Water quality multi-parameter real-time monitoring method based on intelligent sensor and big data

Abstract

The invention provides a water quality multi-parameter real-time monitoring method based on an intelligent sensor and big data, and relates to the technical field of water quality monitoring, and the water quality multi-parameter real-time monitoring method comprises the following steps: collecting water quality multi-parameter data and environment auxiliary data of each monitoring point in real time; extracting water quality parameter characteristics and environment correlation characteristics; constructing a water quality state recognition model; reversely deducing a pollution diffusion source to obtain regional pollution traceability data; constructing a water quality risk assessment index, and judging a risk level; the monitoring effect is evaluated, and an optimization score coefficient is constructed; and correcting the monitoring system. According to the method, the pollution severity and the influence of the sensitive area are quantified, and the multi-level judgment logic is matched, so that the risk level division is more suitable for an actual scene, and the timeliness and scientificity of pollution coping are effectively improved; through quantitative evaluation of the three core indexes, a system operation short board can be accurately positioned, a targeted correction scheme can be formulated, the monitoring system is continuously optimized, and stability, high efficiency and data quality of long-term operation of the monitoring system are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of water quality monitoring technology, and in particular to a method for real-time monitoring of multiple parameters of water quality based on intelligent sensors and big data. Background Technology

[0002] With the increasing global awareness of water resource protection and the continuous improvement of the ecological environment governance system, water quality monitoring, as a core link in water environment management, pollution prevention and control, and ecological security, is seeing its technological application scenarios and demands continuously expand. The advancement of industrialization and urbanization has led to increasingly complex causes of water pollution, with more types of pollutants and more diverse diffusion pathways, placing higher demands on the real-time performance, multi-parameter coverage, and data integration and analysis capabilities of water quality monitoring.

[0003] Water quality monitoring refers to the systematic measurement and evaluation of physical, chemical, and biological indicators in water bodies through a series of technical means. Its technological foundation involves multiple disciplines such as analytical chemistry, environmental engineering, sensor technology, and automation control. Traditional monitoring mainly relies on manual sampling and laboratory analysis, using analytical methods such as spectrometry, chromatography, and electrochemical methods to detect specific parameters.

[0004] Conventional water quality parameter monitoring methods lack a deep integration of environmental correlation characteristics and water quality parameter characteristics, resulting in insufficient accuracy and reliability in water quality status assessment, making it difficult to accurately reflect the actual water quality situation. Furthermore, they cannot accurately trace pollution sources, increasing the difficulty of pollution control. Risk assessment systems have significant shortcomings, often using a single indicator to measure risk levels. They fail to fully quantify the dual core factors of pollution severity and the impact on sensitive areas, and lack secondary verification based on water quality status types and reference adjustments based on historical pollution recurrence rates. This leads to a disconnect between risk level assessment and actual scenarios, resulting in untargeted early warning information that is difficult to guide efficient pollution control efforts. Summary of the Invention

[0005] This invention provides a real-time multi-parameter water quality monitoring method based on intelligent sensors and big data. This addresses the shortcomings of existing technologies, which lack deep integration of environmental correlation characteristics and water quality parameter characteristics, resulting in insufficient accuracy and reliability in water quality status assessment, making it difficult to accurately reflect actual water quality conditions. Furthermore, the inability to accurately trace pollution sources increases the difficulty of pollution control. The risk assessment system also has significant shortcomings, often relying on a single indicator to measure risk levels. This fails to fully quantify the dual core factors of pollution severity and the impact on sensitive areas, and lacks secondary verification based on water quality status types and reference adjustments based on historical pollution recurrence rates. Consequently, risk level determination is disconnected from actual scenarios, and early warning information lacks specificity, making it difficult to guide efficient pollution control efforts.

[0006] On the one hand, this invention provides a method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data, including: S1: Based on the target parameters of water quality monitoring and the environmental parameters of the monitoring area, collect multi-parameter water quality data and environmental auxiliary data of each monitoring point in real time; the multi-parameter water quality data includes pollutant concentration gradient data. S2: Preprocess water quality multi-parameter data and environmental auxiliary data to extract water quality parameter features and environmental correlation features; S3: Construct a water quality status identification model based on water quality parameter characteristics and environmental correlation characteristics to identify water quality status types; S4: Combine pollutant concentration gradient data to infer the source of pollution diffusion and obtain regional pollution source tracing data; S5: Based on water quality status type and regional pollution source tracing data, construct a water quality risk assessment index, determine the risk level, and generate targeted early warning and disposal instructions; S6: Collect water quality data and monitoring system operation data in real time after early warning and response, calculate monitoring accuracy index, response timeliness index and data integrity index, evaluate the monitoring effect, and construct an optimization score coefficient based on the monitoring effect; S7: Correct the monitoring system based on the optimized score coefficient.

[0007] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, step S3, the step of constructing a water quality status identification model includes: S31: Integrate environmental correlation features with water quality parameter features to construct a multi-dimensional feature vector.

[0008] S32: Based on historical water quality sample data, a random forest algorithm is used to train an initial water quality status identification model.

[0009] S33: Optimize the hyperparameters of the initial water quality status identification model using the grid search method to obtain the final water quality status identification model.

[0010] S34: Set three water quality judgment thresholds, input multi-dimensional feature vectors into the final water quality status recognition model, combine historical data for cross-validation, and output water quality status type.

[0011] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, step S4, the step of obtaining regional pollution source tracing data, includes: S41: Based on the pollutant concentration gradient data at each monitoring point, combined with the water flow velocity and direction data in the monitoring area, a diffusion model is used to fit the pollutant diffusion trajectory and infer the spatial range of the initial source of pollution.

[0012] S42: Locate basic information on pollution sources within this spatial range using GIS.

[0013] S43: Based on the basic information of the pollution source, retrieve the pollution source filing file and obtain the pollution source emission record.

[0014] S44: Search the historical pollution treatment database and extract historical event records that are similar to the basic information of the pollution source.

[0015] S45: Integrate basic information on pollution sources, pollution emission records, and historical event records to form regional pollution source tracing data.

[0016] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, step S5, the step of constructing the water quality risk assessment index, includes: S51: Based on the water quality status, extract the degree of pollutant exceedance and the rate of pollution diffusion, and calculate the pollution severity index.

[0017] S52: Combining regional pollution source tracing data, extract the distance between pollution sources and sensitive areas, the ecological importance weight of sensitive areas, and calculate the impact index of sensitive areas.

[0018] S53: Set the weight coefficients for the pollution severity index and the sensitive area impact index, and construct the water quality risk assessment index through a weighted summation formula.

[0019] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, in step S51, the pollution severity index formula is expressed as:

[0020] Wherein, PSPI is the pollution severity index, λ1 is the weighting coefficient for the degree of pollutant exceedance, λ2 is the weighting coefficient for pollution diffusion rate, i is the pollutant type index, and C i S represents the measured concentration of the i-th critical pollutant. i The water quality standard concentration limit for pollutant of category i. V represents the maximum number of times the pollutant exceeds the standard among all monitored pollutants, V represents the actual diffusion rate of the pollutant, and V0 represents the standard threshold for the diffusion rate of the pollutant.

[0021] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, in step S52, the sensitive area influence index formula is expressed as:

[0022] Wherein, SAII is the Sensitive Area Impact Index, n is the total number of affected sensitive areas within the monitoring area, and W j Let D be the ecological importance weight of the j-th sensitive area, and D0 be the safe distance threshold for the sensitive area. jLet be the straight-line distance between the pollution source and the j-th sensitive area.

[0023] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, step S5, the step of determining the risk level, includes: S54: Preset four-level risk level threshold.

[0024] S55: Based on the risk level threshold and the water quality status type, a secondary verification is performed to adjust the risk level and output the change risk information.

[0025] S56: Based on the historical pollution recurrence rate in the regional pollution source tracing data, further adjust the changed risk information and output the final risk level.

[0026] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, step S6, which calculates the monitoring accuracy index, response timeliness index, and data integrity index, includes: S61: Select standard calibration points within the monitoring range and determine the standard values ​​of key water quality parameters.

[0027] S62: Extract the measured values ​​of multiple water quality parameters at the same time point, calculate the relative error of the multiple parameter data, take the average value of the relative errors of all key parameters as the basic value of monitoring accuracy, and calculate the monitoring accuracy index.

[0028] S63: Define response time as the total time from the moment the intelligent sensor collects abnormal water quality data to the release of early warning information and its push to the relevant handling department.

[0029] S64: Set the standard response time and calculate the response time index.

[0030] S65: The theoretical total amount of data collected during the statistical monitoring period. Invalid data within the monitoring period is removed, and the amount of valid data is calculated.

[0031] S66: Calculate the data integrity index based on the theoretical total amount of collected data and the amount of valid data.

[0032] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, step S6, the step of constructing an optimized score coefficient based on the monitoring effect, includes: S67: Normalize the monitoring accuracy index, response timeliness index, and data integrity index to obtain standard index data.

[0033] S68: Based on standard index data, set index weights according to the priority of the monitoring system.

[0034] S69: Calculate the optimization score coefficient by weighted summation based on the index weights.

[0035] According to the real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided by the present invention, step S7, the step of correcting the monitoring system, includes: S71: If the optimized score coefficient is not up to standard, break down the contribution ratio of each index in the standard index data to pinpoint the root cause.

[0036] S72: Develop corrective measures to address the root causes.

[0037] S73: Based on the correction scheme, determine the correction intensity according to the difference between the optimized score coefficient and the qualified threshold of the corresponding index, and recalculate the optimized score coefficient until the qualified standard is reached.

[0038] This invention provides a real-time multi-parameter water quality monitoring method based on intelligent sensors and big data. By collecting multiple water quality parameters and environmental auxiliary data in real time and performing in-depth feature extraction, combined with a water quality status identification model optimized using random forest algorithms and grid search, the accuracy and reliability of water quality status identification are improved. Through the integrated application of pollutant concentration gradients, water flow characteristics, and GIS technology, combined with multi-dimensional information integration from pollution source registration files and historical event records, precise location and tracing of pollution sources are achieved, providing strong support for pollution control. By constructing a dual index quantifying the severity of pollution and the impact on sensitive areas, coupled with a multi-level judgment logic of four-level risk thresholds, secondary verification of water quality status, and adjustment of historical recurrence rates, the risk level classification is made more aligned with actual scenarios. The generation of targeted early warnings and disposal instructions effectively improves the timeliness and scientific nature of pollution response. Through the quantitative evaluation of three core indices—monitoring accuracy, response timeliness, and data integrity—combined with optimized score coefficients formed by weight allocation, the system's operational shortcomings can be accurately identified and targeted correction plans can be formulated, continuously optimizing the monitoring system and ensuring its long-term stability, efficiency, and data quality. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a real-time water quality multi-parameter monitoring method based on intelligent sensors and big data provided in an embodiment of the present invention; Figure 2This is a flowchart illustrating the calculation of the monitoring accuracy index, response timeliness index, and data integrity index in this embodiment of the invention; Figure 3 This is a flowchart illustrating the construction of optimized score coefficients based on monitoring results in an embodiment of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0042] Example: The following is combined Figures 1-3 This invention describes a real-time water quality multi-parameter monitoring method based on intelligent sensors and big data.

[0043] like Figures 1-3 As shown in the figure, the real-time monitoring method for multiple water quality parameters based on intelligent sensors and big data provided in this embodiment of the invention includes: S1: Determine the target parameters for water quality monitoring and the environmental parameters of the monitoring area, and collect multi-parameter water quality data and environmental auxiliary data of each monitoring point in real time. The water quality monitoring targets include key water quality indicators such as pH value, dissolved oxygen (DO), ammonia nitrogen (NH3-N), chemical oxygen demand (COD), total phosphorus (TP), total nitrogen (TN), and heavy metals. Environmental parameters of the monitoring area include water temperature, air pressure, wind speed, wind direction, rainfall, and topography. Environmental auxiliary data include meteorological data, hydrological data, and geographic information data. Meteorological data (temperature, air pressure, etc.) are acquired using standard-compliant automatic weather stations. Hydrological data (water flow velocity, flow direction, etc.) are measured using an acoustic Doppler current profiler. Geographic information data (topography, etc.) is retrieved from the geographic information public service platform. Data acquisition uses high-precision intelligent sensors, with a collection frequency of once per hour. In sensitive areas, this can be adjusted to once every 15 minutes. Each monitoring point must be evenly distributed in the upstream, midstream, and downstream of the monitoring area, covering key locations such as water sources, sewage outlets, and ecologically sensitive areas, with a number of no less than five.

[0044] S2: Preprocess the multi-parameter water quality data and environmental auxiliary data, and extract water quality parameter features and environmental correlation features by combining big data analysis. Preprocessing steps include: removing outliers using the 3σ criterion and filling missing values ​​using linear interpolation. The Z-score standardization formula is used to unify the data dimensions, expressed as X'=(X-μ) / σ, where X' is the standardized data of the multi-parameter water quality data and environmental auxiliary data, X is the original data of the multi-parameter water quality data and environmental auxiliary data, μ is the mean, and σ is the standard deviation. Wavelet transform algorithm is used to remove sensor noise interference.

[0045] Water quality parameter characteristics include statistical features such as the mean, variance, peak value, and trend rate of change of each indicator. Environmental correlation characteristics include the correlation coefficient between water temperature and dissolved oxygen, the coupling coefficient between wind speed and pollutant diffusion, and the negative correlation between rainfall and ammonia nitrogen concentration.

[0046] S3: Construct a water quality status identification model based on water quality parameter characteristics and environmental correlation characteristics to identify water quality status types.

[0047] Step S3, the steps for constructing the water quality status identification model include: S31: Integrate environmental correlation features with water quality parameter features to construct a multi-dimensional feature vector. The multi-dimensional feature vector has ≥15 dimensions and specifically includes pH mean, dissolved oxygen variance, ammonia nitrogen peak, COD trend change rate, water temperature-dissolved oxygen correlation coefficient, wind speed coupling coefficient, etc. Redundant features are removed through feature selection algorithm, and core features with a contribution of ≥10% are retained.

[0048] The steps for removing redundant features using a feature selection algorithm include: The system pre-defines feature selection targets and redundancy judgment criteria, clearly defines redundant features based on water quality status differentiation, sets a core retention threshold of ≥10% for single feature contribution, and selects evaluation indicators such as random forest Gini importance and mutual information.

[0049] The combined process of "statistical initial screening - algorithmic fine screening" is executed. First, variance filtering and correlation coefficient matrix are used to remove low-fluctuation / high-collinearity features. Then, random forest importance ranking and recursive feature elimination are used to eliminate quantified contribution. Cross-validation is used to screen redundant candidates.

[0050] Based on the contribution threshold and business logic, core features are retained, and redundancy is eliminated to construct a multi-dimensional feature vector with ≥15 dimensions. The effect of cross-validation on improving the accuracy of water quality status identification is verified, and a simplified feature set serving status identification is output.

[0051] S32: Based on historical water quality sample data, a random forest algorithm is used to train the initial water quality status identification model. The historical water quality sample data must cover water quality data from the monitored area over the past 5 years, with a sample size of ≥1000 groups. Each sample group contains a feature vector and a corresponding water quality status label, including excellent, acceptable, and exceeding standards, and the samples must evenly cover all three statuses. The initial parameters of the random forest algorithm are set as follows: 100 decision trees, a maximum depth of 15 layers, a minimum number of sample splits of 2, and a minimum number of leaf nodes of 1.

[0052] S33: The hyperparameters of the initial water quality state identification model are optimized using a grid search method to obtain the final water quality state identification model. The parameter search range for the grid search includes 50-200 decision trees, a maximum depth of 5-25, and a minimum number of sample splits of 2-10. The accuracy of 5-fold cross-validation is used as the evaluation index to select the optimal combination of hyperparameters. For example, when the number of decision trees is 150, the recall rate of the model for the out-of-standard state is improved by 5%; a maximum depth of 20 layers can better capture the nonlinear interaction between water temperature, dissolved oxygen, and ammonia nitrogen; and a minimum number of sample splits of 4 can effectively filter out random fluctuations. Finally, the optimal combination of 150 decision trees, a maximum depth of 20 layers, and a minimum number of sample splits of 4 is determined, making the model more accurate in distinguishing between the three types of states.

[0053] S34: Set three water quality assessment thresholds, input multi-dimensional feature vectors into the final water quality status recognition model, and cross-validate using historical data to output the water quality status type. The three water quality assessment thresholds are based on the "Surface Water Environmental Quality Standard" (GB3838-2002), including Level 1 (Excellent), Level 2 (Qualified), and Level 3 (Exceeding Standard). Cross-validation uses 5-fold cross-validation, requiring a model accuracy ≥92% to ensure the overall reliability of the distinction between excellent, qualified, and exceeding standards. The recall rate for exceeding standards must be ≥90% to avoid missing pollution events; otherwise, return to step S32 for retraining. The water quality standard concentration limits are directly adopted from the corresponding categories in mandatory standards such as the "Surface Water Environmental Quality Standard" (GB3838-2002) and the "Groundwater Quality Standard" (GB / T14848-2017). For characteristic pollutants without standards, the limits are derived with reference to the "Technical Guidelines for the Formulation of Water Quality Standards for Human Health" (HJ837-2017), and the corresponding limit levels are selected in combination with the water body function of the monitoring area (drinking water source, fishery water, etc.).

[0054] S4: By combining pollutant concentration gradient data from multiple water quality parameters, the source of pollution diffusion can be deduced to obtain regional pollution source tracing data.

[0055] Step S4, the steps for obtaining regional pollution source tracing data, include: S41: Based on the pollutant concentration gradient data at each monitoring point, combined with the water flow velocity and direction data of the monitoring area, a diffusion model is used to fit the pollutant diffusion trajectory, and the spatial range of the initial pollution source is inferred. The pollutant concentration gradient data is calculated using the concentration difference between adjacent monitoring points and is expressed as follows:

[0056] Where H is the pollutant concentration gradient, K m and K s Let m be the pollutant concentration at adjacent monitoring points m and s, and L be the distance between adjacent monitoring points m and s.

[0057] Water flow velocity and direction data are collected in real time through velocity and direction sensors, or by accessing historical average data from a GIS hydrological database; the diffusion model adopts the Eulerian-Lagrange diffusion model, the core formula of which is expressed as:

[0058] Where C(x,y,t) represents the pollutant concentration at spatial coordinates (x,y) at time t, and Q is the pollutant emission amount, i.e., the total mass of pollutants emitted by the pollution source per unit time. D xy The lateral diffusion coefficient describes the ability of pollutants to diffuse laterally in water flow, reflecting the influence of factors such as water turbulence on pollutant diffusion. x u y ...

[0059] S42: Locate basic information of pollution sources within this spatial range using GIS. Access the GIS spatial database and filter pollution sources by latitude and longitude. Basic information includes the source name, type, geographical coordinates, number and location of discharge outlets, and main types of pollutants emitted. Pollution source types are classified according to the "Classification Management Directory of Stationary Pollution Source Discharge Permits," including industrial pollution sources (subdivided into chemical, metallurgical, printing and dyeing, food processing, etc.), agricultural pollution sources (livestock and poultry farming, aquaculture, fertilizer application, etc.), domestic pollution sources (urban sewage treatment plants, direct discharge outlets of domestic sewage, etc.), and other pollution sources (solid waste storage sites, transportation pollution, etc.), defined based on the pollution source's production process, types of pollutants emitted, and source attributes.

[0060] S43: Based on basic pollution source information, retrieve the pollution source filing files from the environmental protection department to obtain pollution source emission records. The environmental protection department's filing files are retrieved through the government data sharing platform. The emission records include pollution source codes, permitted emission amounts, actual emission concentrations over the past year, emission periods, and the operational status of pollution control facilities.

[0061] S44: Retrieve historical pollution treatment databases to extract historical event records similar to the basic information of pollution sources. Search criteria are set as follows: pollution source type similarity ≥ 80%, consistent main pollutant types, and seasonal deviation of pollution occurrence ≤ 1 month. Historical event records include pollution occurrence time, pollution degree, treatment measures, treatment cycle, treatment effect, and recurrence in the past 3 years. The historical pollution treatment database is constructed by integrating emergency response archives of relevant departments for sudden environmental incidents, pollution control case databases of research institutions, and pollution prevention and control practice records of industry associations. Data sources include publicly available data from the National Environmental Monitoring Center, data filed with local relevant departments' emergency management platforms, and pollution control research results published in core journals. The database is updated quarterly. The method for obtaining pollution source type similarity is as follows: extract the four-level hierarchical feature information of the target pollution source to be traced, and construct a four-level hierarchical feature library of "major category - intermediate category - subcategory - feature attribute" based on the classification system of the "Classification Management Directory of Discharge Permits for Fixed Pollution Sources". Weights were assigned to the four levels of features, and matching scores were calculated. The weight allocation was determined based on the accuracy of the features in representing the pollution source type, including: Category weight W1 = 0.1, Sub-category weight W2 = 0.2, Sub-category weight W3 = 0.3, and Feature attribute weight W4 = 0.4. The categories include industrial pollution sources, agricultural pollution sources, domestic pollution sources, and other pollution sources. Sub-categories include industrial pollution sources such as chemical, metallurgical, printing and dyeing, and food processing; and agricultural pollution sources such as livestock and poultry farming, aquaculture, and fertilizer application. Sub-categories include, for example, subdividing the chemical category into petrochemical, coal chemical, and fine chemical industries. Feature attributes include key attributes such as production process type, core production equipment, types of raw and auxiliary materials, pollutant generation stages, and emission outlet type.

[0062] The feature matching degree is calculated layer by layer. A perfect match is scored as 1, and an incomplete match is scored from 0 to 1 according to the degree of matching. The final similarity calculation formula is:

[0063] In the formula, P represents the similarity of pollution source types, and W... i M represents the weight of the feature at level i. i denoted as the matching score for the i-th level feature.

[0064] The classification of M1 is based on the "major categories" in the "List of Classification Management of Discharge Permits for Stationary Pollution Sources", namely industrial, agricultural, domestic and other pollution sources. If the major category of the target pollution source is consistent with that of the historical pollution source, then M1=1; if the major categories are inconsistent, then M1=0.

[0065] The classification of M2 is based on the "medium category" in the "List of Classification Management of Discharge Permits for Stationary Pollution Sources", such as "chemical, metallurgy, and printing and dyeing" under the industrial category. If the target is consistent with the medium category of the historical pollution source, then M2=1; if the medium category is inconsistent, then M2=0.

[0066] The classification of M3 is based on industry segmentation standards, such as "petrochemical / coal chemical / fine chemical" under the chemical industry category. If the subcategories are completely identical, then M3=1; if the subcategories are different but belong to the same category and the processes / pollutants are highly related, then the scores are assigned according to the preset subcategory similarity matrix.

[0067] Methods for presetting subclass similarity matrices include: The feature dimensions that have the most direct impact on the similarity of pollution source types were selected, including the similarity of production processes, the overlap of pollutants, the consistency of pollution-generating links, and the similarity of raw and auxiliary materials.

[0068] Based on the contribution of each dimension to the similarity of pollution source types, different weights are assigned, specifically: production process similarity is 0.4, pollutant overlap is 0.3, consistency of pollution-generating links is 0.2, and raw and auxiliary material similarity is 0.1.

[0069] For each pair of subcategories, scores are assigned to four dimensions based on data sources such as industry standards, environmental filing data, and pollution cases, with a score range of 0-1, where 1 indicates a perfect match.

[0070] Taking petrochemical and coal chemical industries as examples: Production process similarity: 0.7 (both involve hydrocarbon processing, but the reaction pathways are different); Pollutant overlap: 0.8 (both emit VOCs and petroleum pollutants); Consistency of pollution-generating processes: 0.6 (some pollution-generating processes overlap); Raw material similarity: 0.5 (the raw materials are crude oil and coal, respectively, with low overlap).

[0071] For each pair of subclasses, a weighted sum is used to obtain their similarity score, as shown in the formula:

[0072] Finally, the subclass similarity matrix is ​​obtained and presented in tabular form, as shown in Table 1: ; Every 1-2 years, in conjunction with the revision of the "List of Classified Management of Discharge Permits for Stationary Pollution Sources" and the emergence of new industry processes, the subcategories and corresponding scores are supplemented or adjusted.

[0073] The feature attribute matching score M4 is calculated as follows:

[0074] The following are some situations where the similarity of pollution source types is 80%: Scenario 1: Major category, intermediate category, and subcategory are completely matched (M1=M2=M3=1), and the feature attribute matching degree is ≥75%.

[0075] Scenario 2: Major category and intermediate category are completely matched (M1=M2=1), subclass matching degree is ≥80%, and feature attribute matching degree is ≥90%.

[0076] To avoid the problem of "similar types but inconsistent pollution characteristics" caused by simple quantitative scoring, a mandatory matching rule for key features is set as an additional basis for similarity ≥80%, including: the main pollutant types emitted by the target pollution source and the historical pollution source are consistent; the pollution generation process of the target pollution source and the historical pollution source is similar to ≥80%; and for industrial pollution sources, the core principles of the production process are consistent.

[0077] S45: Integrate basic information on pollution sources, pollution emission records, and historical event records to form regional pollution source tracing data. The information fusion uses a weighted average method, assigning weights to key information. The weights are: pollution source type 0.3, emission concentration 0.25, historical recurrence rate 0.2, treatment effectiveness 0.15, and treatment facility status 0.1, ultimately forming a structured source tracing data report.

[0078] S5: Integrate water quality status type and regional pollution source tracing data to construct a water quality risk assessment index, determine the risk level, and generate targeted early warning and disposal instructions.

[0079] Step S5, the steps for constructing the water quality risk assessment index, include: S51: Based on the water quality status, extract the degree of pollutant exceedance and the pollution diffusion rate, and calculate the pollution severity index, expressed by the formula:

[0080] Wherein, PSPI is the pollution severity index, λ1 is the weighting coefficient for the degree of pollutant exceedance, λ2 is the weighting coefficient for pollution diffusion rate, i is the pollutant type index, and C i S represents the measured concentration of the i-th critical pollutant. i The water quality standard concentration limit for pollutant of category i. V represents the maximum exceedance multiple among all monitored pollutants, V0 represents the actual diffusion rate of the pollutant, and V0 represents the standard threshold for the diffusion rate of the pollutant. The selection of key water quality parameters is based on the core indicators of the "Surface Water Environmental Quality Standard" (GB3838-2002) and the pollution characteristics of the monitoring area. Required parameters include pH, dissolved oxygen, ammonia nitrogen, total phosphorus, total nitrogen, and chemical oxygen demand (COD). Optional parameters are added according to the type of regional pollution (e.g., lead and cadmium are added to areas with heavy metal pollution, and volatile organic compounds are added to areas with organic pollution).

[0081] S52: Combining regional pollution source tracing data, extract the distance between pollution sources and sensitive areas, the ecological importance weight of sensitive areas, and calculate the sensitive area impact index. The formula is as follows:

[0082] Wherein, SAII is the Sensitive Area Impact Index, n is the total number of affected sensitive areas within the monitoring area, and W j Let D be the ecological importance weight of the j-th sensitive area, and D0 be the safe distance threshold for the sensitive area. j denoted as , representing the straight-line distance between the pollution source and the j-th sensitive area. The safe distance threshold for sensitive areas is determined based on the type of sensitive area (drinking water source protection zone, nature reserve, aquaculture zone, etc.), combined with regulations and standards such as the "Technical Specification for Delineation of Drinking Water Source Protection Zones" (HJ338-2018) and the "Regulations on Nature Reserves," through simulation using a pollutant migration and diffusion model. The safe distance threshold is ≥1000m for primary drinking water source protection zones, ≥800m for core areas of nature reserves, and ≥500m for general sensitive areas.

[0083] S53: Set the weighting coefficients for the pollution severity index and the sensitive area impact index, and construct a water quality risk assessment index using a weighted summation formula, expressed as follows:

[0084] Where R is the water quality risk assessment index, and λ PSPI , λ SAII These are the weighting coefficients for the pollution severity index and the sensitive area impact index, respectively, and λ PSPI +λ SAII =1, the water quality risk assessment index R ranges from 0 to 3, and the larger the value, the higher the risk.

[0085] Step S5, the steps for determining the risk level include: S54: Preset four risk level thresholds, including low risk, medium risk, high risk, and extremely high risk.

[0086] The specific thresholds for Level 4 risk are divided as follows: The water quality is stable, pollutants have not significantly exceeded the standards, the risk of diffusion is low, sensitive areas are minimally affected, and no active intervention is required. Therefore, it is set as low risk, and the corresponding water quality risk assessment index range is 0≤R<0.75.

[0087] There are signs of localized pollutant exceedances or accelerated spread. Sensitive areas may be slightly affected. Monitoring and preliminary investigation need to be strengthened. The risk level is set at medium risk, with a corresponding water quality risk assessment index range of 0.75≤R<1.5.

[0088] If pollutants exceed standards severely or spread rapidly, sensitive areas face significant threats, requiring the activation of emergency response and pollution source verification. This area is classified as high-risk, with a corresponding water quality risk assessment index range of 1.5 ≤ R < 2.25.

[0089] If pollutants are severely exceeding standards and spreading out of control, sensitive areas (such as drinking water sources and nature reserves) are facing significant threats and require immediate emergency response measures, the risk level is classified as extremely high, with a corresponding water quality risk assessment index range of 2.25≤R≤3.

[0090] S55: Perform a secondary verification based on the water quality status type, adjust the risk level, and output the changed risk information. If the water quality status type is Level 1, adjust to low risk regardless of the R value; if the water quality status type is Level 2, maintain the original level if R ≤ 0.5, and upgrade to Level 1 if R > 0.5; if the water quality status type is Level 3 (exceeding standards), upgrade to Level 1 but not exceeding extremely high risk if R ≥ 0.4, and maintain the original level if R < 0.4.

[0091] S56: Based on the historical pollution recurrence rate in the regional pollution source tracing data, further adjust the changed risk information and output the final risk level. The formula for calculating the historical pollution recurrence rate is:

[0092] Where δ is the historical pollution recurrence rate, A is the number of times the pollution source has recurred in the past 3 years, and M is the total number of monitoring sessions in the past 3 years.

[0093] The adjustment rules include: For a recurrence rate ≥30%, the risk level is increased by one level, but not exceeding extremely high risk. For a recurrence rate ≤10% and <30%, the risk level remains unchanged. For a recurrence rate <10%, the risk level is decreased by one level, but not lower than low risk.

[0094] S6: Real-time collection of water quality data and monitoring system operation data after early warning and response; calculation of monitoring accuracy index, response timeliness index, and data integrity index; evaluation of monitoring effectiveness; and construction of optimization scoring coefficients based on monitoring effectiveness. Monitoring system operation data is collected through the built-in operation status monitoring module of the monitoring equipment, including sensor operating voltage, data transmission delay, equipment operating temperature, calibration cycle, fault alarm records, etc., with the collection frequency consistent with that of water quality data.

[0095] The steps for calculating the monitoring accuracy index include: S61: Select standard calibration points within the monitoring area and determine the standard values ​​of key water quality parameters. The standard calibration points shall be evenly distributed in the monitoring area, covering the upstream, midstream, downstream and near key pollution sources, with no fewer than 5 points; the standard values ​​shall be determined in the laboratory using the national standard method specified in the "Surface Water Environmental Quality Standard" (GB3838-2002), with a parallel sample error ≤5%.

[0096] S62: Extract the measured values ​​of multiple water quality parameters at the same time point, calculate the relative error of the multiple parameter data, expressed by the formula: Relative Error = |Measured Value - Standard Value| / Standard Value × 100%; Take the average value of the relative errors of all key parameters as the baseline value of monitoring accuracy, and calculate the monitoring accuracy index. Baseline value of monitoring accuracy = Sum of relative errors of all parameters / m, where m is the total number of key parameters, and m≥5; Monitoring accuracy index = 1 - Baseline value of monitoring accuracy, usually set at a threshold of ≥0.95 for qualified monitoring accuracy; if it is lower than this threshold, the monitoring accuracy is considered unqualified.

[0097] The steps for calculating the response time index include: S63: Define response time as the total time from the moment the intelligent sensor collects abnormal water quality data to the release of early warning information and its push to the relevant handling department.

[0098] S64: Set a standard response time limit and calculate the response time limit index; Response time limit index = Standard response time limit / Actual response time limit. When the actual response time limit ≤ Standard response time limit, the response time limit index ≥ 1, and it is judged as qualified. When the actual response time limit > Standard response time limit, the response time limit index < 1, and it is judged as unqualified. The standard response time limit is set according to the graded response requirements of the "National Emergency Response Plan for Sudden Environmental Incidents". The standard response time limit for general sudden environmental incidents (low to medium risk) is ≤ 2 hours, and the standard response time limit for major and above sudden environmental incidents (high and extremely high risk) is ≤ 1 hour.

[0099] The steps for calculating the data integrity index include: S65: Theoretical total data volume collected within the monitoring period. Theoretical total data volume = number of monitoring points × number of key parameters × collection frequency × monitoring duration. Invalid data within the monitoring period is removed, and the amount of valid data is calculated. Invalid data is judged based on values ​​exceeding a reasonable range, data marked as missing, and data marked as sensor malfunction. Typically, a data integrity threshold of ≥98% is set; data below this threshold is considered unqualified for integrity.

[0100] S66: Based on the theoretical total amount of collected data and the amount of valid data, calculate the data integrity index, which is expressed as: Data integrity index = valid data amount / theoretical total amount of collected data × 100%. Usually, the data integrity qualification threshold is set to ≥98%. If it is lower than this threshold, the data integrity is judged to be unqualified.

[0101] Step S6, the step of constructing the optimized score coefficient based on the monitoring results, includes: S67: Normalize the monitoring accuracy index, the response timeliness index, and the data integrity index to obtain standard index data. The formula is: Standard index = (Original index - Minimum index value) / (Maximum index value - Minimum index value); where the monitoring accuracy index ranges from 0 to 1, the response timeliness index ranges from 0 to 2, and the data integrity index ranges from 0 to 100%; after normalization, all standard indices are mapped to the 0-1 interval.

[0102] S68: Based on the aforementioned standard index data, set index weights according to the monitoring system priority. The monitoring system refers to a complete water quality monitoring technology system consisting of a sensing layer, transmission layer, data layer, and application layer, with core functions covering the entire process of data acquisition, transmission, processing, analysis, and early warning. The monitoring system priority is determined through a combination of the analytic hierarchy process (AHP) and expert scoring, with the specific steps as follows: Construct a priority evaluation index system, including data reliability, monitoring timeliness, and early warning accuracy.

[0103] Invite 5-10 experts in water environment monitoring and automation control to construct a judgment matrix by comparing the importance of each indicator pairwise using the 1-9 scale method.

[0104] After passing the consistency test (CR<0.1), the weights of each indicator are obtained by calculating the eigenvector.

[0105] Adjustments are made based on the needs of the monitoring scenario. For example, drinking water source monitoring systems prioritize data reliability, while industrial wastewater discharge monitoring systems prioritize monitoring timeliness. The index weighting rules are as follows: based on the monitoring system priority evaluation results, the core priority indicators are assigned higher weights. After adjustments based on the monitoring scenario needs, the monitoring accuracy index corresponds to data reliability (core priority), with a weight ω. a=0.5; Response timeliness index corresponds to monitoring timeliness (high priority), weight ω b =0.3; Data integrity index corresponds to early warning accuracy (high priority), weight ω c =0.2, the weights satisfy ω a +ω b +ω c =1, which can be fine-tuned according to the specific monitoring scenario (e.g., in remote areas, the weight of response time can be reduced to 0.2, and the weight of data integrity can be increased to 0.3).

[0106] S69: Based on the index weights, the optimization score coefficient is calculated using a weighted summation method. A passing threshold for the optimization score is typically set at ≥0.96. Scores below this threshold are considered unqualified, and the optimization process must be initiated. The formula is expressed as:

[0107] Where S is the optimization score coefficient, α1, α2, and α3 are the monitoring accuracy index, response timeliness index, and data integrity index, respectively, and ω a ω b ω c These are the weighting coefficients for the monitoring accuracy index, response timeliness index, and data integrity index, respectively. Among them, the monitoring accuracy index is the core indicator, with a weight ω. a =0.5; Response timeliness index is the key indicator, with a weight ω b =0.3; Data integrity index is the basic indicator, with a weight ω c =0.2; the weights satisfy ω a +ω b +ω c =1, and can also be fine-tuned according to actual monitoring needs.

[0108] S7: Correct the monitoring system based on the optimized score coefficient. The steps include: S71: If the optimized score coefficient is unqualified, analyze the contribution ratio of each index in the standard index data to pinpoint the root cause. A low monitoring accuracy index indicates sensor malfunction or calibration deviation. A low response timeliness index indicates data transmission delay or cumbersome early warning procedures. A low data integrity index indicates blind spots in sensor deployment or unstable transmission links. The contribution ratio formula is expressed as:

[0109] For example, if the optimization score coefficient S=0.886, the contribution percentage of α1 is approximately 55.3% (0.5×0.98) / 0.886×100%), the contribution percentage of α2 is approximately 22.4% (0.3×0.665) / 0.886×100%), and the contribution percentage of α3 is approximately 22.3% (0.2×0.9827) / 0.886×100%). If the standard value of α2 is the lowest at 0.665 and the contribution percentage is low, the root cause is insufficient response time.

[0110] S72: Develop corrective measures to address root causes, including triggering self-tests and replacement alerts for sensor malfunctions, recalibrating based on standard values ​​for calibration deviations, optimizing transmission priorities for transmission delays, simplifying approval processes for cumbersome early warning procedures, adding monitoring points to deployment blind spots, and strengthening signal enhancement devices for unstable transmission links.

[0111] S73: Determine the correction intensity based on the difference between the optimized score coefficient and the corresponding index's pass threshold, and recalculate the optimized score coefficient until the pass standard is met. The specific correction intensity is categorized as: minor deviation, moderate deviation, and severe deviation. Minor deviations are corrected using standard methods, such as simplifying one approval step or calibrating a single sensor; moderate deviations are corrected using enhanced methods, such as optimizing transmission priority, simplifying approval steps, or replacing two or fewer faulty sensors; severe deviations are corrected using comprehensive methods, such as replanning the monitoring point layout, upgrading the transmission network, and replacing all low-precision sensors. After correction, data must be collected again for a monitoring duration of ≥24 hours, and the optimized score coefficient must be recalculated until the optimized score coefficient S≥0.96.

[0112] In summary, the real-time multi-parameter water quality monitoring method based on intelligent sensors and big data improves the accuracy and reliability of water quality status identification by collecting multiple water quality parameters and environmental auxiliary data in real time and extracting features in depth. Combined with a water quality status identification model optimized using random forest algorithms and grid search, this method enhances the accuracy and reliability of water quality status identification. By integrating pollutant concentration gradients, water flow characteristics, and GIS technology, along with multi-dimensional information integration from pollution source registration files and historical event records, the method achieves precise location and tracing of pollution sources, providing strong support for pollution control. The method utilizes a dual index construction that quantifies the severity of pollution and the impact on sensitive areas, coupled with a multi-level judgment logic that includes four-level risk thresholds, secondary verification of water quality status, and adjustment based on historical recurrence rates. This makes risk level classification more aligned with real-world scenarios, and the generation of targeted early warnings and response instructions effectively improves the timeliness and scientific rigor of pollution response. Through quantitative evaluation of three core indices—monitoring accuracy, response timeliness, and data integrity—and the optimized score coefficient formed by weighted allocation, the method can accurately identify system operational shortcomings and formulate targeted correction plans, continuously optimizing the monitoring system and ensuring its long-term stability, efficiency, and data quality.

[0113] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data, characterized in that, include: S1: Based on the target parameters of water quality monitoring and the environmental parameters of the monitoring area, collect multi-parameter water quality data and environmental auxiliary data of each monitoring point in real time; the multi-parameter water quality data includes pollutant concentration gradient data. S2: Preprocess the water quality multi-parameter data and the environmental auxiliary data to extract water quality parameter features and environmental correlation features; S3: Construct a water quality status identification model based on the water quality parameter characteristics and the environmental correlation characteristics to identify the water quality status type; S4: Combine the pollutant concentration gradient data to deduce the source of pollution diffusion and obtain regional pollution source tracing data; S5: Based on the water quality status type and the regional pollution source tracing data, construct a water quality risk assessment index, determine the risk level, and generate targeted early warning and disposal instructions; S6: Collect water quality data and monitoring system operation data after early warning and response in real time, calculate monitoring accuracy index, response timeliness index and data integrity index, evaluate the monitoring effect, and construct an optimization score coefficient based on the monitoring effect; S7: Correct the monitoring system based on the optimized score coefficient.

2. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 1, characterized in that, Step S3, the steps for constructing the water quality status identification model include: S31: The environmental correlation features are fused with the water quality parameter features to construct a multi-dimensional feature vector; S32: Based on historical water quality sample data, the initial water quality status identification model is trained using the random forest algorithm; S33: Optimize the hyperparameters of the initial water quality state identification model using the grid search method to obtain the final water quality state identification model; S34: Set a three-level water quality judgment threshold, input the multi-dimensional feature vector into the final water quality status recognition model, cross-validate with historical data, and output the water quality status type.

3. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 1, characterized in that, Step S4, the steps for obtaining the regional pollution source tracing data, include: S41: Based on the pollutant concentration gradient data at each monitoring point, combined with the water flow velocity and direction data in the monitoring area, a diffusion model is used to fit the pollutant diffusion trajectory and infer the spatial range of the initial source of pollution. S42: Locate basic information on pollution sources within this spatial range using GIS; S43: Based on the aforementioned basic information about the pollution source, retrieve the pollution source registration file and obtain the pollution source emission records; S44: Search the historical pollution treatment database and extract historical event records that are similar to the basic information of the pollution source; S45: The basic information of the pollution source, the emission records of the pollution source, and the historical event records are integrated to form the regional pollution source tracing data.

4. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 1, characterized in that, Step S5, the step of constructing the water quality risk assessment index includes: S51: Based on the water quality status type, extract the degree of pollutant exceedance and the pollution diffusion rate, and calculate the pollution severity index; S52: Based on the aforementioned regional pollution source tracing data, extract the distance between pollution sources and sensitive areas, the ecological importance weight of sensitive areas, and calculate the impact index of sensitive areas; S53: Set the weight coefficients for the pollution severity index and the sensitive area impact index, and construct the water quality risk assessment index through a weighted summation formula.

5. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 4, characterized in that, In step S51, the pollution severity index formula is expressed as: ； Wherein, PSPI is the pollution severity index, λ1 is the weighting coefficient for the degree of pollutant exceedance, λ2 is the weighting coefficient for pollution diffusion rate, i is the pollutant type index, and C i S represents the measured concentration of the i-th critical pollutant. i The water quality standard concentration limit for pollutant of category i. V represents the maximum number of times the pollutant exceeds the standard among all monitored pollutants, V represents the actual diffusion rate of the pollutant, and V0 represents the standard threshold for the diffusion rate of the pollutant.

6. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 4, characterized in that, In step S52, the formula for the sensitive area influence index is expressed as: ； Wherein, SAII is the Sensitive Area Impact Index, n is the total number of affected sensitive areas within the monitoring area, and W j Let D be the ecological importance weight of the j-th sensitive area, and D0 be the safe distance threshold for the sensitive area. j Let be the straight-line distance between the pollution source and the j-th sensitive area.

7. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 1, characterized in that, Step S5, the steps for determining the risk level include: S54: Preset four-level risk level thresholds; S55: Based on the risk level threshold and the water quality status type, a secondary verification is performed to adjust the risk level and output the changed risk information; S56: Based on the historical pollution recurrence rate in the regional pollution source tracing data, further adjust the changed risk information and output the final risk level.

8. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 1, characterized in that, Step S6, which involves calculating the monitoring accuracy index, response timeliness index, and data integrity index, includes the following steps: S61: Select standard calibration points within the monitoring range and determine the standard values ​​of key water quality parameters; S62: Extract the measured values ​​of the water quality multi-parameter data at the same time point, calculate the relative error of the multi-parameter data, take the average value of the relative errors of all key parameters as the basic value of monitoring accuracy, and calculate the monitoring accuracy index; S63: Define response time as the total time from the moment the smart sensor collects abnormal water quality data to the release of early warning information and its push to the relevant handling department; S64: Set the standard response time and calculate the response time index; S65: The theoretical total amount of data collected during the statistical monitoring period; invalid data within the monitoring period is removed, and the amount of valid data is counted. S66: Calculate the data integrity index based on the theoretical total amount of collected data and the amount of effective data.

9. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 8, characterized in that, Step S6, the step of constructing the optimized score coefficient based on the monitoring effect, includes: S67: Normalize the monitoring accuracy index, the response timeliness index and the data integrity index to obtain standard index data; S68: Based on the standard index data, set the index weights according to the monitoring system priority; S69: Calculate the optimization score coefficient by weighted summation based on the index weights.

10. The method for real-time monitoring of multiple water quality parameters based on intelligent sensors and big data according to claim 9, characterized in that, Step S7, the steps for correcting the monitoring system include: S71: If the optimized score coefficient is not up to standard, break down the contribution ratio of each index in the standard index data to locate the root cause. S72: Develop a corrective plan to address the root cause; S73: Based on the correction scheme, determine the correction intensity according to the difference between the optimized score coefficient and the qualified threshold of the corresponding index, and recalculate the optimized score coefficient until the qualified standard is reached.

Images