Method and system for predicting water consumption of secondary water supply pool and regulation and storage method and system

Abstract

The invention belongs to the technical field of urban water supply, and relates to a secondary water supply pool water consumption prediction method and system and a secondary water supply pool water consumption regulation and storage method and system.The prediction method comprises the steps that environmental parameters and historical accumulated water consumption of water supply areas of all pools are obtained; missing time and date features are added into the historical environment parameters and the historical accumulated water consumption, a data set is formed after the accumulated water consumption is missed, and the data set is divided into a training set, a verification set and a test set; constructing a water consumption prediction model by taking historical water consumption, historical environmental parameters and predicted time point environmental parameters as input and taking predicted time point water consumption as output on the basis of a time series large model TimeGPT; setting parameters of the water consumption prediction model, and training and verifying the water consumption prediction model through the training set and the verification set; and inputting the test set into the water consumption prediction model to obtain the water consumption at the prediction time point. The regulation and storage method is based on predicted water consumption and real-time water supply pool water level. The water consumption can be accurately predicted, and water supply optimization regulation and storage are achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban water supply, and in particular, relates to a water consumption prediction method and system for a secondary water supply pool, and a water storage method and system. Background Art

[0002] With the rapid development of society and the continuous increase in water demand, urban water supply systems are facing increasingly severe challenges. Although the overall water supply service capacity of some cities can meet the regional water demand, water supply shortages are still common during peak water use periods, and even cause serious impacts on residents' lives. In addition, the urban water supply network has been in operation for many years. Due to aging and damage, the pipeline resistance coefficient has increased significantly, especially during peak water use periods, the pressure drop problem of the water supply network is more prominent. One of the root causes of these problems is that the regulation and storage capacity of the existing water supply system is limited, with surplus water supply during off-peak periods and insufficient water supply during peak periods.

[0003] At present, the urban water supply network mainly adopts two water supply methods: direct water supply from the municipal pipeline network and secondary regulation and pressurized water supply. Among them, the secondary regulation and pressurized water supply is mainly for users in medium and high-rise buildings. The most common form is the combined water supply of the pool and the variable frequency pump. Water supply forecasting is one of the core links of secondary regulation and pressurized water supply. The change of water supply is affected by many factors, including seasonal changes, weather conditions, holidays, emergencies, etc. The interaction between these factors makes the prediction of water supply particularly complicated. Traditional water supply forecasting methods such as linear regression often have poor prediction accuracy when facing complex and changeable water supply systems and cannot meet the needs of water supply systems.

[0004] Water tank regulation is another core link of secondary regulation and pressurized water supply. Due to the water storage function of the water tank, it theoretically has a high peak regulation capacity. However, according to research, in the existing secondary water supply system, the water tank is often in a "on-the-go" state where the water intake and water consumption are basically equal, and it is almost impossible to achieve effective regulation of water volume. This results in the water tank not only being unable to meet the regulation and storage needs during peak water use, but may also further aggravate the water supply pressure problem of the municipal pipe network, affecting the overall water supply efficiency and residents' water use experience. Summary of the invention

[0005] In view of the above-mentioned problems of poor water consumption prediction accuracy in the prior art, the present invention provides an agent-based secondary water supply pool water consumption prediction method and system, and a water supply and storage method and system, which can accurately predict water consumption, and scientifically and rationally perform dynamic optimization of water replenishment based on the predicted water consumption combined with the liquid level of the water supply pool, thereby achieving "peak shaving and valley filling" of water volume, reducing the load during the peak period of urban water use, and ensuring the stability and reliability of residents' water use.

[0006] In order to achieve the above object, the first aspect of the present invention provides a method for predicting water consumption of a secondary water supply pool, the steps of which are: Data acquisition steps: Obtain environmental parameters and historical cumulative water consumption of each water tank water supply area. Environmental parameters include environmental parameters at the prediction time point and historical environmental parameters; Missing filling steps: add missing time to historical environmental parameters and historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval; add date features to historical environmental parameters and historical water consumption, and convert them into digital date features through unique hot encoding; add missing cumulative water consumption to historical environmental parameters and historical water consumption; Dataset generation steps: The missing historical environmental parameters and historical cumulative water consumption are used to form a dataset, and the dataset is divided into a training set, a validation set, and a test set; Model construction steps: Based on the time series large model TimeGPT, the water consumption prediction model is constructed with historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as inputs and water consumption at the predicted time point as output; Model training and verification steps: set water consumption prediction model parameters, train the water consumption prediction model through the training set, and verify the trained water consumption prediction model through the verification set; Prediction step: Input the test set into the water consumption prediction model to obtain the water consumption at the predicted time point.

[0007] In some embodiments, in the missing filling step, the specific method of adding the missing cumulative water consumption is: deleting the erroneous data in the historical cumulative water consumption, and using the random forest algorithm to fill the missing cumulative water consumption of the historical cumulative water consumption.

[0008] In some embodiments, in the model training and verification steps, the specific method for verifying the water consumption prediction model is: using a rolling window method to evaluate the performance of the water consumption prediction model in different time periods. If the evaluation result fails, retrain the water consumption prediction model until the evaluation result passes.

[0009] In a second aspect of the present invention, a secondary water supply pool water consumption prediction system is provided, which is used to implement the secondary water supply pool water consumption prediction method described in the first aspect of the present invention, comprising: Data acquisition module, to obtain the environmental parameters, historical environmental parameters and historical cumulative water consumption of each pool water supply area at the predicted time point; The missing filling module adds missing time to the historical environmental parameters and historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval; adds date features to the historical environmental parameters and historical water consumption, and converts them into digital date features through unique hot encoding; adds missing cumulative water consumption to the historical environmental parameters and historical water consumption; The data set generation module forms a data set with the missing historical environmental parameters and historical cumulative water consumption, and divides the data set into a training set, a validation set, and a test set; The model building module is based on the time series large model TimeGPT, takes historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and takes the water consumption at the predicted time point as output to build a water consumption prediction model; Model training and verification module, setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set; The prediction module inputs the test set into the water consumption prediction model to obtain the water consumption at the predicted time point.

[0010] In some embodiments, a model optimization module is further included, wherein the model optimization module includes: A data monitoring module is used to monitor the actual water consumption at the predicted time point in real time; The new training set generation module adds the actual water consumption at the prediction time point monitored in real time and the environmental parameters at the prediction time point into the training set to form a new training set, so as to train the water consumption prediction model through the new training set and optimize the water consumption prediction model.

[0011] In a third aspect of the present invention, a method for regulating water supply in a secondary water supply pool is provided, the steps of which are as follows: Data acquisition steps: Obtain environmental parameters and historical cumulative water consumption of each water tank water supply area. Environmental parameters include environmental parameters at the prediction time point and historical environmental parameters; Missing filling step: adding missing time to the historical environmental parameters and historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval, and adding missing cumulative water consumption to the historical environmental parameters and historical water consumption; Dataset generation steps: The missing historical environmental parameters and historical cumulative water consumption are used to form a dataset, and the dataset is divided into a training set, a validation set, and a test set; Model construction steps: Based on the time series large model TimeGPT, the water consumption prediction model is constructed with historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as inputs and water consumption at the predicted time point as output; Model training and verification steps: set water consumption prediction model parameters, train the water consumption prediction model through the training set, and verify the trained water consumption prediction model through the verification set; Prediction step: input the test set into the water consumption prediction model to obtain the water consumption at the predicted time point; Liquid level measurement steps: Real-time measurement of water level in the water supply tank; Judgment steps: determine the water consumption period according to the predicted water consumption at the time point; determine whether the water level of the water supply pool during the peak water consumption period is lower than the set minimum water level threshold of the pool; Water replenishment adjustment steps: During the valley period and low peak period of water usage, replenish water into the water supply pool until the water level in the water supply pool reaches the set maximum water level of the pool, and stop replenishing water; during the peak water usage period, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, replenish water into the water supply pool.

[0012] In some embodiments, in the determination step, the water use period is determined based on a sliding window and a threshold determination algorithm, and the specific method is as follows: Parameter setting steps: Set the sliding window size to , the peak threshold is , the low peak threshold is ; Average water consumption calculation steps: The sliding window traverses the water consumption data sequence at the entire prediction time point. At each window position, the water consumption data in the sliding window is expressed as: ,

[0013] In the formula, Indicates A sliding window, Indicates Water consumption in a certain period of time; Calculate the average water usage for the period within the sliding window:

[0014] Water consumption period judgment step: If the average water consumption in the current sliding window , it enters the peak detection mode. In the peak detection mode, it moves to the next sliding window and calculates the average water consumption of the next sliding window. If the average water consumption of multiple consecutive sliding windows exceeds the peak threshold, the time corresponding to these sliding windows is regarded as a continuous water consumption peak period. If multiple consecutive water consumption peak periods are detected, these consecutive water consumption peak periods are merged into one water consumption peak period, and the previous one with the largest water consumption and the longest duration is selected. The peak period is the final water consumption peak period. ≥4; if the average water consumption in the current sliding window , it enters the low-peak detection mode. In the low-peak detection mode, it moves to the next sliding window and calculates the average water consumption of the next sliding window. If the average water consumption of multiple consecutive sliding windows is lower than the low-peak threshold, the time corresponding to these sliding windows is a continuous low-peak water consumption period. If multiple consecutive low-peak water consumption periods are detected, these consecutive low-peak water consumption periods are merged into one low-peak water consumption period, and the previous one with the largest water consumption and the longest duration is selected. The low-peak period is the final water consumption low-peak period. ≥4; the rest of the water usage period is the off-peak period.

[0015] In some embodiments, in the missing filling step, the specific method of adding the missing cumulative water consumption is: deleting the erroneous data in the historical cumulative water consumption, and using the random forest algorithm to fill the missing cumulative water consumption of the historical cumulative water consumption.

[0016] In some embodiments, in the model training and verification steps, the specific method for verifying the water consumption prediction model is: using a rolling window method to evaluate the performance of the water consumption prediction model in different time periods. If the evaluation result fails, retrain the water consumption prediction model until the evaluation result passes.

[0017] In a fourth aspect, the present invention provides a water supply and storage system for a secondary water supply tank, which is used to implement the water supply and storage method for a secondary water supply tank described in the third aspect of the present invention, comprising: Data acquisition module, to obtain the environmental parameters, historical environmental parameters and historical cumulative water consumption of each pool water supply area at the predicted time point; A missing filling module adds missing time to historical environmental parameters and historical cumulative water consumption, so that the time of the time series exists continuously according to the set time interval, and adds missing cumulative water consumption to historical environmental parameters and historical water consumption; The data set generation module forms a data set with the missing historical environmental parameters and historical cumulative water consumption, and divides the data set into a training set, a validation set, and a test set; The model building module is based on the time series large model TimeGPT, takes historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and takes the water consumption at the predicted time point as output to build a water consumption prediction model; Model training and verification module, setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set; The prediction module inputs the test set into the water consumption prediction model to obtain the water consumption at the predicted time point; The liquid level sensor is installed in the water supply tank and is used to measure the water level of the water supply tank in real time; The judgment module determines the water consumption period according to the predicted water consumption at the time point; determines whether the water level of the water supply pool during the peak water consumption period is lower than the set minimum water level threshold of the pool; The water replenishment regulation module replenishes water to the water supply pool during valley and off-peak periods until the water level in the water supply pool reaches the set maximum water level of the pool, and then stops replenishing water; during peak water usage periods, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, water is replenished to the water supply pool.

[0018] In some embodiments, the water replenishment regulation module includes: The water inlet valve is installed on the water inlet pipe of the water supply tank; The controller is connected to the water inlet valve. During the valley period and the low peak period, the water inlet valve is controlled to open to replenish water in the water supply pool until the water level in the water supply pool reaches the set maximum water level of the pool, and the water inlet valve is controlled to close to stop replenishing water. During the peak water usage period, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, the water inlet valve is controlled to open to replenish water in the water supply pool.

[0019] Compared with the prior art, the advantages and positive effects of the present invention are: (1) The water consumption prediction method and system for secondary water supply pools provided by the present invention builds a water consumption prediction model based on the time series large model TimeGPT, and performs training and verification. It uses the transfer learning ability of the time series large model TimeGPT and its adaptability to time series data, combined with water consumption data environmental parameters (such as temperature, humidity, precipitation, snowfall, solar radiation intensity, etc.) and date characteristics (such as weekdays, weekends, holidays), and uses the water consumption model after training and verification to accurately predict the future water consumption of each water supply pool. Compared with traditional prediction methods, the prediction method and system of the present invention can accurately reflect the water consumption change trend of the water supply pool, providing a more reliable basis for water supply and storage decision-making.

[0020] (2) The water supply and storage method and system of the secondary water supply pool provided by the present invention accurately divides the water consumption predicted by the water consumption prediction model constructed based on the time series large model TimeGPT into peak water consumption period, flat valley period and low water consumption period, and dynamically optimizes the water supply pool to replenish water in combination with the water level of the water supply pool. During the flat valley period and the low peak period, the water inlet valve of the water supply pool is opened to replenish water in the water supply pool to ensure that the water supply pool reaches the full water level before the peak water consumption period. During the peak water consumption period, the water inlet valve is closed, and the water stored in the water supply pool is used to supply users, effectively avoiding the phenomenon of "grabbing water" between the water supply pool and the municipal pipeline network during peak periods. This dynamic adjustment mechanism realizes the "peak shaving and valley filling" of water volume, significantly reduces the load of urban water supply during peak periods, and ensures the stability and reliability of residents' water use.

[0021] (3) The water supply and storage method and system for the secondary water supply pool provided by the present invention monitors the water level of the water supply pool in real time through a liquid level sensor, and remotely controls the opening and closing of the water inlet valve through a controller according to the water supply level and predicted water consumption, thereby realizing refined management of the water supply system. This not only improves the water supply efficiency, but also reduces the operation and maintenance costs, providing strong technical support for urban water supply management.

[0022] (4) The secondary water supply tank water supply and storage method and system provided by the present invention are also provided with a human-machine operation panel and an alarm module, wherein the human-machine operation panel serves as a bridge for interaction between the controller and the pump station management personnel. Through the human-machine operation panel, the management personnel can set the selection and switching of local operation modes, including manual operation mode and automatic operation mode, with flexible control. In the event of an abnormal situation (for example, water level exceeding the limit, equipment failure, etc.), an alarm is issued through the alarm module so that the operation and maintenance personnel can take timely measures to ensure water supply safety, thereby improving the emergency response capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flowchart of a method for predicting water consumption of a secondary water supply pool according to an embodiment of the present invention; Figure 2 It is a structural block diagram of the water consumption prediction system for the secondary water supply pool according to an embodiment of the present invention; Figure 3 It is a flowchart of the water supply and storage method of the secondary water supply pool according to an embodiment of the present invention; Figure 4 This is a flowchart of determining a water use period based on a sliding window and a threshold determination algorithm according to an embodiment of the present invention; Figure 5 It is a structural block diagram of the water supply and storage system of the secondary water supply tank described in an embodiment of the present invention.

[0024] In the figure, 1. Data acquisition module, 2. Missing filling module, 3. Data set generation module, 4. Model construction module, 5. Model training and verification module, 6. Prediction module, 7. Model optimization module, 701. Data monitoring module, 702. New training set generation module, 8. Liquid level sensor, 9. Judgment module, 10. Water replenishment adjustment module, 1001. Water inlet valve, 1002. Controller, 11. Human-machine operation panel, 12. Alarm module. DETAILED DESCRIPTION

[0025] The present invention is described in detail below by way of exemplary embodiments. However, it should be understood that elements, structures, and features in one embodiment may also be beneficially combined in other embodiments without further description.

[0026] In view of the problems of poor water consumption prediction accuracy and weak regulation and storage capacity in the existing urban secondary regulation and pressurized water supply, the present invention provides a water consumption prediction method and system for a secondary water supply pool, and a water supply regulation and storage method and system. The water consumption is predicted by a water consumption prediction model constructed based on an LSTM neural network structure, and the predicted water consumption is combined with the current water consumption and the operating state of the water pump characteristic curve to perform water supply and storage. The water consumption can be accurately predicted, and the operating state of the water pump can be scientifically and reasonably adjusted according to the predicted water consumption combined with the water pump characteristic curve to ensure efficient and stable operation of the water pump, realize water supply optimization and storage, improve water supply and storage efficiency, reduce the energy consumption and failure rate of the water pump, and improve the reliability and stability of the water supply system. The following is a detailed description of a water consumption prediction method and system for a secondary water supply pool, and an optimized regulation and storage method and system provided by the present invention in conjunction with the accompanying drawings.

[0027] See also Figure 1 The first embodiment of the present invention provides a method for predicting water consumption of a secondary water supply pool, the steps of which are: S1. Data acquisition step: Obtain environmental parameters of the water supply area of ​​each pool and historical cumulative water consumption. Environmental parameters include environmental parameters at the prediction time point and historical environmental parameters.

[0028] Specifically, the historical cumulative water consumption of each water supply area of ​​the water pool is monitored by the digital information system (such as SCADA) of the water management department (which can be the cumulative water consumption of 1 hour, 30 minutes, or 15 minutes). In some embodiments, the daily 15-minute cumulative water consumption data of each water supply area of ​​the water pool for two years is obtained.

[0029] Specifically, the environmental parameters are weather data, including: daily maximum temperature, daily minimum temperature, daily average temperature, humidity, rainfall, water volume, and solar radiation intensity. Weather data is selected from weather data with the same granularity as water consumption data downloaded from professional weather websites (such as: Amap API, Xinzhi Weather API, Hefeng Weather API, OpenWeather, AccuWeather, Visual Crossing, etc.).

[0030] S2, missing filling steps: add missing time to historical environmental parameters and historical cumulative water consumption to make the time of the time series exist continuously according to the set time interval; add date features to historical environmental parameters and historical water consumption, and convert them into digital date features through one-hot encoding; add missing cumulative water consumption to historical environmental parameters and historical water consumption.

[0031] It should be noted that, considering the impact of holidays on urban water consumption, adding date features to historical environmental parameters and historical water consumption and converting them into digital date features can improve the accuracy and reliability of the water consumption prediction model when training the water consumption prediction model later.

[0032] Specifically, the specific method of adding missing cumulative water consumption is to delete the erroneous data in the historical cumulative water consumption (such as repeated flow values ​​that appear continuously, etc.), and use the random forest algorithm to fill the missing cumulative water consumption in the historical cumulative water consumption. The stability and consistency of the cumulative water consumption distribution are ensured by filling the missing values, ensuring the integrity of the cumulative water consumption.

[0033] S3, data set generation step: the historical environmental parameters and historical cumulative water consumption after missing filling form a data set, and the data set is divided into a training set, a validation set and a test set.

[0034] Specifically, when dividing the data set, most of the data is used as training data to train the water supply flow prediction model constructed later, and the data in the training set, validation set and test set are not repeated.

[0035] Specifically, in some embodiments, the data of the training set and the validation set are the non-missing cumulative water consumption and its corresponding weather data and date features, and the data of the test set are the filled cumulative consumption missing values ​​and its corresponding weather data and date features. The training set, validation set, and test set can also be divided according to a set ratio, but the filled cumulative consumption missing values ​​and their corresponding weather data and date features are not included in the training set and the validation set, so as to ensure the accuracy and reliability of the water consumption prediction model when the water consumption prediction model is trained later.

[0036] S4. Model construction steps: Based on the time series large model TimeGPT, with historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and water consumption at the predicted time point as output, a water consumption prediction model is constructed.

[0037] It should be noted that changes in various environmental factors in the city directly or indirectly affect changes in water consumption. Selecting weather data and historical water consumption that have the greatest impact on water consumption as input, and predicting water consumption as output to build a water consumption prediction model can accurately reflect the trend of water consumption changes.

[0038] S5, model training and verification steps: setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set.

[0039] Specifically, when training the water consumption prediction model, the water consumption prediction models of multiple water supply pools can be trained at the same time, and the model can be fine-tuned on the set data set to enhance the model's function, so that the model can adapt to the subtle differences of different time series data, maximize the performance indicators of the water consumption prediction model, and obtain a water consumption prediction model with the water consumption characteristics of each water supply pool and water supply area. When fine-tuning, set the number of iterations and specify the loss function to minimize the prediction error.

[0040] Specifically, in some embodiments, the forecast method is used to set the number of iterations for model optimization, and the number of iterations for fine-tuning the generation of a new data set is specified; the corresponding string representing the loss function is passed to the finetune_loss parameter of the forecast method, "DEFAULT", "MAE", "MSE", "RMSE", "MAPE", and "SMAPE", to obtain the indicator improvement ratio of different loss functions compared with the default loss function. Selecting a loss function with a higher indicator improvement can maximize the accuracy of the prediction. The comparison results of the indicator improvement of different loss functions compared with the default loss function are shown in Table 1. When the iterative training number finetune_loss is set to 10, using MAE as the loss function can improve the indicator by 3.05% compared with using the default loss function. It should be noted that the number of iterations is adjusted according to the model performance. If the number of iterations is too large, it may lead to too long training time and overfitting of the model.

[0041] Table 1

[0042] Specifically, in some embodiments, the specific method for verifying the water consumption prediction model is: using a rolling window method to evaluate the performance of the water consumption prediction model in different time periods, and if the evaluation result fails, retraining the water consumption prediction model until the evaluation result passes. Performing performance evaluation by cross-validation can improve the accuracy and reliability of the water consumption prediction model.

[0043] S6. Prediction step: input the test set into the water consumption prediction model to obtain the predicted water consumption at the prediction time point.

[0044] In some embodiments, the above-mentioned secondary water supply pool water consumption prediction method also includes a model optimization step: real-time monitoring of the water consumption at the prediction time point in the pipe network SCADA system at the prediction time point, adding the real-time monitored water consumption at the prediction time point, the environmental parameters at the prediction time point, and the date characteristics at the prediction time point to the training set to form a new training set, and training the water consumption prediction model through the new training set to optimize the water consumption prediction model. Adding real-time data to the water consumption model training data, continuously optimizing the water consumption prediction model, realizing the closed loop of the entire water consumption prediction process, and ensuring the timeliness of the water consumption prediction model.

[0045] It should be noted that, unlike the traditional method of directly using the predicted value of the previous time point as the predicted input value of the next time point, when the data of this application is input into the water consumption prediction model, the predicted value of the previous time point is not directly used as the input value of the next time point. Instead, when the time point reaches the predicted time point, the water consumption, environmental parameters, and date characteristics monitored in real time at that time point are added to the training set for training. After the water consumption prediction model is optimized, the prediction of the next time point is entered. This will not cause error accumulation and improve the prediction accuracy.

[0046] The above-mentioned secondary water supply pool water consumption prediction method of the embodiment of the present invention constructs a water consumption prediction model based on the time series large model TimeGPT, and trains and verifies it. It uses the transfer learning ability of the time series large model TimeGPT and its adaptability to time series data, combined with the water consumption data environmental parameters (such as temperature, humidity, precipitation, snowfall, solar radiation intensity, etc.) and date characteristics (such as weekdays, weekends, holidays). The water consumption model after training and verification can accurately predict the future water consumption of each water supply pool, and can accurately reflect the trend of water consumption changes, providing a more reliable basis for water supply and storage decisions.

[0047] See also Figure 2 The second embodiment of the present invention provides a secondary water supply pool water consumption prediction system, which is used to implement the secondary water supply pool water consumption prediction method described in the first aspect of the present invention, including: Data acquisition module 1, obtains the environmental parameters, historical environmental parameters and historical cumulative water consumption of each water supply area at the predicted time point; Missing filling module 2 adds missing time to historical environmental parameters and historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval; adds date features to historical environmental parameters and historical water consumption, and converts them into digital date features through one-hot encoding; adds missing cumulative water consumption to historical environmental parameters and historical water consumption; Dataset generation module 3, which forms a data set with the missing historical environmental parameters and historical cumulative water consumption, and divides the data set into a training set, a validation set, and a test set; Model building module 4, based on the time series large model TimeGPT, takes historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and takes the water consumption at the predicted time point as output to build a water consumption prediction model; Model training and verification module 5, setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set; Prediction module 6 inputs the test set into the water consumption prediction model to obtain the water consumption at the predicted time point.

[0048] In some embodiments, the secondary water supply pool water consumption prediction system further includes a model optimization module 7, and the model optimization module includes: The data monitoring module 701 is used to monitor the actual water consumption at the predicted time point in real time; The new training set generation module 702 adds the actual water consumption at the prediction time point monitored in real time, the environmental parameters at the prediction time point, and the date characteristics at the prediction time point to the training set to form a new training set, so as to train the water consumption prediction model through the new training set and optimize the water consumption prediction model.

[0049] Add real-time data to the water consumption model training data, continuously optimize the water consumption prediction model, realize the closed loop of the entire water consumption prediction process, and ensure the timeliness of the water consumption prediction model. It should be noted that, unlike the traditional method of directly using the predicted value of the previous time point as the predicted input value of the next time point, when the data of this application is input into the water consumption prediction model, the predicted value of the previous time point is not directly used as the input value of the next time point. Instead, when the time point arrives at the predicted time point, the water consumption, environmental parameters, and date characteristics monitored in real time at that time point are input into the training set for training. After optimizing the water consumption prediction model, the prediction of the next time point is entered. This will not cause the accumulation of errors and improve the prediction accuracy.

[0050] The above-mentioned secondary water supply pool water consumption prediction system in this embodiment builds a water consumption prediction model based on the time series large model TimeGPT, and trains and verifies it. It uses the transfer learning ability of the time series large model TimeGPT and its adaptability to time series data, combined with the water consumption data environmental parameters (such as temperature, humidity, precipitation, snowfall, solar radiation intensity, etc.) and date characteristics (such as weekdays, weekends, holidays). The water consumption model after training and verification can accurately predict the future water consumption of each water supply pool, and can accurately reflect the trend of water consumption changes, providing a more reliable basis for water supply and storage decisions.

[0051] See also Figure 3 The third aspect of the present invention provides a method for regulating water supply in a secondary water supply pool, the steps of which are as follows: S1. Data acquisition step: Obtain environmental parameters of the water supply area of ​​each pool and historical cumulative water consumption. Environmental parameters include environmental parameters at the prediction time point and historical environmental parameters.

[0052] Specifically, the historical cumulative water consumption of each water supply area of ​​the water pool is monitored by the digital information system (such as SCADA) of the water management department (which can be the cumulative water consumption of 1 hour, 30 minutes, or 15 minutes). In some embodiments, the daily 15-minute cumulative water consumption data of each water supply area of ​​the water pool for two years is obtained.

[0053] Specifically, the environmental parameters are weather data, including: daily maximum temperature, daily minimum temperature, daily average temperature, humidity, rainfall, water volume, and solar radiation intensity. Weather data is selected from weather data with the same granularity as water consumption data downloaded from professional weather websites (such as: Amap API, Xinzhi Weather API, Hefeng Weather API, OpenWeather, AccuWeather, Visual Crossing, etc.).

[0054] S2, missing filling steps: add missing time to historical environmental parameters and historical cumulative water consumption to make the time of the time series exist continuously according to the set time interval; add date features to historical environmental parameters and historical water consumption, and convert them into digital date features through one-hot encoding; add missing cumulative water consumption to historical environmental parameters and historical water consumption.

[0055] It should be noted that, considering the impact of holidays on urban water consumption, adding date features to historical environmental parameters and historical water consumption and converting them into digital date features can improve the accuracy and reliability of the water consumption prediction model when training the water consumption prediction model later.

[0056] Specifically, the specific method of adding missing cumulative water consumption is to delete the erroneous data in the historical cumulative water consumption (such as repeated flow values ​​that appear continuously, etc.), and use the random forest algorithm to fill the missing cumulative water consumption in the historical cumulative water consumption. The stability and consistency of the cumulative water consumption distribution are ensured by filling the missing values, ensuring the integrity of the cumulative water consumption.

[0057] S3, data set generation step: the historical environmental parameters and historical cumulative water consumption after missing filling form a data set, and the data set is divided into a training set, a validation set and a test set.

[0058] Specifically, when dividing the data set, most of the data is used as training data to train the water supply flow prediction model constructed later, and the data in the training set, validation set and test set are not repeated.

[0059] Specifically, in some embodiments, the data of the training set and the validation set are the non-missing cumulative water consumption and its corresponding weather data and date features, and the data of the test set are the filled cumulative consumption missing values ​​and its corresponding weather data and date features. The training set, validation set, and test set can also be divided according to a set ratio, but the filled cumulative consumption missing values ​​and their corresponding weather data and date features are not included in the training set and the validation set, so as to ensure the accuracy and reliability of the water consumption prediction model when the water consumption prediction model is trained later.

[0060] S4. Model construction steps: Based on the time series large model TimeGPT, with historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and water consumption at the predicted time point as output, a water consumption prediction model is constructed.

[0061] It should be noted that changes in various environmental factors in the city directly or indirectly affect changes in water consumption. Selecting weather data and historical water consumption that have the greatest impact on water consumption as input, and predicting water consumption as output to build a water consumption prediction model can accurately reflect the trend of water consumption changes.

[0062] S5, model training and verification steps: setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set.

[0063] Specifically, when training the water consumption prediction model, the water consumption prediction models of multiple water supply pools can be trained at the same time, and the model can be fine-tuned on the set data set to enhance the model's function, so that the model can adapt to the subtle differences of different time series data, maximize the performance indicators of the water consumption prediction model, and obtain a water consumption prediction model with the water consumption characteristics of each water supply pool and water supply area. When fine-tuning, set the number of iterations and specify the loss function to minimize the prediction error.

[0064] Specifically, in some embodiments, the forecast method is used to set the number of iterations for model optimization, and the number of iterations for fine-tuning the generation of a new data set is specified; the corresponding string representing the loss function is passed to the finetune_loss parameter of the forecast method, "DEFAULT", "MAE", "MSE", "RMSE", "MAPE", and "SMAPE", to obtain the indicator improvement ratio of different loss functions compared with the default loss function. Selecting a loss function with a higher indicator improvement can maximize the accuracy of the prediction. As shown in Table 1, when the iterative training number finetune_loss is set to 10, using MAE as the loss function can improve the indicator by 3.05% compared with using the default loss function. It should be noted that the number of iterations is adjusted according to the model performance. If the number of iterations is too large, it may lead to too long training time and overfitting of the model.

[0065] Specifically, in some embodiments, the specific method for verifying the water consumption prediction model is: using a rolling window method to evaluate the performance of the water consumption prediction model in different time periods, and if the evaluation result fails, retraining the water consumption prediction model until the evaluation result passes. Performing performance evaluation by cross-validation can improve the accuracy and reliability of the water consumption prediction model.

[0066] S6. Prediction step: input the test set into the water consumption prediction model to obtain the predicted water consumption at the prediction time point.

[0067] S7, liquid level measurement step: real-time measurement of the water level of the water supply tank.

[0068] S8, judgment step: judging the water use period according to the water consumption at the predicted time point; judging whether the water level of the water supply pool during the peak water use period is lower than the set minimum water level threshold of the pool.

[0069] Specifically, see Figure 4 , based on the sliding window and threshold judgment algorithm to determine the water use period, the specific method is: S81, parameter setting step: set the sliding window size to , the peak threshold is , the low peak threshold is .

[0070] S82, average water consumption calculation step: The sliding window traverses the water consumption data sequence at the entire prediction time point. At each window position, the water consumption data in the sliding window is expressed as: ,

[0071] In the formula, Indicates A sliding window, Indicates Water consumption in a certain period of time; Calculate the average water usage for the period within the sliding window:

[0072] S83, water consumption period determination step: if the average water consumption of the current sliding window , it enters the peak detection mode. In the peak detection mode, it moves to the next sliding window and calculates the average water consumption of the next sliding window. If the average water consumption of multiple consecutive sliding windows exceeds the peak threshold, the time corresponding to these sliding windows is regarded as a continuous water consumption peak period. If multiple consecutive water consumption peak periods are detected, these consecutive water consumption peak periods are merged into one water consumption peak period, and the previous one with the largest water consumption and the longest duration is selected. The peak period is the final water consumption peak period. ≥4; if the average water consumption in the current sliding window , it enters the low-peak detection mode. In the low-peak detection mode, it moves to the next sliding window and calculates the average water consumption of the next sliding window. If the average water consumption of multiple consecutive sliding windows is lower than the low-peak threshold, the time corresponding to these sliding windows is a continuous low-peak water consumption period. If multiple consecutive low-peak water consumption periods are detected, these consecutive low-peak water consumption periods are merged into one low-peak water consumption period, and the previous one with the largest water consumption and the longest duration is selected. The low-peak period is the final water consumption low-peak period. ≥4; the rest of the water usage period is the off-peak period.

[0073] S9, water replenishment adjustment steps: During the valley period and low peak period of water usage, replenish water into the water supply pool until the water level in the water supply pool reaches the set maximum water level of the pool, and stop replenishing water; during the peak water usage period, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, replenish water into the water supply pool.

[0074] Specifically, the maximum water level of the pool is set according to the position of the pool overflow and the height of the water inlet. Water is replenished to the maximum water level and then replenishment is stopped. The minimum water level threshold is set according to the outlet height and the secondary water supply pump protection liquid level (the pool must not be "drained" and water supply to mid- and high-level users must not be cut off).

[0075] The water supply and storage method of the secondary water supply pool provided by the present invention accurately divides the water consumption predicted by the water consumption prediction model constructed based on the time series large model TimeGPT into peak water consumption period, flat valley period and low water consumption period, and dynamically optimizes the water supply pool in combination with the water level of the water supply pool. During the flat valley period and the low peak period, the water inlet valve of the water supply pool is opened to replenish water in the water supply pool to ensure that the water supply pool reaches the full water level before the peak water consumption period. During the peak water consumption period, the water inlet valve is closed, and the water stored in the water supply pool is used to supply users, which effectively avoids the phenomenon of "grabbing water" between the water supply pool and the municipal pipeline network during peak periods. This dynamic adjustment mechanism realizes the "peak shaving and valley filling" of water volume, significantly reduces the load of urban water supply during peak periods, and ensures the stability and reliability of residents' water use.

[0076] See also Figure 5 In a fourth aspect, the present invention provides a secondary water supply tank water supply and storage system, which is used to implement the secondary water supply tank water supply and storage method described in the third aspect of the present invention, comprising: Data acquisition module 1, obtains the environmental parameters, historical environmental parameters and historical cumulative water consumption of each water supply area at the predicted time point; The missing filling module 2 adds missing time to the historical environmental parameters and the historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval, and adds the missing cumulative water consumption to the historical environmental parameters and the historical water consumption; Dataset generation module 3, which forms a data set with the missing historical environmental parameters and historical cumulative water consumption, and divides the data set into a training set, a validation set, and a test set; Model building module 4, based on the time series large model TimeGPT, takes historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and takes the water consumption at the predicted time point as output to build a water consumption prediction model; Model training and verification module 5, setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set; Prediction module 6, inputs the test set into the water consumption prediction model to obtain the water consumption at the predicted time point; The liquid level sensor 8 is arranged in the water supply tank and is used to measure the water level of the water supply tank in real time; The judgment module 9 judges the water use period according to the water consumption at the predicted time point; judges whether the water level of the water supply pool during the peak water use period is lower than the set minimum water level threshold of the pool; The water replenishment regulation module 10 replenishes water into the water supply pool during the valley period and the low peak period until the water level in the water supply pool reaches the set maximum water level of the pool, and then stops replenishing water; during the peak period of water use, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, water is replenished into the water supply pool.

[0077] In some embodiments, the water replenishment regulation module 10 includes: The water inlet valve 1001 is arranged on the water inlet pipe of the water supply tank; Controller 1002 is connected to the water inlet valve. During the valley period and the off-peak period, the water inlet valve is controlled to open to replenish water in the water supply pool until the water level in the water supply pool reaches the set maximum water level of the pool, and the water inlet valve is controlled to close to stop replenishing water. During the peak water usage period, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, the water inlet valve is controlled to open to replenish water in the water supply pool.

[0078] On the one hand, the controller controls the water replenishment in the water supply pool according to the judged water usage period, and on the other hand, it also directly controls the water replenishment in the water supply pool according to the water level of the pool. Specifically, when the water level of the pool is lower than the set minimum water level threshold of the pool during the peak water usage period, the water inlet valve is controlled to open to ensure that the water level of the pool will not be too low, so as not to affect the normal water supply demand.

[0079] In some embodiments, the secondary water supply tank water supply and storage system also includes a human-machine operation panel 11 and an alarm module 12, wherein the human-machine operation panel 11 is connected to the controller 1002 and the water inlet valve 1001, and the alarm module 12 is connected to the controller 1002. The human-machine operation panel serves as a bridge for interaction between the controller and the pump station manager. Through the human-machine operation panel, the manager can set the selection and switching of the local operation mode, including the manual operation mode and the automatic operation mode, with flexible control. In the event of an abnormal situation (for example, water level exceeding the limit, equipment failure, etc.), an alarm is issued through the alarm module so that the operation and maintenance personnel can take timely measures to ensure water supply safety and improve emergency response capabilities.

[0080] Specifically, the manual operation mode is: the operator can directly control the opening and closing degree of the water inlet valve through the virtual button or knob on the human-machine operation panel to achieve precise adjustment of the water supply flow. This mode is suitable for system debugging, emergency treatment or precise control under specific working conditions, providing great flexibility and control for operation and maintenance personnel.

[0081] The automatic operation mode is: the operator switches to the automatic operation mode through the human-machine operation panel, and the controller takes over the control. The controller automatically adjusts the opening of the water inlet valve according to the judgment result of the judgment module and the real-time monitored water level information to keep the water level within a reasonable range.

[0082] Specifically, in some embodiments, the controller adopts PLC, and the human-machine operation panel adopts a high-resolution touch screen design.

[0083] Specifically, in some embodiments, a main control module is also included, which serves as the brain of the system and is used to control the automatic operation of the entire system according to the set workflow, that is, automatically control data collection, add missing values, build and train models, predict water consumption, determine water use time periods and pool water levels, and adjust water replenishment. The water consumption in each time period in the future, water use time period analysis, pool water level, and water inlet control strategy (i.e., water inlet valve opening strategy) are uploaded to the cloud platform through the wireless communication module and sent to the manager's mobile terminal.

[0084] The secondary water supply tank water supply and storage system described in the above embodiment of the present invention includes an Agents system, and the Agents system includes a main control module, a prediction sub-agent, a scheduling sub-agent, and a control sub-agent. Among them, the prediction sub-agent is composed of a model construction module, a model training and verification module, a model optimization module and a prediction module, and is used to predict water consumption. The scheduling sub-agent is a judgment module, which is used to judge the water use period and the water level of the tank. The control sub-agent is a water replenishment adjustment module for controlling water replenishment adjustment. The main control module serves as the brain of the Agents system, which is used to control the prediction sub-agent, the scheduling sub-agent, and the control sub-agent according to the set workflow. While realizing their own functions, the three sub-agents communicate with the other sub-agents and cooperate with each other to form a hierarchical dynamic flexible water replenishment management mode. The workflow is: step 1: read the pool-related parameters (highest water level, lowest water level) set by the user and the historical water consumption data of the pump station and the pool water supply area in each period. Step 2: Call fill_gaps function to fill in the missing time points of data, call Visual Crossing, date_features function to add exogenous variables (i.e. weather data and date features) to the historical water consumption data, use random forest algorithm (RFA) to fill in the missing values ​​of water consumption, and obtain a data set that can be used for prediction model training. Step 3: Input the training data set into the prediction sub-agent (FA), the prediction sub-agent (FA) creates a time series with a set time interval (for example: 15 minutes) for the next day, uses the method of step 2 to add exogenous variables for each future time point, calls the TimeGPT model to predict the water consumption of each pool in each period, obtains the water consumption data for each period and inputs it into the scheduling sub-agent (SA), and the water consumption data and water consumption data analysis for each period are transmitted to the cloud platform and the mobile terminal of the manager through the wireless communication module. Step 4: The water consumption data for each period is input into the scheduling sub-agent (SA), and the scheduling sub-agent (SA) determines the water use period based on the sliding window and threshold judgment algorithm. Step 5: Input the determined water use period into the control sub-agent (CA), which generates the water inlet valve opening strategy, controls the water inlet valve opening according to the water inlet valve opening strategy, and transmits the water inlet valve opening strategy to the main control module, which uploads it to the cloud platform and the mobile terminal of the manager. In the skill design of the Agents system, the main control module calls the API interface related to each sub-agent by calling the tool. The API interface is the key external resource of the Agents system. The Agents system can access and utilize online services and information, and can also add the API of the open source toolkit, which is helpful for developing other functions (such as gapier).

[0085] The secondary water supply tank water supply and storage system provided by the present invention accurately divides the water consumption predicted by the water consumption prediction model constructed based on the time series large model TimeGPT into peak water consumption period, flat valley period and low water consumption period, and dynamically optimizes the water supply tank to replenish water in combination with the water level of the water supply tank. During the flat valley period and the low peak period, the water tank inlet valve is opened to replenish water in the water supply tank to ensure that the water supply tank reaches the full water level before the peak water consumption period. During the peak water consumption period, the water inlet valve is closed, and the water stored in the water supply tank is used to supply users, which effectively avoids the phenomenon of "grabbing water" between the water supply tank and the municipal pipeline network during peak periods. This dynamic adjustment mechanism realizes the "peak shaving and valley filling" of water volume, significantly reduces the load during the peak period of urban water supply, and ensures the stability and reliability of residents' water use.

[0086] The above embodiments are used to explain the present invention rather than to limit the present invention. Any modification and change made to the present invention within the spirit of the present invention and the protection scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A method for predicting water consumption in a secondary water supply pool, characterized in that: The steps are: Data acquisition steps: Obtain environmental parameters and historical cumulative water consumption of each water tank water supply area. Environmental parameters include environmental parameters at the prediction time point and historical environmental parameters; Missing filling steps: add missing time to historical environmental parameters and historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval; add date features to historical environmental parameters and historical water consumption, and convert them into digital date features through unique hot encoding; add missing cumulative water consumption to historical environmental parameters and historical water consumption; Dataset generation steps: The missing historical environmental parameters and historical cumulative water consumption are used to form a dataset, and the dataset is divided into a training set, a validation set, and a test set; Model construction steps: Based on the time series large model TimeGPT, the water consumption prediction model is constructed with historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as inputs and water consumption at the predicted time point as output; Model training and verification steps: set water consumption prediction model parameters, train the water consumption prediction model through the training set, and verify the trained water consumption prediction model through the verification set; Prediction step: Input the test set into the water consumption prediction model to obtain the water consumption at the predicted time point.

2. The method for predicting water consumption of a secondary water supply pool according to claim 1, characterized in that: In the missing filling step, the specific method of adding the missing cumulative water consumption is: deleting the erroneous data in the historical cumulative water consumption, and using the random forest algorithm to fill the missing cumulative water consumption of the historical cumulative water consumption.

3. The method for predicting water consumption of a secondary water supply pool according to claim 1, characterized in that: In the model training and verification steps, the specific method for verifying the water consumption prediction model is: using a rolling window method to evaluate the performance of the water consumption prediction model in different time periods. If the evaluation result fails, the water consumption prediction model is retrained until the evaluation result passes.

4. A secondary water supply pool water consumption prediction system, used to implement the secondary water supply pool water consumption prediction method according to any one of claims 1 to 3, characterized in that: include: Data acquisition module, to obtain the environmental parameters, historical environmental parameters and historical cumulative water consumption of each pool water supply area at the predicted time point; The missing filling module adds missing time to the historical environmental parameters and historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval; adds date features to the historical environmental parameters and historical water consumption, and converts them into digital date features through unique hot encoding; adds missing cumulative water consumption to the historical environmental parameters and historical water consumption; The data set generation module forms a data set with the missing historical environmental parameters and historical cumulative water consumption, and divides the data set into a training set, a validation set, and a test set; The model building module is based on the time series large model TimeGPT, takes historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and takes the water consumption at the predicted time point as output to build a water consumption prediction model; Model training and verification module, setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set; The prediction module inputs the test set into the water consumption prediction model to obtain the water consumption at the predicted time point.

5. A method for regulating water supply in a secondary water supply pool, characterized in that: The steps are: Data acquisition steps: Obtain environmental parameters and historical cumulative water consumption of each water tank water supply area. Environmental parameters include environmental parameters at the prediction time point and historical environmental parameters; Missing filling step: adding missing time to the historical environmental parameters and historical cumulative water consumption so that the time of the time series exists continuously according to the set time interval, and adding missing cumulative water consumption to the historical environmental parameters and historical water consumption; Dataset generation steps: The missing historical environmental parameters and historical cumulative water consumption are used to form a dataset, and the dataset is divided into a training set, a validation set, and a test set; Model construction steps: Based on the time series large model TimeGPT, the water consumption prediction model is constructed with historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as inputs and water consumption at the predicted time point as output; Model training and verification steps: set water consumption prediction model parameters, train the water consumption prediction model through the training set, and verify the trained water consumption prediction model through the verification set; Prediction step: input the test set into the water consumption prediction model to obtain the water consumption at the predicted time point; Liquid level measurement steps: Real-time measurement of water level in the water supply tank; Judgment steps: determine the water consumption period according to the predicted water consumption at the time point; determine whether the water level of the water supply pool during the peak water consumption period is lower than the set minimum water level threshold of the pool; Water replenishment adjustment steps: During the valley period and low peak period, replenish water into the water supply pool until the water level in the water supply pool reaches the set maximum water level of the pool, and stop replenishing water; during the peak water usage period, if the water level in the water supply pool is lower than the set minimum water level threshold of the pool, replenish water into the water supply pool, otherwise do not replenish water.

6. The water supply and storage method of the secondary water supply pool as claimed in claim 5, characterized in that: In the judgment step, the water use period is judged based on the sliding window and threshold judgment algorithm, and the specific method is as follows: Parameter setting steps: Set the sliding window size to , the peak threshold is , the low peak threshold is ; Average water consumption calculation steps: The sliding window traverses the water consumption data sequence at the entire prediction time point. At each window position, the water consumption data in the sliding window is expressed as: ， In the formula, Indicates A sliding window, Indicates Water consumption in a certain period of time; Calculate the average water usage for the period within the sliding window: Water consumption period judgment step: If the average water consumption in the current sliding window , it enters the peak detection mode. In the peak detection mode, it moves to the next sliding window and calculates the average water consumption of the next sliding window. If the average water consumption of multiple consecutive sliding windows exceeds the peak threshold, the time corresponding to these sliding windows is regarded as a continuous water consumption peak period. If multiple consecutive water consumption peak periods are detected, these consecutive water consumption peak periods are merged into one water consumption peak period, and the previous one with the largest water consumption and the longest duration is selected. The peak period is the final water consumption peak period. ≥4; if the average water consumption in the current sliding window , it enters the low-peak detection mode. In the low-peak detection mode, it moves to the next sliding window and calculates the average water consumption of the next sliding window. If the average water consumption of multiple consecutive sliding windows is lower than the low-peak threshold, the time corresponding to these sliding windows is a continuous low-peak water consumption period. If multiple consecutive low-peak water consumption periods are detected, these consecutive low-peak water consumption periods are merged into one low-peak water consumption period, and the previous one with the largest water consumption and the longest duration is selected. The low-peak period is the final water consumption low-peak period. ≥4; the rest of the water usage period is the off-peak period.

7. The water supply and storage method of the secondary water supply pool as claimed in claim 5, characterized in that: In the missing filling step, the specific method of adding the missing cumulative water consumption is: deleting the erroneous data in the historical cumulative water consumption, and using the random forest algorithm to fill the missing cumulative water consumption of the historical cumulative water consumption.

8. The water supply and storage method of the secondary water supply pool as claimed in claim 5, characterized in that: In the model training and verification steps, the specific method for verifying the water consumption prediction model is: using a rolling window method to evaluate the performance of the water consumption prediction model in different time periods. If the evaluation result fails, the water consumption prediction model is retrained until the evaluation result passes.

9. A water supply and storage system for a secondary water supply pool, used to implement the water supply and storage method for a secondary water supply pool as claimed in any one of claims 5 to 8, characterized in that: include: Data acquisition module, to obtain the environmental parameters, historical environmental parameters and historical cumulative water consumption of each pool water supply area at the predicted time point; A missing filling module adds missing time to historical environmental parameters and historical cumulative water consumption, so that the time of the time series exists continuously according to the set time interval, and adds missing cumulative water consumption to historical environmental parameters and historical water consumption; The data set generation module forms a data set with the missing historical environmental parameters and historical cumulative water consumption, and divides the data set into a training set, a validation set, and a test set; The model building module is based on the time series large model TimeGPT, takes historical water consumption, historical environmental parameters, and environmental parameters at the predicted time point as input, and takes the water consumption at the predicted time point as output to build a water consumption prediction model; Model training and verification module, setting water consumption prediction model parameters, training the water consumption prediction model through the training set, and verifying the trained water consumption prediction model through the verification set; The prediction module inputs the test set into the water consumption prediction model to obtain the water consumption at the predicted time point; The liquid level sensor is installed in the water supply tank and is used to measure the water level of the water supply tank in real time; The judgment module determines the water consumption period according to the predicted water consumption at the time point; determines whether the water level of the water supply pool during the peak water consumption period is lower than the set minimum water level threshold of the pool; The water replenishment regulation module replenishes water to the water supply pool during valley and off-peak periods until the water level in the water supply pool reaches the set maximum water level of the pool, and then stops replenishing water; during peak water usage periods, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, water is replenished to the water supply pool.

10. The secondary water supply pool water supply and storage system according to claim 9, characterized in that: The water replenishment regulation module comprises: The water inlet valve is installed on the water inlet pipe of the water supply tank; The controller is connected to the water inlet valve. During the valley period and the low peak period, the water inlet valve is controlled to open to replenish water in the water supply pool until the water level in the water supply pool reaches the set maximum water level of the pool, and the water inlet valve is controlled to close to stop replenishing water. During the peak water usage period, when the water level in the water supply pool is lower than the set minimum water level threshold of the pool, the water inlet valve is controlled to open to replenish water in the water supply pool.

Images