Artificial intelligence-based flood infiltration purification system

Abstract

The application relates to the technical field of flood purification treatment, and discloses a flood infiltration purification system based on artificial intelligence. A flood data acquisition unit of the system acquires flood physical properties and chemical components in real time, extracts multi-dimensional water quality indexes, identifies abnormal pollution concentration areas through a neural network, and generates a pollution load distribution diagram of an infiltration medium; an infiltration layer state analysis unit calls high-load area data, identifies infiltration layer blockage and saturation segments in combination with historical infiltration efficiency, and generates an infiltration efficiency decay interval; a purification medium regulation unit extracts purification efficiency and regeneration states according to the decay interval, judges medium failure conditions, and generates a purification medium dynamic adjustment strategy; and a purification process execution unit calls regulation parameters, detects pressure changes and pollutant interception amounts, identifies interlayer synergistic action strength, and generates a purification medium backwashing instruction to control the start and stop of a high-pressure backwashing pump of a purification layer after reaching the standard.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of flood purification treatment, in particular to a flood infiltration purification system based on artificial intelligence. BACKGROUND

[0002] After a flood disaster, a large amount of floodwater carries various pollutants such as silt, organic matter, and heavy metals. If the floodwater is directly discharged or utilized, it is easy to cause damage to the surrounding water body and soil environment. Therefore, flood purification is an important part of post-disaster environmental restoration. In current mainstream flood purification technologies, infiltration purification is widely used due to its low cost and relatively simple operation. The core of infiltration purification is to achieve pollutant removal through physical interception, chemical adsorption, and biological degradation of the floodwater by multiple layers of infiltration media.

[0003] The existing infiltration purification system has obvious limitations in actual operation. The monitoring of the traditional system on the water quality of the floodwater is mostly limited to a single dimension or a fixed time point, making it difficult to obtain the dynamic changes of the physical properties and chemical components of the floodwater at the inflow end in real time, and unable to identify abnormal areas of pollution concentration in a timely manner. This leads to an unclear distribution of pollution load of the infiltration media, and some areas are blocked due to excessive accumulation of pollutants, while the purification capacity of other areas is not fully utilized, resulting in uneven overall purification efficiency.

[0004] The existing system lacks scientific basis for judging the state of the infiltration layer, and only relies on experience to determine whether the medium needs to be replaced or cleaned. It is difficult to accurately obtain the real-time changes of the medium porosity and water flow penetration rate, and to identify the blocked and saturated segments of the infiltration layer. Maintenance is often performed only after the medium completely fails, resulting in substandard effluent water quality during a significant decline in purification efficiency. At the same time, the purification medium control of the traditional system is mostly in a fixed mode and lacks dynamic adjustment capability. It is unable to develop priority strategies based on the decline rate of the purification efficiency of the medium in different areas. When some of the media fail, the entire system needs to be shut down for maintenance, affecting the continuity of purification.

[0005] The synergy of adjacent purification layers in the existing system is not taken into account in the regulation and control, the interlayer synergy strength cannot be detected, it is difficult to determine whether the optimal purification state is reached, the generation of backwashing instructions lacks accurate basis, and the situation of untimely backwashing or excessive backwashing often occurs, which not only wastes energy but also shortens the service life of the infiltration medium, further limiting the application effect of the infiltration purification system in flood treatment. The existing flood infiltration purification devices, such as post-disaster temporary water purification stations, riverway infiltration purification systems, and centralized flood treatment equipment in villages and towns, generally have problems such as uneven purification efficiency, lagging medium maintenance, and rigid regulation and control mode. The post-disaster temporary water purification station relies on fixed infiltration medium and cannot adapt to the dynamic changes of flood pollutants in real time, which easily leads to local blockage and water quality fluctuations; the riverway infiltration purification system lacks interlayer synergy monitoring and cannot accurately control the backwashing time, resulting in energy waste and medium loss; the centralized treatment equipment in villages and towns needs to replace the failed medium, which seriously affects the continuity of purification. The flood infiltration purification system based on artificial intelligence disclosed in the present application adapts to the above-mentioned existing devices, solves the core technical problems through intelligent monitoring, dynamic regulation and control, and synergy optimization, and improves the purification efficiency and operation stability of the existing devices. SUMMARY

[0006] The present application aims to provide a flood infiltration purification system based on artificial intelligence to solve the problems raised in the background art.

[0007] To achieve the above-mentioned purpose, the present application provides a flood infiltration purification system based on artificial intelligence, which comprises a flood inflow end, an infiltration layer and a purification layer, and further comprises:

[0008] The flood data acquisition unit obtains the physical properties and chemical components of the flood inflow end in real time, extracts multi-dimensional water quality indicators and suspended matter distribution states, identifies abnormal pollution concentration areas through a neural network model, and generates an infiltration medium pollution load distribution map;

[0009] The infiltration layer state analysis unit calls the high load area marked in the infiltration medium pollution load distribution map of the infiltration layer, obtains the medium porosity and water flow penetration rate of the corresponding area, identifies the blocked and saturated segments of the current infiltration layer in combination with historical infiltration efficiency data, and generates an infiltration efficiency decay interval;

[0010] The purification medium regulation and control unit extracts the water quality purification efficiency and medium regeneration state of the position in the continuous two purification cycles according to the physical position indicated by the infiltration efficiency decay interval, calculates the gap between the actual purification effect and the expected purification target, judges whether the medium is in a failed state, selects the purification efficiency decline rate of the failed medium area as the priority determination basis, and generates a dynamic adjustment strategy for the purification medium;

[0011] The purification process execution unit calls the control parameters set in the purification medium dynamic adjustment strategy, detects the pressure change of water flow through the medium and the pollutant interception amount in real-time purification process, and identifies the synergistic effect strength between adjacent purification layers. If the synergistic effect strength is higher than the system preset threshold, the synergistic purification readiness state is recorded, and a purification medium backwashing instruction is generated to control the start and stop of the high-pressure backwashing pump of the purification layer.

[0012] Preferably, the percolation medium pollution load distribution map includes pollution load level coding, medium blockage risk coefficient, and spatial distribution thermodynamic map. The percolation efficiency decay interval includes pore blockage rate, hydraulic breakthrough delay time length, and purification capacity decay gradient. The purification medium dynamic adjustment strategy includes medium replacement priority level, purification intensity adjustment amplitude, and control strategy effective time sequence. The purification medium backwashing instruction includes backwashing pressure set value, medium layer disturbance frequency, and water flow backwashing time length.

[0013] Preferably, the generation process of the percolation medium pollution load distribution map specifically includes:

[0014] The flood data acquisition unit monitors the flow rate and pollutant concentration of the flood inflow end in real time, extracts the settling characteristics of different time particulate matters and the diffusion trajectory of chemical substances, calculates the pollutant aggregation degree and medium adsorption saturation degree based on the deviation value of real-time monitoring data and the standard water quality model, and generates a set of pollution load distribution parameters;

[0015] Based on the pollutant aggregation degree and medium adsorption saturation degree in the set of pollution load distribution parameters, the physical structure parameters of the multi-layer percolation medium are combined to calculate the load rate of each medium unit under the current hydrological condition, identify the corresponding relationship between the load rate and the medium capacity limit, and generate a medium unit overload risk index;

[0016] According to the medium unit overload risk index, the medium unit whose risk index exceeds the preset critical value is determined, the corresponding unit position and pollutant type are associated and mapped, and a percolation medium pollution load distribution map is generated.

[0017] Preferably, the generation process of the percolation efficiency decay interval specifically includes:

[0018] The percolation layer state analysis unit calls the high-risk medium unit identified in the percolation medium pollution load distribution map of the percolation layer, obtains the initial porosity, real-time water flow breakthrough rate and saturation critical time of the corresponding unit in the purification cycle, calculates the pore shrinkage rate and water flow breakthrough delay time length respectively, and generates a medium performance degradation data set.

[0019] Based on the pore shrinkage rate and water flow penetration delay time in the medium performance degradation data set, the total amount of pollutant interception and adsorption efficiency distribution of the corresponding medium unit in the period are called to identify the medium plugging degree and hydraulic conductivity performance decline level, and the medium state degradation information is obtained;

[0020] According to the medium state degradation information, the plugging development rate of different position points in the performance degradation data set is combined to calculate the medium performance attenuation characteristic value, identify the spatial range of the attenuation abnormal area in the purification process, and generate the infiltration performance attenuation interval.

[0021] Preferably, the generation process of the purification medium dynamic adjustment strategy specifically includes:

[0022] The purification medium regulation unit extracts the effluent water quality index and medium regeneration efficiency of the unit in two consecutive purification periods according to the medium unit number indicated by the infiltration performance attenuation interval, and obtains the medium performance lag information by combining the deviation value between the actual effluent water quality and the design purification standard;

[0023] Based on the medium performance lag information, it is judged whether the performance lag value exceeds the medium failure threshold, the medium unit number with performance failure is screened, and the water quality purification efficiency decline rate of the corresponding unit is extracted, sorted according to the decline rate value, and a medium replacement priority sequence is generated;

[0024] The sorting results in the medium replacement priority sequence are called to sequentially configure enhanced purification measures for the medium unit, adjust the purification strength of the medium unit within the system operation load range, record the unit number and the adjusted purification parameters, and generate the purification medium dynamic adjustment strategy.

[0025] Preferably, the generation process of the purification medium backwashing instruction specifically includes:

[0026] The purification process execution unit calls the medium regulation parameters recorded in the purification medium dynamic adjustment strategy to detect the pressure gradient change and pollutant removal efficiency of water flow through each medium layer in the real-time purification process, identify the water pressure linkage effect between adjacent medium layers, and generate an interlayer synergistic effect strength index;

[0027] According to the linkage effect values of adjacent three layers of medium in the interlayer synergistic effect strength index, it is judged whether the linkage effect is higher than the system preset synergistic threshold, and if the judgment condition is met, it is marked as a synergistic purification ready state, and the medium unit number and state parameters are associated to generate a purification medium backwashing instruction to control the opening of the blowdown valve.

[0028] Preferably, the system further includes a purification performance evaluation unit:

[0029] The purification performance evaluation unit calls the purification cycle marked in the purification medium backwashing instruction, screens the effluent water quality trajectory in the final purification stage, compares the pollutant removal efficiency with the purification time length, and if the efficiency continues to improve without reaching a stable state, calculates the time length required to reach the target purification degree and updates the end control timing to obtain the end purification extension control result.

[0030] Preferably, the generation process of the end purification extension control result specifically comprises:

[0031] The purification performance evaluation unit calls the purification cycle number marked in the purification medium backwashing instruction, screens the water quality monitoring data in the final purification stage in the corresponding cycle, extracts the continuous efficiency sequence and time node data from the start of the final stage to the standard effluent, and generates the end purification trajectory sequence.

[0032] Based on the efficiency change trend in the end purification trajectory sequence, the water quality improvement rate in the final purification stage is analyzed, the efficiency increment value of the last segment of the sequence is extracted, and compared with the corresponding duration of the final stage, if the efficiency continues to rise and does not enter the stable interval, the additional time length required to reach the designed purification standard is calculated, and the end purification additional time interval is generated.

[0033] The system required extension time value in the end purification additional time interval is called, the real-time end purification control cycle is updated, the originally set purification termination time point is corrected, the end purification control interval is redefined, and the end purification extension control result is obtained.

[0034] Preferably, the system further comprises a dynamic scheduling management unit:

[0035] The dynamic scheduling management unit integrates the full-process data of the infiltration medium pollution load distribution map, the infiltration efficiency decay interval, the purification medium dynamic adjustment strategy and the purification medium backwashing instruction, establishes a mapping relationship library of medium state and control instruction, optimizes the iterative update of purification parameters through a self-adaptive learning algorithm, and generates a system self-optimization scheduling strategy.

[0036] Preferably, the implementation process of the system self-optimization scheduling strategy comprises:

[0037] The dynamic scheduling management unit monitors the real-time changes of water quality indicators and medium state in the purification system, compares the difference degree of historical operation data and current performance parameters, and when the difference degree exceeds the adaptive tolerance, starts the parameter correction mechanism, outputs the power regulation instruction of the high-pressure water pump to control the operation mode of the water flow distributor according to the corresponding relationship between the medium load rate and the purification efficiency.

[0038] Compared with the prior art, the beneficial effects of the present application are:

[0039] The system set flood data acquisition unit can obtain the physical properties and chemical composition of the flood inflow end in real time, extract multi-dimensional water quality indicators and suspended matter distribution state, and identify pollution concentration abnormal areas with the help of a neural network model, and finally generate a leaching medium pollution load distribution map. This process breaks the limitations of traditional single-dimensional, fixed-time-point monitoring, allowing staff to clearly understand the pollution load of the leaching medium in different areas, including pollution load level coding, medium blockage risk coefficient and spatial distribution heat map, so that targeted intervention measures can be taken in advance for high-load areas to avoid local blockage affecting the overall purification process and ensure the balance of system purification efficiency.

[0040] The leaching layer state analysis unit calls the high-load area in the leaching medium pollution load distribution map of the leaching layer, obtains the medium porosity and water flow penetration rate of the corresponding area, and identifies the current leaching layer blockage and saturation segment based on historical penetration efficiency data, generating a leaching efficiency decay interval containing porosity blockage rate, hydraulic penetration delay time, and purification capacity decay gradient. This unit realizes accurate judgment of the state of the leaching layer, no longer relying on experience, but based on the combination of real-time data and historical data, accurately capturing the medium performance degradation process, allowing staff to timely grasp the specific area and degree of leaching layer efficiency decay, providing clear direction for subsequent medium regulation, and avoiding the problem of substandard effluent water quality caused by maintenance after the medium completely fails.

[0041] The purification medium regulation unit extracts the water quality purification efficiency and medium regeneration state in the continuous two purification periods of the physical location indicated by the leaching efficiency decay interval, calculates the difference between the actual purification effect and the expected purification target, judges whether the medium is invalid, and selects the purification efficiency decline rate of the invalid medium area as the priority judgment basis, generating a purification medium dynamic adjustment strategy containing medium replacement priority level, purification intensity adjustment amplitude, and regulation strategy effective timing. This dynamic adjustment mode changes the disadvantages of traditional fixed regulation and shutdown maintenance, and can develop differentiated strategies according to the failure degree and efficiency decline rate of different media. When some media need to be maintained, it does not affect the normal operation of other areas, ensuring the continuity of system purification, while through priority sorting, the medium area with fast efficiency decline is processed first, minimizing the impact on the overall purification effect.

[0042] The purification process execution unit calls the regulation and control parameters in the purification medium dynamic adjustment strategy, detects the pressure change and pollutant interception amount of water flow through the medium in the real-time purification process, identifies the synergy strength between adjacent purification layers, and records the synergy purification readiness state and generates a purification medium backwashing instruction when the synergy strength is higher than the system preset threshold. This unit includes the interlayer synergy into the regulation and control category, ensures that the system operates in the optimal synergy state, and generates the backwashing instruction based on the actual synergy effect and real-time data, including the backwashing pressure set value, medium layer disturbance frequency, and water flow backwashing time length, which not only avoids the aggravation of medium blockage caused by untimely backwashing, but also prevents energy waste and medium loss caused by excessive backwashing, prolongs the service life of the infiltration medium, and at the same time guarantees the stability of the purification effect. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 A timing diagram of the artificial intelligence-based flood infiltration purification system described in the present application;

[0044] Figure 2 A flowchart for displaying the constituent elements such as the infiltration medium pollution load distribution diagram;

[0045] Figure 3 A flowchart for generating the infiltration efficiency decay interval;

[0046] Figure 4 A system schematic diagram of the artificial intelligence-based flood infiltration purification system described in the present application.

[0047] In the figure: 1, flood inflow end, 2, infiltration layer, 3, purification layer. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0049] Please refer to Figure 1The application provides a flood infiltration purification system based on artificial intelligence, which comprises multiple intelligent units to realize real-time purification of flood. The system first acquires the physical properties and chemical components of the flood inflow end by a flood data acquisition unit, extracts multi-dimensional water quality indexes such as turbidity, chemical oxygen demand and suspended matter distribution state, analyzes the spatial distribution characteristics of pollutants in the water flow by using a convolutional neural network model, identifies the abnormal area of pollution concentration and generates a pollution load distribution map of the infiltration medium. The infiltration layer state analysis unit calls the high load area marked in the above distribution map, combines the medium porosity sensor and the water flow breakthrough rate monitoring data, compares the historical penetration efficiency curve to judge the blockage and saturation segments of the current infiltration layer, and outputs the infiltration efficiency decay interval. The purification medium regulation unit extracts the water quality purification efficiency and medium regeneration state data of the position in the continuous two purification cycles according to the physical position indicated by the decay interval, determines whether the medium is invalid by calculating the deviation value of the actual purification effect and the expected purification target, and generates a dynamic adjustment strategy of the purification medium according to the priority basis of the purification efficiency decline rate. The purification process execution unit calls the regulation parameters in the adjustment strategy, monitors the water flow pressure change and the pollutant interception amount in the real-time purification process, and simultaneously evaluates the synergistic effect intensity between adjacent purification layers 3. When the synergistic effect intensity is higher than the preset threshold, the synergistic purification ready state is triggered, and a backwashing instruction of the purification medium is generated to control the start-stop timing of the high-pressure backflushing pump of the purification layer 3, so as to complete the whole-process adaptive purification.

[0050] Embodiment 1: refer to Figure 2 The implementation of the flood data acquisition unit depends on the pre-deployed distributed sensor network at the flood inflow end. The network is composed of multiple types of sensor nodes, including ultrasonic flow meters, optical turbidity meters, ion-selective electrodes and spectral analysis probes. These sensors are embedded in the form of an array into the wall of a water channel or pipeline to synchronously collect physical property data such as flow rate, water temperature and turbidity, and chemical component data such as heavy metal ion concentration, organic pollutant content and nitrogen and phosphorus indicators at a set sampling frequency. The sensor nodes are connected to the data collector through an industrial bus. The collector amplifies, filters and digitizes the original signals, which are then uploaded to the central processing server through the edge computing gateway. The server runs a data preprocessing program to eliminate outliers and compensate for environmental interference (such as the influence of temperature on electrode measurement), thereby obtaining standardized multi-dimensional water quality index sequences. The extraction of the suspended matter distribution state adopts a combination of image recognition and laser scattering technology. An underwater high-speed camera continuously captures water flow video streams, and a background difference algorithm is used to identify the motion trajectories of suspended particulate matter. A laser particle size analyzer simultaneously determines the particle size distribution of particulate matter in real time. The system fuses image trajectory data and particle size data to calculate the spatial concentration gradient and settling velocity curve of suspended matter, forming a suspended matter distribution state matrix.

[0051] The neural network model adopts a convolutional neural network structure for identifying the abnormal area of pollution concentration. The input layer of the model receives the pre-processed multi-dimensional water quality index matrix and the suspended matter distribution state matrix. The input data is first normalized to have a value range of 0 to 1. The convolutional layer uses multiple convolution kernels of different sizes for feature extraction. The pooling layer uses a max-pooling operation to reduce the data dimension and retain significant features. The fully connected layer flattens the feature map obtained after convolution and pooling, and inputs it into the classifier for abnormal area judgment. The classifier outputs the pollution probability value of each monitoring grid. The grid with a probability value exceeding the set threshold is marked as an abnormal area of pollution concentration. The model training stage uses a data set of historical flood events, which includes normal water quality areas and labeled data of known pollution events. The network weights are optimized by the back propagation algorithm, so that the model can accurately distinguish between background concentration and abnormal pollution.

[0052] The generation of the leach medium pollution load distribution map is based on the abnormal area results identified by the neural network model. The system calculates two key parameters, namely the pollution aggregation degree and the medium adsorption saturation degree. The pollution aggregation degree is obtained by analyzing the spatial and temporal variation coefficients of each water quality index in the abnormal area. The area with a high variation coefficient is considered as an aggregation hotspot. The medium adsorption saturation degree is calculated according to the ratio of the theoretical adsorption capacity of the medium to the current adsorption amount. The current adsorption amount is obtained by integrating the concentration difference between the inlet and outlet water and the cumulative treated water volume. These two parameters together constitute the pollution load distribution parameter set. The system then calls the physical structure parameter database of the multi-layer leach medium, which stores information such as the material, thickness, specific surface area, porosity, and designed adsorption capacity of each medium unit. For each medium unit, the system calculates its real-time load rate, which is the percentage of the current adsorption amount to the designed adsorption capacity, based on the current hydrological conditions such as flow and flow rate. By comparing the load rate with the preset medium capacity threshold, the system generates a medium unit overload risk index. The system associates the units with risk index exceeding the threshold with their spatial coordinates and the dominant pollutant type, and uses the geographic information system engine to render and generate the leach medium pollution load distribution map. The map displays the pollution load levels of different color codes in a visual form (such as green for low load, yellow for medium load, and red for high load), and superimposes the contour lines of the medium blockage risk coefficient and the heat map reflecting the spatial distribution of pollution load.

[0053] The implementation of the leaching layer state analysis unit obtains the real-time state parameters of the medium units marked as high load areas in the pollution load distribution map from the medium state monitoring subsystem by calling the unit code. The porosity monitoring uses a non-invasive gamma-ray attenuation method. The sensor emits gamma rays to penetrate the medium layer, and the real-time porosity is calculated by detecting the attenuation value of the ray intensity. The water flow penetration rate is calculated by the water head loss measured by the differential pressure sensor installed at the water inlet and outlet of the unit combined with the Darcy formula. The saturation critical time is the time point predicted based on the current pollutant load rate and the medium adsorption kinetics model. The system calculates the pore shrinkage rate, which is the difference between the current porosity and the porosity of the last period divided by the time interval; and the water flow penetration delay time, which is the difference between the actual penetration time and the theoretical clean medium penetration time. These two sets of data constitute the core content of the medium performance degradation data set. The generation of the leaching efficiency attenuation interval depends on the in-depth analysis of the medium performance degradation data set. The system extracts the pore shrinkage rate and the water flow penetration delay time from the data set, and calls the accumulated total amount of pollutants intercepted by the corresponding medium unit in the current purification period (obtained by integrating the continuous monitoring of the inlet and outlet water concentrations) and the adsorption efficiency distribution data (obtained by fitting the isothermal adsorption model). The degree of medium clogging is quantified by the ratio of the pore shrinkage rate to the initial porosity, and the level of hydraulic conductivity performance decline is represented by the ratio of the water flow penetration delay time to the theoretical penetration time. The system further calculates the performance attenuation characteristic value of each medium unit, which is a weighted comprehensive index. The weights are allocated according to the importance of the pollutant type and the purification target. The system identifies the cluster of continuous medium units whose attenuation characteristic values exceed the critical threshold in the three-dimensional space model, draws its spatial boundary, and generates the leaching efficiency attenuation interval. The interval data body contains the average pore clogging rate of all units in the interval, the maximum hydraulic penetration delay time, and the purification capacity decay gradient map reflecting the decay of the purification capacity along the water flow direction.

[0054] The formulation of the dynamic adjustment strategy of the purification medium begins with the specific list of medium unit numbers indicated by the filtration efficiency decay interval. The system extracts the detailed records of the effluent water quality indicators of these units in the last two consecutive complete purification cycles from the water quality history database, including but not limited to chemical oxygen demand removal rate, ammonia nitrogen removal rate, total phosphorus removal rate, and the concentration of specific toxic and harmful substances. At the same time, the system retrieves the medium regeneration state log of these units, which records the performance recovery data after each backwashing, chemical cleaning or biological activation. The system compares the actual effluent water quality indicators with the design purification standards item by item, calculates the absolute deviation and relative deviation, and synthesizes a comprehensive performance lag index by weighting. When this index exceeds the failure threshold dynamically adjusted according to the medium type and the running time, the medium unit is judged to be in a failure state. For all the medium units judged to be in a failure state, the system further analyzes the change curve of the water purification efficiency of each unit over time, and calculates the purification efficiency decline rate of each unit in the last cycle by using the linear regression method. Subsequently, the system sorts all the failed units according to the decline rate values from high to low to generate a medium replacement priority sequence. Within the range allowed by the current running load of the system, the system allocates appropriate enhanced purification measure resources to each unit according to the priority sequence. For the highest priority unit, it may allocate stronger purification intensity; for the slightly lower priority unit, it may take the conventional intensity adjustment.

[0055] Example 2: see Figure 3, the infiltration layer state analysis unit first accesses the infiltration medium pollution load distribution map generated by the flood data acquisition unit and stored in the central database, which identifies the risk level of different spatial positions in the entire infiltration bed in a grid structure. The analysis algorithm built-in the unit automatically filters out all medium unit codes marked as "high risk" in the map, which correspond to specific positions in the physical structure of the infiltration layer. Subsequently, the system sends data call instructions to the sensor network deployed near these high-risk units, which consists of porosity detection sensors, high-precision pressure transducers, and timing trigger modules. The porosity detection sensor uses the principle of sound wave reflection to calculate real-time porosity data inside the medium by emitting sound waves and receiving echoes reflected from the surface of the medium particles. The pressure transducer is installed at the inlet and outlet of each high-risk unit to continuously monitor the pressure loss of the water flow through the unit, and combined with the known thickness of the medium layer, the actual penetration rate of the water flow is calculated using fluid mechanics principles. The saturation critical time parameter is not obtained directly by measurement, but is predicted by the current pollution load rate of the unit (from the pollution load distribution map) and the adsorption kinetics model of this type of medium pre-stored in the database, which considers factors such as pollutant concentration, water flow velocity, and remaining adsorption capacity of the medium. After obtaining the three basic parameters of initial porosity, real-time water flow penetration rate, and saturation critical time, the system starts the performance degradation calculation program. The calculation of the porosity shrinkage rate is by comparing the real-time porosity obtained in the current measurement period with the baseline porosity of the unit recorded in the system database at the end of the last complete purification period (which can be regarded as the initial porosity of this period), and dividing the difference between the two measurement points by the actual time interval between the two measurement points, thereby obtaining an average rate value reflecting the change of porosity with time. The calculation of the water flow penetration delay time is more complex, which needs to establish a reference baseline: the system will call the theoretical hydraulic conductivity coefficient of the medium unit in a new or completely regenerated state, combined with the current operating hydraulic load, to calculate the theoretical water flow penetration time; then compare the actual measured penetration time with this theoretical time, and the difference is the penetration delay time. These two calculation results: porosity shrinkage rate and water flow penetration delay time, together with the original measurement data they are based on, are packaged and time-stamped and labeled with unit code tags, forming a structured record, which is stored in a special medium performance degradation data set. This data set continuously adds new records as the purification period progresses, forming the performance change history trajectory of each unit.

[0056] The system further calls the total amount of pollutants trapped by the high-risk medium unit of interest in the current purification cycle from the operation log, which is calculated by the difference between the upstream water concentration and the downstream water concentration after flowing through the unit, multiplied by the cumulative amount of water processed. At the same time, the system analyzes the adsorption efficiency distribution of the unit for different types of pollutants, which is achieved by fitting the change curve of its water concentration and comparing it with the standard adsorption isotherm model. Based on these data, the system conducts two quantitative assessments: the degree of medium clogging is compared with the critical shrinkage rate preset according to the type of medium, and its relative severity is expressed in percentage; the level of hydraulic conductivity performance decline is represented by the proportion of the delay time of water flow penetration to the theoretical penetration time, and the higher the proportion, the more serious the decline. The evaluation results of the two aspects are integrated into the medium state degradation information, which not only contains numerical results, but also contains the judgment of the main pollutant type causing degradation.

[0057] The final generation of the leaching performance decay interval is a process of spatial clustering and boundary identification. The system calls the medium state degradation information of all high-risk units and reads the continuous records of the clogging development rate of each unit in the performance degradation data set. A comprehensive "medium performance decay characteristic value" is calculated for each unit, which is a dimensionless index that integrates pore shrinkage rate, penetration delay time, clogging degree, and conductivity performance decline level, etc. The weight of each variable is adjusted according to the type of pollutants it handles and the purification priority set by the system. After calculation, the system spatially interpolates the decay characteristic values of all units in the three-dimensional digital model to generate a continuous decay characteristic field. Then, the system applies spatial analysis algorithms to identify the connected regions whose characteristic values exceed the preset threshold in this characteristic field, and these regions are judged as "decay anomaly areas". The algorithm accurately outlines the boundary range of these anomaly areas in the three-dimensional space of the purification process, and the final generated leaching performance decay interval is the digital definition of these spatial ranges. The interval data body contains key parameters such as the average pore clogging rate, the maximum hydraulic penetration delay time, and the purification capacity decline gradient calculated along the water flow direction of all units in the interval.

[0058] The purification medium dynamic adjustment strategy is executed by the purification medium regulation unit, which formulates in dependence on the seepage performance decay interval results. The unit first analyzes the decay interval data and extracts the list of unique numbers of medium units with abnormal performance decline. The system queries the water quality historical database for the detailed effluent water quality archives of these units in the last complete purification cycle (cycle N) and the previous cycle (cycle N-1), including the minute-level or hour-level monitoring values of key indicators such as chemical oxygen demand and ammonia nitrogen; at the same time, the medium regeneration records in the two cycles are retrieved from the equipment management log, and the performance recovery rate after each regeneration is calculated. Then, the system compares the actual effluent water quality data of each unit in the two cycles with the purification standard limit value point by point, calculates the comprehensive deviation value, which reflects the degree of performance lag of the medium unit using the weighted Euclidean distance algorithm. Then, the deviation value of each unit is compared with the dynamic adjustment medium failure threshold value, and if the deviation value exceeds the current applicable threshold value, the unit is determined to be failed and the number is added to the “failed unit list”.

[0059] For each failed medium unit entering the list, the system needs to further evaluate the urgency of its performance deterioration, which is the key to determining the processing priority. The system focuses on the changes in water purification efficiency of the unit in the farthest cycle (cycle N). By fitting the efficiency curve through linear regression method, the slope of the efficiency change with time, i.e. the purification efficiency decline rate, is calculated. The faster the decline rate, the faster the performance of the unit is deteriorating, and the more urgent the intervention is needed. The system calculates the purification efficiency decline rate for each failed unit in the list, and then ranks them in descending order according to the rate value, thereby generating a medium replacement priority sequence. This sequence clearly indicates which units need the most immediate attention. Finally, the system allocates appropriate regulation resources and formulates specific enhanced purification measures for each failed unit in the sequence according to the generated medium replacement priority sequence and the current actual operation load of the purification system (such as total treatment water volume, available energy consumption, reagent dosing capacity, etc.). For the units ranked at the front of the sequence with the highest priority, strong regulation measures may be allocated. For lower priority units, regular intensity adjustments may be taken. The system accurately records the number of each regulated unit and the specific regulation parameters allocated to it, and all these instructions and parameters are integrated into a structured and executable purification medium dynamic adjustment strategy. The physical processing structure of the system is in turn flood inflow end 1, seepage layer 2 and purification layer 3, and the flood data collected by the sensors will flow through each structure in turn to complete the purification treatment. See Figure 4, the system schematic diagram of the flood infiltration purification system based on artificial intelligence, the left side is the flood inflow end 1, used for introducing the flood to be purified. The flood inflow end 1 is connected with the infiltration layer 2 on the right side, the infiltration layer 2 is used for physically intercepting and preliminarily purifying the flood, and sensors for monitoring the medium porosity and water flow penetration rate are arranged in the layer. The infiltration layer 2 is connected with the purification layer 3 on the right side, the purification layer 3 is used for deep purification of the infiltrated flood to improve the water quality standard rate, and sensors for monitoring the water flow pressure change and the pollutant interception amount are arranged between the purification layer 3 and the infiltration layer 2 and at the water outlet end of the purification layer. At the same time, the system icon indicates the correlation of each functional unit and the physical structure, wherein the flood data acquisition unit is connected with the sensors of the flood inflow end 1 and the infiltration layer 2 to obtain data, the infiltration layer state analysis unit, the purification medium regulation unit and the purification process execution unit form a data interaction and control link in sequence with the infiltration layer 2 and the purification layer 3, and the cooperative working relationship between the system physical structure and the functional units is clearly embodied.

[0060] In example 3, after the purification process execution unit is started, the purification medium dynamic adjustment strategy file generated and pushed by the purification medium regulation unit is called from the system instruction queue. The file is structured data, containing the medium unit number list and the regulation parameters that need to be adjusted, such as the upper limit of the specific filter layer backwashing pressure, the threshold value of the target pollutant interception amount of the activated carbon adsorption unit, the nutrient solution dosing rate of the biofilm carrier, etc. The unit instruction analysis module reads the parameters and converts them into control signals recognizable by the underlying actuators. Subsequently, the unit activates the real-time monitoring of the sensor network of each medium layer, the pressure sensor array is sampled at a high frequency (such as 100 times per second), and the small pressure fluctuations when the water flow passes through different media are captured, and the pressure data are recorded in a differential form, and the real-time pressure gradient change is calculated. The pollutant interception amount monitoring is realized by taking samples from the water outlet of each medium unit at regular intervals for analysis by an online water quality analyzer, comparing the water concentrations before and after the water inlet and outlet, and combining the instantaneous flow data to calculate the pollutant interception mass per unit time.

[0061] After collecting the continuous pressure gradient change sequence and the pollutant interception amount data, the system begins to evaluate the interaction between adjacent purification layers. The evaluation process focuses on the water pressure linkage effect, that is, how the change of the resistance of the upper layer medium affects the hydraulic load of the lower layer medium. The system establishes a time series data set for each pair of adjacent upper and lower layers (for example, layer A and layer B), which includes the pressure gradient through layer A and the pressure gradient through layer B recorded at the same timestamp. To quantify their synergistic relationship, the system calculates a parameter called "interlayer synergistic effect strength indicator". This indicator is obtained by analyzing the dynamic correlation of the two pressure gradient sequences within a certain time window. Specifically, for a time window containing n sampling points, the indicator can be characterized as:

[0062] ;

[0063] in: and They represent the times at time 1 and 2 respectively. Based on the pressure gradient values ​​of layer A and layer B and These are the average values ​​of their respective pressure gradients within the time window. The main part on the right-hand side of the formula calculates the Pearson correlation coefficient between the two pressure gradient sequences, reflecting the synchronicity of their changes. The formula also includes... The term introduces a peak time based on the cross-correlation coefficient. The penalty factor It represents the time delay in the response of one sequence to another. This is the total length of the analysis time window. Thus, The value depends not only on the similarity of pressure changes (correlation coefficient) but also on the timeliness of the response; the smaller the delay, the closer the penalty factor is to 0. The closer the value is to the correlation coefficient itself, the greater the lag. The value decreases accordingly. This calculation process is repeated for all adjacent media layer pairs in the system (e.g., layers AB, BC, CD, etc.).

[0064] After generating the synergistic effect strength index between adjacent layers, the system enters the state determination phase. The determination logic is to check whether there are three consecutive media layers in the current purification process, where the synergistic effect strength between each pair is higher than the system's preset synergistic threshold. This threshold is a configurable parameter, usually set based on the historical best operating data of the media combination, such as 0.8. The system scans the media layer sequence, finds a combination of three consecutive layers that meets the conditions, marks it and its contained media units as "synergistic purification ready state," and records the unit number, synergistic effect strength value, and state mark timestamp. The purification media backwash command is generated based on the "synergistic purification ready state" determination result. The command generation module creates a backwash command for each media unit in the ready state (usually the middle layer of the three-layer combination or the layer that needs the most cleaning). This is a specific control command set, containing key operating parameters: backwash pressure setpoint, determined by a combination of real-time assessment of the media unit's blockage degree and the upper limit of mechanical strength safety; media layer disturbance frequency, i.e., the frequency of pressure or airflow pulse application during backwashing, optimized based on historical data of the media and contaminants; and water flow backwash duration, which depends on the estimated contaminant load and expected cleaning effect. These parameters are encapsulated into standard control commands, which are sent to field actuators via industrial networks to control high-pressure backwash pumps and drain valves, enabling precise and automatic cleaning of specific media layers and ensuring efficient recovery and continuous operation of the purification process.

[0065] Embodiment 4: The purification performance evaluation unit plays a role of closed-loop evaluation and optimization in the system operation, and its workflow starts with receiving the backwash instruction of the purification medium from the purification process execution unit, which is a sign of the end of a purification cycle and internally embeds the cycle number information triggering evaluation. The unit retrieves all water quality monitoring raw data after the entire purification process enters the final stage of the corresponding cycle from the central historical database according to this number. The definition of the final purification stage is usually based on preset rules. The retrieved data includes the analysis results of the effluent water samples collected at fixed time intervals, the concentration values of key indicators such as chemical oxygen demand, ammonia nitrogen, and total suspended solids, and the real-time pollutant removal efficiency calculated. These data points are arranged in chronological order to form a continuous time sequence from the start of the final stage to the end of the cycle, i.e., the end purification trajectory sequence. This sequence clearly records the dynamic process of water quality indicators changing over time in the last period of the cycle.

[0066] To accurately analyze the end purification effect, the evaluation unit needs to further process the end purification trajectory sequence. The core data sequence retrieved by the system for the final stage of a certain purification cycle shows that the original termination time of this cycle is the 480th minute, and the system sets the standard as the outlet chemical oxygen demand concentration falling below 50 mg / L. The monitoring data shows that from the 450th minute, the chemical oxygen demand concentration is still fluctuating and decreasing, with the value gradually decreasing from 68.5 mg / L to 50 mg / L, and the removal efficiency correspondingly slowly rising from 92.01% to 100%, as shown in Table 1.

[0067] Table 1: Water quality purification trajectory data for the final stage of purification cycle P-20240926-03

[0068]

[0069] Based on the above trajectory sequence, the purification performance evaluation unit starts the trend analysis algorithm. The algorithm first examines the efficiency increase of the last segment of the sequence, i.e., calculates the difference between the efficiency values of adjacent time points and observes the change rule. From the data log, it can be seen that after the original end time point of the cycle, the chemical oxygen demand removal efficiency continues to increase monotonically and does not show a plateau or downward trend. The built-in "stable interval" judgment logic in the system starts to work, which is usually based on historical big data statistics, such as requiring the fluctuation range of the efficiency value to be less than a certain threshold in consecutive time intervals to be considered as entering a stable state. The efficiency increase of the last segment of the current sequence is obviously beyond the fluctuation range required by the stable interval. The system determines that at the forced end of the purification cycle, the purification process has not reached a stable state, and the water quality improvement rate is still positive, meaning that the purification potential has not been fully released.

[0070] After determining that the efficiency is not yet stable and is still rising, the system needs to calculate the additional runtime required to reach the design purification standard. The design purification standard is specifically embodied here as a stable effluent chemical oxygen demand concentration of less than fifty milligrams per liter. The evaluation unit uses a prediction algorithm, such as linear regression or exponential fitting based on the last section of data points, to extrapolate the curve of the chemical oxygen demand concentration over time. The algorithm calculates the time point at which the predicted curve intersects the concentration standard line. Through calculation, the predicted compliance time point may be at five hundred and five minutes. The time difference from the original cycle termination time to the predicted compliance time is the additional end purification time interval that the system needs to extend, which in this example may be twenty-five minutes. This additional time interval is a key decision parameter, which quantifies the additional time that the current cycle needs to run in order to achieve the expected effluent water quality.

[0071] In Example 5, the dynamic scheduling management unit serves as the central coordinating mechanism of the system, and its physical carrier is a high-performance industrial server that establishes real-time data connections with the floodwater data acquisition unit, the infiltration layer state analysis unit, the purification medium regulation unit, and the purification process execution unit through a high-speed data bus. The core of the internal operation of the unit is a relational database management system, which builds a data model named the medium state and regulation instruction mapping relationship library. This model takes the unique number of each medium unit as the primary key, for example, assigning an identifier to each grid after dividing the infiltration bed into grids. The database establishes a dynamically updated record for each medium unit, which contains the latest reported state parameters of the unit, such as the overload risk indicators of the medium unit extracted from the pollution load distribution map, the purification capacity decay gradient obtained from the infiltration efficiency decay interval, the purification intensity adjustment amplitude recorded from the dynamic adjustment strategy, and the medium layer disturbance frequency captured from the backwashing instruction. The record also associates all the regulation instructions that have been issued to the unit in the past and the effect evaluation summary after their execution. This structure enables the system to trace back to any medium unit under a specific state parameter, and determine what kind of regulation instruction will bring about what kind of performance change.

[0072] The implementation of the adaptive learning algorithm relies on the continuous analysis of the historical operation data accumulated in the mapping relationship library. The algorithm adopts an optimization method based on time difference learning, and the learning goal is to match the optimal purification parameter combination for various medium states. The algorithm regards the operation of the system as a sequential decision-making process, and each decision cycle is regarded as a time step. At each time step, the algorithm observes the current state of the system, which is composed of the key state parameter vectors of all medium units. The algorithm selects an action from the possible action set for execution, which is the adjustment of the system purification parameters, such as adjusting the power setting value of the high-pressure water pump or changing the opening mode of the water flow distributor. After the action is executed, the system will move to a new state and generate an immediate reward signal, which is calculated according to the control target, for example, the reward value can be proportional to the improvement of the overall removal rate of pollutants and inversely proportional to the increase of energy consumption. The core of the algorithm is to learn a value function that estimates the long-term cumulative reward expectation value of executing a specific action in a specific state.

[0073] The triggering and generation of the system self-optimization scheduling strategy rely on real-time monitoring and difference calculation. The dynamic scheduling management unit continuously receives water quality index data and medium state data from the underlying sensor network. The system maintains a set of performance parameters for the current operation period, such as the average purification efficiency so far, the energy consumption per unit of water, etc. The system retrieves the best performance parameter sequence of the system's past operation under similar inflow load and similar environmental conditions from the historical database as a reference benchmark. The difference calculation uses the weighted Manhattan distance method to weight and sum the absolute values of the differences between each dimension of the current performance parameter vector and the historical best parameter vector. The weights are allocated according to the importance of each parameter. The system pre-sets an adaptive tolerance threshold, which can also be dynamically adjusted according to the stable and fluctuating periods of system operation. When the calculated real-time difference exceeds the threshold, it indicates that the current operating state deviates significantly from the historical optimal state, and the parameter correction mechanism is immediately activated.

[0074] After the parameter correction mechanism is started, the system first analyzes the main cause of the deviation. It will query the mapping relationship database, focus on analyzing the corresponding relationship between the current medium load rate and the current purification efficiency, and compare it with the load rate and efficiency relationship curve in the historical data. If it is found that the purification efficiency at the current load rate interval is significantly lower than the historical same period level, it is determined that the purification parameters in this area need to be optimized. The system generates specific control instructions according to the current recommended strategy of the adaptive learning algorithm. The generated instructions are specific and executable. These instruction sets constitute the system self-optimization scheduling strategy for the current period. The strategy is issued to the corresponding executor, and the new state data and effect data produced after the executor executes are recorded back to the mapping relationship database for the next round of learning and optimization, thereby forming a continuous improvement closed loop. For example, in the process of processing high-suspended flood, the system may monitor that although the overall purification efficiency meets the standard, the energy consumption is higher than the historical best level. After triggering optimization through difference calculation, the adaptive algorithm may try to fine-tune the backflushing frequency and water distribution mode to find the lowest energy consumption operating point while ensuring efficiency.

[0075] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based flood infiltration and purification system, comprising a flood inflow end, an infiltration layer, and a purification layer, characterized in that, It also includes the following components: The flood data acquisition unit acquires the physical characteristics and chemical composition of the flood inflow end in real time, extracts multi-dimensional water quality indicators and suspended solids distribution status, identifies areas of abnormal pollution concentration through a neural network model, and generates a pollution load distribution map of the infiltration medium. The infiltration layer state analysis unit calls the high-risk media unit marked in the infiltration medium pollution load distribution map of the infiltration layer, obtains the three parameters of the corresponding unit in the purification cycle: initial porosity, real-time water flow penetration rate and saturation critical time, calculates the pore shrinkage rate and water flow penetration delay time respectively, and generates a media performance degradation dataset. Based on the pore shrinkage rate and water flow penetration delay in the media performance degradation dataset, the total amount of pollutants retained and the distribution of adsorption efficiency of the corresponding media unit within the cycle are called to identify the degree of media blockage and the level of decline in hydraulic conductivity, and to obtain media state degradation information. Based on the media state degradation information, combined with the blockage development rate at different locations in the performance degradation dataset, the media efficiency decay characteristic value is calculated, the spatial range of the decay anomaly zone in the purification process is identified, and the percolation efficiency decay interval is generated. The purification medium control unit extracts the water purification efficiency and medium regeneration status of the physical location indicated by the infiltration efficiency decay range in two consecutive purification cycles, calculates the gap between the actual purification effect and the expected purification target, determines whether the medium is in a failure state, selects the purification efficiency decline rate corresponding to the failure medium area as the priority judgment basis, and generates a dynamic adjustment strategy for the purification medium. The purification process execution unit calls the media control parameters recorded in the purification media dynamic adjustment strategy, detects the pressure gradient change and pollutant removal efficiency of water flow through each media layer during real-time purification, identifies the water pressure linkage effect between adjacent media layers, that is, how the change of upper media resistance affects the hydraulic load of lower media, and generates an inter-layer synergistic effect strength index. Based on the linkage effect value of the three adjacent media in the interlayer synergy strength index, it is determined whether the linkage effect is higher than the preset synergy threshold of the system. If the determination condition is met, it is marked as a synergistic purification ready state, and the media unit number and status parameters are associated to generate a purification media backwashing command to control the opening of the sewage valve.

2. The artificial intelligence-based flood infiltration and purification system according to claim 1, characterized in that, The percolation media pollution load distribution map includes pollution load level coding, media blockage risk coefficient, and spatial distribution heat map. The percolation efficiency decay range includes pore blockage rate, hydraulic penetration delay time, and purification capacity decay gradient. The purification media dynamic adjustment strategy includes media replacement priority level, purification intensity adjustment range, and control strategy activation sequence. The purification media backwashing command includes backwashing pressure set value, media layer disturbance frequency, and water flow backwashing duration.

3. The artificial intelligence-based flood infiltration and purification system according to claim 1, characterized in that, The process of generating the percolation medium contamination load distribution map specifically includes: The flood data acquisition unit monitors the flow velocity and pollutant concentration at the flood inflow end in real time, extracts the sedimentation characteristics of particulate matter and the diffusion trajectory of chemical substances at different times, and calculates the pollutant aggregation degree and media adsorption saturation based on the deviation value between real-time monitoring data and standard water quality model, generating a set of pollution load distribution parameters. By combining the pollutant aggregation degree and media adsorption saturation of the pollution load distribution parameter set with the physical structure parameters of the multilayer infiltration media, the load rate of each media unit under the current hydrological conditions is statistically analyzed, the correspondence between the load rate and the media tolerance is identified, and an overload risk index of the media unit is generated. Based on the overload risk index of the media unit, media units whose risk index exceeds the preset critical value are identified, and the corresponding unit locations are mapped with the types of pollutants to generate a pollutant load distribution map of the percolation media.

4. The artificial intelligence-based flood infiltration and purification system according to claim 1, characterized in that, The generation process of the dynamic adjustment strategy for the purification medium specifically includes: The purification media control unit extracts the effluent water quality indicators and media regeneration efficiency of the unit in two consecutive purification cycles based on the media unit number indicated by the percolation efficiency decay range. Combined with the deviation value between the actual effluent water quality and the design purification standard, the media performance hysteresis information is obtained. Based on the media performance hysteresis information, it is determined whether the performance hysteresis value exceeds the media failure threshold, the media unit number of the performance failure is screened, and the water purification efficiency decline rate of the corresponding unit is extracted. The units are sorted according to the decline rate value to generate a media replacement priority sequence. The sorting results in the media replacement priority sequence are called, and enhanced purification measures are configured for the media units in sequence. The purification intensity of the media units is adjusted within the system operating load range. The unit number and the adjusted purification parameters are recorded to generate a dynamic adjustment strategy for the purification media.

5. The artificial intelligence-based flood infiltration and purification system according to claim 1, characterized in that, It also includes a purification performance evaluation unit: The purification efficiency evaluation unit calls the purification cycle marked by the backwashing instruction of the purification medium, screens the effluent water quality trajectory of the final purification stage, compares the pollutant removal efficiency with the purification time, and if the efficiency continues to improve but has not reached a stable state, calculates the time required to reach the target purification level and updates the end control sequence to obtain the end purification continuous control result.

6. The artificial intelligence-based flood infiltration and purification system according to claim 5, characterized in that, The process of generating the end-of-pipe purification continuation control result specifically includes: The purification efficiency evaluation unit calls the purification cycle number marked in the backwashing instruction of the purification medium, filters the water quality monitoring data of the final purification stage under the corresponding cycle, extracts the continuous efficiency sequence and time node data from the start time of the final stage to the effluent meeting the standard, and generates the final purification trajectory sequence. Based on the efficiency change trend in the terminal purification trajectory sequence, the water quality improvement rate in the terminal purification stage is analyzed, the efficiency increase value of the last segment of the sequence is extracted, and it is compared with the duration corresponding to the last segment. If the efficiency continues to rise and has not entered the stable range, the supplementary time required to reach the design purification standard is calculated, and the terminal purification supplementary time interval is generated. The system calls the required extension time value in the end-of-pipe purification compensation time interval, updates the real-time end-of-pipe purification control cycle, corrects the originally set purification termination time point, redefines the end-of-pipe purification control interval, and obtains the end-of-pipe purification continuity control result.

7. The artificial intelligence-based flood infiltration and purification system according to claim 1, characterized in that, It also includes a dynamic scheduling management unit: The dynamic scheduling and management unit integrates the entire process data of the percolation media pollution load distribution map, percolation efficiency decay range, purification media dynamic adjustment strategy, and purification media backwashing command. It establishes a mapping relationship library between media status and control commands, optimizes the iterative update of purification parameters through adaptive learning algorithm, and generates a system self-optimizing scheduling strategy.

8. The artificial intelligence-based flood infiltration and purification system according to claim 7, characterized in that, The implementation process of the system's self-optimizing scheduling strategy includes: The dynamic scheduling management unit monitors the real-time changes in water quality indicators and media status in the purification system, compares the differences between historical operating data and current performance parameters, and activates the parameter correction mechanism when the differences exceed the adaptive tolerance. Based on the correspondence between media load rate and purification efficiency, it outputs power adjustment commands for the high-pressure water pump to control the operation mode of the water flow distributor.

Images