An intelligent water quality monitoring device based on the Internet of Things

Abstract

The present invention discloses an intelligent water quality monitoring device based on the Internet of Things, which relates to the technical field of environmental monitoring. Through a deep learning model, the system can analyze complex non-linear relationships in water quality data, accurately predict the trend of water quality changes, and identify potential water quality problems in advance. The dynamic weight calculation module can dynamically adjust the importance of water quality parameters according to real-time data and environmental changes, improving the monitoring accuracy and adaptability. The intelligent feedback mechanism automatically adjusts the sampling frequency and sensor position when a warning signal is triggered to ensure accurate monitoring of water quality changes. The Internet of Things technology supports the real-time transmission of water quality data and analysis results to the cloud, and managers can view the water quality status, historical trends, and warning records through a remote monitoring interface, improving the response ability and management efficiency of the system. The intelligent water quality monitoring device of the present invention has the characteristics of high precision, high intelligence, and strong adaptability.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and specifically to an intelligent water quality monitoring device based on the Internet of Things. Background Art

[0002] With the increasingly serious environmental pollution problems, water quality monitoring has become an important link in protecting water resources and maintaining ecological balance. Traditional water quality monitoring methods have problems such as low monitoring accuracy, slow response, and weak data processing capabilities. Especially when facing the water quality monitoring of large-scale waters, traditional technologies often cannot achieve efficient and accurate data collection and dynamic analysis. Therefore, how to combine technologies such as the Internet of Things, big data analysis, and deep learning to improve the intelligent level of water quality monitoring has become a technical problem to be solved urgently.

[0003] In the prior art, Chinese invention patent CN115792158B discloses a method and device for realizing dynamic water quality monitoring based on the Internet of Things. The device collects water quality data through a water chemical index detector and uses a water quality monitoring model for analysis. Although this technology can effectively realize water quality monitoring, its intelligent level is relatively low, mainly relying on rules and models set manually, and it is difficult to achieve adaptive adjustment. Especially when water quality data is abnormal, the response of the system is relatively lagging, and it cannot be adjusted in real time according to the dynamic changes of the environment, resulting in possible large errors in the monitoring results.

[0004] In addition, Chinese invention patent CN114845260B discloses a hydrological monitoring data acquisition system based on the Internet of Things, which realizes fixed-point and patrol monitoring of rivers through unmanned patrol devices and hydrological monitoring base stations, and combines a remote monitoring center for data analysis. Although this technology realizes the remote transmission and analysis of data through the Internet of Things, it mainly focuses on hydrological data and ignores the dynamic monitoring and prediction of water quality. This system also fails to combine advanced deep learning technologies to accurately predict and give intelligent feedback on water quality changes.

[0005] Although the above designs improve the efficiency and real-time performance of data collection through Internet of Things technology, there are still certain limitations, such as insufficient monitoring accuracy, inability to automatically adapt to different environmental conditions, failure to conduct intelligent early warning and dynamic adjustment, etc. Summary of the Invention

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and propose an intelligent water quality monitoring device based on the Internet of Things to solve the above problems.

[0007] The object of the present invention is achieved through the following technical solutions: An intelligent water quality monitoring device based on the Internet of Things, comprising: a sensor module, responsible for collecting original water quality data, including: a data acquisition unit, used for regularly collecting water quality data, and the water quality data includes various parameters such as but not limited to pH value, dissolved oxygen, turbidity, ammonia nitrogen, temperature, etc.; a data connection unit, used for transmitting the collected original water quality data to a central processing unit for subsequent processing;

[0008] A data cleaning and standardization module, which receives the original water quality data transmitted by the sensor module and performs cleaning and standardization processing, including: an adaptive cleaning algorithm unit, through a machine learning algorithm, based on the historical water quality data trend and real-time changes, dynamically adjusts the data cleaning rules to remove noise; an anomaly detection unit, which uses a machine learning model to real-time identify outliers in the original water quality data and perform denoising processing; an adaptive threshold adjustment unit, according to environmental conditions and data characteristics, dynamically adjusts the cleaning threshold, converts the cleaned water quality data into a unified format and timestamp, ensures data consistency and facilitates subsequent packaging;

[0009] A data backup and Token sequence generation module, which receives the cleaned water quality data and converts it into an analyzable sequence, including: a data backup unit, which backs up the cleaned water quality data according to time and spatial information, and each data packet contains various water quality parameters and the collection time; a Token sequence generation unit, which combines multiple data packets in chronological order and spatial position order to generate a continuous Token sequence, and the Token sequence contains multi-dimensional water quality parameter information and is formed according to data changes and spatial positions; a data transfer unit, which transmits the generated Token sequence to a large model analysis module for in-depth analysis;

[0010] A large model analysis module, which receives the Token sequence, extracts water quality parameter features and analyzes the correlation relationship, including: a deep learning analysis unit, based on a deep neural network (DNN) and a self-attention mechanism (Transformer), performs multi-level correlation analysis on the Token sequence to capture the complex non-linear relationship between water quality parameters; a time series modeling unit, using a recurrent neural network (RNN) of long short-term memory (LSTM) or gated recurrent unit (GRU), models the Token sequence to extract the long-term trend and short-term fluctuations of water quality changes; an adaptive weighting mechanism unit, calculates the correlation degree of each water quality parameter through the self-attention mechanism and weights it, and adjusts the parameter importance in real time according to data changes and environmental conditions to generate weighted feature data;

[0011] The dynamic weight calculation module receives the weighted feature data from the large model analysis module and calculates dynamic weights, including: a dynamic weight calculation unit that automatically assigns dynamic weights to each water quality parameter according to the real-time water quality data analysis results, reflecting its influence degree in the current environment; a weight adjustment unit that dynamically corrects the water quality parameter weights in combination with real-time data input and environmental variable changes to ensure the accuracy of the analysis results.

[0012] The prediction and early warning module receives the dynamic weights and feature data and conducts water quality prediction and anomaly early warning, including: a water quality prediction unit that uses a deep learning method based on a large model, combines historical water quality data and real-time monitoring data, predicts the long-term water quality change trend, and considers the impacts of climate change, seasonal factors, and pollution sources; a potential problem identification unit that identifies potential anomalies based on real-time water quality data and prediction results and generates early warning signals; an early warning response unit that automatically issues an alarm according to the early warning signals and initiates adjustment measures to optimize the data collection frequency and sampling locations to cope with anomalies.

[0013] The intelligent feedback mechanism module adjusts the monitoring strategy and optimizes data collection according to the early warning response, including: a feedback optimization unit that automatically adjusts the sampling frequency and sensor locations to improve the water quality monitoring accuracy when an early warning is triggered; a data collection optimization unit that increases the sampling frequency and enhances the data collection intensity during abnormal events to provide more detailed real-time information; a dynamic strategy adjustment unit that adaptively adjusts the sensor configuration, sampling mode, and analysis period based on the data analysis results and environmental changes.

[0014] The adaptive cleaning algorithm unit in the data cleaning and standardization module is based on a machine learning model and dynamically adjusts cleaning parameters (such as thresholds, rules, anomaly point recognition sensitivity) using historical water quality data and real-time changes to ensure the accuracy of the cleaned water quality data and its adaptability to environmental changes.

[0015] The Token sequence generation unit in the data backup and Token sequence generation module packs the cleaned water quality data into a multi-dimensional Token sequence according to time series and spatial location information to ensure data continuity and spatial relevance and provide input for large model analysis.

[0016] The deep learning analysis unit in the large model analysis module extracts multi-level features of the Token sequence through DNN and Transformer, and the time series modeling unit captures the water quality change trend through LSTM or GRU to generate weighted feature data for subsequent weight calculation and prediction.

[0017] The dynamic weight calculation unit in the dynamic weight calculation module calculates and assigns dynamic weights to each water quality parameter according to real-time water quality data and environmental variables, and the weight adjustment unit dynamically corrects the weights according to data changes to improve the adaptability and accuracy of the analysis.

[0018] The water quality prediction unit in the prediction and early warning module combines historical data and real-time data to predict the water quality trend through a deep learning model. The potential problem identification unit generates early warning signals based on the prediction results to ensure the timely identification of potential anomalies.

[0019] The feedback optimization unit and data acquisition optimization unit in the intelligent feedback mechanism module adjust the sampling frequency and data acquisition intensity according to the early warning signals. The dynamic strategy adjustment unit adaptively optimizes the monitoring strategy to ensure accurate monitoring during anomalies.

[0020] The device also includes a support database module, which includes: a high-dimensional data storage unit that stores the cleaned water quality data, Token sequences, and weighted feature data, supporting multi-dimensional indexing and efficient query; a calculation result storage unit that stores the intermediate results of large model analysis and weight calculation for fast retrieval and real-time analysis; a data retrieval optimization unit that realizes fast data access through efficient storage and indexing technologies and supports the real-time processing of large-scale water quality data.

[0021] The device also includes an Internet of Things support unit, which includes: a data transmission unit that transmits water quality data and analysis results to the cloud in real time through Internet of Things technology; a remote monitoring unit that supports viewing the water quality status, historical trends, and early warning records through a remote interface to enhance the system interactivity.

[0022] The beneficial effects of the present invention are:

[0023] 1. The water quality prediction and early warning module of the present invention combines historical data and real-time data through a deep learning model, and can accurately predict the water quality change trend and identify potential water quality problems in advance. For example, the system can predict the change trends of water quality parameters such as dissolved oxygen concentration and turbidity in the next few hours or days. When it is predicted that some parameters are about to exceed the safety threshold, the system will generate early warning signals. Through timely early warnings, managers can take measures in advance to avoid the further deterioration of water quality problems. By using a deep neural network (DNN) and a Transformer self-attention mechanism, the present invention can simultaneously analyze the complex non-linear relationships between multiple water quality parameters. During the process of water quality change, the system can comprehensively consider various factors (such as climate change, seasonal fluctuations, etc.) to provide more accurate water quality predictions.

[0024] 2. The dynamic weight calculation module in the present invention can adjust the importance of each water quality parameter in real time according to water quality data and environmental conditions (such as precipitation, temperature, etc.). For example, in the dry season, the changes in water temperature and dissolved oxygen may have a greater impact on water quality, so the weights of water temperature and dissolved oxygen will be correspondingly increased. In the rainy season, the system will automatically increase the monitoring weights of parameters such as turbidity and ammonia nitrogen. By dynamically adjusting the weights, the system can better adapt to environmental changes, improve the accuracy and applicability of monitoring. Through the feedback optimization unit and the data acquisition optimization unit, the system can automatically adjust the sampling frequency and sensor position according to the warning signal. When water quality anomalies or emergencies occur, the water quality monitoring system can respond quickly and improve the data acquisition accuracy. For example, in the case of water quality anomalies, the system will automatically increase the data sampling frequency and mobilize more sensors for detailed monitoring to ensure accurate recording of water quality changes.

[0025] 3. By combining deep learning algorithms and real-time data analysis, the present invention can achieve accurate prediction and automatic warning of water quality anomalies. When the water quality anomaly trend is predicted, the system will automatically issue an alarm and can take a series of automated response measures (such as adjusting the sampling frequency, changing the sensor position, etc.) to minimize manual intervention and improve the response efficiency. Through the dynamic strategy adjustment unit, the system can adaptively optimize the monitoring strategy according to real-time monitoring data and environmental changes. For example, in the dry season, the system may strengthen the monitoring of water temperature and chemical components, while in the precipitation season, it will strengthen the monitoring of water flow changes and turbidity. This adaptive monitoring ability ensures the efficiency and sustainability of water quality monitoring.

[0026] 4. By converting water quality data into Token sequences and analyzing them through a large model, the system can efficiently extract important features in water quality data and conduct accurate analysis based on deep learning models. This makes water quality data processing more efficient, especially in large-scale water quality monitoring systems, where it can quickly process a large amount of data to ensure timely analysis results. Through the high-dimensional data storage unit, the system can efficiently store water quality data, Token sequences, and weighted feature data, and support multi-dimensional indexing and real-time query, which provides strong support for the management and query of large-scale water quality monitoring data and ensures that the system can handle a large amount of data and high-frequency real-time analysis requirements.

[0027] 5. The present invention uses Internet of Things technology to achieve real-time transmission of water quality data, and sends water quality monitoring results and warning signals to the cloud management platform in real time. It supports managers to view information such as water quality status, historical trends, and warning records through a remote interface. Managers can view water quality monitoring data regardless of geographical location, improving the flexibility and response speed of management. Through the Internet of Things support unit, the system can be easily integrated with other devices and platforms, supporting remote control and cross-platform operations. Managers can adjust the working mode of the system as needed, flexibly configure monitoring parameters and sensor layouts to ensure the scalability and compatibility of the water quality monitoring system.

[0028] 6. Through deep learning analysis and dynamic adjustment, the present invention can comprehensively monitor water quality changes. Especially in key scenarios such as water source areas, drinking water quality, and industrial wastewater treatment, it can detect water quality anomalies in a timely manner and give early warnings to ensure the safety of water sources and the compliance of water quality. In case of emergencies or abnormal water quality changes, the system can respond quickly, automatically adjust the monitoring strategy and enhance data collection efforts. Through the warning and feedback mechanism, the system can effectively reduce the risks of environmental pollution and water quality pollution, ensuring the sustainable utilization of water sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is the system architecture diagram of the present invention;

[0030] Figure 2 is the system interaction diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0031] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] It should be noted that in the following solutions, the orientation concepts of "left", "right", "up", "down", "front", "back", "inner", and "outer" are all relative directions, and will not be listed one by one here.

[0033] Embodiment 1:

[0034] As Figure 1 and Figure 2 shown, this embodiment provides an Internet of Things-based intelligent water quality monitoring device. The system mainly consists of a sensor module, a data cleaning and standardization module, a data backup and Token sequence generation module, and a database support module. The specific structure is as follows:

[0035] The sensor module includes:

[0036] Data acquisition unit: This unit is responsible for regularly collecting water quality data. The collected water quality parameters include pH value, dissolved oxygen, turbidity, ammonia nitrogen, temperature, etc. This module transmits the collected data to the central processing unit through wireless communication.

[0037] Data connection unit: This unit is responsible for transmitting the collected raw water quality data to the central processing unit for subsequent processing, ensuring the real-time transmission and storage of data after collection.

[0038] The data cleaning and standardization module includes:

[0039] Adaptive cleaning algorithm unit: This unit adopts an adaptive data cleaning algorithm based on machine learning, dynamically adjusts the data cleaning rules, and automatically identifies and removes noise and outliers by analyzing the trends of historical water quality data and real-time data changes. Especially in the case of seasonal and environmental changes, the cleaning rules will be adjusted accordingly to ensure data quality.

[0040] Anomaly detection unit: This unit uses a machine learning model to real-time identify outliers in water quality data, such as abnormal fluctuations in dissolved oxygen values, etc., and performs denoising processing to ensure the reliability of the final data.

[0041] Adaptive threshold adjustment unit: According to environmental conditions (such as temperature changes, rainfall, etc.) and water quality characteristics, this unit can dynamically adjust the cleaning threshold. For example, when the water temperature is high in summer, it automatically adjusts the cleaning threshold of turbidity, enabling the cleaning operation to adapt to changes in water quality data.

[0042] Data format standardization unit: Converts the cleaned water quality data into a unified format and adds timestamps to ensure data consistency and facilitate subsequent analysis and storage.

[0043] The data backup and Token sequence generation module includes:

[0044] Data backup unit: The cleaned water quality data is backed up according to time and space information to form raw data packets. Each data packet contains multiple water quality parameters and records the timestamp of data collection for subsequent serialization processing.

[0045] Token sequence generation unit: Multiple data packets are combined according to the time order and spatial position order to generate a continuous Token sequence. Each Token represents a water quality data point or related water quality parameter. Through the relationship between data changes and spatial positions, a Token sequence of multi-dimensional information is formed. These Token sequences will be passed as inputs to the subsequent large model analysis module.

[0046] Data Transfer Unit: The generated Token sequence will be transmitted to the large model analysis module for further in-depth analysis.

[0047] The support database module includes:

[0048] High-Dimensional Data Storage Unit: The cleaned water quality data, Token sequence, and weighted feature data will be stored in the support database, supporting multi-dimensional indexing and efficient querying. These data will be used for subsequent analysis and model training.

[0049] Calculation Result Storage Unit: Stores the intermediate results of large model analysis and weight calculation, facilitating quick retrieval and real-time analysis to ensure the efficient processing of large-scale water quality data.

[0050] Data Retrieval Optimization Unit: Through efficient storage and indexing technologies, it enables fast data access, supporting the fast querying and processing of real-time data.

[0051] Working Process

[0052] Water Quality Data Collection:

[0053] The sensor module regularly collects water quality data through the data collection unit. The collected water quality parameters include information such as pH value, dissolved oxygen, turbidity, ammonia nitrogen, temperature, etc.

[0054] The data is transmitted to the central processing unit through the data connection unit, and the process of data cleaning and standardization begins.

[0055] Data Cleaning and Standardization:

[0056] The original water quality data is transmitted to the data cleaning and standardization module, where noise is removed through the adaptive cleaning algorithm unit, and outlier detection is performed.

[0057] According to real-time environmental changes (such as seasonal changes, climate changes, etc.), the cleaning rules and thresholds are automatically adjusted through the adaptive threshold adjustment unit to ensure the accuracy of the data.

[0058] After the data is cleaned, it will be uniformly formatted and appended with a timestamp, ready for backup.

[0059] Adaptive Noise Removal and Outlier Detection Algorithm

[0060] The adaptive cleaning algorithm is divided into two steps: noise removal and outlier detection. First, by calculating the change range of each water quality data point and adjusting the threshold in combination with real-time environmental changes, noise removal and outlier identification are achieved.

[0061] Noise Removal Formula

[0062] The original water quality data is D={d1,d2,…,di}, where each data point d i is the real-time value of water quality parameters (such as pH, dissolved oxygen, etc.).

[0063] To perform noise removal, the change rate R of the data point is defined i as follows:

[0064]

[0065] where R i represents the change rate between consecutive data points. If R i is greater than a certain threshold T noise , it is considered that this data point may be affected by noise and needs to be denoised.

[0066] The threshold T noise is dynamically adjusted according to the seasonal variation S t and the real-time environmental variable E t :

[0067]

[0068] where: σ(D) is the standard deviation of the data set D, reflecting the degree of data fluctuation;

[0069] α is a control factor, determining the sensitivity of the noise threshold;

[0070] f(S t ,E t ) is a dynamic adjustment function based on the seasonal variation S t and the environmental variable E t . For example, when the water temperature is high in summer, the noise threshold can be appropriately increased.

[0071] Outlier detection formula

[0072] Assume that the true value of the water quality data should be within a certain reasonable range. Outlier detection identifies outliers by calculating the degree of deviation of the data point, using the Z-Score method:

[0073]

[0074] where: μ(D) is the mean of the data set D;

[0075] σ(D) is the standard deviation of the data set D.

[0076] If ∣Z i ∣>T anomaly , then d i is determined as an outlier, where T anomaly is the dynamically adjusted outlier detection threshold.

[0077] Similarly, the threshold T anomaly is dynamically adjusted according to the seasonal variation S t and environmental factor E t as follows:

[0078]

[0079] where β is a control factor that determines the sensitivity of outlier detection.

[0080] Data cleaning formula

[0081] After noise removal and outlier detection are completed, the cleaned data D cleaned can be obtained through the following formula:

[0082]

[0083] That is, only when the data points meet the noise removal and outlier detection criteria are they retained.

[0084] Adaptive threshold adjustment algorithm

[0085] In the actual environment, the fluctuations of water quality data are affected by various factors (such as seasonal changes, climate changes, etc.). Therefore, to improve the accuracy of data cleaning, the thresholds T noise and T anomaly must be dynamically adjusted according to these changes.

[0086] Threshold adjustment formula

[0087] For adaptive threshold adjustment, the seasonal variation S t and environmental factor E t affect the volatility of water quality data, and the threshold can be dynamically adjusted through the following formula:

[0088]

[0089] where:

[0090] α and β are adjustment factors that vary dynamically according to environmental conditions;

[0091] f(S t ,E t ) is a adjustment function based on seasonal variation and environmental factors, which may include factors affecting water quality such as weather, temperature, precipitation, etc.

[0092] Through these adaptive adjustment mechanisms, the threshold can be automatically updated as the environmental conditions change, ensuring that the data cleaning rules always match the current environmental state, thereby improving the accuracy and real-time performance of water quality data processing.

[0093] Token sequence generation:

[0094] The cleaned data will be backed up in chronological and spatial order to generate the original data packets, and the Token sequence generation unit will combine multiple data packets into a continuous Token sequence.

[0095] The Token sequence provides water quality data with temporal and spatial correlations for the subsequent large model analysis module.

[0096] To achieve an innovative Token sequence generation process, the present invention proposes a Token sequence generation algorithm for water quality data with temporal and spatial correlations. This algorithm not only considers the time series characteristics of water quality data but also combines the spatial location relationships of water quality data to ensure that the generated Token sequence can fully capture the spatio-temporal dependencies between water quality data.

[0097] Generation of original data packets

[0098] The result of cleaning the water quality data is D cleaned ={d1, d2, …, d n}, where each data point d i contains multiple water quality parameters (such as pH, dissolved oxygen, turbidity, etc.) and timestamp information t i , as well as spatial location information p i =(x i , yi), that is, the geographical location of the water quality data point.

[0099] We first back up the cleaned data in chronological and spatial order to form the original data packet B i , and each data packet contains the water quality data within a time period:

[0100]

[0101] Among them, B i represents the i-th data packet, which contains the continuous water quality data from time t i1 to t im . Each data point d ij contains the water quality parameter value and carries the corresponding timestamp t ij and spatial location p ij .

[0102] Token sequence generation

[0103] Next, multiple data packets are combined in chronological and spatial order through the following innovative formula to generate a Token sequence with spatio-temporal correlations. Each Token represents a slice of water quality data in time and space, ensuring the spatio-temporal continuity of the data.

[0104] Space-time weighted Token generation formula

[0105] To fully reflect the correlation between time and space, we introduce a space-time weighted factor W for each data point i , which is calculated by the following formula:

[0106]

[0107] where: ∥t i −t0∥ is the time difference between the current time t i and the reference time t0, normalized to the maximum time span T max ;

[0108] ∥p i −p0∥ is the spatial distance between the current spatial position p i and the reference position p0, normalized to the maximum spatial span P max ;

[0109] α is a tuning factor that determines the influence of time and space distances on the weighted factor.

[0110] Generate Token sequence

[0111] According to the space-time weighted factor W i , the generated Token sequence T seq consists of multiple consecutive data points. Each Token represents a water quality data point and carries the corresponding space-time weighted information. The specific generation formula is as follows:

[0112]

[0113] where d i is the water quality data point, and W i is the corresponding space-time weighted factor, reflecting the influence of time and space.

[0114] Space-time sliding window processing

[0115] To ensure space-time continuity and the context relevance of data, the present invention adopts a space-time sliding window technique to generate the Token sequence. Specifically, assume that each Token sequence is extracted from the original data packet through a sliding window, and the window size is w t (time window) and w p (space window). The Token sequence within the sliding window is generated by the following formula:

[0116]

[0117] This formula ensures that each token sequence takes into account the temporal and spatial scope within the window, making the generated token sequence have stronger spatio-temporal correlation.

[0118] Token sequence combining spatio-temporal information

[0119] The token sequence generated by the above algorithm not only contains the water quality data itself, but also contains the weighted information of time and space, enabling the subsequent large model analysis to better capture the variation laws of water quality data in different time and space dimensions. The final token sequence T seq has the following characteristics:

[0120]

[0121] Each token (di, Wi) is a combination of water quality data di and the corresponding spatio-temporal weighting factor Wi, reflecting the temporal and spatial correlation of water quality data and providing richer inputs for the large model analysis module.

[0122] Through the token sequence generation algorithm, we can backup water quality data in time and space order and combine spatio-temporal weighting factors to generate a highly correlated token sequence. This method effectively captures the changes in water quality data in time and space, ensuring that the subsequent large model analysis can accurately reflect the dynamic changes of water quality.

[0123] Data storage and query:

[0124] The cleaned water quality data, token sequence and its related feature data will be stored in a supporting database for quick query and retrieval.

[0125] During the large model analysis process, the stored calculation results and intermediate data can be retrieved in real time to provide support for further optimization of the system.

[0126] Through the machine learning-driven adaptive cleaning algorithm, it can automatically identify and process noise and abnormal data under different environmental conditions to ensure the accuracy and reliability of water quality data.

[0127] Achieve real-time dynamic monitoring:

[0128] By adopting the backup and token sequence generation methods of time series and spatial location information, it can continuously and real-time track and analyze the changes of water quality parameters to ensure that the system can adapt to the dynamically changing water quality environment.

[0129] Enhance data processing efficiency:

[0130] The supported database module adopts efficient storage and indexing technologies, greatly improving the efficiency of data storage and query, and ensuring that a large amount of water quality data can be quickly processed in a large-scale water quality monitoring system.

[0131] Adaptive adjustment and optimization of monitoring strategies:

[0132] By dynamically adjusting cleaning rules, thresholds, and data acquisition strategies, this embodiment can optimize the monitoring strategy according to real-time water quality changes and environmental conditions, ensuring the accuracy and timeliness of monitoring.

[0133] Support for subsequent large model analysis:

[0134] The generated Token sequence and standardized data provide a basis for subsequent large model analysis, ensuring the multi-dimensionality and depth of the analysis, and being able to effectively extract potential patterns and trends in water quality data.

[0135] This embodiment provides an Internet of Things-based intelligent water quality monitoring system, which combines technologies such as sensor data acquisition, machine learning-driven adaptive cleaning, real-time data storage and processing, etc., and has good real-time performance, flexibility, and adaptability.

[0136] Embodiment 2:

[0137] As Figure 1 and Figure 2 shown, this embodiment further expands Embodiment 1, combines deep learning analysis, large model training, and dynamic weight calculation modules to construct a more intelligent and accurate water quality monitoring system. The system performs multi-level feature extraction on the Token sequence of water quality data through deep learning models (DNN and Transformer), uses LSTM and GRU to model time series, generates weighted feature data, and dynamically adjusts weights according to real-time water quality data and environmental variables, thereby enhancing the adaptability and accuracy of the system to water quality changes. In addition, through the Internet of Things support unit, the system transmits water quality data and analysis results to the cloud in real time to support remote monitoring and management.

[0138] The deep learning analysis unit of the large model analysis module is based on a deep neural network (DNN) and a self-attention mechanism (Transformer), and performs multi-level feature extraction on the Token sequence. Each Token represents a water quality parameter. Through the self-attention mechanism, complex non-linear relationships between water quality parameters can be captured. The deep neural network (DNN) is used to analyze long-term and short-term trends in water quality data, and the Transformer module can dynamically adjust weights according to the context relationship of water quality data to ensure the accuracy of the analysis.

[0139] The time series modeling unit uses recurrent neural networks (RNNs) such as long short-term memory (LSTM) or gated recurrent unit (GRU) to model the time series of water quality changes. LSTM and GRU can handle long-term dependencies in water quality data, capture long-term trends and short-term fluctuations in water quality changes, and generate weighted feature data, which will be passed as input to the subsequent dynamic weight calculation module.

[0140] Through the joint modeling of DNN, Transformer, and LSTM / GRU, the system extracts multi-dimensional weighted feature data and adjusts the weight of each feature according to the current environment and water quality changes.

[0141] The dynamic weight calculation unit of the dynamic weight calculation module dynamically calculates the weights of water quality parameters based on the weighted feature data generated by the large model analysis module. This unit will analyze water quality data and environmental variables (such as climate change, precipitation, etc.) in real time and adjust the importance of each water quality parameter according to the analysis results. For example, in the dry season, the impact of water temperature may be more significant, so the weight of this parameter will increase; while in the rainy season, turbidity may have a greater impact on water quality, and the system will increase the weight of this parameter.

[0142] The weight adjustment unit dynamically adjusts the weights of each water quality parameter through real-time data input and feedback of environmental changes, ensuring that the analysis model can adapt to changes in environmental conditions in real time. The flexibility of the weight adjustment mechanism enables the system to continuously provide accurate water quality monitoring results under different seasons, climate conditions, and water quality environments.

[0143] The data transmission unit of the Internet of Things support unit uses Internet of Things technology to transmit water quality data and analysis results to the cloud in real time. Through wireless communication technologies such as Wi-Fi or LTE, water quality data can be efficiently transmitted to remote servers or cloud platforms, ensuring the real-time and accuracy of data.

[0144] The remote monitoring unit system provides a remote monitoring interface through which users can view information such as real-time water quality status, historical trends, and warning records. Through this function, managers can timely understand the water quality status and potential anomalies and make corresponding decisions based on water quality changes. The introduction of the remote monitoring unit greatly improves the interactivity and operation convenience of the system, enabling managers to access water quality data anytime and anywhere.

[0145] Working process

[0146] Data collection and cleaning:

[0147] The system regularly collects water quality data through the sensor module, and preprocesses the raw data through the data cleaning and standardization module. The cleaned data will be formatted and transmitted to the data backup and Token sequence generation module for storage and serialization.

[0148] Deep learning analysis and feature extraction:

[0149] The cleaned data is converted into a Token sequence through the Token sequence generation module and input into the large model analysis module. The deep learning analysis unit (DNN and Transformer) first performs multi-level feature extraction on the data to capture the non-linear relationships between various water quality parameters. The time series modeling unit (LSTM / GRU) performs time series modeling on the data to extract the long-term trends and short-term fluctuations of water quality changes, generating weighted feature data.

[0150] Dynamic weight calculation and adjustment:

[0151] The weighted feature data is analyzed through the dynamic weight calculation module. The weights of water quality parameters are dynamically calculated according to real-time water quality data and environmental conditions. The results of weight calculation are fed back to the system in real time to adjust the input of the prediction model and improve the accuracy and adaptability of the prediction.

[0152] Dynamic weight calculation algorithm

[0153] The water quality data consists of multiple parameters P = {p1, p2, …, p m}, where each parameter p i represents a certain feature in the water quality (such as pH, dissolved oxygen, turbidity, etc.). Through the dynamic weight calculation module, the system dynamically adjusts the weight w i of each parameter p i according to the water quality data and environmental conditions (such as climate change, precipitation, temperature, etc.), enabling the system to automatically adapt to different environmental changes.

[0154] Initial weight calculation of water quality parameters

[0155] For each water quality parameter p i , the initial weight can be calculated by the following formula:

[0156]

[0157] where σ(p i ) is the standard deviation of the water quality parameter p i , reflecting the volatility of this parameter, and m is the total number of water quality parameters. Through this formula, the system assigns initial weights to each water quality parameter, and parameters with larger fluctuations will obtain higher initial weights.

[0158] Dynamic Weight Adjustment Affected by Environmental Conditions

[0159] The dynamic weight adjustment takes into account the impact of environmental conditions on water quality parameters. For example, factors such as seasonal temperature changes and precipitation. We define the impact factor E of environmental conditions t to represent the degree of influence of the environment on water quality parameters. This factor can be expressed by the following formula:

[0160]

[0161] Where: T t is the current temperature, T0 is the reference temperature, and T max is the maximum temperature change;

[0162] P t is the current precipitation, P0 is the reference precipitation, and P max is the maximum precipitation;

[0163] γ is an adjustment factor used to balance the impact of different environmental factors on water quality parameters.

[0164] The environmental condition impact factor E t will affect the dynamic weight of each water quality parameter. Specifically, the dynamic weight w i of water quality parameter p i (t) is adjusted by the following formula:

[0165]

[0166] Where, f(E t ) is an adjustment function based on environmental condition E t and can be expressed in the following form:

[0167]

[0168] Where, α and β are adjustment factors used to control the degree of influence of environmental changes on the weight.

[0169] Feedback Mechanism after Weight Adjustment

[0170] When the changes in water quality data and environmental conditions affect the weights of water quality parameters, the new dynamic weight (t) will be fed back to the system and used as the input for the subsequent prediction model. These dynamic weights will be used to adjust the input features of the model to improve the prediction accuracy. This feedback process is expressed by the following formula:

[0171]

[0172] Where, D adjustedIt is the adjusted water quality parameter set, which contains each water quality parameter and its corresponding dynamic weight w i (t).

[0173] Overall flowchart and feedback mechanism

[0174] Data collection and cleaning: Collect water quality data through sensors, and remove noise and outliers through an adaptive cleaning algorithm to ensure the accuracy of the data.

[0175] Initial weight calculation: Calculate the initial weight of each water quality parameter based on the cleaned water quality data .

[0176] Environmental condition analysis: Analyze environmental conditions (such as temperature, precipitation, etc.) in real time, and calculate the environmental condition impact factor E t .

[0177] Dynamic weight adjustment: According to the environmental condition E t and the initial weight , calculate the dynamic weight of each water quality parameter (t).

[0178] Feedback and optimization: Feed the dynamic weight (t) back to the system, adjust the input features of the water quality prediction model, and optimize the prediction results.

[0179] Data transmission and remote monitoring:

[0180] The water quality data and analysis results are transmitted to the cloud in real time through the Internet of Things data transmission unit, ensuring that managers can view the water quality monitoring results remotely. Through the remote monitoring unit, users can view the water quality status and historical trends in real time, receive water quality warning signals in a timely manner, and adjust the monitoring strategy according to the warnings.

[0181] By combining deep learning models such as DNN, Transformer, and LSTM / GRU, the system can deeply mine the complex non-linear relationships in the water quality data, capture the long-term trends and short-term fluctuations of water quality changes, which makes the water quality prediction more accurate and can effectively identify potential water quality problems.

[0182] The dynamic weight calculation module dynamically adjusts the weights of water quality parameters according to real-time data and environmental changes, enabling the system to flexibly adapt to different environmental conditions, such as climate change, rainfall change, etc. This dynamic adaptation ability improves the accuracy and reliability of water quality monitoring.

[0183] The Internet of Things support unit greatly enhances the interactivity and convenience of the system through real-time data transmission and remote monitoring functions. Managers can obtain real-time data and early warning information through a remote interface at any location, enabling them to respond more quickly to water quality changes and take measures to ensure water source security.

[0184] The data backup and Token sequence generation module ensures the efficient storage and processing of water quality data. Through high-dimensional data storage and multi-dimensional indexing technologies, the system can quickly process a large amount of water quality data and support real-time query and analysis.

[0185] This embodiment provides an intelligent water quality monitoring system based on deep learning and dynamic weight calculation. Combining with Internet of Things technology, it can monitor, analyze water quality changes in real time and predict potential problems, greatly improving the intelligent level and response ability of water quality management.

[0186] Embodiment 3:

[0187] As Figure 1 and Figure 2 shown, in this embodiment, based on Embodiment 1 and Embodiment 2, combined with the water quality prediction and early warning module and the intelligent feedback mechanism module, the present invention further strengthens the system's ability to predict water quality changes and emergency response ability. The system analyzes historical and real-time data through a deep learning model to predict future water quality trends and can identify potential water quality problems in advance. Through the early warning system, managers can obtain early warning signals in a timely manner and take measures. The intelligent feedback mechanism can automatically adjust the monitoring strategy after the early warning is triggered to improve the monitoring accuracy and response efficiency.

[0188] The water quality prediction unit of the water quality prediction and early warning module uses deep learning methods, combines historical water quality data and real-time water quality monitoring data to predict water quality trends. Deep learning models (such as LSTM, GRU, etc.) are trained on time series data to analyze the long-term dependence relationship and change trend between water quality parameters. The water quality prediction unit can predict future water quality conditions according to environmental changes (such as seasonal fluctuations, climate changes, etc.).

[0189] The potential problem identification unit is based on the output of the water quality prediction unit. The potential problem identification unit identifies possible water quality anomalies through real-time monitoring of water quality trends. For example, the prediction model may find that the change trends of water temperature, dissolved oxygen or turbidity may exceed the normal range, and at this time the system will generate an early warning signal to notify relevant managers.

[0190] Early warning signal generation When the prediction result shows that the water quality is about to become abnormal, the potential problem identification unit automatically generates an early warning signal and transmits it to the early warning response unit. The early warning signal will trigger subsequent processing and response processes, including measures such as increasing the sampling frequency and adjusting the sampling location.

[0191] When the feedback optimization unit of the intelligent feedback mechanism module is triggered by a warning signal, according to the warning type (such as sudden pollution source, abnormal temperature, etc.), the feedback optimization unit automatically adjusts the data sampling frequency and sensor position. The system will optimize the monitoring strategy according to the urgency of the warning. For example, when the water quality is abnormal, the system will automatically increase the data sampling frequency to ensure that more water quality change information is captured.

[0192] When water quality anomalies occur or a warning is triggered, the data acquisition optimization unit will enhance the data acquisition efforts. For example, when there are abnormal fluctuations in turbidity, the system may increase the sampling frequency of other relevant parameters in the water quality (such as dissolved oxygen) to comprehensively monitor the changes in water quality.

[0193] Nonlinear Feature Extraction

[0194] The water quality data consists of multiple parameters. Assume the water quality parameter data set is D = {p1, p2, …, p m}, where p i represents the i-th water quality parameter (such as pH, dissolved oxygen, turbidity, etc.). The deep learning analysis unit (including DNN and Transformer) takes these parameter data as input and performs feature extraction through a multi-layer neural network.

[0195] DNN Feature Extraction Formula

[0196] Each water quality parameter pi undergoes a non-linear mapping through the multi-layer perceptron (MLP) of the deep neural network. We define the input of the k-th layer as h (k−1) (the initial input is D), and the output is h (k) , and the formula is:

[0197]

[0198] Where: W (k) is the weight matrix of the k-th layer;

[0199] b (k) is the bias term;

[0200] σ(.) is the activation function (such as ReLU, Leaky ReLU, etc.).

[0201] Through the learning of the multi-layer neural network, the model can extract complex non-linear relationships from the water quality data.

[0202] Transformer Feature Extraction Formula

[0203] To capture the dependency relationships among water quality parameters, a self-attention mechanism is adopted for feature extraction. Assume that the input water quality data is processed by an encoder to obtain query (Query), key (Key), and value (Value) vectors Q, K, and V. The self-attention output is calculated through the following formula:

[0204]

[0205] where: d k is the dimension of the key vector;

[0206] softmax is the standard Softmax function, which is used to calculate the attention weights among water quality parameters.

[0207] After multiple layers of self-attention mechanisms, Transformer can capture the complex correlation relationships among water quality parameters.

[0208] Weighted Feature Generation

[0209] The feature vectors h (k) extracted by DNN and Transformer are weighted and combined to generate feature data with spatio-temporal correlations. The weighting process adjusts the importance of each water quality parameter through an adaptive weight factor wi(t). The specific formula is:

[0210]

[0211] where, w i (t) is the weight factor dynamically adjusted by real-time environmental factors, and h i (t) is the feature extracted from the water quality parameter p i by the deep learning model.

[0212] Long-Term Trend and Short-Term Fluctuation Modeling

[0213] To capture the long-term trend and short-term fluctuations of water quality changes, LSTM or GRU is used for time series modeling. Assume that the time series input is Xt={x1, x2, …, xt}, where xi is the water quality data at the i-th moment. LSTM or GRU will predict future water quality changes based on past water quality data, capturing long-term dependencies and short-term fluctuations in the time series.

[0214] LSTM State Update Formula

[0215] The state update of the LSTM cell is carried out through the following formula:

[0216]

[0217] where: i tis the input gate, which determines the importance of the current input information;

[0218] f t is the forget gate, which determines the degree of retention of previous information;

[0219] o t is the output gate, which controls the output at the current time;

[0220] c t is the cell state, which stores long-term dependence information.

[0221] Through the above state update formula of LSTM, the model can capture the long-term trends and short-term fluctuations in water quality data.

[0222] Weighted Feature Generation and Output

[0223] The time series feature h output by the LSTM model t and the weighted feature are combined to generate weighted predicted feature data :

[0224]

[0225] Here, w i (t) is the weight dynamically adjusted by environmental factors, and h i (t) is the time series feature extracted by the LSTM model, representing the change trend of water quality parameters.

[0226] Dynamic Policy Adjustment Unit This unit adaptively adjusts the monitoring policy by analyzing the changes in water quality data and environmental changes in real time, which includes adjusting the sensor configuration, changing the sampling mode, and adjusting the analysis period. For example, in the dry season, the impact of water temperature on water quality may increase, and the system will automatically increase the monitoring frequency of water temperature; while in the rainy season, the system will automatically increase the monitoring of turbidity and ammonia nitrogen in water quality.

[0227] Data Transmission Unit of the Internet of Things Support Unit Through the Internet of Things technology, water quality data and analysis results are transmitted to the cloud in real time to ensure remote monitoring and data sharing. The data transmission unit supports transmitting real-time monitoring results to the management platform through wireless communication technology, improving the remote response ability of the system.

[0228] The Remote Monitoring Unit provides a remote monitoring interface for managers, supporting real-time viewing of information such as water quality status, historical trends, and warning records. Managers can view water quality data through the interface, understand potential anomalies, and make response decisions in advance.

[0229] Working Process

[0230] Water Quality Data Collection and Cleaning:

[0231] The system collects water quality data through the sensor module, including but not limited to multiple water quality parameters such as pH value, dissolved oxygen, turbidity, etc. The data collected by the sensor is transmitted to the central processing unit through the data connection unit and enters the data cleaning and standardization module for preprocessing.

[0232] Water quality prediction and trend analysis:

[0233] The cleaned data is input into the water quality prediction unit, and deep learning models (such as LSTM, GRU, etc.) are used to analyze historical data and real-time data to predict future water quality change trends. For example, the system can predict the changes in water quality parameters within the next few hours or days.

[0234] Based on the prediction results, the potential problem identification unit detects upcoming water quality anomalies according to the water quality trend, such as sudden fluctuations in dissolved oxygen concentration or a sharp rise in turbidity, and generates a warning signal.

[0235] Warning signal generation and feedback optimization:

[0236] When the potential problem identification unit generates a warning signal, the feedback optimization unit will automatically adjust the sampling frequency and sensor location according to the warning type. For example, if it is predicted that the turbidity will be abnormal, the system will immediately increase the monitoring frequency of turbidity and mobilize more sensors to sample in this area.

[0237] The data acquisition optimization unit further increases the sampling intensity to ensure that more detailed water quality change information can be captured in a timely manner, which helps the system to respond more precisely to water quality changes.

[0238] Remote monitoring and response:

[0239] The water quality data and warning signals are transmitted to the cloud in real time through the data transmission unit. The remote monitoring unit provides a real-time interface for managers to support viewing the water quality status, historical trends, and warning records. Managers can make emergency responses in a timely manner according to the warning signals.

[0240] Dynamic adjustment and optimization:

[0241] The dynamic strategy adjustment unit adjusts the monitoring strategy based on data analysis and real-time monitoring results to adapt to different environmental conditions and water quality changes. For example, in the case of a sudden pollution source, the system will give priority to collecting water quality data near the pollution source to quickly respond to abnormal changes.

[0242] Through the water quality prediction function of the deep learning model, the system can predict the water quality trend and potential problems in advance. The prediction model can timely identify the anomalies in water quality changes based on historical data and real-time monitoring data, provide early warnings for managers, and ensure that response measures are taken in a timely manner.

[0243] The feedback optimization unit and the data acquisition optimization unit adjust the sampling frequency and sensor position in real time according to the warning signal, thereby improving the monitoring accuracy. When sudden water quality changes occur, the system can respond quickly to ensure the timeliness and accuracy of data.

[0244] The dynamic strategy adjustment unit adaptively adjusts the monitoring strategy according to the real-time water quality data and environmental changes. This intelligent adjustment mechanism enables the system to flexibly adjust the monitoring strategy according to different seasons, weather changes or emergencies, enhancing the adaptability and flexibility of the system.

[0245] Through the Internet of Things technology, water quality data and warning signals can be transmitted to the cloud in real time. The remote monitoring unit provides a real-time monitoring interface for managers, greatly improving the management efficiency and response speed. Managers can view the water quality status at any time and take emergency measures in a timely manner.

[0246] The intelligent feedback mechanism and dynamic adjustment strategy in this embodiment ensure the comprehensiveness and accuracy of the water quality monitoring process, and can quickly provide corresponding solutions in case of water quality anomalies or emergencies to ensure water quality safety and environmental protection.

[0247] By combining deep learning, intelligent feedback and Internet of Things technology, this embodiment provides an efficient, accurate and dynamically adaptable intelligent water quality monitoring system, greatly enhancing the intelligent level of water quality monitoring and improving the warning ability and monitoring accuracy.

[0248] The above is only the preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be changed within the scope of the concept described herein through the above teachings or the technology or knowledge in related fields. Any changes and variations made by those skilled in the art without departing from the spirit and scope of the present invention shall fall within the protection scope of the appended claims of the present invention.

Claims

1. An intelligent water quality monitoring device based on the Internet of Things, characterized in that Including: A sensor module responsible for collecting raw water quality data, including: a data collection unit for regularly collecting water quality data, where the water quality data includes multiple parameters such as pH value, dissolved oxygen, turbidity, ammonia nitrogen, and temperature; a data connection unit for transmitting the collected raw water quality data to the central processing unit for subsequent processing; A data cleaning and standardization module that receives the raw water quality data transmitted by the sensor module and performs cleaning and standardization processing, including: an adaptive cleaning algorithm unit that dynamically adjusts data cleaning rules to remove noise based on historical water quality data trends and real-time changes through machine learning algorithms; an anomaly detection unit that uses a machine learning model to real-time identify outliers in the raw water quality data and perform denoising processing; an adaptive threshold adjustment unit that dynamically adjusts the cleaning threshold according to environmental conditions and data characteristics, converts the cleaned water quality data into a unified format and timestamp, ensuring data consistency and facilitating subsequent packaging; A data backup and Token sequence generation module that receives the cleaned water quality data and converts it into an analyzable sequence, including: a data backup unit that backs up the cleaned water quality data according to time and spatial information, where each data packet contains multiple water quality parameters and the collection time; a Token sequence generation unit that combines multiple data packets in chronological order and spatial position order to generate a continuous Token sequence, where the Token sequence contains multi-dimensional water quality parameter information and is formed according to data changes and spatial positions; a data transfer unit that transmits the generated Token sequence to the large model analysis module for in-depth analysis; A large model analysis module that receives the Token sequence, extracts water quality parameter features and analyzes the correlation relationship, including: a deep learning analysis unit that performs multi-level correlation analysis on the Token sequence based on a deep neural network and a self-attention mechanism to capture the complex non-linear relationship between water quality parameters; a time series modeling unit that uses a recurrent neural network with long short-term memory or gated recurrent units to model the Token sequence and extract the long-term trend and short-term fluctuations of water quality changes; an adaptive weighting mechanism unit that calculates the correlation degree of each water quality parameter through the self-attention mechanism and weights it, and adjusts the parameter importance in real-time according to data changes and environmental conditions to generate weighted feature data; A dynamic weight calculation module that receives the weighted feature data from the large model analysis module and calculates the dynamic weight, including: a dynamic weight calculation unit that automatically assigns a dynamic weight to each water quality parameter according to the real-time water quality data analysis result, reflecting its influence degree in the current environment; a weight adjustment unit that dynamically corrects the water quality parameter weights in combination with real-time data input and environmental variable changes to ensure the accuracy of the analysis result; The prediction and early warning module receives dynamic weights and feature data for water quality prediction and anomaly early warning, including: a water quality prediction unit that uses deep learning methods based on large models, combines historical water quality data and real-time monitoring data to predict long-term water quality change trends, taking into account climate change, seasonal factors, and pollution source impacts; a potential problem identification unit that identifies potential anomalies based on real-time water quality data and prediction results and generates early warning signals; and an early warning response unit that automatically issues an alarm according to the early warning signal and initiates adjustment measures to optimize the data collection frequency and sampling locations to address anomalies. The intelligent feedback mechanism module adjusts the monitoring strategy and optimizes data collection according to the early warning response, including: a feedback optimization unit that automatically adjusts the sampling frequency and sensor locations when an early warning is triggered to improve the accuracy of water quality monitoring; a data collection optimization unit that increases the sampling frequency and enhances data collection efforts during abnormal events to provide more detailed real-time information; and a dynamic strategy adjustment unit that adaptively adjusts the sensor configuration, sampling mode, and analysis period based on data analysis results and environmental changes.

2. The intelligent water quality monitoring device based on the Internet of Things according to claim 1, characterized in that: The adaptive cleaning algorithm unit in the data cleaning and standardization module is based on a machine learning model and dynamically adjusts the cleaning parameters using historical water quality data and real-time changes to ensure the accuracy of the cleaned water quality data and its adaptability to environmental changes.

3. An intelligent water quality monitoring device based on the Internet of Things according to claim 1, characterized in that: The Token sequence generation unit in the data backup and Token sequence generation module packages the cleaned water quality data into multi-dimensional Token sequences according to time series and spatial location information to ensure data continuity and spatial correlation, providing input for large model analysis.

4. An intelligent water quality monitoring device based on the Internet of Things according to claim 1, characterized in that: The deep learning analysis unit in the large model analysis module extracts multi-level features of the Token sequence through a deep neural network and self-attention mechanism, and the time series modeling unit captures the water quality change trend through long short-term memory or gated recurrent units to generate weighted feature data for subsequent weight calculation and prediction.

5. An intelligent water quality monitoring device based on the Internet of Things according to claim 1, characterized in that: The dynamic weight calculation unit in the dynamic weight calculation module calculates and assigns dynamic weights to each water quality parameter according to real-time water quality data and environmental variables, and the weight adjustment unit dynamically corrects the weights according to data changes to improve the adaptability and accuracy of the analysis.

6. The intelligent water quality monitoring device based on the Internet of Things according to claim 1, wherein: The water quality prediction unit in the prediction and early warning module combines historical data and real-time data to predict the water quality trend through a deep learning model, and the potential problem identification unit generates early warning signals based on the prediction results to ensure the timely identification of potential anomalies.

7. An intelligent water quality monitoring device based on the Internet of Things according to claim 1, characterized in that: The feedback optimization unit and data collection optimization unit in the intelligent feedback mechanism module adjust the sampling frequency and data collection efforts according to the early warning signal, and the dynamic strategy adjustment unit adaptively optimizes the monitoring strategy to ensure accurate monitoring during anomalies.

8. An intelligent water quality monitoring device based on the Internet of Things according to claim 1, characterized in that: The device further includes a support database module, and the support database module includes: a high-dimensional data storage unit that stores the cleaned water quality data, Token sequences, and weighted feature data, supporting multi-dimensional indexing and efficient querying; a calculation result storage unit that stores the intermediate results of large model analysis and weight calculation, facilitating fast retrieval and real-time analysis; and a data retrieval optimization unit that realizes fast data access through efficient storage and indexing technologies, supporting real-time processing of large-scale water quality data.

9. The intelligent water quality monitoring device based on the Internet of Things according to claim 1, characterized in that: The device further includes an Internet of Things support unit, and the Internet of Things support unit includes: a data transmission unit that transmits the water quality data and analysis results to the cloud in real time through Internet of Things technology; and a remote monitoring unit that supports viewing the water quality status, historical trends, and warning records through a remote interface, enhancing the system interactivity.

Images