An artificial intelligence-based method and system for calculating pollutant discharge of a river-entering sewage outlet

Abstract

This invention belongs to the field of artificial intelligence and river discharge outlet control technology, and discloses an AI-based method and system for calculating pollutant discharge from river discharge outlets. The method utilizes a large visual language model and vector retrieval technology to extract and structure key elements such as hydrology and discharge requirements from multi-source text. It intelligently matches the optimal calculation model, constructs differentiated paths based on compliance status and discharge permit information, and completes discharge volume calculations under different scenarios, supporting shared discharge outlet accounting by multiple discharge units. The calculation is performed according to the user-selected fully automatic, semi-automatic, or manual mode, generating a standardized report with a chain of evidence. Data is persistently stored throughout the process and supports traceability. This invention improves calculation efficiency, standardization, and reliability, reduces labor costs, adapts to complex scenarios, and provides technical support for the approval of river discharge outlets.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence and river discharge control technology, and specifically relates to an artificial intelligence-based method and system for calculating pollutant discharge from river discharge outlets. Background Technology

[0002] Although the current "Technical Guidelines for Supervision and Management of Sewage Outlets into Rivers and the Sea" standard provides some general guidelines for calculating discharge volume, it does not specify detailed guidelines for model selection, parameter selection, and calculation method selection. This results in significant difficulties for grassroots approval personnel and third-party compilers in accurately calculating pollutant discharge volumes from sewage outlets into rivers.

[0003] The existing technologies have the following shortcomings: First, the calculation of sewage discharge from river outlets relies on manual extraction of parameters from multiple source documents and manual selection of calculation models. This is labor-intensive, time-consuming, and cumbersome, and has a high technical threshold. Insufficient human experience can easily lead to errors in the selection of technical routes, affecting the rationality of the calculation results. Second, neither manual calculation nor simple AI-assisted calculation can form a complete calculation basis chain. The source of parameters, the logic of model selection, and the calculation process lack traceability, making it difficult to meet the compliance requirements of administrative licensing approval. Third, when existing AI methods are applied to discharge calculation, the lack of multimodal model support and a full-process compliance verification mechanism makes them prone to logical contradictions and result deviations in complex projects with long processes and multiple scenarios, such as receiving water bodies that do not meet standards and multiple sewage discharge units sharing sewage outlets. This makes it difficult to meet the statutory rigor requirements of administrative licensing. Fourth, existing calculation methods are mostly designed for single-type sewage outlets or single water area scenarios, and cannot be adapted to complex scenarios such as multiple sewage discharge units sharing sewage outlets, receiving water bodies that do not meet standards, and special period supervision, resulting in insufficient applicability. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned shortcomings of the prior art and provide a method and system for calculating pollutant discharge from river discharge outlets based on artificial intelligence.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for calculating pollutant discharge from river discharge outlets based on artificial intelligence, comprising the following steps: S1 analyzes multiple original documents, including the demonstration report and monitoring report of the sewage outlet into the river, and extracts key elements of hydrological parameters, sewage outlet discharge requirements and receiving water body functional objectives. After normalization and conflict detection, it outputs a structured parameter set. S2, call the pre-built multimodal model library, which includes hydrodynamic-water quality coupling model, empirical parameter library and historical case library, and intelligently match the optimal calculation model according to hydrological conditions and regulatory requirements; S3. Based on the structured parameter set, determine whether the receiving water body meets the standards, and construct a differentiated calculation path in combination with the discharge permit information of the responsible entity. Perform discharge calculation under the scenarios of receiving water body meeting and not meeting the standards respectively to obtain the daily discharge volume during special periods, the daily discharge volume during non-special periods, and the annual discharge volume. S4, based on the user-specified calculation mode, will route the execution flow to one of the fully automatic, semi-automatic assisted, or manual calculation paths, call the optimal calculation model matched in the multimodal model library, and complete the calculation of key pollutant emissions from sewage outlets into rivers; S5 generates a calculation report that includes the source of parameters, the basis for model selection, the calculation process, and compliance explanations. It also persists the data of the entire calculation process and supports full-process traceability.

[0006] Furthermore, step S1 specifically includes: S11 receives input multi-source raw documents; for application documents, it performs multi-task concurrent parsing of each page of the document through a large visual language model; for argumentation report documents, it locates the target chapter by searching the vector database and calls the large text model to extract key elements step by step. S12 adopts a multi-channel parallel architecture and extracts five key elements—name of responsible entity, month of special period, emission parameters, pollutant discharge permit information, and pollutant concentration—through special prompt instructions. It integrates rule matching and deep learning technologies to identify and label structured elements, and then stores them in a vector database after vectorization. S13, through semantic understanding of large language models and similarity matching of vector embedding, unifies and standardizes the same element that is expressed differently in different documents or chapters into standard field names and units of measurement. If there are conflicting parameters in the extracted elements, they will be automatically marked as conflict pending verification and an early warning will be triggered. The current process will be paused and can only continue to the next step after manual review and confirmation. S14 sets the target water quality value, hydrological boundary conditions, model applicability rules, sewage outlet discharge standards, and spatial coordinate system rules corresponding to the functional categories of the receiving water bodies, for subsequent model matching and parameter calls.

[0007] Furthermore, the preset hydrological boundary conditions include flow thresholds and corresponding guarantee rates for dry season, normal water season, and wet season; the text-based large model is used to identify whether there are special period daily emission regulatory requirements in the demonstration report. If such conditions exist, the hydrological parameters for the dry season will be used for calculating emission limits during special periods, while those for non-special periods will be used for calculating limits during normal periods. If such conditions do not exist, the hydrological parameters for normal periods will be used for both special and non-special periods.

[0008] The default discharge standard for the preset discharge outlet adopts the Class A standard in the "Discharge Standard of Pollutants for Urban Wastewater Treatment Plants" GB18918; when local standards or industry-specific requirements are specified, the corresponding technical parameters will be parsed and adapted first. The preset spatial coordinate system rules uniformly adopt the CGCS2000 coordinate system, automatically verify the latitude and longitude format, unit and precision in the document, initiate spatial semantic reasoning and issue warnings for fuzzy spatial descriptions, and submit for manual review.

[0009] Furthermore, the multimodal model library in step S2 specifically includes: The emission back-calculation model recommendation library covers commonly used mathematical models for rivers, lakes, reservoirs, estuaries and nearshore sea areas. Based on the properties of the receiving water body and the constancy of water flow, the stability of sewage discharge, and the mixing characteristics of the water area, it intelligently recommends zero-dimensional, vertical one-dimensional, planar two-dimensional, vertical two-dimensional, three-dimensional or river network models. A library of recommended models for measuring river pollution carrying capacity intelligently recommends zero-dimensional, one-dimensional, two-dimensional, or one-dimensional models of rivers and estuaries based on river morphology, pollutant mixing characteristics, and tidal influence. The lake (reservoir) pollution carrying capacity calculation model recommendation library intelligently recommends uniform mixing, non-uniform mixing, eutrophication or stratification models based on water surface area, average water depth, water nutrient status and planar morphology. The historical case library stores typical cases of sewage outlet setting up into rivers that have been reviewed and approved by experts, which are used for model selection reference and calculation result verification. The empirical parameter library stores source data in a structured format, including parameter names, value ranges, applicable water body types, applicable model types, and technical specification numbers. During calculations, values ​​extracted from documents are used first, while manually entered values ​​can override extracted values. If neither is available, the default values ​​from the empirical parameter library are used.

[0010] Furthermore, step S3 specifically includes: S31. By semantic retrieval, locate the relevant content of the water ecological environment status survey and analysis in the demonstration report, identify the corresponding water quality target level of each water body section, and determine the standard value of each pollutant according to the strict principle; call the large language model to extract the measured monitoring data of pollutants of each section in the document, compare the measured data with the standard value item by item, determine the exceeding factors and the frequency of exceeding the standard according to the annual average, monthly average and the average of the wet and dry water periods, and generate a structured judgment conclusion. S32. If the responsible entity has obtained a discharge permit and specifies a clear annual discharge limit, the annual discharge limit shall be used as the annual discharge limit, and the daily discharge limit shall be obtained by dividing the annual discharge limit by 365. If there is no discharge permit or no discharge limit is specified, S33 or S34 shall be implemented according to the determination conclusion of S31. S33, when the receiving water body meets the standards, based on the discharge concentration at the sewage outlet. Cp (mg / L) and design daily drainage volume Q p (m³ / d), according to C p × Q p / 10 6 (t / d) Calculate the initial daily discharge; call the calculation model matched in S2, using the hydrological parameters during the dry season as input, substitute the initial daily discharge into the model, and calculate the predicted mixed concentration at the key section of the receiving water body. C and the water quality standard value C s Comparison: If C ≤ C s The initial emissions are then the daily emissions for that specific period; if C > C s Then C ≤ C s To constrain the maximum allowable discharge, the bisection method is used to iteratively calculate the daily discharge for special periods. The result is the daily discharge for special periods. Keeping the model unchanged, the hydrological parameters are switched to the normal water period, and the above process is repeated to obtain the daily discharge for non-special periods. The annual discharge is calculated by adding the product of the number of days in the special period and the daily discharge for the special period to the product of the number of days in the non-special period and the daily discharge for the non-special period. S34. When the receiving water body fails to meet standards, select the actual measurement method, survey and statistical method, or estimation method to calculate the pollutant discharge of other sewage outlets into the river within the calculation range; call the calculation model to calculate the predicted mixed concentration of key sections of the receiving water body, in order to C ≤ C s To constrain the allowable emission concentration of the sewage outlet to be demonstrated in the reverse calculation If the reduction rate exceeds 100%, then the emission data of surrounding sewage outlets will be included in the constraint set, and the reduction and substitution accounting will be performed cyclically until the total emission of all sewage outlets within the calculation range meets the requirement. C ≤ C s The hydrological parameters were switched to dry season and normal water season respectively, and the daily and annual discharges during special periods and non-special periods were calculated using the same method.

[0011] Furthermore, when multiple sewage discharge entities share the same sewage outlet into the river, S31 to S34 are executed first to obtain the total daily and annual discharge of key pollutants from the shared sewage outlet; then it is determined whether the receiving water body does not exceed the standards and whether each responsible entity holds a sewage discharge permit specifying the permitted discharge limit: if the conditions are met, the sum of the permitted annual discharge limits of each responsible entity is used as the total annual discharge of the shared sewage outlet; if the conditions are not met, an intelligent allocation method is selected.

[0012] Furthermore, the intelligent selection method is chosen from the following three methods for allocation: First, the allocation is based on the proportion of wastewater discharge. The deviation between the actual wastewater discharge and the designed wastewater discharge of each wastewater discharge unit is compared. If the deviation exceeds 20%, the proportion is calculated using the average annual wastewater discharge of the past three years. Otherwise, the proportion is calculated using the designed wastewater discharge. The total discharge is then allocated according to the proportion. Second, the allocation is based on the proportion of emissions of key pollutants. The emissions of multiple key pollutants of each polluting unit are statistically analyzed and ranked. The top three emissions are selected to calculate the weighted average proportion, and the total emissions are allocated according to this value. Third, the allocation is based on the proportion of heavy metals and toxic and hazardous substances emitted. Heavy metals and toxic and hazardous substances are included in the category of key pollutants and allocated according to the emission proportion of various substances or the weighted average. If only some polluting units emit a certain type of substance, the responsibility proportion of those polluting units shall not be less than 60%.

[0013] Furthermore, the three calculation paths in step S4 are as follows: S41, when using fully automatic path calculation, the path planning engine driven by the large language model automatically loads the parameters into the matching model and performs calculations based on the mapping relationship between the structured parameter set and the knowledge graph, and automatically generates structured calculation results; S42, when using a semi-automatic assisted calculation path, construct an interactive calculation interface to support users in supplementing or modifying constraints, allowing user input values ​​to overwrite the corresponding parameter fields extracted by S1, recommending an appropriate calculation model based on the updated parameter set, and executing the calculation after user confirmation; S43, when using manual calculation path, constructs an interactive calculation interface, allowing users to manually input constraints, select calculation models, perform calculations step by step, view intermediate results, and finally generate calculation results.

[0014] Furthermore, step S5 specifically includes: integrating the verified calculation results, process data, and model parameters to generate a standardized calculation report containing parameter sources, calculation process, and compliance descriptions, which can be exported as PDF or Word format.

[0015] The calculation report, process data, model parameters and judgment conclusions are persistently stored in a relational database in a structured format. A data index is established with the calculation task ID as the primary key, which supports data traceability, historical query and secondary call for subsequent calculations.

[0016] Based on the same inventive concept, this invention provides an artificial intelligence-based system for calculating pollutant discharge from sewage outlets into rivers, comprising: a document parsing module, a model matching module, a differential calculation module, a path routing module, and a report generation and storage module.

[0017] Furthermore, the document parsing module is configured to parse multi-source original documents containing demonstration reports and monitoring reports on the setting up of sewage outlets into rivers, extract key elements such as hydrological parameters, sewage outlet discharge requirements, and functional objectives of receiving water bodies, and output a structured parameter set after normalization processing and conflict detection.

[0018] The model matching module is configured to call a pre-built multimodal model library, which includes a hydrodynamic-water quality coupling model, an empirical parameter library, and a historical case library, and intelligently matches the optimal calculation model based on hydrological conditions and regulatory requirements.

[0019] Furthermore, the differentiated calculation module is configured to determine whether the receiving water body meets the standards based on the structured parameter set, and to construct a differentiated calculation path in combination with the discharge permit information of the responsible entity, and to perform discharge calculations under the scenarios of receiving water body meeting and not meeting the standards, respectively, to obtain the daily discharge volume during special periods, the daily discharge volume during non-special periods, and the annual discharge volume.

[0020] Furthermore, the path routing module is configured to route the execution flow to one of the fully automatic, semi-automatic assisted, or manual calculation paths according to the calculation mode specified by the user, and call the optimal calculation model matched in the multimodal model library to complete the calculation of the discharge of key pollutants from the sewage outlet into the river.

[0021] Furthermore, the report generation and storage module is configured to generate a calculation report that includes the source of parameters, the basis for model selection, the calculation process, and a compliance description, and to persistently store the data of the entire calculation process, supporting full-process traceability.

[0022] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a text intelligent parsing engine to automatically extract key parameters from multiple source documents such as river discharge outlet setting demonstration reports and monitoring reports. It dynamically matches the most suitable calculation model by combining multiple dimensions such as discharge outlet type and receiving water body function. Relying on a differentiated calculation path mechanism, it realizes compliance calculation in complex scenarios such as receiving water body compliance and non-compliance, and multiple discharge units sharing discharge outlets. Finally, it outputs auditable and traceable discharge calculation results and a complete chain of evidence, which significantly improves the intelligence and standardization level of river discharge outlet setting approval, while lowering the threshold for report preparation and improving preparation efficiency. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for calculating pollutant discharge from a river discharge outlet based on artificial intelligence, according to the present invention. Figure 2 This is a schematic diagram of the composition of an artificial intelligence-based pollutant discharge measurement system for river discharge outlets according to the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of this invention, not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0025] Example 1 like Figure 1 The diagram shows a flowchart of a method for calculating pollutant discharge from river discharge outlets based on artificial intelligence, according to the present invention. The method includes the following steps: S1 analyzes multiple original documents, including demonstration reports and monitoring reports on the setting of sewage outfalls into rivers, extracts key elements such as hydrological parameters, sewage outfall discharge requirements, and functional objectives of receiving water bodies, and outputs a structured parameter set after normalization and conflict detection.

[0026] Step S1 specifically includes: S11 receives input original documents from multiple sources; for application documents, it performs multi-task concurrent parsing of each page of the document using a large visual language model; for argumentation report documents, it locates the target chapter by searching the vector database and calls the large text model to extract key elements step by step.

[0027] S12 adopts a multi-channel parallel architecture and extracts five key elements—name of responsible entity, month of special period, emission parameters, pollutant discharge permit information, and pollutant concentration—through special prompt instructions. It integrates rule matching and deep learning technologies to identify and label structured elements, and then stores them in a vector database after vectorization.

[0028] S13, through semantic understanding of large language models and similarity matching of vector embedding, unifies and standardizes the same element, which is expressed differently in different documents or chapters, into standard field names and units of measurement.

[0029] If there are conflicting parameters in the extracted elements, they are automatically marked as conflict pending verification and an early warning is triggered. The current process is paused and can only continue after manual review and confirmation.

[0030] S14 sets the target water quality value, hydrological boundary conditions, model applicability rules, sewage outlet discharge standards, and spatial coordinate system rules corresponding to the functional categories of the receiving water bodies, for subsequent model matching and parameter calls.

[0031] The preset hydrological boundary conditions include flow thresholds and corresponding guarantee rates for dry season, normal water season, and wet season; the text large model is used to identify whether there are special period daily emission regulatory requirements in the demonstration report: If such conditions exist, the hydrological parameters for the dry season will be used for calculating emission limits during special periods, while those for non-special periods will be used for calculating limits during normal periods. If such conditions do not exist, the hydrological parameters for normal periods will be used for both special and non-special periods.

[0032] The default discharge standard for the preset discharge outlet adopts the Class A standard in the "Discharge Standard of Pollutants for Urban Wastewater Treatment Plants" GB18918; when local standards or industry-specific requirements are specified, the corresponding technical parameters will be parsed and adapted first. The preset spatial coordinate system rules uniformly adopt the CGCS2000 coordinate system, automatically verify the latitude and longitude format, unit and precision in the document, initiate spatial semantic reasoning and issue warnings for fuzzy spatial descriptions, and submit for manual review.

[0033] S2, call the pre-built multimodal model library, which includes hydrodynamic-water quality coupling models, empirical parameter libraries and historical case libraries, and intelligently match the optimal calculation model according to hydrological conditions and regulatory requirements.

[0034] The multimodal model library in step S2 specifically includes: The emission back-calculation model recommendation library covers commonly used mathematical models for rivers, lakes, reservoirs, estuaries and nearshore sea areas. Based on the properties of the receiving water body and the constancy of water flow, the stability of sewage discharge, and the mixing characteristics of the water area, it intelligently recommends zero-dimensional, vertical one-dimensional, planar two-dimensional, vertical two-dimensional, three-dimensional or river network models.

[0035] The river pollution carrying capacity measurement model recommendation library intelligently recommends zero-dimensional, one-dimensional, two-dimensional or estuary one-dimensional models of rivers based on river section morphology, pollutant mixing characteristics and tidal influence.

[0036] The lake (reservoir) pollution carrying capacity calculation model recommendation library intelligently recommends uniform mixing, non-uniform mixing, eutrophication or stratification models based on water surface area, average water depth, water nutrient status and planar morphology.

[0037] The historical case library stores typical cases of sewage outlet setting up into rivers that have been reviewed and approved by experts, which are used for model selection reference and calculation result verification.

[0038] The empirical parameter library stores source data in a structured format, including parameter names, value ranges, applicable water body types, applicable model types, and technical specification numbers. During calculations, values ​​extracted from documents are used first, while manually entered values ​​can override extracted values. If neither is available, the default values ​​from the empirical parameter library are used.

[0039] S3. Based on the structured parameter set, determine whether the receiving water body meets the standards, and construct a differentiated calculation path in combination with the discharge permit information of the responsible entity. Perform discharge calculation under the scenarios of receiving water body meeting and not meeting the standards respectively to obtain the daily discharge volume during special periods, the daily discharge volume during non-special periods, and the annual discharge volume.

[0040] Step S3 specifically includes: S31. By semantic retrieval, locate the relevant content of the water ecological environment status survey and analysis in the demonstration report, identify the corresponding water quality target level of each water body section, and determine the standard value of each pollutant according to the strict principle; call the large language model to extract the measured monitoring data of pollutants of each section in the document, compare the measured data with the standard value item by item, determine the exceeding factors and the frequency of exceeding the standard according to the annual average, monthly average and the average of the wet and dry water periods, and generate a structured judgment conclusion. S32. If the responsible entity has obtained a discharge permit and specifies a clear annual discharge limit, the annual discharge limit shall be used as the annual discharge limit, and the daily discharge limit shall be obtained by dividing the annual discharge limit by 365. If there is no discharge permit or no discharge limit is specified, S33 or S34 shall be implemented according to the determination conclusion of S31. S33, when the receiving water body meets the standards, based on the discharge concentration at the sewage outlet. C p (mg / L) and design daily drainage volume Q p (m³ / d), according to C p × Q p / 10 6 (t / d) Calculate the initial daily discharge; call the calculation model matched in S2, using the hydrological parameters during the dry season as input, substitute the initial daily discharge into the model, and calculate the predicted mixed concentration at the key section of the receiving water body. C and the water quality standard value C s Comparison: If C ≤ C s The initial emissions are then the daily emissions for that specific period; if C > C s Then C ≤ C s To constrain the maximum allowable discharge, the bisection method is used to iteratively calculate the daily discharge for special periods. The result is the daily discharge for special periods. Keeping the model unchanged, the hydrological parameters are switched to the normal water period, and the above process is repeated to obtain the daily discharge for non-special periods. The annual discharge is calculated by adding the product of the number of days in the special period and the daily discharge for the special period to the product of the number of days in the non-special period and the daily discharge for the non-special period. S34. When the receiving water body fails to meet standards, select the actual measurement method, survey and statistical method, or estimation method to calculate the pollutant discharge of other sewage outlets into the river within the calculation range; call the calculation model to calculate the predicted mixed concentration of key sections of the receiving water body, in order to C ≤ C s To constrain the allowable emission concentration of the sewage outlet to be demonstrated in the reverse calculation If the reduction rate exceeds 100%, then the emission data of surrounding sewage outlets will be included in the constraint set, and the reduction and substitution accounting will be performed cyclically until the total emission of all sewage outlets within the calculation range meets the requirement. C ≤ C s The hydrological parameters were switched to dry season and normal water season respectively, and the daily and annual discharges during special periods and non-special periods were calculated using the same method.

[0041] When multiple sewage discharge entities share the same sewage outlet into a river, S31 to S34 are executed first to obtain the total daily and annual discharge of key pollutants from the shared sewage outlet. Then, it is determined whether the receiving water body does not exceed the standards and whether each responsible entity holds a sewage discharge permit specifying the permitted discharge limit. If the conditions are met, the sum of the permitted annual discharge limits of each responsible entity is taken as the total annual discharge of the shared sewage outlet. If the conditions are not met, an intelligent allocation method is selected.

[0042] The intelligent selection method is chosen from the following three options for allocation: First, the allocation is based on the proportion of wastewater discharge. The deviation between the actual wastewater discharge and the designed wastewater discharge of each wastewater discharge unit is compared. If the deviation exceeds 20%, the proportion is calculated using the average annual wastewater discharge of the past three years. Otherwise, the proportion is calculated using the designed wastewater discharge, and the total discharge is allocated according to the proportion.

[0043] Second, the allocation is based on the proportion of key pollutant emissions. The emissions of multiple key pollutants from each polluting unit are statistically analyzed and ranked. The top three emissions are selected to calculate the weighted average proportion, and the total emissions are allocated according to this value.

[0044] Third, the allocation is based on the proportion of heavy metals and toxic and hazardous substances emitted. Heavy metals and toxic and hazardous substances are included in the category of key pollutants and allocated according to the emission proportion of various substances or the weighted average. If only some polluting units emit a certain type of substance, the responsibility proportion of those polluting units shall not be less than 60%.

[0045] S4, based on the user-specified calculation mode, will route the execution flow to one of the fully automatic, semi-automatic assisted, or manual calculation paths, and call the optimal calculation model matched in the multimodal model library to complete the calculation of key pollutant emissions from sewage outlets into rivers.

[0046] The three calculation paths in step S4 are as follows: S41, when using fully automatic path calculation, the path planning engine driven by the large language model automatically loads the parameters into the matching model and performs calculations based on the mapping relationship between the structured parameter set and the knowledge graph, and automatically generates structured calculation results; S42, when using a semi-automatic assisted calculation path, construct an interactive calculation interface to support users in supplementing or modifying constraints, allowing user input values ​​to overwrite the corresponding parameter fields extracted by S1, recommending an appropriate calculation model based on the updated parameter set, and executing the calculation after user confirmation; S43, when using manual calculation path, constructs an interactive calculation interface, allowing users to manually input constraints, select calculation models, perform calculations step by step, view intermediate results, and finally generate calculation results.

[0047] S5 generates a calculation report that includes the source of parameters, the basis for model selection, the calculation process, and compliance explanations. It also persists the data of the entire calculation process and supports full-process traceability.

[0048] Step S5 specifically includes: integrating the verified calculation results, process data and model parameters to generate a standardized calculation report that includes parameter sources, calculation process and compliance descriptions, and supports exporting to PDF or Word format; The calculation report, process data, model parameters and judgment conclusions are persistently stored in a relational database in a structured format. A data index is established with the calculation task ID as the primary key, which supports data traceability, historical query and secondary call for subsequent calculations.

[0049] To further illustrate the complete implementation process of the above method, the following detailed explanation is provided with specific calculation examples.

[0050] The third wastewater treatment plant in a certain city (hereinafter referred to as the "Factory Project") plans to set up a sewage discharge outlet into the Qingshui River, discharging the treated effluent into the river. The user uploads the "Application Form for Setting Up a Sewage Discharge Outlet into the River" and the "Factory Report for Setting Up a Sewage Discharge Outlet into the Qingshui River" to the system. The system focuses on ammonia nitrogen (NH3-N) as the key controlled pollutant and completes the automatic calculation of the entire process according to the method of this invention to verify the calculation logic and efficiency improvement effect.

[0051] The system invokes a large visual language model to perform multi-task concurrent parsing of the application, and simultaneously locates the "Aquatic Ecological Environment Status Survey and Analysis" section in the demonstration report through vector retrieval, extracting and normalizing the following key parameters: The designed daily drainage capacity of the sewage outlet is Qp = 5 × 10 4m³ / d (equivalent to 0.579 m³ / s); The Class A standard of the "Discharge Standard of Pollutants for Urban Wastewater Treatment Plants" (GB18918-2002) shall be implemented, with an ammonia nitrogen discharge concentration Cp = 5 mg / L; The water quality target of the receiving water body (Qingshui River) is Class III, with an ammonia nitrogen standard limit Cs = 1.0 mg / L; The background concentration of ammonia nitrogen in the receiving water body is C0 = 0.20 mg / L; The design flow rate during the dry season is Qh(dry) = 2.0 m³ / s, and the cross-sectional flow velocity is u(dry) = 0.40 m / s; The design flow rate during the normal water season is Qh(normal) = 7.5 m³ / s; The special period is from July 1st to September 30th (92 days in total), and the non-special period is the remaining 273 days.

[0052] During the element normalization stage, the system uses vector embedding cosine similarity calculation to uniformly identify different expressions such as "dry season" and "average flow in the driest month" from the demonstration report as parameters for the dry season. It also identifies "a monitoring section of the Qingshui River" and "the control section of the demonstration river segment" as the same control section, eliminating ambiguity in subsequent model calls. No contradictory parameters were found during conflict detection, and the structured parameter set was written to the runtime cache via an asynchronous interface, automatically proceeding to the next step.

[0053] Based on the extracted water diffusion characteristic parameters (the Qingshui River section under demonstration is approximately 30m wide and has an average depth of approximately 1.5m) and the continuous and stable characteristics of sewage discharge, the system calls the model library matching algorithm to determine that the Qingshui River is a small plain river, where pollutants can achieve lateral uniform mixing over a short distance. It recommends using a zero-dimensional uniform mixing model for back-calculation of discharge volume and writes the model type identifier, applicable basis, and called technical specifications into the runtime cache for reference in subsequent report generation and source tracing. After the user confirms the recommended model on the front end, the system officially calls the model to perform subsequent calculations.

[0054] The system compared the ammonia nitrogen water quality monitoring data of the Qingshui River over the past three years with the Class III water quality standard item by item in the demonstration report, and determined that the receiving water body met the standard (annual average ammonia nitrogen value of 0.35 mg / L, lower than the Class III standard limit of 1.0 mg / L). At the same time, the system detected that the responsible entity had not yet obtained a discharge permit, and could not directly use the permitted discharge limit. Therefore, based on the determination conclusion, the "annual discharge calculation path when the receiving water body meets the standard" (step S33) was triggered, which calculated the daily discharge during special periods using parameters during the dry season and the daily discharge during non-special periods using parameters during the normal water season.

[0055] Step S31: Based on the upper limit of emission concentration Cp = 5 mg / L and the design daily wastewater discharge Qp = 5 × 10 4 m³ / d, calculate the initial daily emissions (unit: t / d) using the following formula: Substituting the values, the initial daily emissions W0 = 5 × 50000 × 10 -6 =0.25t / d.

[0056] Steps S31-S33: Substitute the dry season parameters (Qh=2.0m³ / s) into the zero-dimensional uniform mixing model to calculate the predicted mixing concentration at the control section (unit: mg / L): Substituting Cp = 5 mg / L, Qp = 0.579 m³ / s, C0 = 0.20 mg / L, and Qh(dry) = 2.0 m³ / s, the calculation process is as follows: C=(5×0.579+0.20×2.0) / (0.579+2.0)=(2.895+0.400) / 2.579=3.295 / 2.579=1.278mg / L.

[0057] The predicted concentration C = 1.278 mg / L > Cs = 1.0 mg / L, exceeding the water quality standard. The system then directly calculates the allowable discharge concentration Cp,req (unit: mg / L) from the zero-dimensional model analytical solution without binary iteration. The corresponding formula is: Substituting the parameters, the calculation process is as follows: Cp,req=[1.0×(0.579+2.0)] [0.20 × 2.0] / 0.579 = (2.579) 0.400) / 0.579=2.179 / 0.579=3.763mg / L.

[0058] Daily emissions during special periods are calculated using the following formula (unit: t / d): Substituting Cp,req=3.763mg / L and Qp=50000m³ / d, we get W_t_r = 3.763 × 50000 × 10 -6 =0.188t / d.

[0059] The system switches the hydrological parameters to the normal water level period (Qh(normal) = 7.5 m³ / s), keeps the model and initial discharge concentration Cp = 5 mg / L unchanged, and substitutes them into the zero-dimensional uniform mixing model. The calculation process is as follows: C=(5×0.579+0.20×7.5) / (0.579+7.5)=(2.895+1.500) / 8.079=4.395 / 8.079=0.544mg / L The predicted concentration at the control section is C=0.544mg / L≤Cs=1.0mg / L, which meets the water quality standards. The initial daily discharge is the daily discharge during non-special periods: W_non-special_day = 0.25t / d.

[0060] The system reads the month field for special periods: July 1st to September 30th, Tspecial = 92 days; for non-special periods, Tnon-special = 273 days. The annual ammonia nitrogen discharge from the sewage outlet into the river is calculated using the following formula (unit: t / a): Substituting the values, the calculation process is as follows: W_year = 92 × 0.188 + 273 × 0.25 = 17.296 + 68.250 = 85.5 t / a; The system writes the above calculation results and intermediate parameters into the database in structured JSON format, and automatically generates a standardized calculation report. The report covers the location of the parameter source documents (including report page numbers and field positions), the basis for model selection (small plain river, zero-dimensional uniform mixing), the calculation process of each step, and compliance instructions (referencing GB18918—2002 Class A standard and GB3838—2002 Class III water quality standard), and supports one-click export to Word format.

[0061] This example shows that the entire process from document upload to report generation takes approximately 18 minutes, which is 25 times more efficient than traditional manual calculation (approximately 8-10 hours). All key parameters in the report can be traced back to the specific location in the original document, meeting the compliance requirements of administrative licensing approval and verifying the significant advantages of the method of this invention in terms of calculation efficiency and standardization.

[0062] Example 2 like Figure 2 The diagram shows the composition of an artificial intelligence-based pollutant discharge measurement system for river discharge outlets according to the present invention. The system includes, in sequence, a document parsing module, a model matching module, a differential measurement module, a path routing module, and a report generation and storage module.

[0063] The document parsing module is configured to parse multiple original documents, including demonstration reports and monitoring reports on the setting of sewage outlets into rivers, extract key elements such as hydrological parameters, sewage outlet discharge requirements, and functional objectives of receiving water bodies, and output a structured parameter set after normalization and conflict detection.

[0064] The model matching module is configured to call a pre-built multimodal model library, which includes a hydrodynamic-water quality coupling model, an empirical parameter library, and a historical case library, and intelligently matches the optimal calculation model based on hydrological conditions and regulatory requirements.

[0065] The differentiated calculation module is configured to determine whether the receiving water body meets the standards based on the structured parameter set, and to construct a differentiated calculation path in combination with the discharge permit information of the responsible entity. It then performs discharge calculations under the scenarios of receiving water body meeting and not meeting the standards, respectively, to obtain the daily discharge volume during special periods, the daily discharge volume during non-special periods, and the annual discharge volume.

[0066] The path routing module is configured to route the execution flow to one of the fully automatic, semi-automatic assisted, or manual calculation paths according to the calculation mode specified by the user, and call the optimal calculation model matched in the multimodal model library to complete the calculation of the discharge of key pollutants from the sewage outlet into the river.

[0067] The report generation and storage module is configured to generate a measurement report that includes the source of parameters, the basis for model selection, the calculation process and compliance description, and to persistently store the data of the entire measurement process, supporting full-process traceability.

[0068] To further illustrate the collaborative logic of the above modules during system operation, the specific workflow of each module will be described in detail below based on the computing scenario of Example 1.

[0069] After receiving the "Application Form for Setting Up a Sewage Discharge Outlet into the River" and the "Demonstration Report for Setting Up a Sewage Discharge Outlet into the Qingshui River" uploaded by the user, the document parsing module uses the Visual Language Model (VLM) to perform multi-page concurrent parsing of the application form, and uses the Vector Embedding Model (GLM-4-AirX) to vectorize the document into blocks and store them in the Milvus vector database. Then, it uses semantic retrieval to locate key chapters such as "Investigation and Analysis of the Current Status of the Water Ecological Environment", "Hydrological Characteristics", and "Discharge Concentration", and extracts five key elements in parallel: the name of the responsible entity (the third sewage treatment plant of a certain city), the special period (July-September), and the designed discharge volume (Qp=5×10). 4 m³ / d), pollutant discharge permit information (no permit obtained) and pollutant discharge concentration (ammonia nitrogen Cp=5mg / L).

[0070] During the element normalization stage, the module uses vector embedding cosine similarity calculation to merge "dry season" and "average flow of the driest month" in the demonstration report into the standard field "design flow during the dry season," identifying "a certain monitoring section of the Qingshui River" and "the control section of the demonstration river segment" as the same control section, thus eliminating ambiguity in subsequent model calls. No alarm was triggered during conflict detection, and the structured parameter set was written to the runtime cache via an asynchronous interface and passed to the model matching module.

[0071] The model matching module reads the water diffusion characteristic parameters (river width 30m, average water depth 1.5m, stable flow) from the cache and executes the preset model matching rules: determining that the receiving water body is a small plain river and that pollutants can achieve lateral uniform mixing within a short distance, and selecting a zero-dimensional uniform mixing model based on the model applicability rule library. The module then serializes the model type identifier ("zero-dimensional model"), applicable basis (refer to Appendix A of the "Technical Guidelines for Supervision and Management of Sewage Outfalls into Rivers and Seas") and the list of required parameters (Cp, Qp, C0, Qh, Cs) and sends them to the differentiation calculation module, while displaying the recommendation reasons on the front end for user confirmation or manual model switching.

[0072] After receiving the parameter set and model identifier from the model matching module, the differential calculation module first executes the receiving water body compliance determination sub-process: it calls the large language model to extract the ammonia nitrogen monitoring data of Qingshui River in the past three years from the demonstration report, compares it item by item with the Class III water quality standard (Cs=1.0mg / L), determines that the receiving water body meets the standard (annual average 0.35mg / L) and that the responsible entity does not have a sewage discharge permit, and triggers step S33 calculation path. The module executes the following steps sequentially: Calculate the initial daily discharge W0 = 0.25 t / d using Cp = 5 mg / L and Qp = 0.579 m³ / s; substitute these values ​​into the zero-dimensional model to obtain the predicted concentration C = 1.278 mg / L at the control section during the dry season, which exceeds 1.0 mg / L. After exceeding the standard, analytical back-calculation is initiated, yielding the allowable discharge concentration Cp,req = 3.763 mg / L, and the daily discharge Wspecial day = 0.188 t / d during special periods; after switching to parameters for the normal water period, the predicted concentration C = 0.544 mg / L meets the standard, and the daily discharge Wnon-special day = 0.25 t / d during non-special periods; finally, the annual discharge Wannual = 85.5 t / a is calculated. All intermediate results are written to the runtime cache as key-value pairs and passed to the allocation and accounting module (in this example, there are no multiple discharge units sharing the same data, so the data is directly passed to the report generation and storage module).

[0073] The path routing module routes the execution flow based on the measurement mode selected by the user in the front-end interface. In this example, the user selects the "semi-automatic assisted measurement" mode (step S42): After the system completes automatic parameter extraction and model recommendation, it displays a structured parameter confirmation interface to the user, showing all extracted parameters and the recommended measurement model.

[0074] After the user checks the flow parameters during the dry season item by item, they click the "Confirm Execution" button. The path routing module receives the user's confirmation instruction, transmits the final parameter set to the differential calculation module to trigger the calculation, and displays the intermediate results of each step (initial discharge, predicted concentration, back-calculated allowable concentration, daily discharge and annual discharge for each time period) in real time on the front end. The system supports users to modify any parameter and re-trigger the calculation, ensuring the accuracy of the calculation under human-machine collaboration.

[0075] After receiving the structured results from the completed calculations, the report generation and storage module automatically integrates the following content according to a preset template and generates a standardized calculation report: ① Parameter source, including document name, page number, and field location annotations; ② Model selection basis (zero-dimensional uniform mixing model, suitable for horizontal uniform mixing scenarios in small plain rivers, referencing Appendix A of the "Technical Guidelines for Supervision and Management of Sewage Outfalls into Rivers and Seas"); ③ Complete calculation process, including initial discharge calculation, control section predicted concentration verification, analytical back-calculation of allowable discharge concentration, and annual discharge summation calculation; ④ Compliance statement (referencing GB18918—2002 Class A standard and GB3838—2002 Class III water quality standard). The report supports one-click export to Word format, with key data automatically marked for review by auditors.

[0076] Meanwhile, the module persistently stores the calculation report, process data, model parameters and judgment conclusions in a structured format into a relational database, and establishes a data index with the calculation task ID (such as "TASK-20260315-003") as the primary key, supporting data traceability, historical query and secondary call for subsequent calculations, so as to realize permanent traceability of the entire calculation process.

[0077] The aforementioned modules are linked together asynchronously, with intermediate results passed through runtime caching, achieving seamless data flow and real-time command transmission throughout the entire process from document upload to report generation. This example system took approximately 18 minutes, representing a 25-fold increase in efficiency compared to the traditional manual compilation method (approximately 8-10 hours). All key parameters in the report can be traced back to their specific location in the original document, meeting the compliance requirements of administrative licensing approvals and verifying the effectiveness and reliability of the system in real-world engineering scenarios.

[0078] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for calculating pollutant discharge from river discharge outlets based on artificial intelligence, characterized in that, The method includes the following steps: S1 analyzes multiple original documents, including the demonstration report and monitoring report of the sewage outlet into the river, and extracts key elements of hydrological parameters, sewage outlet discharge requirements and receiving water body functional objectives. After normalization and conflict detection, it outputs a structured parameter set. S2, call the pre-built multimodal model library, which includes hydrodynamic-water quality coupling model, empirical parameter library and historical case library, and intelligently match the optimal calculation model according to hydrological conditions and regulatory requirements; S3. Based on the structured parameter set, determine whether the receiving water body meets the standards, and construct a differentiated calculation path in combination with the discharge permit information of the responsible entity. Perform discharge calculation under the scenarios of receiving water body meeting and not meeting the standards respectively to obtain the daily discharge volume during special periods, the daily discharge volume during non-special periods, and the annual discharge volume. S4, based on the user-specified calculation mode, will route the execution flow to one of the fully automatic, semi-automatic assisted, or manual calculation paths, call the optimal calculation model matched in the multimodal model library, and complete the calculation of key pollutant emissions from sewage outlets into rivers; S5 generates a calculation report that includes the source of parameters, the basis for model selection, the calculation process, and compliance explanations. It also persists the data of the entire calculation process and supports full-process traceability.

2. The method according to claim 1, characterized in that, Step S1 specifically includes: S11 receives input multi-source raw documents; for application documents, it performs multi-task concurrent parsing of each page of the document through a large visual language model; for argumentation report documents, it locates the target chapter by searching the vector database and calls the large text model to extract key elements step by step. S12 adopts a multi-channel parallel architecture and extracts five key elements—name of responsible entity, month of special period, emission parameters, pollutant discharge permit information, and pollutant concentration—through special prompt instructions. It integrates rule matching and deep learning technologies to identify and label structured elements, and then stores them in a vector database after vectorization. S13, through semantic understanding of large language models and similarity matching of vector embedding, unifies and standardizes the same element that is expressed differently in different documents or chapters into standard field names and units of measurement. If there are conflicting parameters in the extracted elements, they will be automatically marked as conflict pending verification and an early warning will be triggered. The current process will be paused and can only continue to the next step after manual review and confirmation. S14 sets the target water quality value, hydrological boundary conditions, model applicability rules, sewage outlet discharge standards, and spatial coordinate system rules corresponding to the functional categories of the receiving water bodies, for subsequent model matching and parameter calls.

3. The method according to claim 2, characterized in that, The preset hydrological boundary conditions include flow thresholds and corresponding guarantee rates for dry season, normal water season, and wet season; the text large model is used to identify whether there are special period daily emission regulatory requirements in the demonstration report: If such conditions exist, the hydrological parameters for the dry season will be used for calculating emission limits during special periods, while those for non-special periods will be used for calculating limits during normal periods. If such conditions do not exist, the hydrological parameters for normal periods will be used for both special and non-special periods.

4. The method according to claim 1, characterized in that, The multimodal model library in step S2 specifically includes: The emission back-calculation model recommendation library covers commonly used mathematical models for rivers, lakes, reservoirs, estuaries and nearshore sea areas. Based on the properties of the receiving water body and the constancy of water flow, the stability of sewage discharge, and the mixing characteristics of the water area, it intelligently recommends zero-dimensional, vertical one-dimensional, planar two-dimensional, vertical two-dimensional, three-dimensional or river network models. A library of recommended models for measuring river pollution carrying capacity intelligently recommends zero-dimensional, one-dimensional, two-dimensional, or one-dimensional models of rivers and estuaries based on river morphology, pollutant mixing characteristics, and tidal influence. The lake (reservoir) pollution carrying capacity calculation model recommendation library intelligently recommends uniform mixing, non-uniform mixing, eutrophication or stratification models based on water surface area, average water depth, water nutrient status and planar morphology. The historical case library stores typical cases of sewage outlet setting up into rivers that have been reviewed and approved by experts, which are used for model selection reference and calculation result verification. The empirical parameter library stores source data in a structured format, including parameter names, value ranges, applicable water body types, applicable model types, and technical specification numbers. During calculations, values ​​extracted from documents are used first, while manually entered values ​​can override extracted values. If neither is available, the default values ​​from the empirical parameter library are used.

5. The method according to claim 1, characterized in that, Step S3 specifically includes: S31. By semantic retrieval, locate the relevant content of the water ecological environment status survey and analysis in the demonstration report, identify the corresponding water quality target level of each water body section, and determine the standard value of each pollutant according to the strict principle; call the large language model to extract the measured monitoring data of pollutants of each section in the document, compare the measured data with the standard value item by item, determine the exceeding factors and the frequency of exceeding the standard according to the annual average, monthly average and the average of the wet and dry water periods, and generate a structured judgment conclusion. S32. If the responsible entity has obtained a discharge permit and specifies a clear annual discharge limit, the annual discharge limit shall be used as the annual discharge limit, and the daily discharge limit shall be obtained by dividing the annual discharge limit by 365. If there is no discharge permit or no discharge limit is specified, S33 or S34 shall be implemented according to the determination conclusion of S31. S33, when the receiving water body meets the standards, based on the discharge concentration at the sewage outlet. C p (mg / L) and design daily drainage volume Q p (m³ / d), according to C p × Q p / 10 6 (t / d) Calculate the initial daily discharge; call the calculation model matched in S2, using the hydrological parameters during the dry season as input, substitute the initial daily discharge into the model, and calculate the predicted mixed concentration at the key section of the receiving water body. C and the water quality standard value C s Comparison: If C ≤ C s The initial emissions are then the daily emissions for that specific period; if C > C s Then C ≤ C s To constrain the maximum allowable discharge, the bisection method is used to iteratively calculate the daily discharge for special periods. The result is the daily discharge for special periods. Keeping the model unchanged, the hydrological parameters are switched to the normal water period, and the above process is repeated to obtain the daily discharge for non-special periods. The annual discharge is calculated by adding the product of the number of days in the special period and the daily discharge for the special period to the product of the number of days in the non-special period and the daily discharge for the non-special period. S34. When the receiving water body fails to meet standards, select the actual measurement method, survey and statistical method, or estimation method to calculate the pollutant discharge of other sewage outlets into the river within the calculation range; call the calculation model to calculate the predicted mixed concentration of key sections of the receiving water body, in order to C ≤ C s To constrain the allowable emission concentration of the sewage outlet to be demonstrated in the reverse calculation If the reduction rate exceeds 100%, then the emission data of surrounding sewage outlets will be included in the constraint set, and the reduction and substitution accounting will be performed cyclically until the total emission of all sewage outlets within the calculation range meets the requirement. C ≤ C s The hydrological parameters were switched to dry season and normal water season respectively, and the daily and annual discharges during special periods and non-special periods were calculated using the same method.

6. The method according to claim 5, characterized in that, When multiple sewage discharge entities share the same sewage outlet into a river, S31 to S34 are executed first to obtain the total daily and annual discharge of key pollutants from the shared sewage outlet. Then, it is determined whether the receiving water body does not exceed the standards and whether each responsible entity holds a sewage discharge permit specifying the permitted discharge limit. If the conditions are met, the sum of the permitted annual discharge limits of each responsible entity is taken as the total annual discharge of the shared sewage outlet. If the conditions are not met, an intelligent allocation method is selected.

7. The method according to claim 6, characterized in that, The intelligent selection method is chosen from the following three options for allocation: First, the allocation is based on the proportion of wastewater discharge. The deviation between the actual wastewater discharge and the designed wastewater discharge of each wastewater discharge unit is compared. If the deviation exceeds 20%, the proportion is calculated using the average annual wastewater discharge of the past three years. Otherwise, the proportion is calculated using the designed wastewater discharge. The total discharge is then allocated according to the proportion. Second, the allocation is based on the proportion of emissions of key pollutants. The emissions of multiple key pollutants of each polluting unit are statistically analyzed and ranked. The top three emissions are selected to calculate the weighted average proportion, and the total emissions are allocated according to this value. Third, the allocation is based on the proportion of heavy metals and toxic and hazardous substances emitted. Heavy metals and toxic and hazardous substances are included in the category of key pollutants and allocated according to the emission proportion of various substances or the weighted average. If only some polluting units emit a certain type of substance, the responsibility proportion of those polluting units shall not be less than 60%.

8. The method according to claim 7, characterized in that, The three calculation paths in step S4 are as follows: S41, when using fully automatic path calculation, the path planning engine driven by the large language model automatically loads the parameters into the matching model and performs calculations based on the mapping relationship between the structured parameter set and the knowledge graph, and automatically generates structured calculation results; S42, when using a semi-automatic assisted calculation path, construct an interactive calculation interface to support users in supplementing or modifying constraints, allowing user input values ​​to overwrite the corresponding parameter fields extracted by S1, recommending an appropriate calculation model based on the updated parameter set, and executing the calculation after user confirmation; S43, when using manual calculation path, constructs an interactive calculation interface, allowing users to manually input constraints, select calculation models, perform calculations step by step, view intermediate results, and finally generate calculation results.

9. The method according to claim 8, characterized in that, Step S5 specifically includes: integrating the verified calculation results, process data and model parameters to generate a standardized calculation report that includes parameter sources, calculation process and compliance descriptions, and supports exporting to PDF or Word format; The calculation report, process data, model parameters and judgment conclusions are persistently stored in a relational database in a structured format. A data index is established with the calculation task ID as the primary key, which supports data traceability, historical query and secondary call for subsequent calculations.

10. A system for calculating pollutant discharge from river discharge outlets based on artificial intelligence, used to execute the method according to any one of claims 1-9, characterized in that, The system includes: a document parsing module, a model matching module, a differential measurement module, a path routing module, and a report generation and storage module; The document parsing module is configured to parse multiple original documents containing demonstration reports and monitoring reports on the setting up of sewage outlets into rivers, extract key elements such as hydrological parameters, sewage outlet discharge requirements, and functional objectives of receiving water bodies, and output a structured parameter set after normalization and conflict detection. The model matching module is configured to call a pre-built multimodal model library, which includes a hydrodynamic-water quality coupling model, an empirical parameter library, and a historical case library, and intelligently matches the optimal calculation model based on hydrological conditions and regulatory requirements. The differentiated calculation module is configured to determine whether the receiving water body meets the standards based on the structured parameter set, and to construct a differentiated calculation path in combination with the discharge permit information of the responsible entity. It performs discharge calculations under the scenarios of receiving water body meeting and not meeting the standards, respectively, to obtain the daily discharge volume during special periods, the daily discharge volume during non-special periods, and the annual discharge volume. The path routing module is configured to route the execution flow to one of the fully automatic, semi-automatic assisted, or manual calculation paths according to the calculation mode specified by the user, and call the optimal calculation model matched in the multimodal model library to complete the calculation of the discharge of key pollutants from the sewage outlet into the river. The report generation and storage module is configured to generate a measurement report that includes the source of parameters, the basis for model selection, the calculation process and compliance description, and to persistently store the data of the entire measurement process, supporting full-process traceability.

Images