Water supply and drainage automatic control system with intelligent control

Abstract

The invention discloses a water supply and drainage automatic control system with intelligent control. Firstly, an acquisition module acquires real-time water source data every a preset time period T; secondly, the intelligent control unit inputs the collected water source data to the trained LSTM neural network model, and the LSTM neural network model performs feature extraction; the extracted water source feature vectors are input into a sparse Gaussian process model, and the sparse Gaussian process model outputs system operation parameters of each water supply and drainage automatic control system in a future time period; and finally, the control execution module adjusts the valve opening degree, the water pump power, the water pressure in the pipeline and the pipeline water flow speed in the corresponding time period according to the system operation parameters output by the sparse Gaussian process model. According to the method, the LSTM neural network is introduced to perform feature extraction of the time sequence data, fine adjustment of water source supply scheduling can be realized in combination with the sparse Gaussian process model, and the user satisfaction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water source scheduling, and specifically to a water supply and drainage automatic control system with intelligent control. Background Art

[0002] In the construction of modern urban infrastructure, the water supply and drainage system, as one of the key urban operation guarantee systems, undertakes the function of providing stable water supply and efficient drainage for various water needs such as residents, industry, and agriculture. However, traditional water supply and drainage systems usually adopt timing or manual control methods, which makes it difficult to respond to complex and changing water demand and water quality changes in real time. At the same time, with the acceleration of climate change and urbanization, the volatility and unpredictability of water demand are also gradually increasing, and the management of water supply and drainage systems faces many challenges. Therefore, the introduction of intelligent control technology to achieve efficient and flexible management of water supply and drainage systems has become a technical problem that needs to be solved urgently.

[0003] The control mode of the traditional water supply and drainage system mainly relies on fixed scheduling rules and manual adjustments. Since the water supply parameters cannot be adjusted in real time according to the actual situation, this mode is prone to insufficient water supply pressure during peak water use periods and waste of resources during low water use periods. In addition, water quality monitoring often lags behind the actual situation, making it difficult to identify and deal with sudden abnormal water quality problems in a timely manner. Once emergencies such as water pollution and pipeline leakage occur, the system's emergency response efficiency is low, posing potential risks to urban water supply safety and residents' water use experience.

[0004] With the rapid development of technologies such as the Internet of Things (IoT), big data, and artificial intelligence (AI), it has become a trend to combine these technologies with water supply and drainage systems to achieve intelligent control. IoT technology can realize the interconnection of sensor devices, allowing the system to collect relevant data on water quality, water flow, liquid level and external environment in real time, so that the control system can obtain more comprehensive information input. Artificial intelligence, especially deep learning algorithms such as long short-term memory networks (LSTM), can show superior performance in analyzing time series data, and predict future system state changes by learning historical data and real-time feedback. This capability enables the system to predict and preventively dispatch water consumption peaks and emergencies in advance, greatly improving the intelligence level of the system.

[0005] However, some current intelligent control schemes still face challenges in practical applications. For example, when processing long-term series data, the neural network model is not sensitive to the influence of factors such as seasons and rising temperatures when performing neural network training calculations and predictions, resulting in inaccurate prediction results. In addition, the urban water supply system involves multiple control terminals, and the adjustment parameters of different control terminals are different. The control parameters in the south and the north are also quite different. The control requirements of different regions in the same city are also different. Therefore, the control of water sources and water volume involves a wide range of areas and complex contents. The traditional centralized control mode cannot adapt well to the complex and changing environment and needs. Therefore, how to use advanced algorithms and intelligent control technologies to efficiently analyze the massive real-time data collected, and timely predict and adjust the system operation parameters is still a key problem in the intelligent upgrade of water supply and drainage systems. In addition, the existing dispatching water source control system has the disadvantages of slow response and waste of resources. This system not only reduces the water supply safety and resource utilization efficiency, but also cannot effectively respond to changing climate conditions and emergencies, and does not provide strong technical support for the construction of intelligent infrastructure in modern cities. In view of the above problems, it is necessary to design a water supply and drainage automatic control system with intelligent control, combined with intelligent control technology, to improve the accuracy of water source control prediction and the water resource utilization efficiency of the system. Summary of the invention

[0006] In view of the above problems mentioned in the prior art, the present invention provides a water supply and drainage automatic control system with intelligent control; including a water source unit, a pipeline system, an intelligent control unit, a water quality monitoring module, a water flow control module, a liquid level monitoring module, a fault detection and alarm module and a human-computer interaction interface; the acquisition module obtains real-time water source data every preset time period T; the water source data includes water quality data, liquid level data, external environmental parameters and system operation parameter data; the intelligent control unit inputs the collected water source data into a trained LSTM neural network model, and the LSTM neural network model performs feature extraction; and the extracted water source feature vector is input into a sparse Gaussian process model, and the sparse Gaussian process model outputs the system operation parameters of each water supply and drainage automatic control system in the future time period; the control execution module adjusts the valve opening, water pump power, water pressure in the pipeline, and water flow velocity in the pipeline in the corresponding time period according to the system operation parameters output by the sparse Gaussian process model. The present invention introduces an LSTM neural network to extract the features of time series data, and combined with the sparse Gaussian process model, it can realize fine adjustment of water supply scheduling and improve user satisfaction.

[0007] The present application provides a water supply and drainage automatic control system with intelligent control, including: a water source unit, a pipeline system, an intelligent control unit, a water quality monitoring module, a water flow control module, a liquid level monitoring module, a fault detection and alarm module and a human-computer interaction interface;

[0008] The acquisition module acquires real-time water source data every preset time period T; the water source data includes water quality data, liquid level data, external environmental parameters and system operation parameter data;

[0009] The intelligent control unit inputs the collected water source data into the trained LSTM neural network model, which performs feature extraction; and inputs the extracted water source feature vector into the sparse Gaussian process model, which outputs the system operation parameters of each water supply and drainage automatic control system in the future time period;

[0010] Control execution, according to the system operating parameters output by the sparse Gaussian process model, adjusts the valve opening, water pump power, water pressure in the pipeline, and water flow rate in the pipeline within the corresponding time period.

[0011] Preferably, the intelligent control unit includes a central processing unit, a data acquisition module, a communication module and an execution control module: the water supply and drainage automatic control system forms a distributed intelligent control network through the Internet of Things platform, and multiple water supply and drainage automatic control systems share data through the Internet of Things platform to achieve cross-regional intelligent scheduling.

[0012] Preferably, the water quality data includes pH value, water temperature, and dissolved oxygen content; the liquid level data includes the liquid level height of the water storage tank and the pipeline; the external environmental parameters include the air temperature, humidity, and air pressure at the water source; and the system operation parameter data includes valve opening, water pump power, water pressure in the pipeline, and water flow rate.

[0013] Preferably, the LSTM neural network performs feature extraction, including:

[0014] Time series input: Receive time series water source data including water quality data, liquid level data, external environmental parameters and system operation parameters, and combine real-time and historical water source data to capture short-term and long-term change trends;

[0015] Input gate processing: The input gate identifies the information about the system status change at the current time point and adds it to the system memory; the information about the system status change at the current time point includes: the air temperature rise and the water demand change curve;

[0016] Forget gate processing: Based on the data at the previous time point and the current input, the forget gate is used to decide whether to retain or discard historical information, thereby optimizing the effectiveness of system memory.

[0017] f t =σ(W f *[h t-1 ,x t ]+a t *b f )

[0018] Among them, f tis the forget gate activation value, which is used to control the amount of information retained from the previous cell state in the current time step, and its value range is between 0 and 1; σ is the tanh activation function; W f is the weight matrix of the forget gate, which is used for the linear transformation of the current input and the hidden state of the previous time step; h t-1 represents the hidden state of the previous time step; x t Represents the input data of the current time step; a t is the adaptive weight factor of the temperature at time t. When the temperature rises, the water demand increases. t increase; b f The bias vector of the forget gate is used to adjust the offset of the forget gate output;

[0019] Cell state update: The processing results of the input gate and the forget gate are integrated to dynamically update the cell state of the system to capture long-term and short-term system characteristics;

[0020] Output gate generates hidden state: The output gate generates hidden state based on the current input, previous state and cell state. The hidden state includes the peak and trough periods of water use, abnormal characteristics of water quality fluctuations, that is, the magnitude of pH value or dissolved oxygen content exceeding the set threshold;

[0021] Feature vector generation: Summarize the hidden states into a source feature vector.

[0022] Preferably, the water source characteristic vector includes water use pattern characteristics, water quality fluctuation characteristics, environmental impact characteristics and system regulation response characteristics; water use pattern characteristics include peak water use time period, valley water use time period, average flow rate and peak and valley periodic frequency; water quality fluctuation characteristics include pH value change amplitude; environmental impact characteristics include the relationship between air temperature and water flow rate change, and humidity-water quality change relationship; system regulation response characteristics include response delay time.

[0023] Preferably, the sparse Gaussian process model outputs system operating parameters of each system in a future time period, including: receiving a water source feature vector extracted by an LSTM neural network; selecting an induction point and establishing a sparse Gaussian process model; based on the input water source feature vector, using a radial basis kernel function to calculate the similarity between the input feature and the induction point to generate a covariance matrix, and constructing a joint probability distribution between the input feature and the induction point based on the covariance matrix; by splitting the joint probability distribution and combining the input feature data, using Bayesian reasoning to calculate the conditional expectation and conditional covariance, to obtain the conditional probability distribution of the system state in the future time period; calculating the expected value of the system operating parameters in the future time period according to the conditional probability distribution, and outputting the system operating parameters in the future time period, including valve opening, water pump power, pipeline water pressure and water flow rate.

[0024] Preferably, the aggregating the hidden states into the water source feature vector includes taking the average value of the hidden states of multiple time steps through average pooling to generate the water source feature vector reflecting the global characteristics.

[0025] Preferably, the selecting of induction points and establishing of a sparse Gaussian process model includes randomly selecting induction points or selecting induction points by calculation using a K-means clustering algorithm based on historical data.

[0026] Preferably, before inputting the collected water source data into the trained LSTM neural network, the method further includes:

[0027] Data cleaning: including removing outliers and filling missing values, identifying abnormal data through the set threshold range, and filling missing data using interpolation or regression models;

[0028] Data normalization: standardize the water source data to the same feature space and remove high-frequency noise through Gaussian filtering algorithm;

[0029] Time series resampling: The water source data is resampled according to the preset time interval T to ensure the time series consistency of the input data.

[0030] Preferably, the a t is the temperature adaptive weight factor. When the temperature rises, the water demand increases. t Increase, Among them, a0 is the basic value of the weight factor, Z is the current real-time temperature, T0 is the reference temperature; k is the temperature sensitivity coefficient, and different seasons correspond to different set values.

[0031] The present invention provides a water supply and drainage automatic control system with intelligent control, which can achieve the following beneficial technical effects:

[0032] 1. The water supply and drainage automatic control system of the present invention first acquires real-time water source data every preset time period T by the acquisition module; secondly, the intelligent control unit inputs the collected water source data into the trained LSTM neural network model, and the LSTM neural network model performs feature extraction; and the extracted water source feature vector is input into the sparse Gaussian process model, and the sparse Gaussian process model outputs the system operation parameters of each water supply and drainage automatic control system in the future time period; finally, the control execution module adjusts the valve opening, water pump power, water pressure in the pipeline, and water flow velocity in the pipeline in the corresponding time period according to the system operation parameters output by the sparse Gaussian process model. The present invention introduces the LSTM neural network to extract the features of time series data, and combined with the sparse Gaussian process model, it can realize the refined adjustment of water supply scheduling and improve user satisfaction. The introduction of the LSTM neural network for feature extraction of time series data can effectively overcome the defects of traditional neural networks in long time series processing and capture the dynamic characteristics of the system. Secondly, combined with the sparse Gaussian process model, by selecting representative induction points, the calculation complexity is simplified while the prediction accuracy is improved, which is suitable for control scenarios with high real-time requirements.

[0033] 2. The forget gate processing of the present invention determines whether to retain or discard historical information based on the data at the previous time point and the current input through the forget gate, thereby optimizing the effectiveness of the system memory. By introducing the temperature adaptive weight factor during the forget gate processing, the system can automatically adjust the response sensitivity to changes in water demand according to changes in the external environment, ensuring accurate water supply scheduling under different climatic conditions.

[0034] 3. The present invention forms a distributed intelligent control network through the Internet of Things platform through the water supply and drainage automatic control system. Multiple water supply and drainage automatic control systems share data through the Internet of Things platform to achieve cross-regional intelligent scheduling. The intelligent water supply and drainage automatic control system solution can achieve global intelligent scheduling through advanced machine learning algorithms and Internet of Things technology, overcoming the shortcomings of traditional systems such as slow response and waste of resources. The system not only improves water supply safety and resource utilization efficiency, but also can effectively respond to changing climate conditions and emergencies, providing strong technical support for the construction of intelligent infrastructure in modern cities. Through this system, urban water supply and drainage managers can achieve more efficient, flexible and intelligent management, reduce waste of water resources, and improve overall operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 It is a schematic diagram of the control process steps of a water supply and drainage automatic control system with intelligent control of the present invention;

[0037] Figure 2 It is a schematic diagram of a water supply and drainage automatic control system with intelligent control according to the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] Embodiment 1:

[0040] In view of the above problems mentioned in the prior art, in order to solve the above technical problems, the present invention provides a water supply and drainage automatic control system with intelligent control, the control steps of the automatic control system are as follows: Figure 1 As shown, it includes: first, the acquisition module obtains real-time water source data every preset time period T; secondly, the intelligent control unit inputs the collected water source data into the trained LSTM neural network model, and the LSTM neural network model performs feature extraction; and the extracted water source feature vector is input into the sparse Gaussian process model, and the sparse Gaussian process model outputs the system operation parameters of each water supply and drainage automatic control system in the future time period; finally, the control execution module adjusts the valve opening, water pump power, water pressure in the pipeline, and water flow rate in the pipeline in the corresponding time period according to the system operation parameters output by the sparse Gaussian process model.

[0041] Water quality monitoring modules, water flow control modules, liquid level monitoring modules and environmental parameter sensors are installed in each water source node, pipeline system and water storage facility. Each sensor is connected to the intelligent control platform through the Internet of Things module and can upload data in real time. The acquisition module automatically obtains real-time water source data every preset time period T (for example, 5 minutes). The collected data includes: water quality data: pH value, water temperature, dissolved oxygen content, etc.; liquid level data: liquid level height in the water storage tank and pipeline; external environmental parameters: air temperature, humidity, and air pressure at the water source; system operation parameters: such as valve opening, water pump power, pipeline water pressure, and real-time water flow rate. Data preprocessing Before inputting the collected water source data into the LSTM neural network, the system first preprocesses the data, including data cleaning (removing outliers and filling missing data), data normalization (standardizing data of different units and magnitudes to the same feature space) and time series resampling (ensuring the consistency of time series data).

[0042] LSTM neural network model: The preprocessed water source data is input into the trained LSTM neural network model. The LSTM network can analyze the short-term and long-term change trends of time series data, including but not limited to the following features: water use pattern features identify peak water use time periods and valley time periods, and capture the periodic frequency of water use; water quality fluctuation features detect abnormal fluctuations in water quality parameters, such as a sharp rise or fall in pH; analyze the impact of changes in external temperature and humidity on water demand, and extract the associated features between the environment and the system state; analyze the system's regulation behavior under different conditions based on historical operation records, so as to predict and optimize future control strategies. The LSTM neural network extracts features from the data of each time step to generate a hidden state representing the current state of the system, and then aggregates these hidden states through a pooling method (such as average pooling) to generate a water source feature vector. In some embodiments, the neural network uses GPU algorithms for parallel processing.

[0043] The water source feature vector input is extracted from the LSTM neural network and input into the sparse Gaussian process model (SGP). The SGP model reduces the computational complexity by selecting a small number of representative induction points, and uses a kernel function to evaluate the similarity between the input features and the induction points to generate a covariance matrix. Establishing the joint probability distribution Through the covariance matrix, the SGP model establishes the joint probability distribution between the current system state and the future state. The conditional probability distribution is calculated by Bayesian reasoning, the expected state of the system in the future time period is predicted, and the system operating parameters (such as future valve opening, pump power, water pressure and water flow rate) are output.

[0044] Temperature adaptive adjustment: The SGP model combines the temperature adaptive weight factor to automatically adjust the output sensitivity of the model according to the real-time temperature changes. For example, when the temperature rise is detected, the system increases the value of the temperature adaptive weight factor to improve the response to the water flow, so that the system can quickly respond to the increase in water demand caused by the high temperature environment. The control execution module that receives the prediction results formulates corresponding adjustment strategies based on the system operation parameters output by the sparse Gaussian process model. Valve control: When it is predicted that the water demand will increase in the future time period, the valve opening will be adjusted in advance to ensure that the water flow meets the demand; during the low water consumption period, the valve opening will be automatically reduced to save water resources. Water pump power adjustment: During the peak water consumption period, the water pump power will be automatically increased to increase the water supply pressure; during the low water consumption period, the power will be reduced to save energy. Pipeline pressure and flow rate adjustment: According to the predicted water flow demand, the system adjusts the water pressure and flow rate in the pipeline in real time to maintain the stability of the entire pipeline system and avoid the problem of excessive or low internal pressure in the pipeline.

[0045] When the system detects that water quality data (such as pH value and dissolved oxygen content) exceeds the normal threshold, the control execution module immediately starts the alarm mechanism, triggers the fault detection and alarm module, issues a warning signal and switches to the backup water source to ensure the safety of water supply. In emergency situations such as extreme weather (such as continuous high temperature or heavy rain), the system shares information in real time through the Internet of Things platform, and automatically coordinates data with water supply and drainage systems in other regions to achieve cross-regional intelligent scheduling to ensure the stability of the overall water supply system. The intelligent control unit can not only analyze and process current data in real time, but also feed back new data to the LSTM model to continuously optimize the prediction accuracy of the model. For example, through the dynamic adjustment of the temperature adaptive weight factor, the system can continuously optimize the prediction of the impact of seasonal changes on water demand. The system uploads historical data and operation records to the cloud server, and further optimizes the parameter settings of the LSTM and SGP models through big data analysis, so that the system can more accurately predict and adjust water supply strategies.

[0046] The present invention provides a water supply and drainage automatic control system with intelligent control; Figure 2 As shown, it includes: a water source unit, a piping system, an intelligent control unit, a water quality monitoring module, a water flow control module, a liquid level monitoring module, a fault detection and alarm module and a human-computer interaction interface;

[0047] In some embodiments, the water source unit includes an underground water source, a reservoir, a water storage tank or other water supply facilities to provide basic water supply for the system. The water outlet of the water source unit is connected to the main water supply pipeline, and the water source is transported to the user end through the pipeline system. The water source unit is physically connected to the pipeline system, and the water output is adjusted by control devices such as valves and water pumps. The water quality monitoring module in the water source unit detects the water quality in real time and transmits the information to the intelligent control unit through a data line or wireless signal.

[0048] The pipeline system consists of the main water supply pipeline, branch pipelines, valves and flow control devices, etc., and is responsible for delivering water from the water source unit to the user end. The pipeline system is directly connected to the water source unit and controls the water flow through valves. Sensor devices in the pipeline (such as pressure sensors, flow meters, etc.) communicate with the intelligent control unit to ensure real-time monitoring of the water pressure and water flow status of the pipeline.

[0049] The intelligent control unit is the core component of the system, responsible for intelligent management such as data collection, analysis and processing, and execution control. Central Processing Unit (CPU): The intelligent control unit is equipped with a high-speed central processing unit that can process large amounts of data in real time and execute intelligent algorithms (such as LSTM neural networks and sparse Gaussian process models). Memory and storage devices are used to store collected data, historical records, model parameters, and system operation logs to ensure that the system can continue to learn and optimize. The communication module includes wired and wireless communication devices, such as Ethernet interfaces, Wi-Fi modules, LoRa wireless modules, 4G / 5G modules, etc., to ensure that the intelligent control unit can transmit data with each module and the Internet of Things platform. Connection with water quality monitoring module, flow control module, and liquid level monitoring module: Data is exchanged through wired (such as RS485 bus, Modbus protocol) or wireless communication (such as Wi-Fi, Zigbee), and the intelligent control unit receives real-time data collected by each module and analyzes and processes these data. Connection with external devices and cloud platforms: Through the communication module, the intelligent control unit can upload the collected data and processing results to the cloud server, which is convenient for remote monitoring, data analysis, and model optimization. In addition, the system supports remote control through the Internet, and each device can be interconnected.

[0050] The water quality monitoring module includes pH sensor, dissolved oxygen sensor, temperature sensor, etc., which are used to detect the water quality status. It is connected to the intelligent control unit via wired (RS485 interface) or wireless (Zigbee, Bluetooth, etc.). After data collection, it is uploaded in real time so that the system can analyze and judge the water quality.

[0051] The water flow control module includes electric valves, flow meters, water pumps and pressure sensors. The electric valves and flow meters are used to accurately control and monitor water flow, while the water pumps adjust the water supply pressure. The flow meters and pressure sensors are connected to the intelligent control unit via data cables. The control signals of the electric valves and water pumps are directly sent by the intelligent control unit to ensure that the system can achieve accurate water supply under different requirements.

[0052] Liquid level monitoring module liquid level sensors (such as ultrasonic level gauges, pressure level sensors, etc.) are used to monitor the water level in the water tank and pipeline. The sensors of the liquid level monitoring module are connected to the intelligent control unit via RS485 bus or wireless communication (such as NB-IoT), ensuring that the system can grasp the water storage situation in real time and automatically adjust the water supply according to the water level situation.

[0053] The fault detection and alarm module includes a variety of sensors and alarm devices, such as water quality abnormality alarm, pipeline leakage sensor, power failure alarm device, etc. The communication module of the intelligent control unit is connected to each sensor. Once the intelligent control unit detects abnormal information, it will immediately trigger the alarm device and upload the alarm information to the cloud platform for management personnel to view.

[0054] The human-machine interface touch screen display or other control panel is used for system parameter setting and status monitoring. Users can view real-time data and historical records through the interface and manually control system equipment. The human-machine interface is connected to the intelligent control unit via wired (such as USB interface, serial port) or wireless, and displays the operating status of the system in real time, supporting users to perform manual operation and remote control.

[0055] The IoT platform consists of cloud servers, databases, remote control terminals, etc., which are responsible for the centralized storage, analysis and processing, remote management and monitoring of system data. Each intelligent control unit is connected to the IoT platform through a communication module (such as 4G / 5G, Wi-Fi), and data is securely transmitted through encryption protocols to ensure system stability and data security.

[0056] Distributed intelligent control network Multiple water supply and drainage automatic control systems share data through the Internet of Things platform to form a distributed intelligent control network. Each system can receive and use data from other systems to achieve unified scheduling and intelligent management of resources. Each system is connected to the Internet of Things platform through the Internet (such as VPN dedicated line or public network), and the cloud computing function of the platform is used to achieve cross-regional intelligent scheduling and collaborative control. The water source unit is connected to the pipeline system, and the pipeline system is connected to the water quality monitoring module, water flow control module, and liquid level monitoring module. Each module communicates with the intelligent control unit in two directions. The intelligent control unit is connected to the Internet of Things platform through the communication module to achieve remote management, monitoring and data analysis. The human-computer interaction interface is directly connected to the intelligent control unit, and the user can view the system status and manually control the device through the interface.

[0057] The acquisition module acquires real-time water source data every preset time period T; the water source data includes water quality data, liquid level data, external environmental parameters and system operation parameter data; in some embodiments, the acquisition module is responsible for acquiring various real-time data from multiple sensors and measuring devices, and transmitting them to the intelligent control unit for further analysis and processing. The hardware configuration of the acquisition module includes: the water quality sensor interface is used to connect various water quality detection equipment, such as pH sensor, temperature sensor, dissolved oxygen sensor, etc. The liquid level sensor interface is used to connect the liquid level detection equipment in the water storage tank and pipeline, such as ultrasonic liquid level meter, pressure liquid level sensor. The external environment monitoring interface is used to receive data from external environment sensors, including parameters such as air temperature, humidity, and air pressure. The system operation parameter interface connects the flow meter, pressure sensor, water pump status monitoring equipment, valve opening sensor, etc. in the pipeline system to obtain the current operating status of the system. Preset time interval T: The system sets a collection time interval T during initialization, such as 5 minutes or 15 minutes. This time interval can be adjusted according to the actual application scenario to meet different monitoring needs. For example, during peak water usage, the data collection interval can be shortened to once every 1 minute; during stable water usage, the interval can be extended to once every 30 minutes. The intelligent control unit can dynamically adjust the data collection interval T based on the current system operating status and the trend of historical data changes, for example, automatically increasing the data collection frequency when abnormal water quality changes are detected.

[0058] Water quality data collection, pH value is monitored in real time through pH sensors installed in pipelines and water storage tanks, and data is transmitted to the acquisition module in the form of analog signals or digital signals. Water temperature temperature sensors are embedded in key water flow nodes to collect water temperature data in real time to help determine water quality stability and changes in water use environment. Dissolved oxygen content Dissolved oxygen sensors are installed in water source units and pipelines to ensure that water quality meets water use standards, especially for agricultural and industrial water use.

[0059] The acquisition module obtains data from each water quality sensor at a preset time interval T and transmits it to the intelligent control unit through a data interface (such as RS485, Modbus protocol). After the water quality data is uploaded, the intelligent control unit performs a preliminary analysis of the data to identify any potential water quality abnormalities. If an abnormal pH fluctuation is detected, the system will immediately trigger the fault alarm module and adjust the water supply strategy. The liquid level height of the water tank is monitored by an ultrasonic level meter installed in the water tank to monitor the real-time water level of the water tank, and the data is transmitted to the acquisition module via wireless or wired communication. Pressure level sensors are installed on the main water supply pipe and branch pipes to monitor the water flow and the water level in the pipeline to ensure stable water supply. The acquisition module obtains water level data from the liquid level sensor according to the preset time interval T, and detects the water level of the water tank and pipeline in real time.

[0060] Liquid level data is transmitted to the intelligent control unit via wireless (such as LoRa, NB-IoT) or wired communication (such as RS232, RS485). The intelligent control unit determines whether to turn on or off the water pump based on the current liquid level information to maintain a stable water supply for the system. If the liquid level in the water tank is detected to be too low, the system will start the water replenishment process; if the liquid level is too high, the system will appropriately reduce the water supply of the water pump. The temperature sensors installed near the key nodes are used to collect external environmental parameters to monitor the temperature changes in the water source and water supply area in real time. The humidity sensor is used to monitor the environmental humidity of the water source to analyze the impact of the external climate on water demand. The air pressure sensor is used to monitor the changes in atmospheric pressure to help the system determine the trend of water resource use. The acquisition module obtains real-time environmental data from the external environmental sensor at intervals of T, and these data are transmitted to the intelligent control unit through the wireless module. According to the changes in the current environmental conditions, the sensitivity to the system operating parameters is automatically adjusted. For example, in hot weather, the system will increase the power of the water pump and increase the water supply. The valve position sensor is used to obtain the valve opening status in real time to ensure that the system can adjust the water flow according to actual needs. The current working status of the water pump is obtained from the water pump power monitoring device, including information such as start, stop, and power output. The real-time water pressure in the pipeline is monitored by the pressure sensor to ensure the stability of the water flow in the pipeline. The flow meter installed in the pipeline records the water flow rate in real time to help the system determine the current water use situation.

[0061] The acquisition module acquires the system operation status data at a set interval T, and transmits the data to the intelligent control unit via a data line or wireless communication. The intelligent control unit determines whether it is necessary to adjust the pump power or valve opening based on the real-time water flow velocity and water pressure data to achieve precise water supply control. If a decrease in pipeline water pressure is detected, the system automatically increases the output power of the pump to ensure stable water flow at the user end. The acquisition module ensures that data can be transmitted to the intelligent control unit quickly and accurately through standardized communication protocols (such as RS485, Modbus, LoRa, NB-IoT, etc.). After receiving the collected data, the intelligent control unit performs real-time analysis and processing, generates system feature vectors, and predicts future states through LSTM neural networks and sparse Gaussian process models, and automatically adjusts the operating parameters of each control device.

[0062] The intelligent control unit inputs the collected water source data into the trained LSTM neural network model, and the LSTM neural network model performs feature extraction; and inputs the extracted water source feature vector into the sparse Gaussian process model, and the sparse Gaussian process model outputs the system operation parameters of each water supply and drainage automatic control system in the future time period; the intelligent control unit is the core of the entire system, responsible for receiving the real-time data transmitted by the acquisition module, and analyzing, predicting and controlling the data through intelligent algorithms. Its hardware configuration includes: the central processing unit (CPU) is responsible for executing data processing and control algorithms to ensure efficient computing performance. Memory and storage devices are used to store historical data, model parameters, real-time data buffers and system operation logs. The communication module supports wired (such as Ethernet, RS485) and wireless (such as Wi-Fi, 4G / 5G, LoRa) communications to ensure data exchange with each module and the Internet of Things platform.

[0063] The intelligent control unit receives various water source data from the acquisition module in real time through the communication module, including water quality data (water temperature, pH value, dissolved oxygen), liquid level data, external environmental parameters (air temperature, humidity) and system operation parameters (such as water flow rate, valve opening, pump power). Data cleaning removes outliers and fills in missing data to ensure the integrity and accuracy of input data. Standardize data in different units to the same numerical range (such as 0 to 1) to improve the learning effect of the neural network model. Time series resampling: Resample the data to a uniform time interval T, such as once every 5 minutes, to ensure the continuity and consistency of time series data.

[0064] Input data to LSTM model: The preprocessed time series data is input to the LSTM neural network model. LSTM performs data analysis and feature extraction through its unique gating mechanism (input gate, forget gate, output gate). Short-term feature capture: LSTM neural network can capture real-time change trends from the data of the last few time steps. For example, during the high temperature period in summer, LSTM can detect the short-term increase in water consumption. Long-term feature recognition: LSTM neural network can effectively identify long-term features such as long-term water use patterns and environmental change trends through its memory units (cell states). For example, LSTM can learn the peak water use time period and seasonal changes every week. Generate water source feature vector: Hidden state aggregation: LSTM model generates hidden states at each time step. These hidden states represent the key feature information of the system at the current moment. Pooling and aggregation: The hidden states of all time steps are aggregated by average pooling or maximum pooling to generate a water source feature vector as the input of the sparse Gaussian process model in the next step.

[0065] Water source feature vector input: The water source feature vector generated from the LSTM neural network is input into the sparse Gaussian process model (SGP). The SGP model simplifies the computational complexity and improves the prediction efficiency by selecting a small number of representative induction points. Establishing joint probability distribution: Similarity calculation SGP calculates the similarity between the input features and the induction points through the radial basis kernel function (RBF Kernel) to generate a covariance matrix. This matrix reflects the relationship between the current system state and the future state. Joint probability distribution is based on the covariance matrix. The SGP model establishes the joint probability distribution of the current features and the future system state. Bayesian reasoning and conditional probability distribution: The Bayesian reasoning SGP model calculates the conditional probability distribution based on the input water source feature vector through Bayesian reasoning, and derives the expected value and variance of the future system state. Output system operation parameter SGP model predicts the system operation parameters in the future time period based on the conditional probability distribution, such as valve opening, water pump power, pipeline water pressure and water flow rate. For example, when it is predicted that water consumption will reach a peak in a certain period of time in the future, the system will adjust the valve opening and water pump power in advance to ensure stable water supply.

[0066] The intelligent control unit dynamically adjusts the temperature adaptive weight factor based on the real-time temperature data obtained by the external environment monitoring module to ensure that the system can respond to changes in water demand caused by temperature changes. For example, in hot summer weather, when the temperature T is significantly higher than the baseline temperature, the temperature adaptive weight factor increases, and the system automatically increases the response sensitivity to water flow and pump power to quickly respond to the increase in water demand. Feedback to LSTM and SGP models: The temperature adaptive weight factor will be adjusted in real time during the system control process, and the input data of LSTM and SGP will be corrected, so that the intelligent control unit can still maintain accurate prediction and regulation under different climate conditions.

[0067] The intelligent control unit generates corresponding control instructions based on the system operating parameters in the future time period output by the SGP model. Automatic adjustment of valve opening When it is predicted that the water demand will increase, the system will open more valves in advance to increase the water supply; when the demand decreases, the system will appropriately close the valves to reduce water waste. Pump power adjustment During peak water use periods, the intelligent control unit automatically increases the pump power to ensure sufficient pipeline pressure; during low periods, the system reduces power to save energy. The system continuously monitors the current water quality, water flow and external environmental parameters, and compares the real-time data with the predicted results to ensure the accuracy of the prediction and the stability of the control. If there is a deviation between the real-time monitoring data and the predicted results of the SGP model, the intelligent control unit will immediately adjust the control strategy to ensure the flexibility and accuracy of the system operation.

[0068] The control execution adjusts the valve opening, water pump power, water pressure in the pipeline, and water flow rate in the pipeline within the corresponding time period according to the system operating parameters output by the sparse Gaussian process model. In some embodiments, the control execution module includes the following key hardware devices: Electric valve: installed at the key nodes of the pipeline system, used to adjust the on-off and flow rate of the water flow. Variable speed water pump: supports adjusting the power output of the water pump to achieve dynamic control of the water pressure in the pipeline. Pressure sensor: used to monitor the water pressure in the pipeline in real time to ensure that the water flow is within the appropriate pressure range. Flow meter: installed on the main water supply pipeline, monitors the water flow rate in real time, and helps the system detect the current water use status.

[0069] The sparse Gaussian process model (SGP) calculates the operating parameters of each system in the future time period based on the input water source feature vector, and passes these parameters to the control execution module. Typical operating parameters include: valve opening: the opening percentage of the electric valve in the future time period; pump power: the output power setting value of the pump, in kilowatts (kW); pipeline water pressure: the target water pressure in the pipeline, in megapascals (MPa); water flow rate: the target water flow rate in the pipeline, in meters per second (m / s). The intelligent control unit analyzes the system operating parameters output by the SGP model and converts them into specific control instructions. For example, the system predicts that the water demand will increase significantly in the next 10 minutes, so the valve opening is set to 80% and the pump power is increased to 90% of the rated power. Automatic adjustment mechanism According to the output of the SGP model, the system can adjust the opening of the electric valve in real time. For example, the system detects that the peak water use period is coming. In order to avoid insufficient water supply, the valve opening on the main water supply pipeline is increased to 80% in advance to ensure sufficient water flow.

[0070] The control execution module sends an opening setting signal to the electric valve, instructing the valve to automatically adjust to the target opening value. After the electric valve is adjusted to the specified opening, the current state is fed back to the intelligent control unit through the position sensor to ensure that the actual operation is consistent with the setting. If the real-time monitoring data shows that the water flow rate changes beyond expectations, the system can immediately issue a new instruction for fine-tuning. For example, if it is detected that the valve opening is 80% and the water supply is still insufficient, the system will continue to adjust to 85% or even higher. Operation of variable-speed water pumps: Variable-speed water pumps can flexibly adjust the output power according to changes in water supply demand. For example, the SGP model predicts that water demand will peak in the next 15 minutes, and the system will set the pump power to 90% of the rated output in advance to cope with the upcoming high-demand period.

[0071] The power setting control execution module adjusts the frequency controller of the water pump and sets the target power output according to the prediction results. The real-time monitoring water pump power sensor will continuously monitor the current output power. If it detects that the power does not meet the standard or exceeds the set range, the system will automatically correct it to ensure that the water pump operates in an efficient working state. Energy-saving mode During the low water consumption period, the system will reduce the output power of the water pump and adjust the power to 50% or less to save energy. Automatic adjustment of pipeline pressure Through the pressure sensor installed on the pipeline, the system can monitor the water pressure changes in real time. If the SGP model predicts that the water demand in the future time period will increase significantly, the system will increase the pipeline pressure in advance by adjusting the valve and the water pump power to ensure stable water supply.

[0072] The target water pressure setting control execution module sets the target water pressure to be maintained in the pipeline based on the parameters output by the SGP model. The system precisely regulates the pressure in the pipeline by controlling the output power of the water pump or adjusting the valve opening. For example, when the pipeline pressure sensor detects that the pressure is lower than the target value, the system will increase the water pump power or further open the valve to increase the pipeline water pressure. Water flow control based on demand prediction: The intelligent control unit uses the future water flow demand forecast output by the SGP model to adjust the water flow rate in advance. For example, when the system detects that the water demand is reduced at night, it will reduce the water flow rate to save resources. The flow meter data analysis control execution module combines the real-time flow meter data to determine whether the current water flow rate is consistent with the target speed. Real-time adjustment If the current water flow rate is higher than the target value, the system will reduce the valve opening or reduce the water pump output; if it is lower than the target value, the system will increase the valve opening and water pump power accordingly.

[0073] The automatic abnormal situation handling system will continuously compare the real-time monitoring data with the SGP model prediction. If it detects abnormal pipeline water pressure, excessive pump power, or water flow rate deviating from the normal range, the system will automatically trigger the abnormal handling mechanism. When the fault alarm system detects a pump failure, pipeline leakage or other emergency, it will immediately start the alarm module and transmit the abnormal information to the management personnel's control terminal. Automatic switching to backup equipment If the main water pump stops working due to a fault, the system will automatically switch to the backup water pump to ensure uninterrupted water supply. Remote manual intervention If the system detects an emergency situation beyond the automatic control range, the management personnel can manually intervene through the remote control interface to adjust key parameters such as valve opening and pump power. Responding to sudden peak water use During the high temperature in summer, the SGP model predicts that the water demand on a certain day will reach a peak at 3 pm. The system adjusts the valve opening to 85% in advance and sets the pump power to 90% to ensure sufficient water supply pressure to prevent users from being short of water. Energy-saving control During the low-peak period at night, the SGP model detects that the water demand is significantly reduced, and the control execution module automatically reduces the valve opening to 40% and adjusts the water pump power to 50% to reduce unnecessary energy consumption.

[0074] In some embodiments, the intelligent control unit includes a central processing unit, a data acquisition module, a communication module and an execution control module: the water supply and drainage automatic control system forms a distributed intelligent control network through the Internet of Things platform, and multiple water supply and drainage automatic control systems share data through the Internet of Things platform to achieve cross-regional intelligent scheduling. Central Processing Unit (CPU) Each automatic control system is equipped with a central processing unit for data processing, operation parameter calculation and execution control. The communication module includes an Ethernet interface, a Wi-Fi module, a 4G / 5G wireless module, and a LoRa wireless communication device, which is used to upload local data to the Internet of Things platform and receive scheduling instructions from the platform. Each system monitors water quality, water flow, liquid level, external environmental parameters, etc. through sensors, and transmits data to the intelligent control unit. The core of the cloud server Internet of Things platform is responsible for storing and analyzing data from each automatic control system. The data exchange interface is used for data sharing and communication between water supply and drainage automatic control systems in different regions, and supports multiple communication protocols (such as MQTT, HTTP, CoAP, etc.) to ensure the security and efficient transmission of data. Remote control terminal managers access the Internet of Things platform through terminal devices (such as computers, tablets, and mobile phones) to view the system status in real time and perform remote control and management.

[0075] Each water supply and drainage automatic control system collects data (such as water quality parameters, flow, liquid level, ambient temperature, etc.) in real time through sensors installed in pipelines, water storage tanks, water source units, etc., and uploads these data to the Internet of Things platform through the communication module. Data transmission uses encryption protocols (such as SSL / TLS) to ensure data security and privacy protection. For example, an automatic control system sends water quality data and flow data to the cloud server every 5 minutes, and the platform can receive and store these data in real time. The cloud server of the Internet of Things platform classifies, processes and stores the received data to form a database. These data include real-time monitoring of environmental parameters, system operation status and historical records for further analysis and prediction. Based on the data received from each area, the platform conducts a global analysis, identifies the water supply and drainage needs of each area, and generates an overall resource usage table.

[0076] In some embodiments, the formation of a distributed intelligent control network is achieved through data sharing and synchronization of the Internet of Things platform, and each automatic control system can share data with each other. For example, the automatic control system of a certain area A can view the water supply of the neighboring area B in real time. If the water demand of A suddenly increases, the water source availability of B can be learned from the system of B, and the excess water source can be dispatched. After receiving the request for an increase in water demand in area A, the platform will query the current water source status of area B and coordinate the water supply of the two areas through an intelligent scheduling algorithm (such as a resource optimization algorithm). The Internet of Things platform executes an intelligent scheduling algorithm based on data collected from various areas. For example, a sparse Gaussian process model is used to predict the future water demand of each area, so as to adjust the water source allocation and scheduling strategy in advance. Cross-regional water supply coordination: When the platform detects that a certain area is about to enter a peak water demand, it can allocate water sources from neighboring areas in advance to ensure stable water supply during peak hours. For example, if the system of area A predicts that there will be a large-scale water demand in the next hour, and its own water source is insufficient, the system will request support from area B through the Internet of Things platform, and the platform will arrange an appropriate amount of water source to be allocated to area A based on the availability of water sources in B. When water consumption peaks in multiple areas, the platform will dynamically balance the load based on global data and prediction models. For example, the platform can adjust the pump power and pipeline valve opening of each area to appropriately distribute the water supply pressure in the high-pressure area to the low-pressure area to ensure the overall water supply pressure is balanced. The platform can set priority strategies, for example, to ensure that domestic water takes precedence over industrial water, to ensure that the basic needs of residents can be met even when resources are tight.

[0077] The IoT platform identifies system anomalies by analyzing the data uploaded by each region in real time. For example, if the water quality sensor in a certain area detects pollution, the platform will immediately generate an alarm and notify the control center and managers of the relevant area. The platform will also synchronize the abnormal information to other relevant areas through the data sharing mechanism, and notify the neighboring areas to prepare for water allocation. In the event of an emergency (such as a pipeline rupture, a large-scale power outage, severe water pollution, etc.), the platform can automatically coordinate cross-regional resource scheduling. For example, if the main water supply pipeline in area A is suddenly damaged, the system will automatically assign the backup pipeline and water source support in area B to ensure that the water supply in area A is not interrupted. Managers can intervene in the emergency response process through the remote control terminal, manually adjust the scheduling strategy of each area, and quickly solve the emergency problem. Through the control terminal of the IoT platform, managers can view the operating status of each area in real time, including water source utilization, water supply pressure, water quality data, etc. The platform interface provides a dynamic operation diagram of each area, so that managers can intuitively understand the water supply and drainage situation of the entire system and discover potential problems in time. Managers can send control instructions to the automatic control system of each area through the platform based on real-time monitoring data. For example, if water consumption in a certain area continues to decline, some valves can be remotely closed or pump power can be reduced to reduce energy consumption. The platform supports cross-regional collaborative control. If the water supply demand in an industrial park suddenly increases, managers can issue instructions through the platform to allow systems in other areas to temporarily increase water supply to meet demand.

[0078] In some embodiments, a city is divided into multiple water supply areas, each of which is managed by an independent water supply and drainage automatic control system. Through real-time monitoring and data sharing, the IoT platform finds that area A has a significant peak in water consumption in the morning from Monday to Friday. The platform dispatches water from areas B and C to area A in advance to ensure stable water supply during the morning peak hours. At the same time, during the low period on weekends, the platform reduces the pump power in area A and reduces the water supply to save energy. Some water supply facilities in a certain area were damaged due to floods caused by heavy rains. The IoT platform automatically identified the water supply interruption in the area and assigned backup pipelines in other areas to provide water supply to the area through intelligent scheduling algorithms. Managers monitor the emergency response process in real time through remote terminals to ensure that users' water use is not interrupted.

[0079] In some embodiments, the water quality data includes pH value, water temperature, and dissolved oxygen content; the liquid level data includes the liquid level height of the water storage tank and the pipeline; the external environmental parameters include the air temperature, humidity, and air pressure at the water source; the system operation parameter data includes valve opening, water pump power, water pressure in the pipeline, and water flow rate.

[0080] In some embodiments, LSTM is a recurrent neural network (RNN) that can effectively process time series data. The core of its design is to manage the flow of information through a special gating mechanism (input gate, forget gate and output gate), which solves the problem of information forgetting and gradient disappearance in traditional RNN. The following are the main structural components of LSTM: The input gate controls how much of the current input information is added to the cell state. In the water supply and drainage automatic control system, the input gate receives the collected water quality data (such as pH value, water temperature, dissolved oxygen), liquid level data (such as water level height of the storage tank), external environmental parameters (such as air temperature, humidity) and system operation parameters (such as valve opening, water pump power), etc., and filters out important input information. By adjusting the weight of the input gate, the system can selectively pay attention to environmental changes and system state changes. For example, when the external temperature rises, the system will give priority to recording the change information to cope with the increase in water demand. The forget gate determines the historical information to be retained or forgotten from the cell state, ensuring that the system can remember important trends for a long time and remove irrelevant information. The forget gate determines which historical data needs to be retained based on the current input data and the previous hidden state. For example, if the water quality data for the past few days remains stable, the system can "forget" this unchanged data through the forget gate, so as to focus on the new real-time data and improve the accuracy of prediction. The core storage unit of the cell state LSTM stores and dynamically updates important information for a long time through the adjustment of the input gate and the forget gate. The cell state combines historical water use data, current water quality conditions, and external environmental conditions to record information that reflects long-term trends. For example, the cell state can store data patterns during peak water use periods each week, and combine current real-time data to predict future water supply needs. The output gate determines the information extracted from the cell state and generates the hidden state of the current time step. The output gate extracts the current important feature information (such as peak water use, abnormal water quality characteristics, etc.) based on the current water source data and cell state, and passes the hidden state to subsequent time steps, helping the system to continuously capture short-term and long-term changes in time series.

[0081] In some embodiments, LSTM realizes long-term dependency analysis of time series data through its unique structure. Its specific workflow is as follows: Time series data input: input multi-step water source data (such as hourly water consumption, temperature, liquid level, etc. in the last 24 hours) into the LSTM neural network. Input gate, forget gate and output gate operation: Input gate: analyze the data of the current time step, filter out key information, and add it to the cell state; forget gate: combine the current and previous hidden states to determine which historical information needs to be retained and which information needs to be discarded; output gate: extract feature information from the updated cell state and generate the hidden state of the current time step. Feature vector generation: After each time step operation, the system can output a complete water source feature vector, representing the overall features extracted by LSTM from the time series for subsequent analysis and control.

[0082] In some embodiments, the LSTM workflow of the intelligent water supply and drainage automatic control system

[0083] Step 1: Real-time data input. The acquisition module collects water source related data at preset time intervals (such as every 5 minutes), including water quality data (pH value, water temperature, dissolved oxygen), liquid level data, external environmental parameters (air temperature, humidity) and system operation parameters (valve opening, water pump power, flow rate, etc.), and transmits these data to the intelligent control unit through the communication module. The intelligent control unit organizes the received real-time data into a multidimensional time series as the input of LSTM.

[0084] Step 2: Feature extraction, input gate processing: The LSTM input gate identifies key features based on the data of the current time step and combines historical trends. For example, in a high temperature environment in summer, LSTM will recognize the information of rising temperature and automatically adjust the model weights to make the system pay more attention to the changes in water consumption caused by rising temperature. Forget gate operation: The forget gate combines the current input and the hidden state of the previous time step to determine which historical information should be retained. For example, if the water quality data has remained stable in the past few days, LSTM will automatically reduce its attention to these non-fluctuating data and focus on new real-time water quality data. Cell state update: The operation results of the input gate and forget gate are used to update the cell state of LSTM to capture short-term and long-term change trends. The cell state saves the water use pattern, water quality fluctuations and external environmental influences of the system in the past period of time. Output gate generates hidden state: The output gate extracts key features from the cell state and generates hidden states for feature expression of the current time step. For example, if LSTM detects that there are obvious abnormal fluctuations in the current pH value and dissolved oxygen content, the system will immediately identify this feature and pass it to the subsequent processing module.

[0085] Step 3: Feature vector generation: The hidden states generated at each time step are aggregated through the average pooling method to obtain an overall water source feature vector. This feature vector can represent the current comprehensive state of the system, including peak and trough time periods of water use, abnormal water quality fluctuation characteristics, and significant changes in the external environment.

[0086] During the high temperature period in summer, the system's LSTM model receives gradually rising temperature and pump power data. Through the joint operation of the input gate and the forget gate, the LSTM can identify the correlation trend between rising temperature and water consumption, and predict the possible peak water consumption in the next few hours in advance. The system transmits this information to the intelligent control unit through the extracted feature vector, automatically adjusts the pump power and valve opening, and ensures that sufficient water supply is prepared in advance during the peak water consumption period to avoid water supply shortage. In the daily operation of the system, the LSTM detects short-term fluctuations in pH and dissolved oxygen content in the water quality data, which exceed the normal threshold. Through the long-term and short-term data analysis of the LSTM, the system recognizes that these fluctuations may cause abnormal water quality. The LSTM extracts this feature and feeds it back to the intelligent control unit, triggering the system's abnormal response mechanism. The control unit automatically adjusts the valve, switches to the backup water source, and activates the alarm module to notify the management personnel to conduct inspections and maintenance.

[0087] In some embodiments, the LSTM neural network performs feature extraction, including:

[0088] Time series input: Receive time series water source data including water quality data, liquid level data, external environmental parameters and system operation parameters, and combine real-time and historical water source data to capture short-term and long-term change trends;

[0089] Input gate processing: The input gate identifies the information about the system status change at the current time point and adds it to the system memory; the information about the system status change at the current time point includes: the air temperature rise and the water demand change curve;

[0090] Forget gate processing: Based on the data at the previous time point and the current input, the forget gate is used to decide whether to retain or discard historical information, thereby optimizing the effectiveness of system memory.

[0091] f t =σ(W f *[h t-1 ,x t ]+a t *b f )

[0092] Among them, f t is the forget gate activation value, which is used to control the amount of information retained from the previous cell state in the current time step, and its value range is between 0 and 1; σ is the tanh activation function; Wf is the weight matrix of the forget gate, which is used for the linear transformation of the current input and the hidden state of the previous time step; h t-1 represents the hidden state of the previous time step; x t Represents the input data of the current time step; a t is the adaptive weight factor of the temperature at time t. When the temperature rises, the water demand increases. t increase; b f The bias vector of the forget gate is used to adjust the offset of the forget gate output;

[0093] Cell state update: The processing results of the input gate and the forget gate are integrated to dynamically update the cell state of the system to capture long-term and short-term system characteristics;

[0094] Output gate generates hidden state: The output gate generates hidden state based on the current input, previous state and cell state. The hidden state includes the peak and trough periods of water use, abnormal characteristics of water quality fluctuations, that is, the magnitude of pH value or dissolved oxygen content exceeding the set threshold;

[0095] Feature vector generation: Summarize the hidden states into a source feature vector.

[0096] In some embodiments, the water source characteristic vector includes water use pattern characteristics, water quality fluctuation characteristics, environmental impact characteristics and system regulation response characteristics; water use pattern characteristics include peak water use time period, valley water use time period, average flow rate and peak and valley periodic frequency; water quality fluctuation characteristics include pH value change amplitude; environmental impact characteristics include the relationship between air temperature and water flow rate change, humidity-water quality change relationship; system regulation response characteristics include response delay time.

[0097] In some embodiments, the sparse Gaussian process model outputs system operating parameters of each system in a future time period, including: receiving a water source feature vector extracted by an LSTM neural network; selecting an induction point and establishing a sparse Gaussian process model; based on the input water source feature vector, using a radial basis kernel function to calculate the similarity between the input feature and the induction point to generate a covariance matrix, and constructing a joint probability distribution between the input feature and the induction point based on the covariance matrix; by splitting the joint probability distribution and combining the input feature data, using Bayesian reasoning to calculate the conditional expectation and conditional covariance, to obtain the conditional probability distribution of the system state in the future time period; calculating the expected value of the system operating parameters in the future time period according to the conditional probability distribution, and outputting the system operating parameters in the future time period, including valve opening, water pump power, pipeline water pressure and water flow rate.

[0098] In some embodiments, the structure of the sparse Gaussian process model is as follows: The core of the sparse Gaussian process model of induction points is to select a small number of induction points to approximate the distribution of the original data points, thereby reducing the computational complexity. These induction points act as "reference points" to help the model perform fast and efficient reasoning and calculations without losing prediction accuracy. In the water supply and drainage automatic control system, the induction points can be selected by analyzing the key states in the historical data. For example, the system can select states representing peak water use, valley water use, and abnormal water quality as induction points, thereby improving the prediction accuracy of the model under these key states.

[0099] The kernel function is used to calculate the similarity between the input data and the induced points to generate a covariance matrix. Commonly used kernel functions include radial basis functions, which can capture the smooth change trend of the input data. The radial basis kernel function helps the system identify the similarity of different features in the water source feature vector. For example, the system uses the kernel function to calculate the similarity between the current water flow velocity characteristics and the historical water peak characteristics to predict whether similar peak demand will occur in the future.

[0100] Covariance Matrix The covariance matrix constructs the joint probability distribution of data and captures the correlation between features by measuring the similarity between input features and induced points. The covariance matrix can effectively capture the correlation of complex features such as water quality changes and water demand fluctuations. For example, when the external ambient temperature rises, the covariance matrix can identify the relationship between rising temperatures and increased water demand based on previous correlation data.

[0101] In some embodiments, the induction point is selected: the sparse Gaussian process model selects representative induction points by analyzing historical data. Feature representative points are usually selected through algorithms such as K-means clustering, or induction points are set according to water use conditions during key periods (such as peaks, valleys, and abnormal conditions). These induction points can effectively reduce the amount of calculation of the model while maintaining a comprehensive expression of the input data features. Input feature vector and kernel function calculation: the system inputs the water source feature vector generated by the LSTM neural network into the SGP model, and uses the radial basis kernel function to calculate the similarity between the input data and the induction point. The output of the kernel function generates a covariance matrix to describe the correlation between the input feature vector and the induction point. Establish a joint probability distribution: Based on the covariance matrix, the SGP model establishes a joint probability distribution between the input feature vector and the induction point. This distribution captures the correlation between the current system state and the future state. For example, the system can predict future water quality changes and water demand by analyzing the current water quality and water use pattern and matching it with similar historical patterns. Through Bayesian reasoning, the SGP model combines the current input feature vector to calculate the conditional probability distribution of the system in the future time period. Bayesian reasoning can update predictions of future states based on existing information. The conditional probability distribution contains the expected value and variance of the system state in the future time period, which is used to characterize the uncertainty and confidence of the prediction. Based on the conditional probability distribution, the SGP model calculates the expected value of the system operating parameters in the future time period. Specifically, it includes valve opening, pump power, pipeline water pressure and water flow rate. These parameters can help the intelligent control unit make adjustments in advance to adapt to future changes in water supply demand. For example, when it is predicted that the peak water consumption will come in the next few hours, the system will increase the pump power and valve opening in advance to ensure stable water supply.

[0102] In some embodiments, the SGP workflow of the intelligent water supply and drainage automatic control system is as follows:

[0103] Step 1: Extract the water source feature vector from LSTM. The intelligent control unit processes the collected water source data through the LSTM neural network to generate a comprehensive water source feature vector, which contains the status information of the system at the current moment, including water quality, flow rate, ambient temperature, etc.

[0104] Step 2: Select the induction point. The system selects the induction point by analyzing the key states in the historical data. For example, the system will use the peak and trough periods of water consumption in the past as representative points to ensure the accuracy of the model prediction at these key moments. The selection process can be performed using the K-means clustering algorithm. For example, the peak water consumption period from 3 to 5 pm every day in the past week will be set as the induction point.

[0105] Step 3: Calculate similarity and establish a covariance matrix. The SGP model uses the radial basis kernel function to calculate the similarity between the current water source feature vector and the induction point and generate a covariance matrix. For example, the current water use pattern is highly similar to the historical water use peak, and the covariance matrix will reflect this relationship. Through the covariance matrix, SGP can understand the correlation between the current state and the historical state, thereby generating a prediction of the future state.

[0106] Step 4: Bayesian reasoning and conditional probability distribution calculation Based on the covariance matrix, the SGP model uses Bayesian reasoning and combines the input water source feature vector to calculate the conditional probability distribution in the future time period. This distribution indicates the possible changes in system operating parameters and prediction confidence in each future time period. For example, through Bayesian reasoning, the system can predict the water supply demand in the next 3 hours and use this prediction as the control instruction of the system.

[0107] Step 5: Output system operating parameters. The SGP model calculates the expected value of the future system operating parameters based on the conditional probability distribution and passes it to the control execution module. Valve opening: The SGP model predicts that water consumption will increase in a certain period in the future. The control unit sets the valve opening to 80% in advance based on the output results. Pump power: It is predicted that there will be a peak in water consumption within the next hour. The system will increase the pump power to 90% of the rated power to ensure sufficient water supply. Pipeline water pressure: According to the prediction results, increase the pipeline pressure in advance to prevent the water supply pressure from dropping during peak hours. Water flow rate: By adjusting the flow rate and pressure, ensure that the water flow rate in the future time period meets user needs.

[0108] The system predicts that area A will enter the peak water consumption period from 3 to 6 pm. Through the analysis of the SGP model, it is identified that the water consumption in area A will increase significantly. In response to this demand, the intelligent control unit adjusts the valve opening to 85% and sets the pump power to the highest output level to ensure water supply pressure. At the same time, the system allocates excess water resources from other areas through the IoT platform to ensure balanced water supply during peak hours. Water consumption valley energy saving control During the late night water consumption valley period, the SGP model detects that water demand will be at its lowest level in the next few hours. Based on the prediction results, the system automatically adjusts the valve opening to 30% and reduces the pump power to 50% to reduce energy consumption. At the same time, the system continuously monitors real-time data. If it detects that the water flow rate exceeds the valley range, the system can immediately readjust the operating parameters.

[0109] In some embodiments, aggregating the hidden states into a water source feature vector includes taking an average of hidden states of multiple time steps through average pooling to generate a water source feature vector that reflects global characteristics.

[0110] In some embodiments, the selecting of induction points and establishing of a sparse Gaussian process model includes randomly selecting induction points or selecting induction points by calculating using a K-means clustering algorithm based on historical data.

[0111] In some embodiments, before inputting the collected water source data into the trained LSTM neural network, the method further includes:

[0112] Data cleaning: including removing outliers and filling missing values, identifying abnormal data through the set threshold range, and filling missing data using interpolation or regression models;

[0113] Data normalization: standardize the water source data to the same feature space and remove high-frequency noise through Gaussian filtering algorithm;

[0114] Time series resampling: The water source data is resampled according to the preset time interval T to ensure the time series consistency of the input data.

[0115] In some embodiments, the a t is the temperature adaptive weight factor. When the temperature rises, the water demand increases. t Increase, Among them, a0 is the basic value of the weight factor, Z is the current real-time temperature, T0 is the reference temperature; k is the temperature sensitivity coefficient, and different seasons correspond to different set values. The temperature adaptive weight factor is a dynamic adjustment parameter introduced by the system when predicting water demand. It aims to adjust the system's sensitivity to water demand forecasts according to real-time changes in external temperature. When the temperature rises, it usually leads to an increase in water demand, so it increases with the rise in temperature. This enables the system to automatically increase the water supply during periods of higher temperatures, adjust the valve opening and water pump power in advance, and ensure that users' water needs in high temperature environments are fully met.

[0116] By introducing the temperature adaptive weight factor, the system can automatically adjust the water supply strategy according to the real-time temperature changes, improving the accuracy and response speed of the prediction. For example, in hot summer weather, the system will automatically increase the water supply when the temperature rises to avoid water shortage caused by temperature changes. When the temperature is low and the water demand decreases, the system will reduce accordingly, reducing the water supply and pump power, saving energy and reducing the system operation cost. Setting the temperature sensitivity coefficient is a key parameter for the adjustment range of the temperature sensitivity coefficient. The specific setting method includes: Based on historical data analysis: By analyzing the changes in water consumption under different temperature conditions in the past, the relationship between temperature and water demand is determined. Based on these historical data, the system regresses a suitable sensitivity coefficient k. For example, if the analysis finds that the water consumption increases by 5% for every 1°C increase, then k can be set to 0.05. Seasonal adjustment: The sensitivity coefficients corresponding to different seasons are different. For example, in high temperatures in summer, the sensitivity coefficient k is set to a higher value (such as 0.08) to ensure that the system responds quickly when the temperature rises; in winter, k is set to a lower value (such as 0.02) to reduce overreaction to temperature changes and avoid waste of resources.

[0117] The present invention provides a water supply and drainage automatic control system with intelligent control, which can achieve the following beneficial technical effects:

[0118] 1. The water supply and drainage automatic control system of the present invention first acquires real-time water source data every preset time period T by the acquisition module; secondly, the intelligent control unit inputs the collected water source data into the trained LSTM neural network model, and the LSTM neural network model performs feature extraction; and the extracted water source feature vector is input into the sparse Gaussian process model, and the sparse Gaussian process model outputs the system operation parameters of each water supply and drainage automatic control system in the future time period; finally, the control execution module adjusts the valve opening, water pump power, water pressure in the pipeline, and water flow rate in the pipeline in the corresponding time period according to the system operation parameters output by the sparse Gaussian process model. The present invention introduces the LSTM neural network to extract the features of time series data, and combined with the sparse Gaussian process model, it can realize the refined adjustment of water supply scheduling and improve user satisfaction. First, the introduction of the LSTM neural network for feature extraction of time series data can effectively overcome the defects of traditional neural networks in long time series processing and capture the dynamic characteristics of the system. Secondly, combined with the sparse Gaussian process model, by selecting representative induction points, the calculation complexity is simplified, and the prediction accuracy is improved, which is suitable for control scenarios with high real-time requirements.

[0119] 2. The forget gate processing of the present invention determines whether to retain or discard historical information based on the data at the previous time point and the current input through the forget gate, thereby optimizing the effectiveness of the system memory. By introducing the temperature adaptive weight factor during the forget gate processing, the system can automatically adjust the response sensitivity to changes in water demand according to changes in the external environment, ensuring accurate water supply scheduling under different climatic conditions.

[0120] 3. The present invention forms a distributed intelligent control network through the Internet of Things platform through the water supply and drainage automatic control system. Multiple water supply and drainage automatic control systems share data through the Internet of Things platform to achieve cross-regional intelligent scheduling. The intelligent water supply and drainage automatic control system solution can achieve global intelligent scheduling through advanced machine learning algorithms and Internet of Things technology, overcoming the shortcomings of traditional systems such as slow response and waste of resources. The system not only improves water supply safety and resource utilization efficiency, but also can effectively respond to changing climate conditions and emergencies, providing strong technical support for the construction of intelligent infrastructure in modern cities. Through this system, urban water supply and drainage managers can achieve more efficient, flexible and intelligent management, reduce waste of water resources, and improve overall operating efficiency.

[0121] The above is a detailed introduction to a water supply and drainage automatic control system with intelligent control. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the core idea of ​​the present invention. At the same time, for those skilled in the art, according to the ideas and methods of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A water supply and drainage automatic control system with intelligent control, characterized in that: include: Water source unit, piping system, intelligent control unit, water quality monitoring module, water flow control module, liquid level monitoring module, fault detection and alarm module and human-computer interaction interface; The acquisition module acquires real-time water source data every preset time period T; the water source data includes water quality data, liquid level data, external environmental parameters and system operation parameter data; The intelligent control unit inputs the collected water source data into the trained LSTM neural network model, which performs feature extraction; and inputs the extracted water source feature vector into the sparse Gaussian process model, which outputs the system operation parameters of each water supply and drainage automatic control system in the future time period; Control execution, according to the system operating parameters output by the sparse Gaussian process model, adjusts the valve opening, water pump power, water pressure in the pipeline, and water flow rate in the pipeline within the corresponding time period.

2. The water supply and drainage automatic control system with intelligent control according to claim 1 is characterized in that: The intelligent control unit includes a central processing unit, a data acquisition module, a communication module and an execution control module: the water supply and drainage automatic control system forms a distributed intelligent control network through the Internet of Things platform, and multiple water supply and drainage automatic control systems share data through the Internet of Things platform to achieve cross-regional intelligent scheduling.

3. The water supply and drainage automatic control system with intelligent control according to claim 1 is characterized in that: The water quality data includes pH value, water temperature, and dissolved oxygen content; the liquid level data includes the liquid level height of the water storage tank and the pipeline; the external environmental parameters include the air temperature, humidity, and air pressure at the water source; the system operation parameter data includes valve opening, water pump power, water pressure in the pipeline, and water flow rate.

4. The water supply and drainage automatic control system with intelligent control according to claim 1 is characterized in that: The LSTM neural network performs feature extraction, including: Time series input: Receive time series water source data including water quality data, liquid level data, external environmental parameters and system operation parameters, and combine real-time and historical water source data to capture short-term and long-term change trends; Input gate processing: The input gate identifies the information about the system status change at the current time point and adds it to the system memory; the information about the system status change at the current time point includes: the air temperature rise and the water demand change curve; Forget gate processing: Based on the data at the previous time point and the current input, the forget gate is used to decide whether to retain or discard historical information, thereby optimizing the effectiveness of system memory. f t =σ(W f *[h t-1 ,x t ]+a t *b f ) Among them, f t is the forget gate activation value, which is used to control the amount of information retained from the previous cell state in the current time step, and its value range is between 0 and 1; σ is the tanh activation function; W f is the weight matrix of the forget gate, which is used for the linear transformation of the current input and the hidden state of the previous time step; h t-1 represents the hidden state of the previous time step; x t Represents the input data of the current time step; a t is the adaptive weight factor of the temperature at time t. When the temperature rises, the water demand increases. t increase; b f The bias vector of the forget gate is used to adjust the offset of the forget gate output; Cell state update: The processing results of the input gate and the forget gate are integrated to dynamically update the cell state of the system to capture long-term and short-term system characteristics; Output gate generates hidden state: The output gate generates hidden state based on the current input, previous state and cell state. The hidden state includes the peak and trough periods of water use, abnormal characteristics of water quality fluctuations, that is, the magnitude of pH value or dissolved oxygen content exceeding the set threshold; Feature vector generation: Summarize the hidden states into a source feature vector.

5. The water supply and drainage automatic control system with intelligent control according to claim 4 is characterized in that: The water source characteristic vector includes water use pattern characteristics, water quality fluctuation characteristics, environmental impact characteristics and system regulation response characteristics; water use pattern characteristics include peak water use time period, valley water use time period, average flow rate and peak and valley periodic frequency; water quality fluctuation characteristics include pH value change amplitude; environmental impact characteristics include the relationship between air temperature and water flow rate change, and humidity-water quality change relationship; system regulation response characteristics include response delay time.

6. The water supply and drainage automatic control system with intelligent control according to claim 1 is characterized in that: The sparse Gaussian process model outputs system operation parameters of each system in a future time period, including: receiving a water source feature vector extracted by an LSTM neural network; selecting an induction point and establishing a sparse Gaussian process model; based on the input water source feature vector, using a radial basis kernel function to calculate the similarity between the input feature and the induction point to generate a covariance matrix, and constructing a joint probability distribution between the input feature and the induction point based on the covariance matrix; by segmenting the joint probability distribution and combining the input feature data, using Bayesian reasoning to calculate the conditional expectation and conditional covariance, to obtain the conditional probability distribution of the system state in the future time period; calculating the expected value of the system operation parameters in the future time period according to the conditional probability distribution, and outputting the system operation parameters in the future time period, including valve opening, water pump power, pipeline water pressure and water flow velocity.

7. The water supply and drainage automatic control system with intelligent control according to claim 1 is characterized in that: The step of aggregating the hidden states into a water source feature vector includes taking an average value of the hidden states of multiple time steps through average pooling to generate a water source feature vector reflecting global characteristics.

8. The water supply and drainage automatic control system with intelligent control according to claim 1 is characterized in that: The induction points are selected to establish a sparse Gaussian process model, including randomly selecting induction points or selecting induction points by calculating using a K-means clustering algorithm based on historical data.

9. The water supply and drainage automatic control system with intelligent control according to claim 1, characterized in that: Before inputting the collected water source data into the trained LSTM neural network, the method further includes: Data cleaning: including removing outliers and filling missing values, identifying abnormal data through the set threshold range, and filling missing data using interpolation or regression models; Data normalization: standardize the water source data to the same feature space and remove high-frequency noise through Gaussian filtering algorithm; Time series resampling: The water source data is resampled according to the preset time interval T to ensure the time series consistency of the input data.

10. The water supply and drainage automatic control system with intelligent control according to claim 4, characterized in that: The a t is the temperature adaptive weight factor. When the temperature rises, the water demand increases. t Increase, Among them, a0 is the basic value of the weight factor, Z is the current real-time temperature, T0 is the reference temperature; k is the temperature sensitivity coefficient, and different seasons correspond to different set values.

Images