A high-speed service area water-saving management system and method based on the Internet of Things

Abstract

The application discloses a kind of based on high-speed service area water-saving management system and method of Internet of Things, belong to intelligent water resource management technical field.It aims at solving the operation optimization problem of existing service area water management system, including: obtaining historical water consumption, water quality data, meteorological information, passenger flow, regeneration water pool and rainwater storage tank liquid level and using pre-trained water consumption load prediction model to process, obtain water consumption load prediction value;Water quality data are compared with pre-stored water quality threshold under different water consumption scenarios, obtain warning information;Data extraction is carried out on water consumption load prediction value, and the predicted water consumption is substituted into the operation cost objective function to solve, and the optimal solution vector of the minimum cost of each water consumption subsystem water supply scheduling is used to generate control instruction;Control instruction occurs to each controller to manage the operation state of each water consumption subsystem execution equipment.The application can realize on-demand water supply and accurate resource allocation of high-speed service area under multiple factors change.

Description

Technical Field

[0001] This invention relates to a water-saving management system and method for high-speed service areas based on the Internet of Things, belonging to the field of intelligent water resource management technology. Background Technology

[0002] With increasing water scarcity and stricter environmental protection requirements, water-saving and unconventional water utilization technologies are being promoted in public buildings and park facilities. Highway service areas, as a crucial component of transportation infrastructure, utilize water for various purposes, including public restrooms, catering, green space irrigation, and road cleaning. Their water load is significantly affected by holidays and peak travel periods, resulting in large short-term fluctuations. Statistics show that highway service areas treat approximately 30–150 tons of wastewater daily, with annual wastewater discharge ranging from 66 million to 330 million tons; in Jiangsu Province alone, service areas generate approximately 4,920–24,600 tons of wastewater per day. Currently, most service areas rely primarily on municipal water supply, supplemented by traditional measures such as low-flow equipment and timed irrigation. While some stations have rainwater harvesting or wastewater reuse facilities, these are often decentralized and experience-based, lacking continuous monitoring, forecasting, and dynamic scheduling of overall water volume and quality. This leads to low utilization rates of reclaimed water and rainwater, and is prone to supply-demand imbalances, water quality risks, and equipment idling.

[0003] In recent years, the Internet of Things and data-driven methods have been applied in the field of smart water management, but there are still some challenges in adapting them to service area scenarios: First, the stations are widely distributed and the network conditions in some sections are unstable, so the system needs to have the ability to communicate across multiple links and to be localized in the event of a network outage; Second, reclaimed water needs to meet different water quality requirements for different water use scenarios, and traditional intermittent detection is difficult to support real-time control; Third, passenger flow is random and time-dependent, so it is necessary to predict water load and convert the prediction results into executable scheduling strategies; Fourth, multiple subsystems such as rainwater harvesting, reclaimed water treatment, storage, regulation and distribution coexist, and the system integration and maintainability requirements are high.

[0004] Therefore, there is an urgent need for a smart water-saving management technology solution for highway service areas to achieve real-time perception, water quality assessment and predictive scheduling of water use in multiple stages, and to form a closed-loop control, so as to improve the efficiency of rainwater and reclaimed water utilization, reduce dependence on municipal water supply and operating energy consumption, and improve the precision and reliability of water use management in service areas. Summary of the Invention

[0005] The purpose of this invention is to provide a water-saving management system and method for highway service areas based on the Internet of Things (IoT). This system deeply integrates water use control, environmental sensing, water quality monitoring, and passenger flow data through an IoT architecture. It achieves comprehensive data collection from all elements of the service area, including water sources, water pipes, water usage points, weather, and passenger flow. Combined with a water load prediction model based on a random forest regression algorithm, it performs comprehensive prediction of multi-source data and uses an operating cost objective function to calculate the optimal solution vector for minimizing the water supply scheduling cost of each water-using subsystem. Finally, it generates a set of control instructions for the optimal water supply scheduling cost of each water-using subsystem, which varies with multiple factors such as weather, season, and passenger flow. This enables on-demand water supply and precise resource allocation in highway service areas, avoiding energy waste caused by excessive water supply and optimizing the start-stop strategies of pumps and valves in each water-using system, thereby reducing equipment failures.

[0006] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution.

[0007] In a first aspect, the present invention provides a service area water conservation management method based on the Internet of Things, wherein the management method is executed by a cloud platform and includes:

[0008] Acquire historical water consumption, water quality data, meteorological information, passenger flow, and liquid levels in the reclaimed water tank and rainwater storage tank uploaded by the PLC controller;

[0009] Based on a pre-trained water load prediction model, historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks are processed to obtain water load prediction values; and the water quality data is compared with pre-stored water quality thresholds under different water use scenarios to obtain warning information.

[0010] Data is extracted based on the predicted water load to obtain the predicted water consumption. The predicted water consumption is then substituted into the pre-constructed operating cost objective function to obtain the optimal solution vector for minimizing the water supply scheduling cost of each water subsystem.

[0011] Control commands are generated based on the optimal solution vector of each water supply subsystem with the minimum cost of water supply scheduling and warning information.

[0012] The control commands are sent to the PLC controllers associated with each water subsystem to manage the operating status of the equipment in each water subsystem.

[0013] Optionally, acquire historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of the reclaimed water tank and rainwater storage tank uploaded by the PLC controller, including:

[0014] Flow and pressure information are collected by flow sensors and pressure / differential pressure sensors configured in the water supply pipelines of the toilet flushing subsystem, greening irrigation subsystem, and domestic water subsystem, and the historical water consumption is calculated based on the flow information.

[0015] Water quality information is collected by water quality monitoring instruments configured at preset water quality monitoring nodes to form the water quality data;

[0016] Liquid level information is collected by level sensors configured in the reclaimed water tank and the rainwater storage tank to form the liquid level of the reclaimed water tank and the liquid level of the rainwater storage tank.

[0017] The meteorological information is obtained through a meteorological forecast data interface;

[0018] The passenger flow is obtained through a passenger flow monitoring system or an operational data interface.

[0019] Optionally, the water quality threshold is provided by a water quality threshold table stored on the cloud platform. The water quality threshold table includes at least: a scene identifier field, a water source type field, an indicator field, a threshold type field, a threshold value field, a threshold version number field, and an effective time field; and the threshold record that is in effect is called when comparing thresholds.

[0020] Optionally, the water load prediction model is constructed using a random forest regression prediction model, and the prediction object of the water load prediction model is set as the total water consumption and / or the component water consumption of the service area. The prediction step size of the water load prediction model is a preset time period, which can be any one of 1 hour, 6 hours, or 24 hours. The prediction output is determined to be the predicted value of the water load prediction model within the preset time period Δt. Its unit is m³ / h;

[0021] The random forest regression prediction model consists of M regression decision trees; each regression decision tree obtains a training subset through bootstrapping and randomly draws from the input features when splitting nodes. The candidate features are split using the minimum mean square error as the splitting criterion. A leaf node is generated when the maximum depth or the minimum number of samples in the leaf node is met. The output of the leaf node is the mean of the labels of the samples falling into that leaf node. After training M regression decision trees respectively, the outputs of each regression decision tree are integrated to obtain the water load prediction value.

[0022] Optionally, the training process of the water load prediction model includes:

[0023] Collect historical data on water consumption, water quality, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks;

[0024] Historical data is preprocessed to obtain preprocessed historical data;

[0025] Feature extraction was performed on the preprocessed historical data to obtain historical time-of-use water consumption sequence features, water quality status features, meteorological features, passenger flow features, reclaimed water tank level features, and rainwater storage tank level features.

[0026] Feature vectors are generated based on historical time-of-use water consumption sequence characteristics, water quality characteristics, meteorological characteristics, passenger flow characteristics, reclaimed water tank level characteristics, and rainwater storage tank level characteristics. Supervised learning label values ​​are then assigned to the feature vectors to obtain training sample pairs. The supervised learning label value is the actual water load corresponding to a future preset time period.

[0027] The training sample pairs are divided into training and validation sets according to chronological order.

[0028] The training set is input into the regression model based on the random forest algorithm to train and obtain the initial water load prediction model;

[0029] The validation set is imported into the initial water load prediction model for prediction, and the water load prediction value of the validation set is obtained.

[0030] The prediction error of the actual and predicted water load values ​​associated with the validation set is calculated.

[0031] The initial water load prediction model is updated based on a preset error threshold or a preset number of training iterations. The water load prediction model is stopped when the prediction error between the actual and predicted water load values ​​exceeds the error threshold for K consecutive time periods or the number of training iterations reaches the upper limit. The final water load prediction model is then obtained and used to predict historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks that are uploaded in real time.

[0032] Optionally, the calculation expression for the operating cost objective function is:

[0033]

[0034] In the formula, These are adjustable weighting coefficients, and N is the number of water subsystems; , These represent the actual and predicted water consumption of the i-th subsystem, respectively. , , These are reclaimed water, total water consumption, and rainwater, respectively. This refers to the energy consumption and maintenance costs of the system.

[0035] Optionally, control instructions are generated based on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem, including: performing parameter mapping on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem based on a preset mapping rule to obtain a set of control parameters for water supply scheduling for each water subsystem, wherein the set of control parameters includes: priority of water supply mode for each water subsystem, valve opening percentage, pump station start and stop time, and filter backwashing time and frequency controlled by backwashing quantity.

[0036] Optionally, the expression for the optimal solution vector of the minimum cost water supply scheduling for each water subsystem is:

[0037]

[0038] In the formula, X is the optimal solution vector that minimizes the water supply scheduling cost for each water-using subsystem. Let be the water supply from water source s for the i-th water-using subsystem during the scheduling period t; Let m be the allocation ratio of the water supply method m to the i-th water subsystem, where m includes reclaimed water, rainwater and municipal water. For pump station control quantities, For valve control quantity, To control the dosage for disinfection or purification, For backwashing control of the filter bed;

[0039] The valve opening percentage is controlled by the valve control quantity. The mapping yields the following:

[0040]

[0041] in This represents the valve opening percentage. This is a cutoff function used to limit the control quantity between 0 and 1;

[0042] The start-up and shutdown periods of the pumping station are controlled by the pumping station. The mapping yields the result that satisfies... When a pump start command is generated, A pump stop command is generated at the specified time, in which The preset start / stop threshold is used; the pump frequency setting value is determined by... Mapped to the inverter setpoint using a linear ratio;

[0043] The priority of the water supply method is determined by an allocation ratio. Confirmed, will The water supply priorities are arranged from highest to lowest to form a sequence, and instructions for switching or mixing reclaimed water, rainwater, and municipal water are generated accordingly. When a water quality warning triggers a level one warning, the corresponding priority is forcibly switched to municipal water. and Set to zero and switch to municipal water supply;

[0044] The backwashing time and frequency of the filter are controlled by the backwashing quantity. The mapping yields the following condition: when A backwash action is triggered once within the scheduling cycle, and the backwash duration is set to [time value]. ,in The backwash trigger threshold, This is a mapping function that maps control quantities to backwash duration.

[0045] Optionally, control commands can be generated based on the warning information, including:

[0046] The warning messages are classified according to the preset levels to obtain the warning message level, and a water supply switching command is generated based on the warning message.

[0047] If the warning message is a Level 1 warning, an instruction will be generated to switch the toilet flushing subsystem to the municipal water source;

[0048] If the warning message is a level 2 warning, an instruction to mix water in the toilet flushing subsystem will be generated.

[0049] Optionally, the control commands are sent to the PLC controller associated with each water subsystem to manage the operating status of the equipment in each water subsystem, including:

[0050] The water supply mode priority, valve opening percentage, pump station start / stop time, filter backwash time and frequency are encapsulated into instruction data frames by backwash control quantities and water supply switching instructions, and written into the PLC variable area or Modbus register area according to the preset address mapping relationship. The PLC reads the variables or registers during the scan cycle and outputs control signals to the solenoid valve / proportional valve, frequency converter pump and disinfection / backwash device. At the same time, the PLC writes back the valve opening feedback, pump operating status, warning information and key water quality and quantity data to the status register area for the cloud platform to read, so as to form feedback and be used for correction in the next scheduling cycle.

[0051] Secondly, the present invention provides a water-saving management system for high-speed service areas based on the Internet of Things, comprising:

[0052] Data acquisition unit, water resource supply and demand forecasting unit, control parameter generation unit, and distribution unit;

[0053] The data acquisition unit is used to acquire historical water consumption, water quality data, meteorological information, passenger flow, and liquid levels in the reclaimed water tank and rainwater storage tank uploaded by the controller.

[0054] The water resource supply and demand prediction unit is used to process historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks based on a pre-trained water load prediction model to obtain water load prediction values; and compares the water quality data with pre-stored water quality thresholds under different water use scenarios to obtain warning information.

[0055] The control parameter generation unit is used to extract data based on the predicted water load value, obtain the predicted water consumption, and substitute the predicted water consumption into the pre-constructed operating cost objective function for solution, thereby obtaining the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem; and generate control commands based on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem and warning information.

[0056] The distribution unit is used to send the control commands to the controllers associated with each water subsystem in order to manage the operating status of the execution devices of each water subsystem.

[0057] Optionally, the distribution unit is connected to the PLC controller in each water subsystem;

[0058] Each PLC controller is connected to the data acquisition unit and is used to upload historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of the reclaimed water tank and rainwater storage tank.

[0059] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0060] 1. This invention deeply integrates water use control, environmental sensing, water quality monitoring, and passenger flow data through an Internet of Things (IoT) architecture. It achieves comprehensive data collection from all elements of the service area, including water sources, water pipes, water usage points, weather, and passenger flow. Combined with a water load prediction model based on a random forest regression prediction algorithm, it performs comprehensive prediction of multi-source data and uses an operating cost objective function to calculate the optimal solution vector for minimizing the water supply scheduling cost of each water subsystem. Finally, it generates a set of control instructions for the optimal water supply scheduling cost of each water subsystem, which varies with multiple factors such as weather, season, and passenger flow. Compared with existing solutions, this invention can achieve on-demand water supply and precise resource allocation in highway service areas, avoid energy waste caused by excessive water supply, optimize the start-stop strategies of pumps and valves in each water system, and reduce equipment failures.

[0061] 2. The prediction model of this invention supports multi-scale prediction across multiple time periods, possesses good online learning and updating capabilities, and transforms the water use scheduling results of multiple water sources and scenarios in the service area into executable pump station start / stop, valve opening, water supply switching, and backwashing timing control commands through a closed-loop link of "data acquisition - prediction judgment - objective function solution - control command issuance". Combined with water quality threshold judgment, it forms a graded alarm and water supply adjustment strategy, thereby improving the utilization rate of reclaimed water and rainwater, reducing municipal water intake, and enhancing system operation stability while meeting the water quality safety constraints of reclaimed water.

[0062] 3. Real-time detection of abnormal water quality and automatic generation of warning information; intelligent management of the water supply switching process of the toilet flushing subsystem and reduction of water use failures. Attached Figure Description

[0063] Figure 1 The flowchart shown is a process for water-saving management of high-speed service areas based on the Internet of Things according to the present invention.

[0064] Figure 2 The diagram shown is a structural diagram of the Internet of Things-based water-saving management system for high-speed service areas according to the present invention.

[0065] Figure 3 The diagram shown is a logic diagram of the water management method based on threshold determination and predictive scheduling in the cloud platform of this invention.

[0066] Figure 4 The diagram shown illustrates the data transmission and cloud-edge collaborative processing of this invention.

[0067] Figure 5 The diagram shown is a closed-loop control diagram of the toilet flushing subsystem of the present invention.

[0068] Figure 6 The diagram shown is a schematic of the water load prediction model of the present invention. Detailed Implementation

[0069] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0070] Example 1

[0071] This embodiment provides a service area water conservation management method based on the Internet of Things, including:

[0072] Step 1: Obtain historical water consumption, water quality data, meteorological information, passenger flow, and liquid levels in the reclaimed water tank and rainwater storage tank uploaded by the PLC controller;

[0073] Step 2: Based on the pre-trained water load prediction model, process historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks to obtain water load prediction values; and compare the water quality data with the pre-stored water quality thresholds for different water use scenarios to obtain warning information.

[0074] Step 3: Extract data based on the predicted water load to obtain the predicted water consumption, and substitute the predicted water consumption into the pre-constructed operating cost objective function to obtain the optimal solution vector for minimizing the water supply scheduling cost of each water subsystem.

[0075] Step 4: Generate control commands based on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem and the warning information;

[0076] Step 5: Send control commands to the PLC controllers associated with each water subsystem to control the operating status of the equipment in each water subsystem.

[0077] This implementation example Figure 1 The management method shown is executed by a cloud platform. The cloud platform and PLC controllers communicate via the Internet of Things (IoT) to receive all real-time data uploaded by the PLC controllers of each water subsystem. Each water subsystem PLC controller is connected to the sensors and valves of each execution unit, thereby enabling data acquisition and control of water pipe valves.

[0078] In some possible implementations, historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of the reclaimed water tank and rainwater storage tank uploaded by the PLC controller are obtained, including:

[0079] Flow and pressure information is collected by flow sensors and pressure / differential pressure sensors configured in the water supply pipelines of the toilet flushing subsystem, greening irrigation subsystem, and domestic water subsystem, and historical water consumption is calculated based on the flow information;

[0080] Water quality information is collected by water quality monitoring instruments configured at preset water quality monitoring nodes to form water quality data;

[0081] Liquid level information is collected by level sensors configured in the reclaimed water tank and the rainwater storage tank to form the liquid level of the reclaimed water tank and the liquid level of the rainwater storage tank.

[0082] Obtain meteorological information through meteorological forecast data interfaces;

[0083] Passenger flow can be obtained through passenger flow monitoring systems or operational data interfaces.

[0084] In this embodiment, the accuracy level of the flow sensor is no less than ±0.5%. The flow sensor adopts electromagnetic and ultrasonic flow meters and is installed on toilet flushing pipelines, irrigation main pipes and domestic water main pipes to realize real-time acquisition of instantaneous flow rate (unit L / min) and cumulative water consumption (m³).

[0085] The pressure / differential pressure sensor has a measurement range of 0 to 0.5 MPa and an accuracy of no less than ±0.5%. The differential pressure sensor is placed before and after the pipeline to detect changes in pipeline differential pressure and help identify abnormal conditions such as pipeline blockage and pump station failure.

[0086] The liquid level sensor has a measurement range of 0–5 m and an accuracy of ±1.5%. Installed in rainwater storage tanks, reclaimed water tanks, and pretreatment tanks, the sensor employs ultrasonic or capacitive level gauges to monitor real-time water level changes in the storage tanks and reclaimed water tanks. The online water quality monitoring instrument is installed at key points such as the inlet water inlet, wastewater outlet, and reclaimed water outlet. The instrument can measure the following parameters: COD (Chemical Oxygen Demand, range 0.1–500 mg / L, lower limit 0.1 mg / L); TDS (Total Dissolved Solids, range 0–2000 mg / L); Turbidity (range 0–1000 NTU); pH (range 0–14); DO (Dissolved Oxygen, range 0–20 mg / L). The monitoring instrument must have automatic temperature compensation and online calibration functions to ensure long-term stable operation.

[0087] This embodiment obtains weather, passenger flow, and operational data from other systems through meteorological forecast data interfaces, passenger flow monitoring systems, and operational data interfaces, and effectively integrates multiple factors such as meteorology, season, and passenger flow with real-time data from various water subsystems in the service area.

[0088] The hardware structure in this embodiment is as follows: Figure 2 The system consists of a perception layer, a data acquisition and transmission layer, a cloud platform layer, and an execution layer. The perception layer deploys various sensors to collect water quantity and quality data. The acquired data is transmitted to the transmission layer, enabling bidirectional interaction between local real-time control and cloud data via a PLC / edge gateway. The cloud platform layer handles data storage, machine learning prediction, and optimized scheduling. The execution layer uses field actuators (pumps, valves, disinfection devices, etc.) to perform actual water-saving control actions. The hardware modules work collaboratively to form a closed-loop feedback control system.

[0089] On-site, signals from various sensors in the sensing layer are collected via a PLC controller, and data exchange is completed using fieldbus protocols such as MODBUS RTU. The PLC controller has built-in ladder logic programs and custom algorithms, possessing local logic control capabilities, and can maintain local operation according to preset logic even when the network connection is lost. Data transmission uses multiple communication methods such as NB-IoT, LoRaWAN, 4G / 5G, and wired Ethernet. The sampled data is encrypted before being uploaded to the cloud platform. The system has built-in detection and fault-tolerant reconnection mechanisms to ensure uninterrupted data transmission after a disconnection, and also supports local offline caching. Furthermore, when the network is unavailable, the PLC / edge gateway stores key data (including average flow, maximum flow, alarm logs, etc.) locally and automatically uploads them in batches after the network is restored; it can also trigger local alarms and execute contingency plans (such as temporarily shutting down non-critical water circuits) based on preset thresholds.

[0090] Based on relevant national and industry standards (e.g., GB3838-2002 Surface Water Quality Standard, GB5749-2022 Drinking Water Standard), water quality thresholds are established for different water use scenarios. Real-time reported water quality parameters are compared, and if any exceedances or abnormalities are found (e.g., COD > 50 mg / L, TDS > 500 mg / L, turbidity > 5 NTU), an alarm is triggered and the on-duty personnel are notified via push notification. At the same time, according to preset logic, the allocation of this water source to the corresponding water use scenario can be automatically restricted to ensure water safety.

[0091] Finally, the execution layer targets equipment in various water subsystems, such as toilet flushing subsystems, greening irrigation subsystems, and domestic water subsystems, including toilet flushing solenoid valves, frequency converters, biochemical reactors, ozone / sterilization equipment, and biochemical reactors.

[0092] In some possible implementations, the water quality threshold is provided by a water quality threshold table stored on the cloud platform. The water quality threshold table includes at least: a scene identifier field, a water source type field, an indicator field, a threshold type field, a threshold value field, a threshold version number field, and an effective time field; and the threshold record that is in effect is called when comparing thresholds.

[0093] In this embodiment, the platform layer calls a pre-trained load prediction model to process the multi-source data obtained in step 1 and outputs a predicted water load value. Simultaneously, the platform layer calls a threshold judgment module to read a pre-stored water quality threshold table, compares the real-time water quality data with the corresponding scenario thresholds, and outputs a warning message. The warning message includes at least the alarm level and the alarm-related scenario, used to indicate whether subsequent scheduling strategies should tighten constraints or the water supply mode should be switched.

[0094] The thresholds are derived from a pre-stored water quality threshold table, which corresponds to the water quality boundary conditions for different water use scenarios, making the threshold determination traceable. When the water quality is determined to meet the threshold, it enters the normal strategy branch; when the water quality is determined to exceed the threshold, it enters the alarm strategy branch, forming a split between "compliance - scheduling optimization" and "alarm - strategy switching / mixed water distribution".

[0095] To provide a feasible structural basis for the threshold determination process, a pre-stored water quality threshold table is established using scenarios as indexes and stored in the cloud platform database to provide threshold basis for the threshold determination module. The threshold table includes at least the following fields: scenario identifier field (corresponding to toilet flushing, greening irrigation, road washing, and equipment cooling), water source type field (corresponding to municipal water, rainwater, and reclaimed water), indicator field (including one or more of COD, turbidity, pH, TDS, and DO), threshold type field (including compliance threshold, first-level alarm threshold, and second-level alarm threshold), threshold value field, and applicable condition field (including season, temperature, operating mode, or reuse level). In step 2, the threshold determination module determines the query key based on the current water use scenario and water source type, reads the threshold of the corresponding indicator and compares it with real-time water quality data, and outputs warning information containing alarm level and associated scenario to drive subsequent tightening of scheduling constraints or switching of water supply mode.

[0096] In some possible implementations, the water load prediction model is constructed using a random forest regression prediction model, and the prediction object of the water load prediction model is set as the total water consumption and / or component water consumption of the service area. The prediction step size of the water load prediction model is a preset time period, which can be any one of 1 hour, 6 hours, or 24 hours. The prediction output is determined to be the predicted value of the water load prediction model within the preset time period Δt in the future. Its unit is m³ / h;

[0097] The random forest regression prediction model consists of M regression decision trees; each regression decision tree obtains a training subset through bootstrapping and randomly draws from the input features when splitting nodes. The candidate features are split using the minimum mean square error as the splitting criterion. A leaf node is generated when the maximum depth or the minimum number of samples in the leaf node is met. The output of the leaf node is the mean of the labels of the samples falling into that leaf node. After training M regression decision trees respectively, the outputs of each regression decision tree are integrated to obtain the water load prediction value.

[0098] In this embodiment, the actual water load corresponding to a future preset time period Δt is used as the supervised learning label value. The input feature x(t) at each time step is paired with the corresponding label y(t) to form a training sample pair, forming a training dataset; and the training set and validation set are divided in chronological order.

[0099] During training, M regression decision trees are trained separately, and the outputs of each regression decision tree are integrated to obtain the predicted value of the water load prediction model. The integration method is to take the average value.

[0100]

[0101] in, This represents the output of the j-th regression decision tree, where M is the number of regression decision trees, and Δt represents the predicted output for a preset future time period. For the preset time period, For input features.

[0102] Based on the training results, the importance of the input features is calculated and feature selection is performed. Features with a contribution higher than a preset threshold or ranking in the top K are retained to form the final input feature set. A random forest regression predictor is then retrained on this final input feature set to obtain a water load prediction model based on random forest regression. The output of the water load prediction model is then... As a component of the water resource supply and demand forecast results in step 2, and used to solve the operating cost objective function in step 3, the optimal solution vector for minimizing the water supply scheduling cost of each water-using subsystem is obtained.

[0103] Data is extracted based on the predicted water load values ​​to obtain the predicted water consumption. Specifically, this involves extracting data from the predicted values ​​of the water load forecasting model. The data extraction process is triggered, extracting relevant variables and statistical parameters such as passenger flow, rainfall, and liquid level for the corresponding time period according to the timestamp. The predicted values ​​are then truncated or smoothed according to a threshold, and converted according to a preset time step t to obtain the predicted water consumption. .

[0104] In some possible implementations, the training process for the water load prediction model includes:

[0105] Collect historical data on water consumption, water quality, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks;

[0106] Historical data is preprocessed to obtain preprocessed historical data;

[0107] Feature extraction was performed on the preprocessed historical data to obtain historical time-of-use water consumption sequence features, water quality status features, meteorological features, passenger flow features, reclaimed water tank level features, and rainwater storage tank level features.

[0108] Feature vectors are generated based on historical time-of-use water consumption sequence characteristics, water quality characteristics, meteorological characteristics, passenger flow characteristics, reclaimed water tank level characteristics, and rainwater storage tank level characteristics. Supervised learning label values ​​are then assigned to the feature vectors to obtain training sample pairs. The supervised learning label value is the actual water load corresponding to a future preset time period.

[0109] The training sample pairs are divided into training and validation sets according to chronological order.

[0110] The training set is input into the regression model based on the random forest algorithm to train and obtain the initial water load prediction model;

[0111] The validation set is imported into the initial water load prediction model for prediction, and the water load prediction value of the validation set is obtained.

[0112] The prediction error of the actual and predicted water load values ​​associated with the validation set is calculated.

[0113] The initial water load prediction model is updated based on a preset error threshold or a preset number of training iterations. The water load prediction model is stopped when the prediction error between the actual and predicted water load values ​​exceeds the error threshold for K consecutive time periods or the number of training iterations reaches the upper limit. The final water load prediction model is then obtained and used to predict historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks that are uploaded in real time.

[0114] This embodiment sets the set of parameters to be optimized in the random forest regression predictor. Where M is the number of regression decision trees and D is the maximum depth. The minimum number of samples for a leaf node. The number of candidate features is randomly selected during node splitting; the parameter set is evaluated using grid search or cross-validation. Optimize the algorithm to minimize the error metric on the validation set. The error metric is either the mean absolute percentage error (MAPE) or the root mean square error (RMSE).

[0115] Obtaining the optimal parameter set Then, a random forest regression predictor is trained based on the training set to obtain the model version. The prediction error is calculated on the validation set, and the model version is solidified when the prediction error meets a preset threshold. Used for water load forecasting in step 2.

[0116] In practical applications, the actual water load transmitted back from the PLC is continuously acquired. and compared with water load forecast values Calculate relative deviation

[0117]

[0118] When the relative deviation e(t) exceeds the threshold ε for K consecutive time periods or reaches the preset update cycle, the model is updated: the latest data is added to the training dataset and the training and update operations are re-executed to obtain the updated model version, and the final version of the model is used for water load prediction.

[0119] In addition, the data preprocessing process in this embodiment includes time alignment, missing value imputation, outlier removal and smoothing, and finally obtains historical data that eliminates the dimensional differences between different parameters.

[0120] In some possible implementations, the expression for calculating the objective function of running cost is:

[0121]

[0122] In the formula, These are adjustable weighting coefficients, and N is the number of water subsystems; , These represent the actual and predicted water consumption of the i-th subsystem, respectively. , , These are reclaimed water, total water consumption, and rainwater, respectively. This refers to the energy consumption and maintenance costs of the system.

[0123] In some possible implementations, control instructions are generated based on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem, including: performing parameter mapping on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem based on a preset mapping rule to obtain a set of control parameters for water supply scheduling for each water subsystem, wherein the set of control parameters includes: priority of water supply mode for each water subsystem, valve opening percentage, pump station start and stop time, filter backwash time and frequency controlled by backwashing quantity.

[0124] In some possible implementations, the expression for the optimal solution vector of the minimum cost water supply scheduling for each water subsystem is:

[0125]

[0126] In the formula, X is the optimal solution vector that minimizes the water supply scheduling cost for each water-using subsystem. Let be the water supply from water source s for the i-th water-using subsystem during the scheduling period t; Let be the allocation ratio of the i-th water subsystem to water supply method m, where water supply method m includes reclaimed water, rainwater, and municipal water; For pump station control quantities, For valve control quantity, To control the dosage for disinfection or purification, For backwashing control of the filter bed;

[0127] The valve opening percentage is controlled by the valve control quantity. The mapping yields the following:

[0128]

[0129] in This represents the valve opening percentage. This is a cutoff function used to limit the control quantity between 0 and 1;

[0130] The start and stop times of the pumping station are controlled by the pumping station. The mapping yields the result that satisfies... When a pump start command is generated, A pump stop command is generated at the specified time, in which The preset start / stop threshold is used; the pump frequency setting value is determined by... Mapped to the inverter setpoint using a linear ratio;

[0131] Water supply method priority is determined by allocation ratio Confirmed, will The water supply priorities are arranged from highest to lowest to form a sequence, and instructions for switching or mixing reclaimed water, rainwater, and municipal water are generated accordingly. When a water quality warning triggers a level one warning, the corresponding priority is forcibly switched to municipal water. and Set to zero and switch to municipal water supply;

[0132] The backwashing time and frequency of the filter bed are controlled by the backwashing volume. The mapping yields the following condition: when A backwash action is triggered once within the scheduling cycle, and the backwash duration is set to [time value]. ,in The backwash trigger threshold, This is a mapping function that maps control quantities to backwash duration.

[0133] This implementation example Figure 3 As shown, the water load prediction model outputs the predicted water load value. The optimization scheduling module solves the water supply volume and water supply mode priority of each subsystem based on the operating cost objective function. The scheduling result is encoded into a PLC executable write quantity by the instruction generation module, including valve opening setting, pump start and stop status, water supply mode switching flag and operating mode selection, and sent to the PLC controller to drive the field actuator.

[0134] The on-site PLC will transmit the execution status and key feedback quantities back to the platform, including actual water consumption. Actual valve opening value, pump start / stop status, and tank level changes. When K consecutive time steps meet... When the time comes, online correction or incremental retraining is triggered: the latest data is appended as incremental samples to update the model parameters or compensate for the output bias, and the updated prediction results are used in the next round of scheduling. This forms a closed-loop causal chain of "prediction error - scheduling bias - execution feedback - model correction - rescheduling".

[0135] In some possible implementations, control commands are generated based on the warning information, including:

[0136] The warning messages are classified according to the preset levels to obtain the warning message level, and a water supply switching command is generated based on the warning message.

[0137] If the warning message is a Level 1 warning, an instruction will be generated to switch the toilet flushing subsystem to the municipal water source;

[0138] If the warning message is a level 2 warning, an instruction will be generated to mix and distribute reclaimed water, rainwater, and municipal water from the toilet flushing subsystem.

[0139] This implementation example Figure 5 As shown, taking the toilet flushing subsystem as an example, the platform layer sends the valve opening control quantity and water supply mode switching strategy for toilet flushing to the PLC. The PLC drives the actuator to complete the toilet flushing water supply control. The water quality evaluation result determines the toilet flushing water supply strategy branch: when the water quality evaluation is up to standard, the PLC controls the water supply according to the valve opening setting sent by the platform layer; when a Level I warning is triggered, the platform layer sends a water supply switching strategy, and the PLC executes the water supply mode switching; when a Level II warning is triggered, the platform layer sends a mixed water distribution strategy, and the PLC executes the mixed water distribution control. After the toilet is flushed, the water output detection and status reporting module returns key feedback quantities to the platform layer. The platform layer monitors deviations and triggers cloud platform rescheduling when necessary to ensure that water-saving goals are achieved while meeting safety constraints.

[0140] In some possible implementations, such as Figure 4 The control commands are sent to the PLC controllers associated with each water subsystem to manage the operating status of the equipment in each water subsystem, including:

[0141] The priority of water supply mode, valve opening percentage, pump station start / stop time, and filter backwash time and frequency are encapsulated into instruction data frames by backwash control quantities and water supply switching instructions, and written into the PLC variable area or Modbus register area according to the preset address mapping relationship. The PLC reads variables or registers during the scan cycle and outputs control signals to solenoid valves / proportional valves, variable frequency pumps, and disinfection / backwashing devices. At the same time, the PLC writes back valve opening feedback, pump operating status, warning information, and key water quality and quantity data to the status register area for the cloud platform to read, so as to form feedback and be used for correction in the next scheduling cycle. Data interaction between the control layer and the platform layer is carried out through an encrypted transmission channel. The edge controller / gateway has local caching and breakpoint resume capability, maintains data integrity when the network is abnormal, and retransmits after the network is restored. The control instructions generated by the cloud platform are encoded before being issued, and the control quantities are mapped to write quantities that can be executed by the PLC to realize an executable closed loop of "analysis and decision-making - on-site control". The cloud platform uses a distributed database to store historical sensor data, incoming water quality, reclaimed water utilization rate, rainwater collection volume, passenger flow, meteorological information, etc.; it also supports time series databases for storing high-frequency sensor data to facilitate subsequent time series analysis.

[0142] To ensure the feasibility of "algorithm output - on-site execution," control commands are encoded into control quantities before being issued and mapped to PLC-executable variable writes or register writes. Specifically, the platform layer converts the valve opening setpoint, pump start / stop status, water supply mode switching flag, and reuse processing operation mode from the scheduling results into standardized control words or numerical quantities: the valve opening setpoint is normalized to 0-100% of its range and mapped to a PLC analog setpoint variable; the pump start / stop status and water supply mode switching flag are mapped to PLC digital status variables; and the reuse processing operation mode is mapped to a PLC enumerated variable or mode selection register. Subsequently, the platform layer writes the corresponding address to the PLC via the communication link to achieve real-time control of the actuator. After completing the write parsing, the PLC executes the control action and sends feedback quantities such as actual opening degree, start / stop status, water supply flow rate, and effluent water quality back to the platform layer for deviation calculation and online correction or rescheduling triggering. Control instructions include instruction sequence numbers and verification fields. Before execution, the PLC deduplicates the sequence numbers and performs consistency checks on the verification fields to avoid duplicate execution or abnormal writing caused by communication jitter.

[0143] Example 2

[0144] This embodiment provides a water-saving management system for high-speed service areas based on the Internet of Things, including:

[0145] Data acquisition unit, water resource supply and demand forecasting unit, control parameter generation unit, and distribution unit;

[0146] The data acquisition unit is used to acquire historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of the reclaimed water tank and rainwater storage tank uploaded by the controller;

[0147] The water resource supply and demand forecasting unit is used to process historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks based on a pre-trained water load forecasting model to obtain water load forecast values; and compares the water quality data with pre-stored water quality thresholds under different water use scenarios to obtain warning information.

[0148] The control parameter generation unit is used to extract data based on the water load forecast, obtain the predicted water consumption, and substitute the predicted water consumption into the pre-built operating cost objective function for solution, so as to obtain the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem; and generate control commands based on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem and the warning information.

[0149] The distribution unit is used to send the control commands to the controllers associated with each water subsystem in order to manage the operating status of the execution devices of each water subsystem.

[0150] The data processing procedure of the water load prediction model in this embodiment is as follows: Figure 6 As shown, the water load prediction model in this embodiment differs from the general load prediction model. This embodiment uses the pool level and supply margin as model inputs, takes the fluctuation of passenger flow in the service area and holiday characteristics as key features, and incorporates the prediction output into the water supply optimization solution and PLC execution closed-loop correction mechanism. The above coupling relationship enables the model not only to be used for data fitting, but also to directly act on the water supply control process, thereby obtaining verifiable system performance improvements.

[0151] The prediction model adopted is a water load prediction model based on random forest regression, which is used to output the water load prediction value of the service area for a future preset period under the given multi-source input conditions such as historical water use, water quality, passenger flow, weather and pool liquid level. The prediction value serves as the key input of the optimization scheduling module, which is used to generate the optimal water supply and water supply mode priority of each subsystem, and further generate control instructions for PLC.

[0152] The prediction model based on random forest regression consists of multiple regression decision trees. Through operations such as bootstrapping and candidate feature extraction, the generalization ability of the model can be improved and the diversity of the model can be increased. At the same time, the mean operation is performed in the leaf nodes of multiple decision trees to reduce the sensitivity to outlier samples. After training multiple regression decision trees separately, the water load prediction value of the integrated multiple decision trees is obtained, which reduces the risk of overfitting.

[0153] In some possible implementations, the distribution unit is connected to the PLC controller in each water subsystem;

[0154] Each PLC controller is connected to a data acquisition unit to upload historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of the reclaimed water tank and rainwater storage tank.

[0155] This embodiment, while meeting water quality threshold constraints, increases the proportion of reclaimed water and rainwater in the total water supply and reduces municipal water consumption; by reducing prediction bias and supply-demand mismatch, it reduces the frequency of pump station start-ups and shutdowns and valve operations, thereby improving the stability of the water supply process; by incorporating alarm information into scheduling constraints, it enables restrictions on the use of substandard water sources or switching of water supply methods, reducing the risk of substandard water entering sensitive water use scenarios; under fluctuating water load conditions (such as peak holiday periods), it enables scheduling calculations to output segmented water supply volumes that better match the actual situation, reducing ineffective discharges caused by oversupply or frequent switching caused by undersupply; through online correction or incremental update mechanisms, it enables the model to adaptively update with operational data, reducing long-term drift caused by seasonal changes, changes in passenger flow structure, or changes in pool supply availability, thereby improving the predictive availability and scheduling effectiveness of long-term operation.

[0156] The platform layer uses supply and demand forecasts, water quality warning constraints, and the availability of water in the reservoir as optimization inputs to obtain the optimal solution vector for minimizing the water supply scheduling cost of each water-using subsystem. The weight coefficients in the objective function are adjustable to achieve a comprehensive trade-off between water supply stability, reclaimed water utilization, rainwater utilization, and operating costs. The scheduling results output from the solution are further converted into executable control variables in step 4. These are then written to PLC variables or registers through control instruction encoding, ensuring that the scheduling results can drive the execution layer to complete the actions of valves, pump stations, and treatment units, thereby producing verifiable technical effects.

[0157] In summary, this invention deeply integrates water use control, environmental sensing, water quality monitoring, and passenger flow data through an IoT architecture. It achieves comprehensive data collection across all service areas, including water sources, pipes, water usage points, weather, and passenger flow. Combined with a water load prediction model based on a random forest regression algorithm, it performs comprehensive predictions of multi-source data and uses an operating cost objective function to calculate the optimal solution vector for minimizing the water supply scheduling cost of each water subsystem. Finally, it generates a set of control commands for the optimal water supply scheduling cost of each water subsystem, which varies with multiple factors such as weather, season, and passenger flow. Compared to existing solutions, this invention enables on-demand water supply and precise resource allocation in highway service areas, avoids energy waste caused by excessive water supply, optimizes pump and valve start-stop strategies for each water system, and reduces equipment failures. Secondly, the predictive model of this invention supports multi-scale prediction across multiple time periods, possesses good online learning and updating capabilities, and, through a closed-loop chain of "data acquisition—prediction and judgment—objective function solution—control command issuance," transforms the water usage scheduling results of multiple water sources and scenarios in the service area into executable pump station start / stop, valve opening, water supply switching, and backwashing timing control commands. Combined with water quality threshold determination, it forms a tiered alarm and water supply adjustment strategy, thereby improving the utilization rate of reclaimed water and rainwater, reducing municipal water intake, and enhancing system operational stability while meeting the safety constraints of reclaimed water quality. Finally, this invention can detect abnormal water quality in real time and automatically generate warning information, intelligently managing the water supply switching process of the toilet flushing subsystem.

[0158] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0159] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0160] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0161] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0162] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A water-saving management method for high-speed service areas based on the Internet of Things, characterized in that, The management method is executed by the cloud platform and includes: Acquire historical water consumption, water quality data, meteorological information, passenger flow, and liquid levels in the reclaimed water tank and rainwater storage tank uploaded by the controller; Based on a pre-trained water load prediction model, historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks are processed to obtain water load prediction values; and the water quality data is compared with pre-stored water quality thresholds under different water use scenarios to obtain warning information. Data is extracted based on the predicted water load to obtain the predicted water consumption. The predicted water consumption is then substituted into the pre-constructed operating cost objective function to obtain the optimal solution vector for minimizing the water supply scheduling cost of each water subsystem. Control commands are generated based on the optimal solution vector of each water supply subsystem with the minimum cost of water supply scheduling and warning information. The control commands are sent to the controllers associated with each water subsystem to manage the operating status of the devices in each water subsystem.

2. The service area water-saving management method based on the Internet of Things according to claim 1, characterized in that, Acquire historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of the reclaimed water tank and rainwater storage tank uploaded by the controller, including: Flow and pressure information are collected by flow sensors and pressure / differential pressure sensors configured in the water supply pipelines of the toilet flushing subsystem, greening irrigation subsystem, and domestic water subsystem, and the historical water consumption is calculated based on the flow information. Water quality information is collected by water quality monitoring instruments configured at preset water quality monitoring nodes to form the water quality data; Liquid level information is collected by level sensors configured in the reclaimed water tank and the rainwater storage tank to form the liquid level of the reclaimed water tank and the liquid level of the rainwater storage tank. The meteorological information is obtained through a meteorological forecast data interface; The passenger flow is obtained through a passenger flow monitoring system or an operational data interface.

3. The service area water-saving management method based on the Internet of Things according to claim 1, characterized in that, The water quality threshold is provided by a water quality threshold table stored on the cloud platform. The water quality threshold table includes at least: a scene identifier field, a water source type field, an indicator field, a threshold type field, a threshold value field, a threshold version number field, and an effective time field; and the threshold record that is in effect is called when comparing thresholds.

4. The service area water-saving management method based on the Internet of Things according to claim 1, characterized in that, The water load prediction model is constructed using a random forest regression model. The prediction objects of the water load prediction model are defined as the total water consumption and / or individual water consumption of the service area. The prediction step size of the water load prediction model is a preset time period, which can be any one of 1 hour, 6 hours, or 24 hours. The prediction output is determined to be the predicted value of the water load prediction model within the preset time period Δt. Its unit is m³ / h; The random forest regression prediction model consists of M regression decision trees; each regression decision tree obtains a training subset through bootstrapping and randomly draws from the input features when splitting nodes. The candidate features are split using the minimum mean square error as the splitting criterion. A leaf node is generated when the maximum depth or the minimum number of samples in the leaf node is met. The output of the leaf node is the mean of the labels of the samples falling into that leaf node. After training M regression decision trees respectively, the outputs of each regression decision tree are integrated to obtain the water load prediction value.

5. The service area water-saving management method based on the Internet of Things according to claim 1, characterized in that, The training process of the water load prediction model includes: Collect historical data on water consumption, water quality, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks; Historical data is preprocessed to obtain preprocessed historical data; Feature extraction was performed on the preprocessed historical data to obtain historical time-of-use water consumption sequence features, water quality status features, meteorological features, passenger flow features, reclaimed water tank level features, and rainwater storage tank level features. Feature vectors are generated based on historical time-of-use water consumption sequence characteristics, water quality characteristics, meteorological characteristics, passenger flow characteristics, reclaimed water tank level characteristics, and rainwater storage tank level characteristics. Supervised learning label values ​​are then assigned to the feature vectors to obtain training sample pairs. The supervised learning label value is the actual water load corresponding to a future preset time period. The training sample pairs are divided into training and validation sets according to chronological order. The training set is input into the regression model based on the random forest algorithm to train and obtain the initial water load prediction model; The validation set is imported into the initial water load prediction model for prediction, and the water load prediction value of the validation set is obtained. The prediction error of the actual and predicted water load values ​​associated with the validation set is calculated. The initial water load prediction model is updated based on a preset error threshold or a preset number of training iterations. The water load prediction model is stopped when the prediction error between the actual and predicted water load values ​​exceeds the error threshold for K consecutive time periods or the number of training iterations reaches the upper limit. The final water load prediction model is then obtained and used to predict historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks that are uploaded in real time.

6. The service area water conservation management method based on the Internet of Things according to claim 1, characterized in that, The calculation expression for the objective function of the operating cost is as follows: In the formula, These are adjustable weighting coefficients, and N is the number of water subsystems; , These represent the actual and predicted water consumption of the i-th subsystem, respectively. , , These are reclaimed water, total water consumption, and rainwater, respectively. This refers to the energy consumption and maintenance costs of the system.

7. The service area water conservation management method based on the Internet of Things according to claim 1, characterized in that, The control instructions are generated based on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem. This includes: mapping the parameters of the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem based on a preset mapping rule to obtain the control parameter set for water supply scheduling of each water subsystem. The control parameter set includes: the priority of water supply mode for each water subsystem, valve opening percentage, pump station start and stop time, and filter backwashing time and frequency controlled by backwashing variables. The expression for the optimal solution vector of the minimum cost water supply scheduling for each water subsystem is as follows: In the formula, X is the optimal solution vector that minimizes the water supply scheduling cost for each water-using subsystem. Let be the water supply from water source s for the i-th water-using subsystem during the scheduling period t; Let m be the allocation ratio of the water supply method m to the i-th water subsystem, where m includes reclaimed water, rainwater and municipal water. For pump station control quantities, For valve control quantity, To control the dosage for disinfection or purification, For backwashing control of the filter bed; The valve opening percentage is controlled by the valve control quantity. The mapping yields the following: in This represents the valve opening percentage. This is a cutoff function used to limit the control quantity between 0 and 1; The start-up and shutdown periods of the pumping station are controlled by the pumping station. The mapping yields the result that satisfies... When a pump start command is generated, A pump stop command is generated at the specified time, in which The preset start / stop threshold is used; the pump frequency setting value is determined by... Mapped to the inverter setpoint using a linear ratio; The priority of the water supply method is determined by an allocation ratio. Confirmed, will The water supply priorities are arranged from highest to lowest to form a sequence, and based on this, instructions for switching or mixing water supplies in the order of "reclaimed water - rainwater - municipal water" are generated; when a water quality warning triggers a level one warning, the corresponding priority is forcibly switched. and Set to zero and switch to municipal water supply; The backwashing time and frequency of the filter are controlled by the backwashing quantity. The mapping yields the following condition: when A backwash action is triggered once within the scheduling cycle, and the backwash duration is set to [time value]. ,in This is the backwash trigger threshold. This is a mapping function that maps control quantities to backwash duration.

8. The service area water conservation management method based on the Internet of Things according to claim 1, characterized in that, Based on the warning information, control commands are generated, including: The warning messages are classified according to the preset levels to obtain the warning message level, and a water supply switching command is generated based on the warning message. If the warning message is a Level 1 warning, an instruction will be generated to switch the toilet flushing subsystem to the municipal water source; If the warning message is a level 2 warning, an instruction to mix water in the toilet flushing subsystem will be generated.

9. The service area water-saving management method based on the Internet of Things according to claim 1, characterized in that, The control commands are sent to the controllers associated with each water subsystem to manage the operating status of the equipment in each water subsystem, including: The water supply mode priority, valve opening percentage, pump station start / stop time, filter backwash time and frequency are encapsulated into instruction data frames by backwash control quantities and water supply switching instructions, and written into the PLC variable area or Modbus register area according to the preset address mapping relationship. The PLC reads the variables or registers during the scan cycle and outputs control signals to the solenoid valve / proportional valve, frequency converter pump and disinfection / backwash device. At the same time, the PLC writes back the valve opening feedback, pump operating status, warning information and key water quality and quantity data to the status register area for the cloud platform to read, so as to form feedback and be used for correction in the next scheduling cycle.

10. A water-saving management system for high-speed service areas based on the Internet of Things, characterized in that, include: Data acquisition unit, water resource supply and demand forecasting unit, control parameter generation unit, and distribution unit; The data acquisition unit is used to acquire historical water consumption, water quality data, meteorological information, passenger flow, and liquid levels in the reclaimed water tank and rainwater storage tank uploaded by the controller. The water resource supply and demand prediction unit is used to process historical water consumption, water quality data, meteorological information, passenger flow, and the liquid levels of reclaimed water tanks and rainwater storage tanks based on a pre-trained water load prediction model to obtain water load prediction values; and compares the water quality data with pre-stored water quality thresholds under different water use scenarios to obtain warning information. The control parameter generation unit is used to extract data based on the predicted water load value, obtain the predicted water consumption, and substitute the predicted water consumption into the pre-constructed operating cost objective function for solution, thereby obtaining the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem; and generate control commands based on the optimal solution vector of the minimum cost of water supply scheduling for each water subsystem and warning information. The distribution unit is used to send the control commands to the controllers associated with each water subsystem in order to manage the operating status of the execution devices of each water subsystem.

Images