Resident drinking water health risk assessment method based on big data analysis

Abstract

The present application relates to the technical field of public health and drinking water safety data processing, in particular to a resident drinking water health risk assessment method based on big data analysis, comprising the steps of basic data collection, multi-dimensional label and file generation, individualized management limit value calculation and comprehensive risk comparison and early warning; the method combines terminal water data, pipe network water quality and user physiological files, uses a prediction model to extract environmental parameters, and calculates a user's dynamic threshold adjustment coefficient by weighting to obtain an individualized management limit value; by introducing a transfer learning and sample compensation mechanism to optimize the model parameters for a specific physiological tolerance group, combining a dynamic engine to generate an individualized management threshold, the cumulative amount of pollution exposure is compared with the individualized management limit value for judgment; the present application overcomes the limitations of traditional single fixed judgment standards, solves the problem of low assessment accuracy for special groups, and realizes adaptive drinking water safety control for different characteristic groups.

Description

Technical Field

[0001] This invention relates to the field of public health and drinking water safety data processing technology, specifically a method for assessing the health risks of residents' drinking water based on big data analysis. Background Technology

[0002] As residents pay increasing attention to drinking water health, drinking water health risk assessment has become an important part. However, the existing single judgment standard has failed to establish a dynamic compensation mechanism for people with different physiological tolerance levels; the existing assessment model does not fully consider the significant differences in physiological metabolic capacity, immune system development level and other aspects among different groups; due to the lack of individualized feature collection and integration mechanism, the existing technology cannot support personalized assessment. Meanwhile, for special populations such as infants and young children and patients with chronic diseases, the training accuracy of conventional risk assessment models is low and the generalization ability is poor due to insufficient sample data. In addition, existing assessment results are mostly single judgments of meeting or exceeding the standard, lacking readability and targeted health guidance, resulting in insufficient technical targeting of assessment results and failing to achieve adaptive drinking water safety control for different populations. Summary of the Invention

[0003] The purpose of this invention is to provide a method for assessing the health risks of drinking water for residents based on big data analysis. This addresses the problems of existing technologies where health risk assessment often relies on a single, fixed standard, fails to consider individual physiological and water usage differences, has low accuracy in assessment models for special populations, and lacks personalized guidance value. Specifically, the technical solution of this invention includes the following steps: The system acquires real-time water usage data from smart water purification terminals, pipeline water quality monitoring data, and user attribute data including vital signs such as weight; it extracts the real-time water usage data, pipeline water quality monitoring data, and user attribute data, and aggregates them to generate a user risk profile with tag classification; based on the user risk profile, it determines the individualized management limit for the corresponding user; it calculates the cumulative water quality risk by combining the real-time water usage data, and compares the cumulative water quality risk with the individualized management limit to obtain a comprehensive risk assessment result; and it triggers the corresponding equipment-level early warning strategy based on the comprehensive risk assessment result.

[0004] Preferably, the step of generating a user risk profile including tag classification specifically involves generating and summarizing user attribute tags, historical record tags, water use behavior tags, and water quality status tags. The generation process includes: Age and gender information are extracted from the user attribute data, and risk tolerance levels are divided according to age and gender information to generate user attribute tags; Historical archive information is extracted from the user attribute data, past medical history records are parsed, it is determined whether there are any special medical histories that affect metabolism, and historical archive tags are generated based on the determination results. The average daily water consumption is extracted from the real-time water consumption data, the weight parameter is extracted from the vital signs parameters, and the water consumption ratio is calculated in combination with the weight parameter. Based on the water consumption ratio, the water consumption behavior is divided into a first water consumption ratio interval type below a preset ratio threshold, a second water consumption ratio interval type within a normal ratio threshold, or a third water consumption ratio interval type above a preset ratio threshold, and a water consumption behavior label is generated. Historical water quality fluctuation data are extracted from the pipeline water quality monitoring data, and risk assessment values ​​are assigned based on the historical water quality fluctuation data to generate water quality status labels. The user risk profile is generated by summarizing the user attribute tags, historical record tags, water use behavior tags, and water quality status tags.

[0005] Preferably, the risk assessment index of the smart water purification terminal is obtained by extracting the user's equipment operating environment parameters using a risk prediction model, including the following steps: Water consumption and metabolic data of a sample population whose number reaches a preset sample size threshold are obtained from the basic database as a basic reference dataset, and water consumption profile data of a target population whose number is lower than the preset sample size threshold are obtained as a target population dataset; wherein, the target population dataset contains known real risk markers; A device parameter extraction model is constructed. The device parameter extraction model adopts a multi-layer neural network architecture. Its input layer nodes correspond to multi-dimensional feature dimensions such as real-time water use and pipeline water quality. It is used to extract the device operating environment parameters in the basic reference dataset and the target population dataset. The population matching coefficient is determined by combining the physiological tolerance differences between the target population and the sample population, and the sample missing ratio of the target population dataset is calculated based on the sample number of the target population dataset and the preset sample number threshold, thereby determining the sample compensation coefficient; the device parameter extraction model is jointly trained using the target population dataset containing known real risk markers and the basic reference dataset based on the sample compensation coefficient to obtain the risk prediction model. The risk assessment indicators of the intelligent water purification terminal are output through the risk prediction model.

[0006] Preferably, calculating the sample compensation coefficient includes the following steps: Obtain the number of samples in the target population dataset and compare the number of samples with a preset sample number threshold; determine the sample compensation coefficient based on the comparison result, the sample missing ratio of the target population dataset, and the population matching coefficient.

[0007] Preferably, the individualized management limit for the user is determined by weighting the various category tags in the user's risk profile, including the following steps: Extract the user attribute tags, historical record tags, and water usage behavior tags corresponding to the user from the user risk profile; Calculate the age adjustment coefficient based on the user attribute tags, calculate the health status correction coefficient based on the historical record tags, and calculate the water use behavior compensation coefficient based on the water use behavior tags. The system retrieves the basic water quality limits for equipment from the management database, and obtains a first weighting factor for a preset age adjustment coefficient, a second weighting factor for a preset health status correction coefficient, and a third weighting factor for a preset water use behavior compensation coefficient. Based on the mapping relationship between the age adjustment coefficient, health status correction coefficient, and water use behavior compensation coefficient and their corresponding weighting factors, the system determines the user's threshold adjustment coefficient. Based on the mapping relationship between the basic water quality limits for equipment and the threshold adjustment coefficient, the system determines the individualized management limits for the corresponding user.

[0008] Preferably, the calculation of an age adjustment coefficient based on the user attribute tags, a health status correction coefficient based on the historical record tags, and a water use behavior compensation coefficient based on the water use behavior tags includes the following steps: An age adjustment coefficient is determined based on the matching relationship between the user's actual age and a preset age range; a health status correction coefficient is determined based on whether the user has any special medical history that affects metabolism or makes them susceptible to microorganisms; and a water use behavior compensation coefficient is determined based on the comparison between the user's average daily water consumption and a preset normal water consumption range.

[0009] Preferably, obtaining the first weighting factor of the age adjustment coefficient, the second weighting factor of the health status correction coefficient, and the third weighting factor of the water use behavior compensation coefficient includes the following steps: Set initial values ​​for the first weighting factor, the second weighting factor, and the third weighting factor, and ensure that the sum of each factor is 1; obtain historical maintenance records or disease statistics provided by the regional management department; and dynamically update the first weighting factor, the second weighting factor, and the third weighting factor based on the comparison results between the proportion of the target population whose maintenance records are caused by water quality abnormalities and the preset proportion threshold.

[0010] Preferably, the cumulative water quality risk is calculated by combining the real-time water usage data, and the cumulative water quality risk is compared with the individualized management limit to obtain a comprehensive risk assessment result, including the following steps: Based on the real-time water usage data, the single water usage is extracted, and combined with the target water quality index concentration in the pipeline water quality monitoring data, the single pollution exposure is determined; the vital signs parameters in the user's risk profile are extracted, and based on the multiple single pollution exposures and vital signs parameters within the statistical period, the user's cumulative water quality risk is determined; The cumulative water quality risk is compared with the individualized management limit, and the user's equipment operation and maintenance risk level is determined based on the degree of difference in the comparison, thus generating the comprehensive risk assessment result.

[0011] Preferably, triggering a corresponding equipment-level early warning strategy based on the comprehensive risk assessment result includes the following steps: Based on the comparison results between the accumulated water quality risk and the preset safety and warning limits, the user's equipment operation and maintenance risk level is divided into a safe level, a concern level, or a warning level. According to the determined risk level, strategies are executed to record equipment operation data, issue equipment maintenance instructions to the smart water purification terminal, or issue shutdown maintenance instructions and generate a user drinking water assessment report.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention overcomes the shortcomings of existing technologies that rely on a single fixed standard and cannot adapt to people with different physiological tolerance levels by constructing a multi-dimensional population health risk characteristic model that includes physiological, health, behavioral, and environmental factors, and by combining it with a dynamic threshold calculation engine to generate individualized management thresholds, thereby improving the accuracy of risk assessment. 2. This invention introduces a transfer learning framework, which uses the rich data features of healthy adults to compensate for samples in the target population, and adaptively adjusts the transfer weights according to the proportion of missing samples. By adaptively adjusting the transfer weights, the bias of small sample data on the prediction output of the risk assessment model for special populations is reduced. 3. This invention directly compares real-time exposure levels with individualized thresholds to form a three-tiered early warning strategy of safety, concern, and warning. It can also automatically generate personalized reports containing specific risk factor analyses, providing concrete data references and equipment maintenance guidance for the assessment results. Attached Figure Description

[0013] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0014] Figure 1This is a schematic diagram of the module composition of the Residents' Drinking Water Health Risk Assessment Method based on Big Data Analysis of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] Please see Figure 1 A method for assessing the health risks of drinking water for residents based on big data analysis includes: acquiring real-time water usage data from smart water purification terminals, water quality monitoring data from the pipe network, and user attribute data including vital signs such as weight; extracting the real-time water usage data, water quality monitoring data from the pipe network, and user attribute data, and summarizing them to generate user risk profiles with tag classifications; determining individualized management limits for corresponding users based on the user risk profiles; calculating the cumulative water quality risk by combining real-time water usage data, and comparing the cumulative water quality risk with the individualized management limits to obtain a comprehensive risk assessment result; and triggering corresponding equipment-level early warning strategies based on the comprehensive risk assessment result.

[0017] The steps for generating a user risk profile with categorized tags are as follows: generating and summarizing user attribute tags, historical record tags, water use behavior tags, and water quality status tags. The generation process includes: extracting age and gender information from user attribute data, classifying risk tolerance levels based on age and gender information, and generating user attribute tags; extracting historical record information from user attribute data, determining whether there are any special medical histories that affect metabolism, and generating historical record tags based on the determination results. The average daily water consumption is extracted from real-time water consumption data, and the weight parameter is extracted from the vital signs parameters. The water consumption ratio is calculated in combination with the weight parameter. Based on the water consumption ratio, the water consumption behavior is divided into the first water consumption ratio interval type below the preset ratio threshold, the second water consumption ratio interval type within the normal ratio threshold, or the third water consumption ratio interval type above the preset ratio threshold, and water consumption behavior labels are generated. Historical water quality fluctuation data is extracted from pipeline water quality monitoring data. Risk assessment values ​​are assigned based on historical water quality fluctuation data to generate water quality status labels. User attribute labels, historical record labels, water use behavior labels, and water quality status labels are summarized to generate user risk profiles.

[0018] Using a risk prediction model, the user's equipment operating environment parameters are extracted to obtain the risk assessment index of the smart water purification terminal. The steps include: obtaining water consumption and metabolism data of a sample population with a number of samples reaching a preset sample number threshold from the basic database as a basic reference dataset, and obtaining water consumption record data of a target population with a number of samples below the preset sample number threshold as a target population dataset. The target population dataset contains known real-world risk markers; A device parameter extraction model was constructed, employing a multi-layer neural network architecture. The input layer nodes correspond to multi-dimensional features such as real-time water usage and pipeline water quality, used to extract equipment operating environment parameters from the basic reference dataset and the target population dataset. The multi-layer neural network architecture specifically includes one input layer, three hidden layers, and one output layer. The hidden layers uniformly use the ReLU activation function to avoid gradient vanishing, and the optimizer employs the Adam algorithm. The specific mathematical model formula for the sample compensation coefficient is as follows: ,in The sample compensation coefficient and The population matching coefficient. The preset sample size threshold and Given the number of samples in the target population dataset, the deep mechanism for joint training is to adaptively weight the cross-entropy loss of the target population prediction output and the maximum mean difference loss of the features extracted from the basic reference dataset, and use this weighted average as the total loss function for joint backpropagation iteration. The population matching coefficient is determined by combining the physiological tolerance differences between the target population and the sample population. The sample missing ratio of the target population dataset is calculated based on the sample number of the target population dataset and the preset sample number threshold, and the sample compensation coefficient is determined accordingly. The device parameter extraction model is jointly trained using the target population dataset containing known real risk markers and the basic reference dataset based on the sample compensation coefficient to obtain the risk prediction model.

[0019] This invention is applied to a drinking water health risk assessment scenario involving a smart water purification terminal in a residential household, a cloud management platform, a regional pipeline water quality monitoring system, and a resident health record database. The smart water purification terminal is a water purification device installed in a residential household. It can record at least the single water consumption, average daily water consumption, drinking frequency, drinking time, and the TDS difference between raw water and purified water. It can also receive maintenance instructions or consumable replacement instructions issued by the cloud management platform. The regional pipeline water quality monitoring system consists of water quality sensing devices and their data platform installed in the water supply network or regional monitoring points. The water quality indicators it provides include residual disinfectant, heavy metal ions, turbidity, and microbial indicators. The resident health record database is provided by the regional public health management department or user-authorized data interface, and includes vital signs information such as age, gender, past medical history, estimated basal metabolic rate, and weight. All sensitive health record data directly related to the characteristics of the target individual is obtained in advance based on a legally obtained individual authorization agreement that can be withdrawn at any time. Before formally reading and executing medical history analysis, the cloud platform performs desensitization and differential privacy and other desensitization and security processing on the raw data stream. The scope of core features complies with the minimum necessary principle for privacy information security and has passed data security ethics review. In this invention, the basic data collection refers to the set of raw data required by the cloud management platform before conducting a health risk assessment of residents' drinking water, which comes from three physical or data subjects. When the smart water purification terminal completes a water usage record, the regional pipe network water quality monitoring system forms a water quality monitoring record for the corresponding time period, and the resident health record database contains user attribute records, the cloud management platform receives the real-time water usage data uploaded by the smart water purification terminal through the Internet of Things interface, reads the pipe network monitoring data through the regional pipe network data interface, and reads the user attribute data through the resident health record interface. After the data enters the cloud management platform, the platform matches the fields according to the user number, device number, collection time and residential area, so that the water use behavior, water quality environment and user attributes of the same user in the same statistical period can be merged and processed; in this embodiment, real-time water use data specifically refers to the device operation record formed by the smart water purification terminal during the user's water collection or drinking process, including single water consumption, average daily drinking frequency, drinking time distribution and TDS difference between raw water and purified water. The single water consumption corresponds to the objective water volume collected by the flow meter of the water purification terminal or the metering unit at the device end; the drinking water period corresponds to the water collection time recorded by the local controller of the device; the TDS difference corresponds to the difference between the raw water TDS measurement value and the purified water TDS measurement value; the above data are the actual water volume and water quality-related physical quantities generated during the operation of the smart water purification terminal. In this embodiment, the regional pipeline monitoring data specifically refers to the water quality indicator data collected by water quality monitoring points within the water supply area, including at least disinfectant residue, heavy metal ions, turbidity, and microbial indicators. After receiving this data, the cloud management platform associates the pipeline water quality indicators of the corresponding area with the user's smart water purification terminal data based on the user's residential area or device binding address. Thus, the subsequent risk calculation considers both the user's actual drinking water volume and the actual water quality status of the pipeline network in the user's area. In this invention, user attribute data specifically refers to data related to the user's physiological tolerance, including age, gender, medical history, and physical signs; physical signs are preferably expressed as weight, but may also include an estimated basal metabolic rate; after reading this data, the cloud management platform does not directly output the risk assessment results, but first generates multiple classification labels so that data from different sources can be converted into a unified format used in subsequent calculations. When generating user risk profiles, the cloud management platform first extracts age and gender information from user attribute data. When the user's age and gender fields exist, the platform classifies risk tolerance levels according to age and gender and generates user attribute tags. User attribute tags refer to the physiological characteristic classification results obtained based on age and gender, which are used to represent the basic tolerance differences of users to water quality risk indicators. For example, adults aged 18 to 60 can be classified into the normal risk tolerance level, the elderly over 60 years old into the lower tolerance level, infants aged 0 to 3 years old into the lower tolerance level, and people aged 4 to 17 years old into the transitional tolerance level; for the infants with extremely low tolerance mentioned above, the actual water risk assessment scenario is strictly limited to the indirect exposure scenario of using non-direct drinking water after further boiling the water from the smart water purifier to prepare formula milk powder; the function of this label is to convert text or numerical fields such as age and gender into classification results that can be called upon in subsequent threshold calculations; Then, the cloud management platform extracts historical record information from user attribute data, determines whether there is any special medical history that affects metabolism, and generates historical record tags; historical record tags refer to the health status classification results obtained based on past medical history; If the historical records show a history of chronic kidney disease, the platform will mark it as a special medical history affecting metabolism because this medical history can affect the excretion of water-soluble toxins; if the historical records show a history of gastrointestinal sensitivity, the platform will mark it as a medical history that makes users more susceptible to abnormal microbial indicators; if no such medical history is recorded, it will be marked as no relevant medical history. When a user has multiple medical histories, the platform does not average the medical histories, but retains the classification results that have a more stringent impact on the risk threshold, so as to calculate the health status correction coefficient later; the cloud management platform extracts the average daily water consumption from the real-time water consumption data uploaded by the smart water purification terminal, and calculates the water consumption ratio in combination with the weight parameter; the water consumption ratio refers to the proportional relationship between the user's average daily water consumption and physical parameters such as weight within the statistical period, and is used to determine whether the user's drinking behavior is too low, normal or too high; When the platform obtains multiple single water consumption records within a statistical period, it first sums up the single water consumption within that period, then converts it into an average daily water consumption, and finally divides it by the user's weight to obtain the water consumption ratio. Based on this water consumption ratio, the platform classifies users into a first water consumption ratio range type (below a preset ratio threshold), a second water consumption ratio range type (within a normal ratio threshold), or a third water consumption ratio range type (above a preset ratio threshold), and generates water consumption behavior tags. The preset ratio threshold is 0.02 L / kg·d for the first threshold and 0.05 L / kg·d for the second threshold. The preset normal water consumption range is 0.02 L / kg·d to 0.05 L / kg·d. The preset age range is set to include the normal range of 18 to 60 years old. The preset sample size threshold is 10,000 cases. The safety limit and warning limit are set to 0.5 times and 0.8 times the individualized management limit, respectively. All key parameters are calibrated based on WHO drinking water quality guidelines and historical clinical measurement statistics. Among them, water consumption ratios below the preset first threshold are defined as being below the normal ratio range, water consumption between the preset first threshold and the second threshold is defined as being in the normal ratio range, and water consumption above the preset second threshold is defined as being above the normal ratio range; low-frequency water consumption corresponds to users whose average daily water consumption is below the recommended value, regular water consumption corresponds to users whose average daily water consumption is within the normal recommended range, and water consumption above the upper limit of the preset range corresponds to users whose average daily water consumption is above the normal recommended range. The cloud management platform extracts historical water quality fluctuation data from the pipeline water quality monitoring data, and performs risk assessment and assigns values ​​based on this historical water quality fluctuation data to generate water quality status labels. The water quality status label refers to the environmental risk classification result obtained based on the fluctuation of disinfectant residue, heavy metal ions, turbidity and microbial indicators in the pipeline network of the user's area over a historical period. When the regional pipeline water quality monitoring system provides continuous or time-segmented historical monitoring records, the platform counts whether each water quality indicator exceeds or approaches the basic limit, and assigns risk assessment values ​​based on the number of fluctuations and the amplitude of fluctuations. The specific assignment rules are as follows: multiply the number of fluctuations within the historical period by the first preset weight, and add the maximum fluctuation amplitude multiplied by the second preset weight to obtain the risk assessment score; when the assessment value is lower than the preset risk assessment threshold, it is marked as stable water quality, and when the score is greater than or equal to the preset risk assessment threshold, it is marked as water quality fluctuating significantly; this label allows the objective water quality environment of the user's area to enter the subsequent risk assessment process; the cloud management platform summarizes the user attribute label, historical record label, water use behavior label, and water quality status label according to the user number to form a user risk profile; In this invention, a user risk profile is a structured risk record of a user in a cloud management platform. It includes at least user attribute tags, historical record tags, water use behavior tags, water quality status tags, and corresponding device numbers and area information. The formation of the user risk profile is not the final assessment conclusion, but rather the input data for subsequent risk prediction models, threshold adjustments, and exposure comparisons. In terms of training and using risk prediction models, the cloud management platform or the computing server connected to it obtains water consumption and metabolism data of a sample population whose number reaches a preset sample number threshold from the basic database as a basic reference dataset, and obtains water consumption profile data of the target population whose number is lower than the preset sample number threshold as the target population dataset. The preferred basic reference dataset is a dataset of drinking water exposure and health feedback from healthy adults, which includes drinking water volume, pollutant exposure, health feedback or metabolic-related data. The platform also acquires water usage profile data of target populations with a number of people below the first preset threshold as the target population dataset. Target populations with a number of people below the first preset threshold include infants aged 0 to 3 years, the elderly over 65 years old, or patients with chronic kidney disease, etc., with a small sample size. In this embodiment, the equipment parameter extraction model is a computational model used in the cloud management platform to extract multi-dimensional data correlation features from the basic reference dataset and the target population dataset to form equipment operating environment parameters. Specifically, the cloud management platform first obtains the original records related to the operation of the smart water purification terminal and water quality exposure, such as water consumption, pollutant concentration, TDS difference between raw water and purified water, water quality fluctuations, and metabolic data related to the corresponding user group, as input to the model. Once the basic reference dataset and the target population dataset have completed field mapping, the equipment parameter extraction model reads the aforementioned original records from both datasets as water pollutant exposure features. After internal multi-layer neural network mapping, it outputs a high-dimensional implicit vector used to measure the water quality exposure pattern, namely the equipment operating environment parameters. The extracted equipment operating environment parameters serve as key feature representations for further training of subsequent model network structures. To address the issue of unstable model training due to limited sample size in specific population groups, this invention employs a method of co-training with a basic reference dataset and a target population dataset. The cloud management platform calculates a sample compensation coefficient based on the sample size of the target population dataset and a preset sample size threshold, combined with a population matching coefficient. The sample compensation coefficient is a coefficient used to adjust the degree of influence of the difference between the basic reference data and the target population data in the model training error. The smaller the target population sample, the larger the coefficient, and the more the model needs to utilize the transferable water and metabolism patterns in the basic reference data during training; once the target population sample reaches the threshold, the coefficient can be set to a preset fixed value to reduce the dependence on the basic reference data. Specifically, the cloud management platform first counts the number of samples in the target population dataset, which is recorded as the actual sample size of the target domain; the platform compares this sample size with a preset sample size threshold; the sample size threshold can be set according to the basic sample size required for model training, i.e., the basic sample size threshold. If the number of samples in the target population dataset is less than the threshold, the platform calculates the difference between the threshold and the number of samples, and divides the difference by the sample number threshold to obtain the sample missing ratio. The platform multiplies the sample missing ratio by the population matching coefficient to obtain the sample compensation coefficient. The population matching coefficient is used to represent the similarity in physiological tolerance between the target population with a number below the first preset threshold and healthy adults. The population matching coefficient between the elderly and adults can be set to 0.6, and the population matching coefficient between infants and adults can be set to 0.8. This value is derived from the setting of empirical values ​​for the differences in physiological pharmacokinetics among different populations. If the number of samples in the target population dataset is greater than or equal to the sample number threshold, the cloud management platform will set the sample compensation coefficient to a preset fixed value, preferably 0, so that the training mainly relies on the target population's own samples. During training, the cloud management platform constructs a training error calculation formula using the prediction error value of the target population dataset and the sample difference value based on the sample compensation coefficient, and calculates the total training error of the device parameter extraction model. : In this calculation formula, the risk prediction error Using real risk markers of the target population Compared with the predicted risk results output by the model The cross-entropy loss between them is calculated; at the same time, device operating environment parameters extracted from the basic reference dataset are used. and device operating environment parameters extracted from the target population dataset Calculate the maximum mean difference between the two groups on the feature distribution. And compare this difference with the sample compensation coefficient. Multiplying them yields the final sample difference value; when Once the calculations have been completed according to the number of samples, the computing server uses the above training error as the basis for updating the model weights, and repeatedly performs backpropagation to adjust the device parameters and extract the network parameters of the model until the training error converges and meets the preset training requirements. At this point, the overall model with solidified parameters is the risk prediction model. In this invention, the risk prediction model is a model trained by the above-mentioned training error calculation formula. Its input is the user risk profile and the aforementioned original records corresponding to the smart water purifier terminal. The model first extracts the corresponding equipment operating environment parameters and outputs the risk evaluation index of the smart water purifier terminal based on these parameters. This risk assessment indicator can reflect the degree of impact of equipment operation status on drinking water health risks under the current population classification, health records, water use behavior and water quality environment; its output results serve as one of the bases for subsequent management limit calculation, risk accumulation comparison and equipment classification early warning. Through the above process, the intelligent water purification terminal provides real water consumption and equipment water quality data, the regional pipe network water quality monitoring system provides real water quality indicators, the resident health record database provides real user physiological and medical history information, and the cloud management platform converts these objective data into labels, sample compensation coefficients and risk assessment indicators. This process moves beyond simply using uniform limits for drinking water health risk assessment. Instead, it incorporates user age, medical history, water usage habits, and regional water quality fluctuations into the same calculation process, thus providing a data foundation for subsequent calculations of individualized management limits.

[0020] In a preferred embodiment of the present invention, a threshold adjustment coefficient for the user is obtained by weighted calculation based on multiple classification labels. The threshold adjustment coefficient is then multiplied by a preset basic water quality limit for the equipment to determine the user's management limit, including: Extract user attribute tags, historical record tags, and water use behavior tags from the user risk profile; calculate the age adjustment coefficient based on the user attribute tags, the health status correction coefficient based on the historical record tags, and the water use behavior compensation coefficient based on the water use behavior tags. The system retrieves the basic water quality limits for equipment from the management database, and obtains the first weighting factor of the preset age adjustment coefficient, the second weighting factor of the preset health status correction coefficient, and the third weighting factor of the preset water use behavior compensation coefficient. The basic water quality limits for equipment preset in the management database are strictly anchored to the national standard GB5749-2022 for drinking water. The TDS sensor on the smart water purification terminal is configured with an original monitoring accuracy of ±10ppm. The terminal hardware sends real-time water use data to the cloud at a communication upload frequency of minutes and is equipped with a local flash memory temporary storage anti-loss mechanism that is automatically activated in the event of network interruption. Based on the mapping relationship between the age adjustment coefficient, health status correction coefficient, and water use behavior compensation coefficient and their corresponding weighting factors, the threshold adjustment coefficient for the user is determined. Based on the mapping relationship between the basic water quality limits for equipment and the threshold adjustment coefficient, the individualized management limits for the corresponding user are determined.

[0021] This invention is implemented based on the aforementioned user risk profile that has already been generated; the management limit here refers to the individualized water quality risk assessment limit calculated for the user, the target water quality indicator, and the current smart water purification terminal management scenario; it is the user-specific limit calculated based on the equipment's basic water quality limit, combined with the user's age, medical history, and water usage behavior. When the cloud management platform has obtained the user's risk profile and the management database has saved the equipment water quality baseline limit for a certain water quality indicator, the platform initiates the threshold dynamic adjustment process; the cloud management platform extracts the user's corresponding user attribute tags, historical record tags, and water use behavior tags from the user's risk profile; the user attribute tags are used to calculate the age adjustment coefficient, the historical record tags are used to calculate the health status correction coefficient, and the water use behavior tags are used to calculate the water use behavior compensation coefficient. The above three coefficients are all dimensionless coefficients, and their function is to convert the physiological differences, health differences and drinking habits differences of different users into calculable threshold adjustment amounts; the age adjustment coefficient refers to the water quality tolerance correction value corresponding to different user ages; if the user's age is in the normal range, that is, 18 to 60 years old, the cloud management platform sets the age adjustment coefficient as the first preset benchmark value, which is preferably 1.0. If the user is in the advanced age range, i.e., 60 years or older, the age adjustment factor is calculated using the following formula: in, This is a dimensionless age adjustment factor; The first preset reference value is dimensionless and preferably 1.0; The preset decreasing coefficient is in units of The preferred value is 0.01; The user's actual age, in units of ; The starting age for advanced maternal age is expressed in units of... In this embodiment, the value is 60; Similarly, if the user's age falls within the transition range, i.e., 4 to 17 years old, the calculation is performed using the following formula: in, The second preset reference value is dimensionless and preferably 0.4; This is a preset increment coefficient, in units of The preferred value is 0.04; The highest age limit is defined as , in units of In this embodiment, it is 3; The health status correction coefficient refers to the health tolerance correction value corresponding to different past medical history of users. When the cloud management platform reads the historical file tag, the platform determines whether there is a special medical history that affects metabolism in the tag. If there is a history of chronic kidney disease, because it affects the excretion of water-soluble toxins, the platform sets the health status correction coefficient to the third preset benchmark value, which is preferably 0.5. If a user has a history of gastrointestinal sensitivity, the platform will set the health status correction coefficient to the fourth preset benchmark value, preferably 0.7, because the user is more susceptible to abnormal microbial indicators. If the user has no relevant medical history, the platform will set the health status correction coefficient to the first preset benchmark value of 1.0. When a user has multiple medical histories, such as chronic kidney disease and gastrointestinal sensitivity, the platform will select the lowest coefficient as the health status correction coefficient. For example, when both of the above-mentioned medical histories exist simultaneously, the health status correction coefficient is 0.5 instead of the average value; using a lower coefficient corresponds to stricter management limits, which can avoid reducing the intensity of risk control due to averaging for users with multiple medical histories; the water use behavior compensation coefficient refers to the correction value of drinking habits corresponding to different daily water consumption; when the cloud management platform has calculated the daily water consumption based on multiple water consumption records uploaded by the smart water purification terminal, the platform compares the daily water consumption with the preset lower limit of water consumption and the normal water consumption range; If a user's average daily water consumption is lower than the preset lower limit, it indicates that there is a risk of insufficient drinking water or relative concentration of pollutants in the body. The platform will set the water consumption behavior compensation coefficient to the fifth preset benchmark value, which is preferably 0.9. If the average daily water consumption is within the preset normal water consumption threshold range, the platform will set the water consumption behavior compensation coefficient to the first preset benchmark value of 1.0. If the average daily water consumption exceeds the upper limit of the normal water consumption range, the platform will set the water consumption behavior compensation coefficient to the sixth preset benchmark value, which is preferably 1.1. The normal water consumption range can be determined based on the user's weight and the recommended daily water intake. The platform uses the aforementioned water consumption ratio in the calculation to make the judgment result correspond to the user's vital signs information. The cloud management platform obtains the basic water quality limits for the target water quality indicators from the management database; these basic water quality limits correspond to the preset water quality control limits for the intelligent water purification terminal in the management database. Once the basic water quality limits for the equipment have been obtained, the platform reads the first weighting factor of the age adjustment coefficient, the second weighting factor of the health status correction coefficient, and the third weighting factor of the water use behavior compensation coefficient; the first weighting factor, the second weighting factor, and the third weighting factor refer to the proportion of age factors, health factors, and water use behavior factors in the threshold adjustment coefficient; The platform first sets initial values ​​for the three factors and ensures that their sum is one. The initial values ​​can be set using the analytic hierarchy process, for example, by the management database storing a set of weights that sum to one during the system initialization phase. The cloud management platform obtains historical maintenance records or disease statistics provided by the regional management department. Historical maintenance records can be records of maintenance, filter replacement, or shutdown of smart water purifier terminals due to abnormal water quality; disease statistics can be statistics on gastroenteritis and other diseases related to drinking water quality in the region; after the platform obtains historical maintenance records or disease statistics, the platform will calculate the proportion of the target population whose number of maintenance records caused by abnormal water quality is lower than the first preset threshold; the proportion of the target population whose number is lower than the first preset threshold refers to the proportion of a certain group of people in the maintenance records or disease statistics records caused by abnormal water quality. For example, if the proportion of elderly users in maintenance records caused by water quality abnormalities in a certain area exceeds the preset abnormal fluctuation ratio threshold, the proportion of age factors corresponding to the elderly will increase; the platform will compare the proportion of the target population whose number is lower than the first preset threshold with the preset proportion threshold; if the proportion of the target population whose number is lower than the first preset threshold is higher than the preset proportion threshold, the platform will add the preset weight adjustment value to the weight factor corresponding to the target population whose number is lower than the first preset threshold, while reducing other weight factors so that the sum of the three weight factors is still one; When reducing other weighting factors, the cloud management platform adjusts them according to the existing weighting ratio: the specific weighting adjustment calculation formula is as follows: in, The dimensionless reduction required for the second weighting factor of the health status correction coefficient; This is a preset dimensionless weight adjustment value; This is the current second weighting factor; This is the current third weighting factor; This proportional weight allocation calculation and update mechanism prevents anomalies such as excessive decay or even negative values ​​of a certain weight in multi-cycle management iterations, ensuring the stability and consistency of the weighted evaluation model of the entire management limit. For example, if the proportion of elderly patients with gastroenteritis caused by water quality abnormalities reaches or exceeds the preset proportion threshold, the platform increases the first weight factor, so that the age adjustment coefficient plays a greater role in the threshold adjustment. If the proportion of the target population with a quantity lower than the first preset threshold is lower than or equal to the preset proportion threshold, the platform keeps the initial value unchanged and outputs the corresponding weight factor; after obtaining the three coefficients and three weight factors, the cloud management platform calculates the user's threshold adjustment coefficient using the following formula: in, The final threshold adjustment coefficient is calculated and is dimensionless. These are the first, second, and third weighting factors, respectively, which are dimensionless and whose sum equals 1; Age adjustment factor; A health status correction factor; These are the water use behavior compensation coefficients; all three coefficients mentioned above are dimensionless constants. The prerequisite for this calculation process is that the age adjustment coefficient, health status correction coefficient, water use behavior compensation coefficient, and three weighting factors have all been obtained. After the calculation is completed, the platform obtains a dimensionless threshold adjustment coefficient, which is usually less than, equal to, or slightly greater than 1, and is used to represent the degree of management required for users relative to ordinary adults. The cloud management platform calculates the user's management limit using the following formula: in, Individualized management limits; Basic limits for water quality for equipment; dimensions and Maintain consistency; For example, if a user has a history of chronic kidney disease and a health status correction factor of 0.5, and age or water use behavior factors also cause the threshold adjustment factor to decrease, then their management limit will be lower than the unified basic limit. In this way, under the same water quality index concentration, the user will enter the attention or warning level earlier. The platform will use this management limit as the preliminary assessment result and write it into the user's risk assessment record for subsequent comparison of water quality risk accumulation. In this invention, age, medical history, and water use behavior are considered as three factors in a fixed calculation order when calculating the management limit. The user risk profile generated in the previous step provides tags, which are first converted into three coefficients. These three coefficients are then weighted by a weighting factor to obtain a threshold adjustment coefficient, which is finally multiplied by the device's basic water quality limit to obtain the management limit. Through this sequence, the cloud management platform can transform user differences into water quality management boundaries that can be executed by the smart water purification terminal, avoiding the use of the same water quality judgment limit for different groups of people. Set initial values ​​for the first, second, and third weighting factors, with the sum of each factor being 1; obtain historical maintenance records or disease statistics provided by the regional management department; and dynamically update the first, second, and third weighting factors based on the comparison between the proportion of the target population whose maintenance records are caused by water quality abnormalities and the preset proportion threshold.

[0022] In a preferred embodiment of the present invention, the cumulative water quality risk is calculated by combining real-time water usage data, and the cumulative water quality risk is compared with the management limit to obtain a comprehensive risk assessment result. This includes: extracting the single water usage based on real-time water usage data, and determining the single pollution exposure amount by combining the target water quality index concentration in the pipeline water quality monitoring data; extracting the vital signs parameters in the user's risk profile, and determining the user's cumulative water quality risk based on multiple single pollution exposures and vital signs parameters within the statistical period; the target water quality index is preferably the residual concentration of disinfectant or the concentration of lead ions, and the concentration data is the real-time instantaneous monitoring value obtained by the water quality sensing device. The underlying mathematical logic of the single pollution exposure amount is the product of the actual water usage volume and the instantaneous concentration of the target water quality index. In the final risk accumulation stage, the total amount of all single exposures within the statistical period is obtained by arithmetic accumulation, and then divided by the user's kilogram weight parameter for conversion. The final unit is uniformly defined as mg / kg. The TDS sensor and flow metering module equipped in the intelligent water purification terminal have a factory calibration error allowable range of ±5% and a calibration period of 6 months. The risk prediction model adopts a multi-layer neural network architecture, which includes an input layer, at least two hidden layers, and an output layer. The nodes in the input layer correspond one-to-one with the dimensions of historical water quality multidimensional features. The hidden layers use the ReLU activation function to avoid gradient vanishing. The output layer outputs the predicted continuous value of pollution exposure.

[0023] By using a risk prediction model pre-trained based on historical water quality and health correlation data, the user's pollution exposure is extracted, and then the user's cumulative water quality risk is determined. The system compares the accumulated water quality risk with the individualized management limit, determines the user's equipment operation and maintenance risk level based on the degree of difference in the comparison, and generates a comprehensive risk assessment result. Before entering the comprehensive judgment and comparison step, the system's computing platform will introduce a time decay factor based on a specific sliding window to dynamically couple and correct the lag effect of historical water quality fluctuations accumulated in the early stage, so as to ensure that the water quality concentration value collected at the moment is now collected and the historical water quality status label achieve a strict normalized mapping alignment on the time representation scale.

[0024] Based on the comprehensive risk assessment results, corresponding equipment classification and early warning strategies are triggered, including: based on the comparison results of the cumulative water quality risk with the preset safety limit and early warning limit, the user's equipment operation and maintenance risk level is divided into a safe level, a concern level, or a warning level; according to the determined risk level, strategies such as recording equipment operation data, issuing equipment maintenance instructions to the smart water purification terminal, or issuing shutdown maintenance instructions and generating a user drinking water assessment report are executed respectively.

[0025] This invention is implemented after the management limits have been calculated; the cumulative water quality risk here refers to the unit vital signs parameter exposure of the user within a specified statistical period due to the actual drinking water volume and water quality index concentration; its calculation object comes from the single water consumption of the smart water purification terminal and the water quality index concentration in the regional pipeline monitoring data or equipment end water quality detection data, which belongs to the calculation and processing of objective water volume, pollutant concentration and user vital signs parameter; When a user uses water, the smart water purification terminal records the amount of water used and uploads the device number, user number, time of use, and amount of water used to the cloud management platform. The cloud management platform then obtains the water quality index concentration corresponding to the time of water use from the regional pipeline network monitoring data based on the time of use and the user's location. If the smart water purification terminal provides equipment-side water quality data such as the difference between raw water and purified water TDS, the platform can use it as reference data for the current operating status of the equipment to confirm changes in filter cartridge efficiency. That is, the actual retention rate of the filter cartridge can be checked by the ratio difference between raw water TDS and purified water TDS to help determine the degree of internal risk leakage degradation. For a specific target water quality indicator, the platform obtains the concentration of the water quality indicator and the amount of water used in a single water use, and then proceeds to calculate the single pollution exposure. The single pollution exposure refers to the exposure value obtained by multiplying the concentration of the water quality indicator by the actual amount of water used in a single water use. The cloud management platform extracts water quality index concentrations from pipeline monitoring data or water quality sensor data for the corresponding time period, and combines this with the single water consumption from the smart water purification terminal to calculate the single pollution exposure amount. The calculation formula is as follows: in, This refers to the total amount of pollution exposure in a single instance. For water quality index concentration, This refers to the water consumption per transaction measured at the equipment end. Because the retention efficiency of water purification filter cartridges will decrease as the total load of pollutants is continuously intercepted and accumulated, once the high load index in the raw water is continuously pushed into the filter cartridge with each water use, the pollutants accumulated inside will eventually lead to the risk of releasing pollutants into the user's terminal drinking water. Therefore, using this potential load input for subsequent conversion of vital signs parameters can establish a mapping relationship between the hardware tolerance boundary of the equipment and the physiological tolerance limit of the individual user; within the specified statistical period, the user may use water multiple times; the cloud management platform generates a single pollution exposure record after each water use and saves it according to the user number and statistical period. When the statistical period ends or the platform needs to conduct real-time assessment, the platform extracts all single pollution exposures of users within that statistical period, as well as vital signs from user attribute data, preferably weight, to calculate the cumulative water quality risk, using the following formula: in, This represents the cumulative amount of water quality risk within the statistical period. To specify the total number of water usages within a given statistical period; For the first Concentration of water quality indicators during the first use of water; For the first Water consumption per single use; These are the user's vital signs parameters, in kg. The safety limit is the first product ratio of the management limit, preferably 0.5 × the management limit; the warning limit is the second product ratio of the management limit, preferably 0.8 × the management limit; the first product ratio and the second product ratio correspond to the boundaries of the safety level and the concern level, respectively; when the cumulative amount of water quality risk is less than or equal to the safety limit, the cloud management platform determines that the user is at the safety level; At this time, the platform does not send alarm commands to the smart water purification terminal, nor does it trigger changes in the device indicator lights. It only records the device operation data and the results of this assessment on the cloud management platform. The record includes water consumption, water quality index concentration, water quality risk accumulation, management limits, and judgment time, which are used for subsequent historical statistics and weight updates. When the accumulated water quality risk exceeds the safety limit but is less than or equal to the warning limit, the cloud management platform determines that the user is at the attention level. At this time, the platform generates a maintenance command and sends it to the device's local controller through the data channel connected to the smart water purifier terminal. After receiving the maintenance command, the device's local controller drives the attention indicator light on the device panel to light up, preferably a yellow indicator light, to remind the user that the filter cartridge efficiency may be declining. This level does not directly require shutdown, but will give a clear prompt on the device so that the user can pay attention to the filter cartridge status or arrange maintenance in a timely manner. When the accumulated water quality risk exceeds the warning limit, the cloud management platform determines that the user is at the warning level. At this time, the platform generates a shutdown maintenance command and sends it to the smart water purifier terminal. After receiving the shutdown maintenance command, the local controller of the smart water purifier terminal executes the relevant control for equipment shutdown maintenance and can light up the red indicator light. At the same time, the cloud management platform pushes suggestions for water cut-off and consumable replacement to the user through the mobile application that communicates with the smart water purifier terminal. If the smart water purification terminal is equipped with a filter management function, the maintenance strategy issued by the platform also includes consumable replacement instructions to prompt or trigger the filter replacement process; when the warning level is determined, the cloud management platform calls the text generation template to generate a user drinking water assessment report based on the parameters that triggered the warning; the text generation template does not simply output the conclusion of exceeding the standard, but reads the key parameters that triggered the warning, including the target water quality index, the current water quality risk accumulation, the management limit, the user's age adjustment coefficient, the health status correction coefficient, the water use behavior compensation coefficient, and the label that caused the threshold to be lowered; The platform fills these parameters into a preset text template and generates a drinking water assessment report that includes explanations of the reasons and equipment maintenance recommendations. For example, for users with a history of chronic kidney disease, the report can state that their health status correction factor is low, therefore the management limit for trace heavy metals in the current water quality is lower than that for regular users, and suggest immediately replacing the reverse osmosis filter or suspending drinking water from direct drinking water sources. After the report is generated, the cloud management platform pushes it to the mobile application and saves it together with the results of this comprehensive risk assessment. In a typical household scenario, a user aged 65 or older takes water through a smart water purifier terminal. The device records the amount of water taken each time and uploads it to a cloud management platform. The platform obtains the concentration of heavy metal ions for the corresponding time period based on the pipeline monitoring data of the user's area, and then calculates the cumulative amount of water quality risk within the statistical period based on the user's weight. Since the user is in the advanced age range, the age adjustment coefficient in the aforementioned management limit calculation is less than 1.0. If the user also has a history of chronic kidney disease, the health status correction factor is 0.5, and the final management limit is further reduced; when the current water quality risk accumulation exceeds 0.8 times the management limit, the platform determines it to be a warning level, issues a shutdown maintenance instruction to the smart water purification terminal, and pushes water cut-off and consumable replacement suggestions through the mobile application, while generating a drinking water assessment report explaining that age and medical history have led to the reduction of the management limit. To verify the effectiveness of this assessment method, a system backtesting test set was constructed, comprising 10,000 individuals from the general healthy population and 1,500 individuals from special populations. The test results show that, after adopting a transfer learning architecture with adaptive sample compensation coefficients, the recall rate and F1 score for health risk warnings for special populations remained stable at over 90%. Furthermore, in a parallel comparative experiment with the traditional static assessment strategy using a single fixed national standard limit, the individualized management limit dynamic engine of this invention reduced the false alarm rate of invalid water outages at terminal devices by 35.6% while ensuring the same water quality safety baseline. The end-to-end average latency from system identification of abnormal water quality parameters to the issuance of warning commands at all levels to the terminal was also reduced to within 3 minutes.

[0026] Through this invention, the water intake, water quality index concentration, and user vital signs parameters of the intelligent water purification terminal are used to calculate the actual exposure amount, and individualized management limits are used to determine the graded early warning boundaries. The final output results can directly control the device indicator lights, shutdown maintenance actions, and mobile terminal prompts. This method transforms drinking water health assessment results into equipment maintenance actions for the intelligent water purification terminal, which can reduce the situation of continuing to supply water after the filter cartridge efficiency declines, and provide users with drinking water risk information corresponding to their age, medical history, and drinking behavior.

[0027] Scene 1: Scenario 1 involves a 65-year-old male user weighing 70kg with chronic kidney disease. In this scenario, the system operates by periodically uploading real-time water usage data from the smart terminal, while the network sensors synchronize heavy metal concentrations in real time. Upon receiving information from the cloud regarding the user's age (over 60 years) and any specific medical history affecting metabolism, the system automatically lowers the safety threshold. Data feeder 1 is for terminal water consumption data collection, requiring an online device rate of 0.999 to ensure accurate baseline data for accumulated water quality risks. Data feeder 2 is for network water quality monitoring, requiring a sensor drift rate of less than 0.05 every 6 months to ensure accurate calculation of single pollution exposure levels. Data feeder 3 is for accessing health records, requiring an interface response latency of less than 1 second to ensure accurate threshold adjustment coefficients matching the user's vital signs.

[0028] The calculations based on actual data are as follows. The user's actual age is 65 years old, with the advanced age threshold at 60 years old. The age adjustment coefficient is 1.0 minus 0.01 multiplied by 5, resulting in 0.95. The user suffers from chronic kidney disease, and the health status correction coefficient is 0.5. The average daily water consumption is 2.0L, the weight is 70kg, and the water consumption ratio is 0.028L / kg, falling within the normal range of 0.02L / kg to 0.05L / kg. The water consumption behavior compensation coefficient is 1.0. Assuming the first weighting factor is 0.3, the second weighting factor is 0.4, and the third weighting factor is 0.3, the threshold adjustment coefficient is 0.3 multiplied by 0.95 plus 0.4 multiplied by 0.5 plus 0.3 multiplied by 1.0, totaling 0.785. The basic water quality limit for the equipment is set at 0.0005mg / kg, and the individualized management limit is 0.0005 multiplied by 0.785, resulting in 0.0003925mg / kg. If the heavy metal concentration in the water is 0.015 mg / L, the water consumption is 2.0L, and the single exposure dose divided by body weight (70 kg) results in a cumulative water quality risk of 0.000428 mg / kg. Since 0.000428 is greater than 0.0003925 (ratio approximately 1.09), exceeding the warning limit by 0.8, the system triggers a warning level, issues a shutdown maintenance command, and generates an assessment report.

[0029] Scene 2: Scenario 2 involves a 3-year-old infant weighing 15kg with a history of gastrointestinal sensitivity. Boiled water is used to prepare formula. The system identifies a group with extremely low tolerance and then initiates a transfer learning and sample compensation mechanism tailored to that specific physiological tolerance group. Abundant data from healthy adults is used for target population sample compensation, adaptively adjusting transfer weights to reduce prediction bias caused by small sample sizes. In addition to the same 0.999 online rate requirement as Scenario 1, the data feeder reliability assessment requires the addition of criteria for indirect exposure scenarios beyond direct drinking water, with a false alarm rate below 0.01%.

[0030] The calculations based on actual data are as follows: Actual age 3 years, age adjustment factor is 0.4 + 0.04 multiplied by 0, resulting in 0.4. There is a history of susceptibility to microbial influences, health status correction factor is 0.7. Average daily water consumption is 0.4L, weight is 15kg, water consumption ratio is 0.027L / kg, within the normal range, behavioral compensation factor is 1.0. Threshold adjustment factor is 0.3 multiplied by 0.4 + 0.4 multiplied by 0.7 + 0.3 multiplied by 1.0, summed to 0.7. Individualized management limit is 0.0005 multiplied by 0.7, resulting in 0.00035mg / kg. If the residual disinfectant concentration in the pipeline is 0.01mg / L, water consumption is 0.4L, exposure divided by 15kg, cumulative water quality risk is 0.000267mg / kg. The ratio 0.000266 divided by 0.00035 is approximately 0.76, which is between the safety limit of 0.5 and the warning limit of 0.8, triggering the attention level and illuminating a yellow indicator light to indicate that the filter's efficiency may be declining.

[0031] The comparison table is shown below: The table provides an explanation. Even though the absolute value of the accumulated water quality risk in Scenario 2 is relatively small, the extremely low age coefficient of infants and young children significantly compresses the overall management limit, causing the final risk to still reach the concern level. In Scenario 1, the health coefficient is only 0.5 due to chronic diseases, which greatly lowers the individualized management limit, causing it to quickly exceed the safety threshold and reach the warning level under normal water quality fluctuations. The data comparison confirms that this solution overcomes the limitations of fixed standards and achieves adaptive and precise control of drinking water safety for different population groups.

[0032] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for assessing the health risks of drinking water for residents based on big data analysis, characterized in that, The method includes the following steps: The system acquires real-time water usage data from smart water purification terminals, pipeline water quality monitoring data, and user attribute data including vital signs such as weight; it extracts the real-time water usage data, pipeline water quality monitoring data, and user attribute data, and aggregates them to generate a user risk profile with tag classification; based on the user risk profile, it determines the individualized management limit for the corresponding user; it calculates the cumulative water quality risk by combining the real-time water usage data, and compares the cumulative water quality risk with the individualized management limit to obtain a comprehensive risk assessment result; and it triggers the corresponding equipment-level early warning strategy based on the comprehensive risk assessment result.

2. The method for assessing the health risks of drinking water based on big data analysis according to claim 1, characterized in that, The step of generating a user risk profile that includes tag classification specifically involves generating and summarizing user attribute tags, historical record tags, water use behavior tags, and water quality status tags. The generation process includes: Age and gender information are extracted from the user attribute data, and risk tolerance levels are divided according to age and gender information to generate user attribute tags; Historical archive information is extracted from the user attribute data, past medical history records are parsed, it is determined whether there are any special medical histories that affect metabolism, and historical archive tags are generated based on the determination results. The average daily water consumption is extracted from the real-time water consumption data, the weight parameter is extracted from the vital signs parameters, and the water consumption ratio is calculated in combination with the weight parameter. Based on the water consumption ratio, the water consumption behavior is divided into a first water consumption ratio interval type below a preset ratio threshold, a second water consumption ratio interval type within a normal ratio threshold, or a third water consumption ratio interval type above a preset ratio threshold, and a water consumption behavior label is generated. Historical water quality fluctuation data are extracted from the pipeline water quality monitoring data, and risk assessment values ​​are assigned based on the historical water quality fluctuation data to generate water quality status labels. The user risk profile is generated by summarizing the user attribute tags, historical record tags, water use behavior tags, and water quality status tags.

3. The method for assessing the health risks of drinking water based on big data analysis according to claim 1, characterized in that, The method further includes using a risk prediction model to extract the user's equipment operating environment parameters to obtain risk assessment indicators for the smart water purification terminal, including the following steps: Water consumption and metabolic data of a sample population whose number reaches a preset sample size threshold are obtained from the basic database as a basic reference dataset, and water consumption profile data of a target population whose number is lower than the preset sample size threshold are obtained as a target population dataset; wherein, the target population dataset contains known real risk markers; A device parameter extraction model is constructed. The device parameter extraction model adopts a multi-layer neural network architecture. Its input layer nodes correspond to multi-dimensional feature dimensions such as real-time water use and pipeline water quality. It is used to extract the device operating environment parameters in the basic reference dataset and the target population dataset. The population matching coefficient is determined by combining the physiological tolerance differences between the target population and the sample population, and the sample missing ratio of the target population dataset is calculated based on the sample number of the target population dataset and the preset sample number threshold, thereby determining the sample compensation coefficient; the device parameter extraction model is jointly trained using the target population dataset containing known real risk markers and the basic reference dataset based on the sample compensation coefficient to obtain the risk prediction model. The risk assessment indicators of the intelligent water purification terminal are output through the risk prediction model.

4. The method for assessing the health risks of drinking water based on big data analysis according to claim 3, characterized in that, The calculation of the sample compensation coefficient includes the following steps: Obtain the number of samples in the target population dataset and compare the number of samples with a preset sample number threshold; determine the sample compensation coefficient based on the comparison result, the sample missing ratio of the target population dataset, and the population matching coefficient.

5. The method for assessing the health risks of drinking water based on big data analysis according to claim 2, characterized in that, The individualized management limits for each user are determined by weighting the various category tags in the user's risk profile, including the following steps: Extract the user attribute tags, historical record tags, and water usage behavior tags corresponding to the user from the user risk profile; Calculate the age adjustment coefficient based on the user attribute tags, calculate the health status correction coefficient based on the historical record tags, and calculate the water use behavior compensation coefficient based on the water use behavior tags. The system retrieves the basic water quality limits for equipment from the management database, and obtains a first weighting factor for a preset age adjustment coefficient, a second weighting factor for a preset health status correction coefficient, and a third weighting factor for a preset water use behavior compensation coefficient. Based on the mapping relationship between the age adjustment coefficient, health status correction coefficient, and water use behavior compensation coefficient and their corresponding weighting factors, the system determines the user's threshold adjustment coefficient. Based on the mapping relationship between the basic water quality limits for equipment and the threshold adjustment coefficient, the system determines the individualized management limits for the corresponding user.

6. The method for assessing the health risks of drinking water based on big data analysis according to claim 5, characterized in that, The calculation of an age adjustment coefficient based on the user attribute tags, a health status correction coefficient based on the historical data tags, and a water use behavior compensation coefficient based on the water use behavior tags includes the following steps: An age adjustment coefficient is determined based on the matching relationship between the user's actual age and a preset age range; a health status correction coefficient is determined based on whether the user has any special medical history that affects metabolism or makes them susceptible to microorganisms; and a water use behavior compensation coefficient is determined based on the comparison between the user's average daily water consumption and a preset normal water consumption range.

7. The method for assessing the health risks of drinking water based on big data analysis according to claim 5, characterized in that, Obtaining the first weighting factor of the age adjustment coefficient, the second weighting factor of the health status correction coefficient, and the third weighting factor of the water use behavior compensation coefficient includes the following steps: Set initial values ​​for the first weighting factor, the second weighting factor, and the third weighting factor, and ensure that the sum of each factor is 1; obtain historical maintenance records or disease statistics provided by the regional management department; and dynamically update the first weighting factor, the second weighting factor, and the third weighting factor based on the comparison results between the proportion of the target population whose maintenance records are caused by water quality abnormalities and the preset proportion threshold.

8. The method for assessing the health risks of drinking water based on big data analysis according to claim 1, characterized in that, The cumulative water quality risk is calculated by combining the real-time water usage data, and the cumulative water quality risk is compared with the individualized management limit to obtain a comprehensive risk assessment result, including the following steps: Based on the real-time water usage data, the single water usage is extracted, and combined with the target water quality index concentration in the pipeline water quality monitoring data, the single pollution exposure is determined; the vital signs parameters in the user's risk profile are extracted, and based on the multiple single pollution exposures and vital signs parameters within the statistical period, the user's cumulative water quality risk is determined; The cumulative water quality risk is compared with the individualized management limit, and the user's equipment operation and maintenance risk level is determined based on the degree of difference in the comparison, thus generating the comprehensive risk assessment result.

9. The method for assessing the health risks of drinking water based on big data analysis according to claim 8, characterized in that, Triggering the corresponding equipment-level early warning strategy based on the comprehensive risk assessment results includes the following steps: Based on the comparison results between the accumulated water quality risk and the preset safety and warning limits, the user's equipment operation and maintenance risk level is divided into a safe level, a concern level, or a warning level. According to the determined risk level, strategies are executed to record equipment operation data, issue equipment maintenance instructions to the smart water purification terminal, or issue shutdown maintenance instructions and generate a user drinking water assessment report.

Images