A method and system for predicting and warning the pollution intensity of a river during flood season

Abstract

The application provides a river flood season pollution intensity prediction and early warning method and system, relates to the technical field of water conservancy projects, and the method comprises the following steps: acquiring multi-source data in a basin, and preprocessing the multi-source data; constructing a hydrological water quality model for calculating the hydrological cycle process of the entire basin from the upstream to the downstream; calculating the point source and area source pollution load to obtain water quality concentration; constructing a prediction data set according to the water quality concentration, and training an XGBoost machine learning model by using the prediction data set; and predicting the pollutant concentration in the flood season under the influence of different meteorological conditions and carrying out over-standard concentration early warning by using the trained XGBoost machine learning model. The application can predict the short-term river water quality change peak value in advance by fusing meteorological prediction, a hydrological model and real-time monitoring data, realize early warning of water quality change, and finally integrate into a basin management decision platform to realize early regulation and control of the management department and pre-control of pollution.

Description

Technical Field

[0001] The present invention belongs to the field of water conservancy projects, and in particular relates to a method and system for predicting and warning river pollution intensity during flood season. Background Art

[0002] As a typical complex system of nature, economy, and society, urban and rural areas face water pressures and the resulting environmental risks brought about by rapid population growth and intensified socioeconomic activities. Water pollution and the resulting environmental risks have become key factors hindering the coordinated development of regional economies, societies, and the environment. Despite significant progress in controlling point-source pollution, non-point source pollution and its variability remain profound issues that require urgent resolution. Nitrogen and phosphorus, in particular, have become major pollutants in the Yangtze River Basin, coastal waters, and numerous lakes. Non-point source pollution is gradually becoming a key constraint to the sustained improvement of the aquatic ecosystem.

[0003] Flood season pollution intensity, as a key indicator of changes in pollutant concentrations in rivers or lakes during the flood season (i.e., periods of heavy rainfall), is crucial for assessing the severity of water pollution during this period. However, existing hydrological models and pollution prediction methods often fail to adequately address the nonlinear complexity of non-point source pollution. This limitation is particularly pronounced when hydrological and meteorological conditions significantly influence pollution loads. Therefore, simulations of flood season pollution intensity driven by the "meteorological conditions-pollutant concentration" model are particularly urgent.

[0004] Pollution intensity during flood seasons reflects not only changes in water quality but also the strength of factors such as urban planning, infrastructure development, pollution control measures, and the ability to adapt to climate change. Excessive pollution during flood seasons can severely damage river ecosystems and directly threaten human drinking water safety and ecological health. Particularly in economically developed and densely populated areas, the accelerated pace of urbanization and the continued growth of sewage discharge have exacerbated flood season pollution, making it a major challenge in water environment protection and the development of an ecological civilization.

[0005] Faced with the problem of river pollution during the flood season, the current detection methods often have lags and limitations, making it difficult to detect and effectively respond to sudden pollution incidents in a timely manner. Conducting pollution simulation and early warning research during the flood season is of great significance for scientifically assessing the degree of water pollution, predicting the trend of changes in pollution intensity, and taking effective control measures in advance. Although the common hydrological models today can well simulate the transformation and migration of pollutants within the basin, they are unable to predict the pollution situation during the flood season based on changes in various influencing factors. In addition, the nonlinear relationships between some factors cannot be accurately portrayed by the hydrological models. Machine learning models are powerful tools for mining complex nonlinear data patterns, but the construction of machine learning models requires huge data sets and is not universally applicable to many areas where data is scarce.

[0006] Therefore, the flood season pollution intensity prediction and early warning method based on meteorological conditions and pollutant concentrations has extremely important application value. Summary of the Invention

[0007] In response to the above-mentioned deficiencies in the existing technology, the present invention provides a method and system for predicting and warning of river pollution intensity during flood season. The present invention proposes an innovative prediction and warning method that can utilize refined hydrological models and advanced machine learning technology to overcome the shortcomings of traditional methods and predict flood season pollution intensity in real time and accurately, thereby providing effective early warning support for decision makers, allowing them to take control measures in advance and ensure water environment safety.

[0008] In order to achieve the above objectives, the technical solution adopted by the present invention is: a method for predicting and warning the pollution intensity of a river during flood season, comprising the following steps:

[0009] S1. Obtain multi-source data in the basin and pre-process the multi-source data;

[0010] S2. Construct a hydrological and water quality model to calculate the hydrological cycle process of the entire basin from upstream to downstream;

[0011] S3. Using the constructed hydrological and water quality model, calculate the pollution load of point sources and non-point sources to obtain water quality concentration;

[0012] S4. Build a prediction dataset based on water quality concentration and use the prediction dataset to train the XGBoost machine learning model.

[0013] S5. Use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions and issue warnings for concentrations exceeding the standard, thereby completing the prediction and warning of river pollution intensity during the flood season.

[0014] Beneficial effects of the present invention: The present invention provides a method for predicting and warning of river pollution intensity during flood season based on a combination of mechanism model and machine learning. First, based on an arbitrary hydrological and water quality model, the hydrological and water quality changes at different time and spatial scales of the basin are simulated, and then the machine learning XGBoost model is used to realize the prediction of pollutant concentration during the flood season under the influence of different meteorological conditions and the warning of excessive concentration. The present invention can utilize refined hydrological models and advanced machine learning technology to overcome the shortcomings of traditional methods, and accurately predict the pollution intensity during the flood season in real time, thereby providing effective early warning support for decision makers, taking control measures in advance, and ensuring the safety of the water environment.

[0015] Furthermore, the S1 is specifically:

[0016] S101, obtaining multi-source data within the watershed, and performing cleaning and correction processing on the multi-source data;

[0017] S102, identifying and removing outliers from the multi-source data processed in S101 by using statistical analysis methods, and filling in missing values;

[0018] S103. Based on the multi-source data processed by S102, for the discontinuous time series of the meteorological data and water quality monitoring data, a linear interpolation method is used to insert several new time points between two adjacent dates, and the meteorological data values ​​and water quality concentrations at the new time points are estimated using the trends of the existing data points;

[0019] S104: Based on the data processed in S103, the multi-source data in the watershed are integrated to form a standard data set, wherein the data at each time point in the standard data set includes multi-dimensional information, thereby completing the preprocessing of the multi-source data.

[0020] The beneficial effect of the above further scheme is that the above process is designed to provide high-quality input data for the hydrological and water quality model and ensure the high compatibility and strong flexibility of the model. In this process, the high quality of the original data is ensured through machine learning and data processing methods, laying the foundation for subsequent model calculations and predictions.

[0021] Furthermore, the hydrological cycle process in S2 includes atmospheric water process, soil water process, groundwater process and surface water process, wherein the soil water process follows the soil water balance formula as follows:

[0022]

[0023] in, Indicates the soil water storage on that day. represents the soil water storage on the previous day, Indicates the amount of irrigation water, Indicates the infiltration rate, represents soil evaporation, represents the transpiration of vegetation, Indicates the amount of deep leakage, It means flow in soil;

[0024] Groundwater processes include shallow groundwater processes and deep groundwater processes. The shallow groundwater process follows the shallow groundwater balance formula as follows:

[0025]

[0026] in, Indicates the shallow groundwater storage volume on that day, represents the shallow groundwater storage volume of the previous day, represents the shallow recharge, represents the base flow generation, represents the amount of evaporation from diving, Indicates shallow usage;

[0027] The deep groundwater balance formula followed by the deep groundwater process is as follows:

[0028]

[0029] in, Indicates the amount of deep groundwater storage on that day, represents the amount of deep groundwater storage on the previous day, Indicates the deep recharge amount, Indicates deep usage;

[0030] The pond / wetland water balance equation for surface water storage is as follows:

[0031]

[0032] in, Indicates the amount of pond / wetland storage on that day, Indicates the amount of pond / wetland storage on the previous day. represents the inflow rate, Indicates the flow rate, Indicates rainfall, Indicates the evaporation rate, Indicates the amount of leakage;

[0033] The evolution of water volume in the main river during the surface water process follows the main river water balance formula as follows:

[0034]

[0035] in, Indicates the storage capacity of the main river channel on that day, Indicates the storage volume of the main river channel on the previous day. represents the inflow of the upstream river, Indicates the downstream outflow, represents the riverbank storage volume, Indicates irrigation and consumption, Indicates the evaporation of river water surface, Indicates the amount of leakage.

[0036] The beneficial effect of the above-mentioned further scheme is that by refining the hydrological cycle process into four dynamic subsystems: atmospheric water, soil water, groundwater and surface water, it realizes the refined modeling and multi-level coupling of the water cycle process, and fully depicts the water migration path of "precipitation-infiltration-evapotranspiration-runoff-recharge".

[0037] Furthermore, the expression of the hydrological and water quality model in S2 is as follows:

[0038]

[0039] in, represents the soil water storage, Indicates time, Indicates precipitation, Indicates the evaporation rate, Indicates the infiltration rate, Represents surface runoff.

[0040] The beneficial effects of the above further scheme are: constructing a soil water dynamic equation based on water balance, integrating the physical process of the hydrological cycle with the migration path of pollutants, and realizing the coordinated simulation of hydrological processes and water quality evolution.

[0041] Furthermore, the S3 is specifically:

[0042] S301. Based on the hydrological response unit, a hydrological and water quality model is used to simulate the hydrological process, taking into account the generation, transformation, and migration of water pollutants. The pollutant load and inflow into the river are calculated using the following formula:

[0043]

[0044] in, Indicates the i The pollutant load of each hydrological response unit HRU, N Indicates the number of pollution sources in the hydrological response unit HRU, which includes point source pollution and non-point source pollution. represents the area of ​​the hydrological response unit HRU, Indicates that it comes from j Pollutant concentrations from various pollution sources;

[0045] S302. Based on the changes in pollutant concentration and inflow within each hydrological response unit (HRU), the amount of pollutants entering the river within each HRU is calculated using a hydrological and water quality model. For point source pollution, the pollutant discharge is calculated based on the point source flow and concentration. For non-point source pollution, the amount of pollutants entering the river for each HRU is calculated based on soil erosion, runoff, and pollution load:

[0046]

[0047] in, It represents the total amount of pollutants entering the river in the entire basin, It represents the accumulation of all hydrological response units HRU;

[0048] S303. Calculate the water quality concentration based on the amount of pollutants entering the river in each hydrological response unit (HRU), and complete the calculation of point source and non-point source pollution loads:

[0049]

[0050] in, Indicates water concentration, represents the pollution load, represents the sewage flow rate, Indicates the time conversion factor.

[0051] The beneficial effect of the above further solution is that the present invention uses a hydrological and water quality model to perform a refined simulation of the emission of pollution control factors, so as to effectively obtain the pollutant generation amount of each hydrological response unit HRU.

[0052] Furthermore, the S4 is specifically:

[0053] S401. Using six meteorological data items and Julian dates from historical years as input variables of the XGBoost machine learning model, a permanganate index and total phosphorus concentration prediction dataset was constructed, respectively. The six meteorological data items included daily cumulative rainfall, daily average radiation value, daily maximum temperature, daily minimum temperature, daily average relative humidity, and daily average wind speed.

[0054] S402, performing linear interpolation processing on the six meteorological data items and the Julian date, inserting an average of three new time points between two adjacent Julian dates, and performing linear interpolation processing on each new time point;

[0055] S403, adding the time series data obtained after the linear interpolation process and its corresponding meteorological data, permanganate index, and total phosphorus concentration data to the constructed permanganate index and total phosphorus concentration prediction data set to form a new data set, and using the new data set as the prediction data set;

[0056] S404: Divide the prediction data set obtained in S403 into a training set and a validation set, and use the training set and the validation set to train and validate the XGBoost machine learning model.

[0057] The beneficial effect of the above further scheme is: by constructing a water quality prediction model based on XGBoost machine learning, combining meteorological driving factors with time series characteristics, dynamic prediction of the concentrations of key pollutants (permanganate index, TP) during the flood season is achieved.

[0058] Furthermore, the loss function of the XGBoost machine learning model is expressed as follows:

[0059]

[0060]

[0061] in, represents the loss function of the XGBoost machine learning model, i express i samples, n represents the total number of samples, Represents the predicted value and the true value The error between represents the regularization parameter, which is used to balance the loss function and the regularization term. k Indicates the k A decision tree, K represents the total number of trees in the model, Represents XGBoost machine learning model parameters;

[0062] The expression of the regularization term of the XGBoost machine learning model is as follows:

[0063]

[0064] in, represents the regularization penalty term of a single decision tree, Indicates the number of leaf nodes, represents the leaf node weight, j Represents the index of the leaf node, Indicates the penalty coefficient for the number of leaf nodes.

[0065] The beneficial effect of the above-mentioned further scheme is that through the deep integration of mathematical constraints and algorithmic characteristics, the XGBoost model can ensure prediction accuracy while combining professional rationality and engineering practicality, providing water quality management with an intelligent prediction tool that has both statistical significance and physical interpretability.

[0066] Furthermore, the S5 includes the following steps:

[0067] S501. Use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions.

[0068] S502. Based on the prediction result of S501 and the set warning threshold, trigger the warning mechanism if the safety standard is exceeded;

[0069] S503. Based on the trigger warning mechanism of S502, emergency response measures and decision support are initiated to issue warnings for concentrations exceeding the standard and complete the prediction and warning of the pollution intensity of the river during the flood season.

[0070] The beneficial effect of this further approach is that, by establishing a "forecast-warning-response" decision-making system, it enables proactive identification and precise control of pollution risks during flood season in the river basin. This approach, based on a trained XGBoost model, dynamically analyzes the nonlinear relationship between meteorological factors and pollutant concentrations, outputting pollution intensity forecasts with high spatiotemporal resolution (e.g., hourly resolution), thus overcoming the lag bottleneck of traditional monitoring methods. By setting warning thresholds and automatically triggering alerts based on real-time forecasts, this approach improves emergency response efficiency from "post-event disposal" to "pre-event intervention."

[0071] The present invention provides a river flood season pollution intensity prediction and early warning system, which is used to implement a river flood season pollution intensity prediction and early warning method, including:

[0072] The first processing module is used to obtain multi-source data in the watershed and pre-process the multi-source data;

[0073] The second processing module is used to construct a hydrological and water quality model that calculates the hydrological cycle process of the entire basin from upstream to downstream;

[0074] The third processing module is used to calculate the pollution load of point sources and non-point sources using the constructed hydrological and water quality model to obtain water quality concentration;

[0075] The fourth processing module is used to construct a prediction data set based on the water quality concentration and use the prediction data set to train the XGBoost machine learning model;

[0076] The fifth processing module is used to use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions and issue warnings for concentrations exceeding the standard, thereby completing the prediction and warning of river pollution intensity during the flood season.

[0077] Beneficial effects of the present invention: The present invention provides a method for predicting and warning of river pollution intensity during flood season based on a combination of mechanism models and machine learning. The present invention first simulates hydrological and water quality changes at different temporal and spatial scales within a river basin based on an arbitrary hydrological and water quality mechanism model. It then uses the machine learning XGBoost model to predict pollutant concentrations during the flood season under different meteorological conditions and provide early warnings for concentrations exceeding the standard. The present invention utilizes refined hydrological models and advanced machine learning techniques to overcome the shortcomings of traditional methods and accurately predict pollution intensity during the flood season in real time, thereby providing effective early warning support for decision makers, enabling them to take control measures in advance and ensure water environmental safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 Flow chart of the method of the present invention.

[0079] Figure 2This is the technical roadmap of the present invention.

[0080] Figure 3 It is a schematic diagram of the hydrological cycle structure of the hydrological and water quality model of the present invention.

[0081] Figure 4 Schematic diagram of the division of the hydrological response unit HRU in this embodiment.

[0082] Figure 5 Schematic diagram of the calibration and verification results of the hydrological and water quality model flow and water quality concentration.

[0083] Figure 6 Schematic diagram of water quality concentration training and verification results of the XGBoost machine learning model.

[0084] Figure 7 This is a schematic diagram of the flood season pollution intensity prediction and early warning system in this embodiment.

[0085] Figure 8 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0086] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0087] Example 1

[0088] like Figure 1-Figure 2 As shown, the present invention provides a method for predicting and warning of river pollution intensity during flood season, and its implementation method is as follows:

[0089] S1. Obtain multi-source data within the watershed and pre-process the multi-source data. The implementation method is as follows:

[0090] S101, obtaining multi-source data within the watershed, and performing cleaning and correction processing on the multi-source data;

[0091] S102, identifying and removing outliers from the multi-source data processed in S101 by using statistical analysis methods, and filling in missing values;

[0092] S103. Based on the multi-source data processed by S102, for the discontinuous time series of the meteorological data and water quality monitoring data, a linear interpolation method is used to insert several new time points between two adjacent dates, and the meteorological data values ​​and water quality concentrations at the new time points are estimated using the trends of the existing data points;

[0093] S104: Based on the data processed in S103, the multi-source data in the watershed are integrated to form a standard data set, wherein the data at each time point in the standard data set includes multi-dimensional information, thereby completing the preprocessing of the multi-source data.

[0094] In this embodiment, data collection and preprocessing: Data collection and preprocessing are a crucial part of the present invention, aiming to provide high-quality data support for subsequent hydrological and water quality models and machine learning models.

[0095] Data collection involves gathering relevant basin-wide data from multiple sources, including meteorological data (such as precipitation, temperature, humidity, solar radiation, and wind speed), hydrological data (such as river flow, evaporation, and groundwater levels), water quality monitoring data (such as the permanganate index and total phosphorus concentration), and pollution source data (such as industrial emissions, agricultural non-point source pollution, and urban non-point source pollution). Data typically comes from various monitoring stations, meteorological observatories, and remote sensing imagery. Because the collected data may contain missing data, noisy data, or outliers, effective cleaning and correction are required during the data preprocessing stage. Outliers are identified and removed using statistical analysis methods (such as the Z-Score or IQR method), and missing values ​​are filled using linear interpolation or machine learning methods (such as KNN interpolation) to ensure data integrity and consistency.

[0096] The time series of data may also be discontinuous, especially meteorological data and water quality monitoring data may not be recorded at every time point. To solve this problem, the linear interpolation method is used to insert several new time points between two adjacent dates, and the trends of the existing data points are used to infer the meteorological data values ​​and water quality concentrations at these new time points. In this way, not only are the data gaps filled, but the temporal continuity and accuracy of the data are also ensured. In order to avoid the scale differences between different data sources affecting the effect of the machine learning model, all data are standardized or normalized, usually using Z-Score normalization or Min-Max normalization, so that the range of each feature is the same, thereby improving the training efficiency and stability of the model.

[0097] During the data fusion process, all collected data is integrated to form a unified dataset. The data at each time point contains information from multiple dimensions, such as meteorological data, flow rate, and pollution source data. This data will serve as input features for subsequent hydrological and water quality simulations and machine learning model training. To further improve the model's accuracy and generalization capabilities, feature engineering methods are used to extract meaningful features from the raw data. For example, aggregated data at different time scales (such as daily, monthly, and quarterly average precipitation and temperature) is extracted. Through lagged feature construction, pollutant concentrations or meteorological data from the past few days are used as input for the current forecast to reflect the delayed effects of pollutants. These data preprocessing steps ensure the quality and consistency of the input data, providing strong support for the machine learning model and improving the model's accuracy and generalization capabilities in subsequent forecasts.

[0098] S2. Construct a hydrological and water quality model to calculate the hydrological cycle process of the entire basin from upstream to downstream;

[0099] In this embodiment, the construction and integration of the hydrological and water quality model uses geospatial information technology to comprehensively consider the input of atmospheric precipitation and irrigation to the soil in the hydrological cycle, as well as the output of evaporation, infiltration, and runoff from the soil. The model includes calculations of the atmosphere, soil, rivers, reservoirs, and management measures. The hydrological cycle structure of the hydrological and water quality model is as follows: Figure 3 As shown in the figure, the hydrological cycle of the entire basin from upstream to downstream is calculated. The principles of the water cycle process can be divided into four categories, including atmospheric water process, soil water process, groundwater process and surface water process:

[0100] (1) Atmospheric water processes are mainly divided into rainfall and snowfall. The snowfall process will subsequently affect the water cycle through snow accumulation and snowmelt.

[0101] (2) The soil water process is divided into six processes: infiltration, surface evaporation, vegetation transpiration, soil flow, deep infiltration and irrigation, which follow the soil water balance formula:

[0102]

[0103] in, Indicates the soil water storage on that day, in mm; Indicates the soil water storage on the previous day, in mm; Indicates the amount of irrigation water, in mm; Indicates the infiltration rate in mm; Indicates soil evaporation, in mm; Indicates vegetation transpiration, in mm; Indicates the amount of deep leakage, in mm; Indicates soil flow, unit is mm.

[0104] (3) The hydrological and water quality model has the following characteristics in groundwater simulation: ① Groundwater is divided into shallow groundwater and deep groundwater; ② Only water balance is simulated, not groundwater level; ③ The groundwater between sub-basins is independent of each other, and the influence of lateral runoff is not considered.

[0105] Among them, the main processes of shallow groundwater include recharge, torrent, phreatic evaporation, irrigation and consumption, which follow the shallow groundwater balance formula:

[0106]

[0107] in, Indicates the shallow groundwater storage capacity on that day, in mm; Indicates the shallow groundwater storage capacity of the previous day, in mm; Indicates shallow recharge, in mm; Indicates the base flow generation, in mm; Indicates the amount of evaporation from water, in mm; Indicates shallow usage (irrigation / consumption) in mm.

[0108] The main processes of deep groundwater are divided into three processes: recharge, irrigation and consumption, following the deep groundwater balance formula:

[0109]

[0110] in, Indicates the deep groundwater storage capacity on that day, in mm; Indicates the deep groundwater storage capacity of the previous day, in mm; Indicates the deep recharge volume in mm; Indicates deep usage (irrigation / consumption) in mm.

[0111] (4) The surface water process of the hydrological and water quality model consists of three parts: ① the sub-channel transport process; ② the storage process of reservoirs, ponds, wetlands, and depressions; and ③ the water volume evolution process of the main channel. The sources of surface water include: ① surface runoff generated by the hydrological response unit; ② subsurface flow generated by the hydrological response unit; and ③ base flow generated by the shallow groundwater in the sub-basin.

[0112] There are three pathways for surface water transport in a sub-basin: ① Surface runoff from the hydrological response unit first passes through the sub-channel, resulting in transport losses; ② Part of the remaining surface runoff, interbedded flow, and groundwater baseflow enters the sub-basin's ponds / wetlands, and part enters the sub-basin's main channel; and ③ Outflow from the ponds / wetlands enters the sub-basin's main channel.

[0113] The storage of ponds / wetlands includes inflow, rainfall and evaporation, infiltration, outflow, and consumption (ponds only). Each process follows the pond / wetland water balance formula:

[0114]

[0115] in, Indicates the pond / wetland storage capacity on that day, in m³; Indicates the pond / wetland storage capacity of the previous day, in m³; Indicates the inflow rate in m³; Indicates the flow rate in m³; Indicates rainfall in m³; Indicates evaporation capacity in m³; Indicates leakage volume in m³.

[0116] The evolution of water volume in a major river includes upstream inflow, evaporation from the river surface, riverbank storage, seepage, downstream outflow, consumption, and irrigation. Each process follows the water balance formula for the major river:

[0117]

[0118] in, Indicates the storage capacity of the main river channel on that day, in m³; Indicates the storage volume of the main river channel on the previous day, in m³; Indicates the inflow of the upstream river, in m³; Indicates the downstream outflow, m³; Indicates the riverbank storage volume, in m³; Indicates irrigation and consumption, in m³; Indicates the evaporation of river water surface, in m³; Indicates leakage volume in m³.

[0119] In this embodiment, hydrological mechanism simulation: the hydrological process and pollutant migration and transformation process of the basin are modeled using a hydrological and water quality mechanism model. Assume that the basic expression of the hydrological and water quality model is as follows:

[0120]

[0121] in, represents the soil water storage, Indicates time, Indicates precipitation, Indicates the evaporation rate, Indicates the infiltration rate, represents the amount of surface runoff. This equation describes the changes in water content during the hydrological cycle, taking into account different input and output processes.

[0122] In this embodiment, the simulation of the pollution load during the flood season: the simulation of pollutants is also based on the hydrological and water quality model, and the pollution load model is used for simulation. The common water quality concentration model can be expressed as:

[0123]

[0124] in, Indicates time The pollutant concentration at the time, represents the initial concentration, represents the pollutant degradation rate, Represents the emission of pollutants, and the model simulates the changes of pollutants over time and their degradation process.

[0125] S3. Use the constructed hydrological and water quality model to calculate the pollution load of point sources and non-point sources to obtain the water quality concentration. The implementation method is as follows:

[0126] S301. Based on the hydrological response unit, the hydrological and water quality model is used to simulate the hydrological process, and the generation, transformation and migration of water pollutants are considered;

[0127] S302. Based on the changes in pollutant concentration and inflow within each hydrological response unit (HRU), the amount of pollutants entering the river from each HRU is calculated using a hydrological and water quality model. For point source pollution, the pollutant discharge is calculated based on the point source flow and concentration. For non-point source pollution, the amount of pollutants entering the river from each HRU is calculated based on soil erosion, runoff, and pollution load.

[0128] S303: Calculate the water quality concentration based on the amount of pollutants entering the river in each hydrological response unit (HRU), and complete the calculation of point source and non-point source pollution loads.

[0129] In this example, point and non-point source pollution loads are calculated in a refined manner. Based on hydrological response units (HRUs), a hydrological and water quality mechanism simulation model is constructed to simulate the hydrological processes and pollutant migration and transformation processes of each HRU under different meteorological conditions. Through the model's hydrological cycle and pollutant load calculations, the amount of pollutants entering the river within each HRU (i.e., HRU) is determined. The specific steps are as follows:

[0130] ① Hydrological and water quality model input: Based on the hydrological response unit, the hydrological and water quality model is used to simulate hydrological processes such as precipitation, evaporation, runoff, and infiltration, and consider the generation, transformation, and migration of water pollutants. The pollutant load and inflow into the river are calculated using the following formula:

[0131]

[0132] in, Indicates the i The pollutant load of each hydrological response unit HRU, N Indicates the number of pollution sources in the hydrological response unit HRU, which includes point source pollution and non-point source pollution. represents the area of ​​the hydrological response unit HRU, Indicates that it comes from j Pollutant concentrations from various pollution sources;

[0133] ② Calculation of pollutant load into the river: Based on the changes in pollutant concentration and flow within each hydrological response unit (HRU), the hydrological and water quality model calculates the amount of pollutants entering the river within each HRU. For point source pollution (such as industrial enterprises and sewage treatment plants), pollutant emissions are calculated based on point source flow and concentration. For non-point source pollution (such as agricultural non-point source pollution and urban non-point source pollution), the amount of pollutants entering the river for each HRU is calculated based on soil erosion, runoff, and pollution load. The specific formula for pollutant inflow into the river is as follows:

[0134]

[0135] in, It represents the total amount of pollutants entering the river in the entire basin, It represents the accumulation of all hydrological response units HRU;

[0136] ③ Flood Season Pollution Inflow Data Processing: Calculated pollutant inflow data is processed and used to simulate pollution concentrations in the basin's waters, further assessing trends in pollution intensity during the flood season. Specifically, these pollutant inflows serve as input for the flow-water quality simulation phase. Combined with hydrological and water quality models, these data are used to predict water quality changes within the basin. This data also provides data support for machine learning models, enabling highly accurate predictions of pollution intensity during the flood season.

[0137] The hydrological and water quality model simulates the flow rate by simulating rainfall infiltration, evapotranspiration, soil moisture, groundwater, and runoff for each hydrological response unit using imported meteorological data (daily accumulated rainfall, daily maximum temperature, daily minimum temperature, daily relative humidity, daily average solar radiation, and daily average wind speed). At the same time, pollutant inputs, including point source and non-point source pollution, are set to simulate the entire process from pollutant generation to pollutant entry into the river. The polluted river inflow is calculated using the following formula to obtain water quality data. The model is then calibrated and verified based on the measured data to ensure that the simulation accuracy meets the model requirements:

[0138]

[0139] in, Indicates water quality concentration in mg / L. Indicates the pollution load in kg / d. Indicates sewage flow rate in m³ / d. Indicates the time conversion factor.

[0140] S4. Based on the water quality concentration, a prediction dataset is constructed and used to train the XGBoost machine learning model. The implementation method is as follows:

[0141] S401. Use six meteorological data items and Julian dates from historical years as input variables for the XGBoost machine learning model to construct prediction datasets for permanganate index and total phosphorus concentration, respectively.

[0142] S402, performing linear interpolation processing on the six meteorological data items and the Julian date, inserting an average of three new time points between two adjacent Julian dates, and performing linear interpolation processing on each new time point;

[0143] S403, adding the time series data obtained after the linear interpolation process and its corresponding meteorological data, permanganate index, and total phosphorus concentration data to the constructed permanganate index and total phosphorus concentration prediction data set to form a new data set, and using the new data set as the prediction data set;

[0144] S404: Divide the prediction data set obtained in S403 into a training set and a validation set, and use the training set and the validation set to train and validate the XGBoost machine learning model.

[0145] In this embodiment, six meteorological data items and the Julian date of historical years are used as predictive variables for the XGBoost machine learning model to construct prediction datasets for the permanganate index and TP concentration, respectively. To further expand the data volume of the machine learning model, linear interpolation is performed on the Julian date and the six meteorological data items. An average of three new time points are inserted between two adjacent Julian dates. For each new time point, a linear interpolation method is used to obtain the six meteorological data values ​​and the permanganate index and TP concentration. The interpolated time series data and its corresponding meteorological data and water quality concentration are added to the original dataset to form a new dataset as a prediction dataset for the permanganate index and TP concentration. The prediction dataset is divided into a training set and a validation set in proportion, which are used to train and validate the machine learning model, respectively. Among them, 80% of the dataset is used as the training set and 20% as the validation set. Finally, based on the six meteorological data items for the short term in the future published in the weather forecast, the verified XGBoost machine learning model can obtain the permanganate index, total phosphorus concentration, and pollution intensity values ​​of the control section during the flood season.

[0146] To improve the predictive power of the XGBoost machine learning model and prevent overfitting, this paper uses the objective function of the XGBoost machine learning model to optimize the model's performance. The objective function of the XGBoost machine learning model typically includes a loss function and a regularization term, and its goal is to minimize the following expression:

[0147]

[0148] The loss function is usually in the form of mean square error (MSE) or log loss (Log Loss). The specific choice depends on the nature of the problem and the distribution of the data. In this model, the loss function is:

[0149]

[0150] in, represents the loss function of the XGBoost machine learning model, i express i samples, n represents the total number of samples, Represents the predicted value and the true value The error between represents the regularization parameter, which is used to balance the loss function and the regularization term. k Indicates the k A decision tree, K represents the total number of trees in the model, Represents XGBoost machine learning model parameters;

[0151] The regularization term is used to control the complexity of the model and prevent overfitting:

[0152]

[0153] in, represents the regularization penalty term of a single decision tree, Indicates the number of leaf nodes, represents the leaf node weight, j Represents the index of the leaf node, Indicates the penalty coefficient for the number of leaf nodes.

[0154] The regularization term helps adjust the complexity of the model so that the model can maintain good generalization ability while ensuring accuracy.

[0155] Finally, the model optimizes the objective function through gradient boosting, that is, by iteratively training a series of decision trees, gradually correcting the errors in the previous round of predictions, optimizing the objective function value, and thus achieving higher prediction accuracy.

[0156] In this example, the adaptive learning characteristics of the XGBoost machine learning model enable the model to dynamically adjust to pollution characteristics in different regions, seasons, and meteorological conditions. By validating training sets across different regions and seasons, the model can continuously optimize its predictive capabilities, increasing its adaptability and flexibility in a changing environment.

[0157] S5. Use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions and issue warnings for concentrations exceeding the standard, thereby completing the prediction and warning of river pollution intensity during the flood season. The implementation method is as follows:

[0158] S501. Use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions.

[0159] S502. Based on the prediction result of S501 and the set warning threshold, trigger the warning mechanism if the safety standard is exceeded;

[0160] S503. Based on the trigger warning mechanism of S502, emergency response measures and decision support are initiated to issue warnings for concentrations exceeding the standard and complete the prediction and warning of the pollution intensity of the river during the flood season.

[0161] In this embodiment, risk warning and decision support are key applications of the present invention. They aim to accurately predict changes in pollution intensity during flood seasons, providing timely and scientific decision-making for environmental protection and water quality management departments. After data collection, preprocessing, simulation, and prediction modeling are completed, the resulting flood season pollution intensity forecasts serve directly as input to the decision support system, providing early warning information to relevant departments. By analyzing historical data, real-time monitoring data, and future weather forecasts, combined with the output of machine learning models, the system accurately assesses changing trends in pollutant concentrations during flood seasons and promptly identifies potential pollution risks.

[0162] ① The risk warning process begins by automatically determining whether pollution intensity data predicted by the model, combined with the set warning threshold, exceeds safety standards, triggering the warning mechanism. For example, when the predicted pollution intensity reaches or exceeds the set warning threshold, the system generates a warning report and promptly transmits the warning information to relevant management departments via an automated platform. These warning reports include key information such as predicted pollutant concentrations, the distribution of possible pollution sources, and the warning level, helping decision-makers to take timely response measures, such as strengthening pollution source supervision, initiating emergency response, or issuing public warning information.

[0163] In practical applications, the risk warning system not only relies on historical and real-time data inputs but also dynamically adjusts based on the latest weather forecasts (such as future rainfall and temperature changes). For example, if the weather forecast predicts heavy rainfall in the coming days, the system automatically adjusts the parameters of the warning model to account for the impact of precipitation on pollutant runoff, thereby improving the accuracy and timeliness of the forecast. This adaptive forecasting and adjustment capability enables the early warning system of the present invention to maintain efficient operation under dynamically changing environmental conditions and respond promptly to changing risk factors.

[0164] ② In terms of decision support, this invention not only provides regular pollution intensity forecasts but also offers emergency response recommendations and decision support based on real-time data and forecast results. By combining early warning information with the spatial distribution of pollution sources in the river basin and historical emergency response measures, the system can propose optimized emergency plans and pollution control measures. For example, in areas with high early warning levels, the system may recommend that relevant departments strengthen pollution source monitoring, enhance the operation of water purification facilities, and allocate emergency water quality monitoring resources.

[0165] Furthermore, the risk warning and decision support system can use intelligent analysis to predict the effects of different emergency response plans, helping decision-makers assess the effectiveness of different management measures. By simulating the effects of different policies, decision-makers can select the response measures that best suit the current pollution and resource conditions in the river basin, thereby improving pollution prevention and control efficiency and reducing potential ecological risks and public health hazards.

[0166] By combining high-precision prediction models, real-time data monitoring and weather forecasts, this invention provides strong decision-making support for watershed pollution management. It not only improves the timeliness and accuracy of early warnings, but also provides scientific emergency response recommendations for environmental protection departments, thereby effectively reducing the risk of pollution incidents and ensuring the safety of water resources and public health.

[0167] In summary, this invention utilizes hydrological and water quality models to simulate hydrological elements and pollution loads at river basin outlets, addressing the lack of data for many river basins and providing big data support for machine learning models. This simulation process, based on the integration of hydrological and water quality models with machine learning models, can address the data shortages of traditional mechanistic models and more quickly and accurately predict changes in flood season pollution intensity. In addition to implementing the aforementioned functions, this model can integrate with any common hydrological and water quality models, offering broad adaptability and flexibility. Specifically, this invention supports the following common hydrological and water quality models: the SWAT model (Soil and Water Assessment Tool), the HSPF model (Hydrological Simulation Program—Fortran), the HEC-HMS model (Hydrologic Engineering Center—Hydrologic Modeling System), and the MIKE11 model. This model can be integrated with different hydrological and water quality models, leveraging the strengths of each model to select the most appropriate model for simulating hydrological and water quality processes based on specific needs. Through this integration, the model can adapt to the characteristics of different river basins and provide more accurate flood season pollution intensity forecasts and early warnings.

[0168] The present invention will now be described in further detail with reference to the accompanying drawings.

[0169] 1) Data Processing and Integration: This approach aims to provide high-quality input data for hydrological and water quality models, ensuring high compatibility and flexibility. The system first collects and organizes various data sources within the basin, including digital elevation data, land use data, soil data, meteorological data, pollution source data, and water quality monitoring data. All data undergoes standardization and preprocessing, including interpolation of missing data and correction of outliers, to ensure data integrity and consistency. Throughout this process, machine learning and data processing methods are used to ensure the high quality of the raw data, laying the foundation for subsequent model calculations and predictions. To ensure data compatibility with a variety of common hydrological and water quality models, the present invention unifies and standardizes all data formats, enabling seamless integration with different models such as SWAT, HSPF, and HEC-HMS, ensuring efficient data transfer between these models. Through this data integration approach, data from all sources is merged into a unified dataset that can not only be directly input into various hydrological and water quality models but also provides high-quality training data for machine learning models, further improving prediction accuracy. Through this data integration and processing process, the system has high compatibility with existing mainstream hydrological and water quality models, enabling it to flexibly respond to changes in different river basins and pollution conditions, and improving the system's applicability and flexibility.

[0170] 2) Fine calculation of pollution load discharge of point-area source superposition: The hydrological and water quality model is used to simulate the discharge of pollution control factors in a fine manner to obtain the pollutant generation amount of each hydrological response unit HRU. The division results are as follows: Figure 4 Point sources include industrial pollution, urban domestic sewage, and large-scale livestock and poultry farming. Additional non-point sources include urban non-point sources, rural domestic sewage, and aquaculture pollution. The calculated pollution loads from point-non-point sources entering the river are input into the hydrological and water quality mechanism model as point sources and additional non-point sources, respectively. Point source pollution loads correspond to the Point Source Discharges interface for statistical input, while non-point source pollution loads correspond to management measures in the hydrological and water quality mechanism model and are input into the model in the form of agricultural fertilizers.

[0171] 3) Flow-water quality simulation: The hydrological and water quality change process is simulated using the hydrological and water quality model, and the model is calibrated based on the measured data to achieve the best simulation results. For the simulation of flow and pollution load by the hydrological and water quality model, the Nash efficiency coefficient (NSE) and the certainty coefficient (R 2 ) to evaluate the model calibration effect. 2 >0.60, NSE>0.50, the model flow simulation results are considered "satisfactory"; when R 2 >0.40, NSE>0.35, and the model pollution load simulation results are considered “satisfactory”. Figure 5 The simulated values ​​of flow and water quality concentration obtained by the hydrological and water quality mechanism model and their actual measured values ​​are displayed.

[0172] 4) Construction of a flood season pollution intensity prediction model based on the XGBoost machine learning model: including dataset construction and segmentation, model training, and model performance evaluation. The Julian date and six meteorological data (including daily cumulative rainfall, daily average radiation value, daily maximum temperature, daily minimum temperature, daily average relative humidity, and daily average wind speed) are matched one by one. In order to further expand the data volume of the machine learning model, the Julian date and the six meteorological data are linearly interpolated. An average of three new time points are inserted between two adjacent Julian dates. For each new time point, the linear interpolation method is used to obtain the six meteorological data values, permanganate index, and TP concentration. Specifically, for each pair of adjacent Julian dates and , and the meteorological data or water quality concentration corresponding to that moment and , use the following formula to calculate the interpolated time point Corresponding meteorological data or water quality concentration :

[0173]

[0174] in, Indicates the time point after linear interpolation The corresponding meteorological data value, Julian date The corresponding meteorological data value, Julian date The corresponding meteorological data value.

[0175] The interpolated time series data and its corresponding meteorological data and water quality concentration are added to the original data set to form a new data set as the prediction data set for permanganate index and TP concentration. The data set is divided into two parts, 80% of the data is used as the training set, and 20% of the data is used as the validation set. Secondly, when using the training set data to train the model, the optimal parameter combination is obtained by manual parameter adjustment. The optimal parameters are: learning rate (Eta) is 0.1536, regularization parameter (Gamma) is 0.001, maximum depth of the tree (Max_Depth) is 6, sample sampling rate of each tree (Subsample) is 0.91, feature sampling rate (Colsample_Bytree) is 0.48, number of boosting iterations (Nrounds) is 300, number of early stopping iterations (Early_Stopping_Rounds) is 7, and the remaining parameters are set to the default values ​​in the model; finally, based on the test set data, linear regression is used to evaluate the performance of the flood season pollution intensity prediction model for permanganate index and TP concentration. The results are as follows: Figure 6 shown.

[0176] Based on the six meteorological data in the short term published in the weather forecast, the verified XGBoost model can be used to obtain the permanganate index, total phosphorus concentration and pollution intensity values ​​of the control section during the flood season.

[0177] 5) Prediction and Early Warning System: Based on the calibrated and verified XGBoost machine learning model, this paper designs an automated prediction and early warning system to accurately predict water quality information and assess the pollution intensity during flood season. Figure 7As shown in the figure, this system not only predicts the permanganate index and total phosphorus concentration, but also analyzes and assesses water quality trends in the basin in real time, thereby predicting dynamic changes in pollution intensity. The system receives historical monitoring data from users (such as the permanganate index and total phosphorus concentration) and combines it with meteorological data from weather forecast services (including key meteorological parameters such as precipitation, temperature, humidity, and wind speed) as input variables. It then uses the XGBoost machine learning model to generate high-precision predictions. During system operation, based on the real-time input data and prediction results, the XGBoost machine learning model dynamically assesses pollution intensity during the upcoming flood season. When the system detects that the predicted pollution intensity reaches or exceeds the preset safety threshold, an early warning mechanism is automatically triggered. This mechanism quickly identifies potential pollution risks and sends real-time warning notifications to relevant managers or decision-makers through the system interface, providing pollution intensity forecast data and response recommendations. This process ensures that targeted emergency measures can be implemented promptly when pollution intensity exceeds the standard, effectively ensuring water environmental safety and timely water quality monitoring.

[0178] The system's core innovation lies in its highly adaptable and accurate water quality predictions, achieved through the integration of multi-source data (such as weather forecasts, historical water quality data, and pollution source data) and machine learning. By leveraging machine learning algorithms to analyze water quality trends in real time, the system can flexibly respond to sudden pollution risks in different river basins and under varying meteorological conditions, providing accurate forecasting support and decision-making for water resource management.

[0179] Example 2

[0180] like Figure 8 As shown, the present invention provides a river flood season pollution intensity prediction and warning system, characterized in that the river flood season pollution intensity prediction and warning system is used to execute the river flood season pollution intensity prediction and warning method described in Example 1, including:

[0181] The first processing module is used to obtain multi-source data in the watershed and pre-process the multi-source data;

[0182] The second processing module is used to construct a hydrological and water quality model that calculates the hydrological cycle process of the entire basin from upstream to downstream;

[0183] The third processing module is used to calculate the pollution load of point sources and non-point sources using the constructed hydrological and water quality model to obtain water quality concentration;

[0184] The fourth processing module is used to construct a prediction data set based on the water quality concentration and use the prediction data set to train the XGBoost machine learning model;

[0185] The fifth processing module is used to use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions and issue warnings for concentrations exceeding the standard, thereby completing the prediction and warning of river pollution intensity during the flood season.

[0186] like Figure 8 As shown, the river flood season pollution intensity prediction and warning system provided in the embodiment can execute the technical solution shown in the river flood season pollution intensity prediction and warning method of the above method embodiment. Its implementation principle and beneficial effects are similar and will not be repeated here.

[0187] In this embodiment, the present application can divide functional units based on the method for predicting and warning river flood season pollution intensity. For example, each function can be divided into functional units, or two or more functions can be integrated into a single processing unit. The above-mentioned integrated units can be implemented in the form of hardware or software functional units. It should be noted that the division of units in the present invention is schematic and is only a logical division. In actual implementation, other division methods may be used.

[0188] In this embodiment, the river flood season pollution intensity prediction and warning system includes hardware structures and / or software modules that perform corresponding functions in order to realize the principles and beneficial effects of the river flood season pollution intensity prediction and warning method. It should be readily appreciated by those skilled in the art that, in combination with the various schematic units and algorithm steps described in the embodiments disclosed herein, the present invention can be implemented in the form of hardware and / or a combination of hardware and computer software. Whether a function is executed in a hardware- or computer software-driven manner depends on the specific application and design constraints of the technical solution. Different methods can be used for each specific application to implement the described function, but such implementation should not be considered to be beyond the scope of this application.

Claims

1. A method for predicting and warning river pollution intensity during flood season, characterized in that: The following steps are involved: S1. Obtain multi-source data in the basin and pre-process the multi-source data; S2. Construct a hydrological and water quality model to calculate the hydrological cycle process of the entire basin from upstream to downstream; S3. Using the constructed hydrological and water quality model, calculate the pollution load of point sources and non-point sources to obtain water quality concentration; The S3 is specifically: S301. Based on the hydrological response unit (HRU), a hydrological and water quality model is used to simulate the hydrological process, taking into account the generation, transformation, and migration of water pollutants. The pollutant load and inflow rate are calculated using the following formula: in, Indicates the i The pollutant load of each hydrological response unit HRU, N Indicates the number of pollution sources in the hydrological response unit HRU, which includes point source pollution and non-point source pollution. represents the area of ​​the hydrological response unit HRU, Indicates that it comes from j Pollutant concentrations from various pollution sources; S302. Based on the changes in pollutant concentration and inflow within each hydrological response unit (HRU), the amount of pollutants entering the river within each HRU is calculated using a hydrological and water quality model. For point source pollution, the pollutant discharge is calculated based on the point source flow and concentration. For non-point source pollution, the amount of pollutants entering the river for each HRU is calculated based on soil erosion, runoff, and pollution load: in, It represents the total amount of pollutants entering the river in the entire basin, It represents the accumulation of all hydrological response units HRU; S303. Calculate the water quality concentration based on the amount of pollutants entering the river in each hydrological response unit (HRU), and complete the calculation of point source and non-point source pollution loads: in, Indicates water concentration, represents the pollution load, represents the sewage flow rate, Indicates the time conversion factor; S4. Build a prediction dataset based on water quality concentration and use the prediction dataset to train the XGBoost machine learning model. The S4 is specifically: S401. Using six meteorological data items and Julian dates from historical years as input variables of the XGBoost machine learning model, a permanganate index and total phosphorus concentration prediction dataset was constructed, respectively. The six meteorological data items included daily cumulative rainfall, daily average radiation value, daily maximum temperature, daily minimum temperature, daily average relative humidity, and daily average wind speed. S402, performing linear interpolation processing on the six meteorological data items and the Julian date, inserting an average of three new time points between two adjacent Julian dates, and performing linear interpolation processing on each new time point; S403, adding the time series data obtained after the linear interpolation process and its corresponding meteorological data, permanganate index, and total phosphorus concentration data to the constructed permanganate index and total phosphorus concentration prediction data set to form a new data set, and using the new data set as the prediction data set; S404, dividing the prediction data set obtained in S403 into a training set and a validation set, and using the training set and the validation set to train and validate the XGBoost machine learning model; S5. Use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions and issue warnings for concentrations exceeding the standard, thereby completing the prediction and warning of river pollution intensity during the flood season.

2. The method for predicting and warning river pollution intensity during flood season according to claim 1, characterized in that: The S1 is specifically: S101, obtaining multi-source data within the watershed, and performing cleaning and correction processing on the multi-source data; S102, identifying and removing outliers from the multi-source data processed in S101 by using statistical analysis methods, and filling in missing values; S103. Based on the multi-source data processed by S102, for the discontinuous time series of the meteorological data and water quality monitoring data, a linear interpolation method is used to insert several new time points between two adjacent dates, and the meteorological data values ​​and water quality concentrations at the new time points are estimated using the trends of the existing data points; S104: Based on the data processed in S103, the multi-source data in the watershed are integrated to form a standard data set, wherein the data at each time point in the standard data set includes multi-dimensional information, thereby completing the preprocessing of the multi-source data.

3. The method for predicting and warning river pollution intensity during flood season according to claim 1, characterized in that: The hydrological cycle in S2 includes atmospheric water process, soil water process, groundwater process and surface water process. The soil water balance formula followed by the soil water process is as follows: in, Indicates the soil water storage on that day. represents the soil water storage on the previous day, Indicates the amount of irrigation water, Indicates the infiltration rate, represents soil evaporation, represents the transpiration of vegetation, Indicates the amount of deep leakage, It means flow in soil; Groundwater processes include shallow groundwater processes and deep groundwater processes. The shallow groundwater process follows the shallow groundwater balance formula as follows: in, Indicates the shallow groundwater storage volume on that day, represents the shallow groundwater storage volume of the previous day, represents the shallow recharge, represents the base flow generation, represents the amount of evaporation from diving, Indicates shallow usage; The deep groundwater balance formula followed by the deep groundwater process is as follows: in, Indicates the amount of deep groundwater storage on that day, represents the amount of deep groundwater storage on the previous day, Indicates the deep recharge amount, Indicates deep usage; The pond / wetland water balance equation for surface water storage is as follows: in, Indicates the amount of pond / wetland storage on that day, Indicates the amount of pond / wetland storage on the previous day. represents the inflow rate, Indicates the flow rate, Indicates rainfall, Indicates the evaporation rate, Indicates the amount of leakage; The evolution of water volume in the main river during the surface water process follows the main river water balance formula as follows: in, Indicates the storage capacity of the main river channel on that day, Indicates the storage volume of the main river channel on the previous day. represents the inflow of the upstream river, Indicates the downstream outflow, represents the riverbank storage volume, Indicates irrigation and consumption, Indicates the evaporation of river water surface, Indicates the amount of leakage.

4. The method for predicting and warning river pollution intensity during flood season according to claim 1, characterized in that: The expression of the hydrological and water quality model in S2 is as follows: in, represents the soil water storage, Indicates time, Indicates precipitation, Indicates the evaporation rate, Indicates the infiltration rate, Represents surface runoff.

5. The method for predicting and warning river pollution intensity during flood season according to claim 1, characterized in that: The loss function of the XGBoost machine learning model is expressed as follows: in, represents the loss function of the XGBoost machine learning model, i express i samples, n represents the total number of samples, Represents the predicted value and the true value The error between represents the regularization parameter, which is used to balance the loss function and the regularization term. k Indicates the k A decision tree, K represents the total number of trees in the model, Represents XGBoost machine learning model parameters; The expression of the regularization term of the XGBoost machine learning model is as follows: in, represents the regularization penalty term of a single decision tree, Indicates the number of leaf nodes, represents the leaf node weight, j Represents the index of the leaf node, Indicates the penalty coefficient for the number of leaf nodes.

6. The method for predicting and warning river pollution intensity during flood season according to claim 1, characterized in that: The S5 comprises the following steps: S501. Use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions. S502. Based on the prediction result of S501 and the set warning threshold, trigger the warning mechanism if the safety standard is exceeded; S503. Based on the trigger warning mechanism of S502, emergency response measures and decision support are initiated to issue warnings for concentrations exceeding the standard and complete the prediction and warning of the pollution intensity of the river during the flood season.

7. A river flood season pollution intensity prediction and early warning system, characterized by: The river flood season pollution intensity prediction and early warning system is used to implement the river flood season pollution intensity prediction and early warning method according to any one of claims 1 to 6, comprising: The first processing module is used to obtain multi-source data in the watershed and pre-process the multi-source data; The second processing module is used to construct a hydrological and water quality model that calculates the hydrological cycle process of the entire basin from upstream to downstream; The third processing module is used to calculate the pollution load of point sources and non-point sources using the constructed hydrological and water quality model to obtain water quality concentration; The fourth processing module is used to construct a prediction data set based on the water quality concentration and use the prediction data set to train the XGBoost machine learning model; The fifth processing module is used to use the trained XGBoost machine learning model to predict pollutant concentrations during the flood season under different meteorological conditions and issue warnings for concentrations exceeding the standard, thereby completing the prediction and warning of river pollution intensity during the flood season.

Images