodal sensors to monitor and transmit water pressure and volume in each household in real time can provide a wealth of environmental data and water usage information. Furthermore, as various machine learning algorithms mature, their application in processing large-scale, multi-dimensional data is becoming increasingly widespread. Machine learning algorithms can extract valuable insights and patterns from massive amounts of data.

[0004] However, existing remote monitoring devices are primarily used to count water consumption by individual users for water bill calculation. Traditional methods for regulating water pressure and flow often rely on changing pipe diameters. This approach requires detailed analysis of each household's water usage and the re-laying of pipes, which is not only complex but also challenging to implement. Furthermore, when water usage is abnormal, traditional methods can lead to safety hazards such as pipe bursts, compromising the overall stability and reliability of the water supply system. Summary of the Invention

[0005] In view of the problems that the existing water supply pressure and water volume control rely on changes in pipe diameter, are difficult to construct and cannot cope with sudden water use situations, the present invention provides an intelligent water distribution control method based on time series prediction.

[0006] The present invention provides an intelligent water distribution control method based on time series prediction, comprising:

[0007] Multimodal sensors are used to collect average water consumption and water pressure data for each household every hour. Outliers are identified based on quartiles and adjusted to normal values.

[0008] The average water consumption data and average water pressure data of all households in the previous year after outlier adjustment are used as the original data. The original data of daily and hourly average water consumption and average water pressure of each household in each season of the previous year are calculated based on the original data of each season. The reference data of daily and hourly average water consumption and average water pressure of each household in the current season of the year are calculated based on the average water consumption data and average water pressure data of each household in each hour of the dates that have occurred in the current season of the year after outlier adjustment.

[0009] SARIMA model is used to obtain the predicted data of average hourly water consumption and average water pressure for each household on the second day of the current year based on the original data of average hourly water consumption and average water pressure for each household in each season of the previous year and the reference data of average hourly water consumption and average water pressure for each household in the current season.

[0010] The valve of the water distributor is adjusted based on the predicted data of average hourly water consumption and average water pressure for each household on the second day of the year to achieve intelligent water distribution control.

[0011] According to the intelligent water distribution control method based on time series prediction of the present invention, the method for identifying outliers based on quartiles includes:

[0012] The average water consumption data and average water pressure data of each household in each hour of each season collected by multimodal sensors are sorted from small to large by hour to obtain hourly time series data;

[0013] The first quartile Q1, the third quartile Q3 and the interquartile range IQR are determined from the hourly time series data;

[0014] Determine the lower limit of the average water consumption and average water pressure data for each household per hour as Q1-1.5*IQR; the upper limit is Q3+1.5*IQR; where IQR=Q3-Q1;

[0015] The data of average water consumption or average water pressure per household per hour that exceeds the corresponding lower and upper limits are regarded as outliers.

[0016] According to the intelligent water distribution control method based on time series prediction of the present invention, the method for adjusting abnormal values ​​to normal values ​​is as follows:

[0017] The abnormal values ​​are replaced by the average of all normal values ​​in the average water consumption data or average water pressure data of each household per hour on different days of the same season, and the abnormal values ​​are adjusted to normal values.

[0018] According to the intelligent water diversion control method based on time series prediction of the present invention, the model parameters of the SARIMA model include seasonal cycle, difference order, regression parameter and moving average order;

[0019] The model parameters are determined by autocorrelation graphs and partial autocorrelation graphs, which are obtained by deep learning or grid search.

[0020] According to the intelligent water distribution control method based on time series prediction of the present invention, the method for obtaining the hourly average water consumption forecast data and average water pressure forecast data for each household on the second day of the year using the SARIMA model is as follows:

[0021] The original data of daily and hourly average water consumption and average water pressure of each household in each season of the previous year and the reference data of daily and hourly average water consumption and average water pressure of each household in the current season are input into the SARIMA model to make the initial prediction of the average water consumption data and average water pressure data;

[0022] If the initial forecast result is within the fluctuation range of the box plot generated by the historical data, the initial forecast result will be used as the forecast data for the average hourly water consumption and average water pressure per household on the second day of the year;

[0023] Otherwise, set the input data threshold, delete the data that exceeds the input data threshold in the SARIMA model, input the retained data into the SARIMA model again, and re-forecast the average water consumption data and average water pressure data until the forecast result is within the fluctuation range of the box plot generated by the historical data. The current forecast result is used as the forecast data of the average water consumption and average water pressure per household per hour on the second day of the year.

[0024] According to the intelligent water distribution control method based on time series prediction of the present invention, if the average hourly water consumption forecast for each household on the second day of the year is greater than the average hourly water consumption for each household on the same day, the water distributor valve opening is increased accordingly, thereby increasing the user's water inlet pressure to the average water pressure forecast data;

[0025] If the average hourly water consumption forecast for each household on the second day of the year is less than the average hourly water consumption for each household on the same day, the water distributor valve opening will be reduced accordingly, thereby reducing the user's water inlet pressure to the average water pressure forecast data.

[0026] According to the intelligent water distribution control method based on time series prediction of the present invention, a controller calculates original data and reference data based on the collected average water consumption data and average water pressure data for each household per hour. The SARIMA model is then used to predict the average water consumption and average water pressure forecast data for each household per hour on the second day of the same year. The control instructions for the water distribution valves are calculated based on the average water consumption and average water pressure forecast data for each household per hour on the second day of the same year and the current valve status of each household's water distribution device.

[0027] According to the intelligent water distribution control method based on time series prediction of the present invention, an actuator is used to adjust the opening of the water distributor valve based on the control instruction of the water distributor valve.

[0028] According to the intelligent water distribution control method based on time series prediction of the present invention, the controller, multimodal sensor and actuator transmit data based on a communication network; the communication network's communication protocols include industrial Ethernet, field bus and wireless network.

[0029] According to the intelligent water distribution control method based on time series prediction of the present invention, the controller is powered by an uninterruptible power supply UPS and a backup power supply.

[0030] Beneficial effects of the present invention: The present invention controls the water diversion of urban water by controlling the water diverter valve based on historical water consumption data, improves the water diversion efficiency of the water diverter by online control of the water diverter valve, and can respond to various water use emergencies by adjusting the opening of the water diverter valve, reducing the possibility of danger and ensuring the overall stability and reliability of the water supply system.

[0031] The present invention realizes intelligent water distribution in urban water supply, can optimize water volume scheduling, effectively improve the intelligence level of the water supply system, and optimize the allocation of water supply resources.

[0032] The method proposed in this paper predicts water consumption data on an hourly basis, improving the decision-making and prediction efficiency of smart water distribution. Outliers are generally caused by abnormal user water usage and sensor damage. The method proposed in this paper identifies and adjusts outliers, replacing them with the average value of normal water consumption. This improves prediction accuracy and further enhances the prediction efficiency of smart water distribution. By recycling data at hourly intervals, the prediction model can be continuously optimized, further improving the accuracy of water volume prediction and the overall efficiency of water distribution control. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of the intelligent water distribution control method based on time series prediction described in the present invention. DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0035] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0036] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0037] Specific implementation method 1. Combination Figure 1 As shown, the present invention provides an intelligent water distribution control method based on time series prediction, comprising:

[0038] Multimodal sensors are used to collect average water consumption and water pressure data for each household every hour. Outliers are identified based on quartiles and adjusted to normal values.

[0039] The average water consumption data and average water pressure data of all households in the previous year after outlier adjustment are used as the original data. The original data of daily and hourly average water consumption and average water pressure of each household in each season of the previous year are calculated based on the original data of each season. The reference data of daily and hourly average water consumption and average water pressure of each household in the current season of the year are calculated based on the average water consumption data and average water pressure data of each household in each hour of the dates that have occurred in the current season of the year after outlier adjustment.

[0040] SARIMA model is used to obtain the predicted data of average hourly water consumption and average water pressure for each household on the second day of the current year based on the original data of average hourly water consumption and average water pressure for each household in each season of the previous year and the reference data of average hourly water consumption and average water pressure for each household in the current season.

[0041] The valve of the water distributor is adjusted based on the predicted data of average hourly water consumption and average water pressure for each household on the second day of the year to achieve intelligent water distribution control.

[0042] The data collected by the multimodal sensors can be transmitted to the controller for data analysis and processing. The actual sensor data corresponding to the average water consumption and average water pressure forecast data obtained for each prediction is continuously stored for subsequent predictions.

[0043] The multimodal sensor uses a highly sensitive, multimodal water flow and pressure sensor to ensure measurement accuracy and data transmission stability. The measurement process is based on hourly time periods. The collected water flow data is processed to calculate the average water flow rate within each time period to obtain the hourly average flow rate.

[0044] The measurement data in this embodiment is in hours, for example, average water consumption data within one hour between one o'clock and two o'clock.

[0045] Furthermore, methods for identifying outliers based on quartiles include:

[0046] The average water consumption data and average water pressure data of each household in each hour of each season collected by multimodal sensors are sorted from small to large by hour to obtain hourly time series data;

[0047] The first quartile Q1, the third quartile Q3 and the interquartile range IQR are determined from the hourly time series data;

[0048] Determine the lower limit of the average water consumption and average water pressure data for each household per hour as Q1-1.5*IQR; the upper limit is Q3+1.5*IQR; where IQR=Q3-Q1;

[0049] The data of average water consumption or average water pressure per household per hour that exceeds the corresponding lower and upper limits are regarded as outliers.

[0050] The first quartile Q1 corresponds to the 25th percentile of the time series data, and the third quartile Q3 corresponds to the 75th percentile of the time series data; the gap between the third quartile and the first quartile is called the interquartile range.

[0051] Time series data analysis encompasses a range of methods used to analyze time series data, aiming to extract meaningful statistical information from the data and make predictions about the future. In this implementation, flow data is used for time series analysis. This data is collected from users' hourly water usage throughout the day, and the data for the same time period is analyzed separately.

[0052] In this embodiment, the method for adjusting abnormal values ​​to normal values ​​is:

[0053] The abnormal values ​​are replaced by the average value of all normal values ​​in the average water consumption data or average water pressure data of each household per hour on different days of the same season, and the abnormal values ​​are adjusted to normal values ​​to ensure the continuity and rationality of the data.

[0054] In this embodiment, the calculation and prediction of data are performed based on data from different dates in the same season and the same time period.

[0055] Considering the seasonal cyclical changes in user water consumption, the SARIMA algorithm is used to predict future water consumption, which can support time series data with seasonal components. Four seasonal parameters can be added to the existing SARIMA model. The model parameters of the SARIMA model include seasonal period, difference order, regression parameter, and moving average order.

[0056] The model parameters are determined using autocorrelation plots and partial autocorrelation plots, which are obtained using deep learning or grid search. Cross-validation is used to evaluate the performance of each combination, ultimately selecting the optimal parameters to improve the accuracy of decision-making predictions and further enhance the predictive efficiency of smart water distribution. This implementation collects and predicts water usage data at hourly intervals, enabling data recycling and continuous optimization of the prediction model, further improving the accuracy of water volume predictions and the overall efficiency of the system.

[0057] The autocorrelation graph and partial autocorrelation graph are drawn using the autocorrelation and partial autocorrelation functions, and the parameter size is selected based on the tailing or truncation characteristics.

[0058] Furthermore, the SARIMA model obtains the hourly average water consumption and average water pressure forecast data for each household on the second day of the year as follows:

[0059] The original data of daily and hourly average water consumption and average water pressure of each household in each season of the previous year and the reference data of daily and hourly average water consumption and average water pressure of each household in the current season are input into the SARIMA model to make the initial prediction of the average water consumption data and average water pressure data;

[0060] If the initial forecast result is within the fluctuation range of the box plot generated by the historical data, the initial forecast result will be used as the forecast data for the average hourly water consumption and average water pressure per household on the second day of the year;

[0061] Otherwise, set the input data threshold, delete the data that exceeds the input data threshold in the SARIMA model, input the retained data into the SARIMA model again, and re-forecast the average water consumption data and average water pressure data until the forecast result is within the fluctuation range of the box plot generated by the historical data. The current forecast result is used as the forecast data of the average water consumption and average water pressure per household per hour on the second day of the year.

[0062] The box plot is drawn based on the quartile principle and is generated through historical data, and the boundaries represent the normal fluctuation range of the data.

[0063] In this implementation, hourly water usage is collected throughout the day, and data from the same time period is analyzed separately. Single-step forecasting, using historical data to predict future values, ensures accurate predictions. Actual daily water usage values ​​are then incorporated into the data to predict the next day's value, significantly improving prediction reliability.

[0064] If the prediction results are within the box plot boundaries, the prediction is good and the forecast process is complete. If the prediction results are not within the boundaries, it is considered that there may be abnormal data or the model parameters need to be adjusted. For prediction results that are not within the boundaries, data adjustment and re-forecasting are required. Data with abnormal water consumption (excessive or insufficient) on certain days are eliminated and the remaining data is retrained. The SARIMA model is rebuilt and forecasts are performed.

[0065] Forecast data that exceeds the time limit can be removed as invalid data.

[0066] The intelligent water distribution control of this embodiment can realize personalized services and decision-making services for users, thereby improving the service efficiency and decision-making prediction efficiency of water distribution.

[0067] The steps for identifying, fitting, and testing the SARIMA model are as follows:

[0068] 1) Observe the sequence diagram of the original time series and analyze the basic change characteristics;

[0069] 2) The observation results are converted into correlation diagrams and partial correlation diagrams of stationary series by difference, so as to identify and determine the SARIMA model parameters;

[0070] 3) Estimate model parameters using the maximum likelihood estimation method for the initially established model;

[0071] 4) Use the established model to predict future water consumption data.

[0072] Users' water use habits are related to the season, and the SARIMA model is highly sensitive to periodic properties such as seasonal changes in time series data. The method of the present invention adopts a single-step prediction method, adding the actual water use value of each day to the training set to predict the value of the next day. Only one day is predicted, which greatly improves reliability. For the prediction of water consumption in the first few days of each quarter, historical real data from the same quarter and time point in different years can be added as a training set for training. This increases the richness of the training set and can improve the accuracy of the prediction.

[0073] If the average hourly water consumption forecast for each household on the second day of the year is greater than the average hourly water consumption for each household on the same day, the manifold valve opening will be increased accordingly, thereby increasing the user's water inlet pressure to the average water pressure forecast data;

[0074] If the average hourly water consumption forecast for each household on the second day of the year is less than the average hourly water consumption for each household on the same day, the water distributor valve opening will be reduced accordingly, thereby reducing the user's water inlet pressure to the average water pressure forecast data.

[0075] The opening degree of the water distributor valve is directly related to the user's water inlet pressure. By predicting the water consumption once an hour, the opening degree of the valve can be accurately controlled.

[0076] Furthermore, a controller is used to calculate the original data and reference data based on the collected average water consumption data and average water pressure data for each household per hour; and the SARIMA model is called to predict the average water consumption forecast data and average water pressure forecast data for each household per hour on the second day of the year. Based on the average water consumption forecast data and average water pressure forecast data for each household per hour on the second day of the year and the current valve status of each household's water distributor, calculations are performed to obtain control instructions for the water distributor valves.

[0077] The data collected by multimodal sensors is converted into electrical signals for processing by a controller. As the core component of data processing, the controller receives sensor data and performs calculations and decisions based on predefined control algorithms. The controller can be implemented as a programmable logic controller (PLC), allowing users to program control logic using a programming language.

[0078] The actuator is used to adjust the opening of the manifold valve based on the control command of the manifold valve. The actuator converts the control signal into physical action.

[0079] The controller, multimodal sensors and actuators transmit data based on the communication network to ensure the coordination between various links; the communication network's communication protocols include industrial Ethernet, fieldbus and wireless network.

[0080] In this embodiment, wireless transmission technology can be used for data transmission to ensure reliability and real-time data transmission, avoiding lags in monitoring information due to transmission delays or data loss. Data can be transmitted once an hour. The transmitted data is used to prepare water consumption forecasts for the same time period the next day.

[0081] This embodiment may also include a human-machine interface (HMI): a graphical interface for operators to monitor and control the system, providing real-time data, alarm information, and operation instruction input functions.

[0082] It can also include data storage and management: storage of system operation data, control history, alarm records, etc. This facilitates later analysis, fault diagnosis and system optimization.

[0083] In order to ensure stable power supply of the system and avoid system failure due to power supply problems, the controller is powered by an uninterruptible power supply UPS and a backup power supply.

[0084] The method of the present invention can be implemented through the following modules:

[0085] Data receiving module: used to receive real-time data transmitted by multi-analog sensors;

[0086] Data analysis module: used to analyze the data from the data receiving module and identify abnormal values ​​using the box plot method;

[0087] Data correction module: used to correct abnormal values, the data used is the average value of the same season and time period;

[0088] Model building module: used to build time series models, using the SMRIMA algorithm;

[0089] Result prediction module: used to import user data into the model and output prediction results;

[0090] Decision execution module: used to control the valve of the water distributor according to the prediction results.

[0091] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in conjunction with other described embodiments.

Claims

1. An intelligent water distribution control method based on time series prediction, characterized in that: include: Multimodal sensors are used to collect average water consumption and average water pressure data for each household every hour; The average water consumption data and average water pressure data of each household were used to identify outliers based on quartiles and adjusted to normal values; The average water consumption data and average water pressure data of all households in the previous year after outlier adjustment are used as the original data. The original data of daily and hourly average water consumption and average water pressure of each household in each season of the previous year are calculated based on the original data of each season. The reference data of daily and hourly average water consumption and average water pressure of each household in the current season of the year are calculated based on the average water consumption data and average water pressure data of each household in each hour of the dates that have occurred in the current season of the year after outlier adjustment. SARIMA model is used to obtain the predicted data of average hourly water consumption and average water pressure for each household on the second day of the current year based on the original data of average hourly water consumption and average water pressure for each household in each season of the previous year and the reference data of average hourly water consumption and average water pressure for each household in the current season. The valves of the water distributor are adjusted based on the predicted average hourly water consumption and average water pressure of each household on the second day of the year, achieving intelligent water distribution control. If the average hourly water consumption forecast for each household on the second day of the year is greater than the average hourly water consumption for each household on the same day, the manifold valve opening will be increased accordingly, thereby increasing the user's water inlet pressure to the average water pressure forecast data; If the average hourly water consumption forecast for each household on the second day of the year is less than the average hourly water consumption for each household on the same day, the water distributor valve opening will be reduced accordingly, thereby reducing the user's water inlet pressure to the average water pressure forecast data.

2. The intelligent water distribution control method based on time series prediction according to claim 1 is characterized in that: Methods for outlier identification based on quartiles include: The average water consumption data and average water pressure data of each household in each hour of each season collected by multimodal sensors are sorted from small to large by hour to obtain hourly time series data; Determine the first quartile from hourly time series data , third quartile and interquartile range ; Determine the lower limit of the average water consumption data and average water pressure data for each household per hour ; The upper limit is ;in ; The data of average water consumption or average water pressure per household per hour that exceeds the corresponding lower and upper limits are regarded as outliers.

3. The intelligent water distribution control method based on time series prediction according to claim 2 is characterized in that: The method to adjust outliers to normal values ​​is: The abnormal values ​​are replaced by the average of all normal values ​​in the average water consumption data or average water pressure data of each household per hour on different days of the same season, and the abnormal values ​​are adjusted to normal values.

4. The intelligent water distribution control method based on time series prediction according to claim 3 is characterized in that: The model parameters of the SARIMA model include seasonal period, difference order, regression parameter and moving average order; The model parameters are determined by autocorrelation graphs and partial autocorrelation graphs, which are obtained by deep learning or grid search.

5. The intelligent water distribution control method based on time series prediction according to claim 3 is characterized in that: The method for obtaining the hourly average water consumption and average water pressure forecast data for each household on the second day of the year using the SARIMA model is as follows: The original data of daily and hourly average water consumption and average water pressure of each household in each season of the previous year and the reference data of daily and hourly average water consumption and average water pressure of each household in the current season are input into the SARIMA model to make the initial prediction of the average water consumption data and average water pressure data; If the initial forecast result is within the fluctuation range of the box plot generated by the historical data, the initial forecast result will be used as the forecast data for the average hourly water consumption and average water pressure per household on the second day of the year; Otherwise, set the input data threshold, delete the data that exceeds the input data threshold in the SARIMA model, input the retained data into the SARIMA model again, and re-forecast the average water consumption data and average water pressure data until the forecast result is within the fluctuation range of the box plot generated by the historical data. The current forecast result is used as the forecast data of the average water consumption and average water pressure per household per hour on the second day of the year.

6. The intelligent water distribution control method based on time series prediction according to claim 5 is characterized in that: The controller calculates the original data and reference data based on the collected average water consumption data and average water pressure data of each household per hour; and calls the SARIMA model to predict the average water consumption forecast data and average water pressure forecast data of each household per hour on the second day of the year; based on the average water consumption forecast data and average water pressure forecast data of each household per hour on the second day of the year and the current valve status of each household's water distributor, calculations are performed to obtain the control instructions of the water distributor valve.

7. The intelligent water distribution control method based on time series prediction according to claim 6 is characterized in that: The actuator is used to adjust the opening of the water distributor valve based on the control instruction of the water distributor valve.

8. The intelligent water distribution control method based on time series prediction according to claim 7 is characterized in that: The controller, multimodal sensor and actuator transmit data based on the communication network; the communication protocols of the communication network include industrial Ethernet, field bus and wireless network.

9. The intelligent water distribution control method based on time series prediction according to claim 6 is characterized in that: The controller is powered by an uninterruptible power supply UPS and a backup power supply.

Images