Self-evolution-based water quality adjusting method and device

Abstract

The application provides a water quality adjusting method and device based on self-evolution, relates to the field of water conservancy informatization technology, and comprises the following steps: performing semantic understanding and task analysis according to a current water area state vector and current comprehensive environment data to obtain structured monitoring task description information; automatically constructing corresponding water quality analysis tools according to the structured monitoring task description information to obtain a water quality analysis tool set; performing numerical processing on current comprehensive monitoring data through the water quality analysis tool set to obtain a current water quality numerical result vector; updating the current water area state vector through the current water quality numerical result vector and a current adjusting behavior vector to obtain an updated water area state vector at the next moment; dynamically evolving a self-evolution water quality simulation model according to the updated water area state vector at the next moment to obtain an updated adjusting behavior vector at the next moment, and performing a water quality adjusting task at the next moment according to the updated adjusting behavior vector at the next moment.

Description

Technical Field

[0001] This application relates to the field of water conservancy information technology, and in particular to a water quality regulation method and device based on self-evolution. Background Technology

[0002] With the rapid development of artificial intelligence technology, large language models have shown broad application prospects in the field of water quality regulation. Among the existing technologies, water quality monitoring and decision-making methods based on large language models have emerged, which utilize their semantic understanding and reasoning capabilities to process multi-source heterogeneous water quality data and assist in completing tasks such as water quality assessment, pollution early warning, and generation of regulation and control recommendations.

[0003] Existing water quality regulation methods based on large language models typically use these models directly for numerical computation, leading to a lack of decoupling between semantic reasoning and numerical calculation, resulting in unreliable water quality analysis. Furthermore, they rely on statically predefined water quality analysis tools, which cannot be dynamically generated according to monitoring tasks, resulting in poor adaptability. The lack of a two-way interaction and closed-loop optimization mechanism between water state and regulation decisions makes it difficult to continuously improve regulation strategies based on the effects of regulation, leading to low accuracy and robustness of water quality regulation and hindering the effective implementation of water quality regulation tasks. Summary of the Invention

[0004] To address the aforementioned problems, in a first aspect, the present invention provides a self-evolving water quality regulation method, comprising: Construct a self-evolving water quality simulation model, obtain the current water state vector and current regulation behavior vector generated by the self-evolving water quality simulation model at the current moment, as well as the current comprehensive environmental data of the target water area at the current moment; perform semantic understanding and task parsing based on the current water state vector and current comprehensive environmental data to obtain structured monitoring task description information; Automatically construct corresponding water quality analysis tools based on the structured monitoring task description information to obtain a set of water quality analysis tools; Obtain the current comprehensive monitoring data of the target water area at the current moment, and perform numerical processing on the current comprehensive monitoring data through a set of water quality analysis tools to obtain the current water quality numerical result vector; The current water state vector is updated by using the current water quality numerical result vector and the current regulation behavior vector to obtain the updated water state vector for the next moment. The self-evolving water quality simulation model is dynamically evolved based on the updated water state vector at the next moment to obtain the updated regulation behavior vector at the next moment, and the water quality regulation task at the next moment is carried out based on the updated regulation behavior vector at the next moment.

[0005] In one embodiment, the self-evolving water quality simulation model includes a water area state update unit, a large language model, and multiple water quality monitoring decision-making entities; The water state update unit includes a water quality semantic analysis function, a behavior influence function, and a result mapping function; The water state update unit is used to dynamically update the water state vector. The water state vector includes water state data for each water zone, and the water state data includes water quality data, flow data, water level data, and sewage discharge data. The water quality monitoring decision-making body is used to generate control behaviors based on the water state vector, and the control behaviors of all water quality monitoring decision-making bodies are combined into a control behavior vector.

[0006] In one embodiment, semantic understanding and task parsing are performed based on the current water state vector and current comprehensive environmental data to obtain structured monitoring task description information, including: The current water state vector and current comprehensive environmental data are input into the water quality semantic analysis function for semantic understanding and task parsing to obtain structured monitoring task description information; The structured monitoring task description information includes monitoring objectives, a set of first constraints, and an execution path. The monitoring objectives include water quality improvement targets and risk warning indicators. The set of first constraints includes time constraints, data access constraints, and execution security constraints. The execution path includes the execution order and dependencies of water quality analysis steps.

[0007] In one embodiment, a corresponding water quality analysis tool is automatically constructed based on the structured monitoring task description information to obtain a set of water quality analysis tools, including: The second set of constraints is determined based on the monitoring objectives, the first set of constraints, and the execution path described in the structured monitoring task description information. The second set of constraints includes syntax validity constraints, causal correctness constraints, numerical stability constraints, and interface compliance constraints. The structured monitoring task description information is converted into generation prompts. The large language model generates water quality analysis tools that satisfy the second set of constraints based on the generation prompts. The types of water quality analysis tools include comprehensive water quality indicator index calculation tools, data preprocessing tools, dissolved oxygen saturation calculation tools, pollution source contribution assessment tools, and water quality trend prediction tools. All of these water quality analysis tools constitute the water quality analysis tool set.

[0008] In one embodiment, the current comprehensive monitoring data is numerically processed using a set of water quality analysis tools to obtain a current water quality numerical result vector, including: The current comprehensive monitoring data is input into each water quality analysis tool in the set of water quality analysis tools for numerical mapping, and the current water quality numerical results corresponding to each water quality analysis tool are obtained. All current water quality numerical results are arranged into a vector according to the sequence number of the water quality analysis tool.

[0009] In one embodiment, the current water state vector is updated using the current water quality numerical result vector and the current regulation behavior vector to obtain an updated water state vector, including: The current regulatory behavior vector is mapped to the regulatory influence change vector through the behavior influence function, and the current water quality numerical result vector is mapped to the water quality change vector through the result mapping function. Set the first response coefficient and the second response coefficient, and obtain the random perturbation term at the current time. The water state change vector is calculated based on the first response coefficient, the change vector of the control effect, the second response coefficient, the water quality change vector, and the random disturbance term. Add the current water state vector to the water state change vector to obtain the updated water state vector.

[0010] In one embodiment, after performing the water quality regulation task for the next time step based on the updated regulation behavior vector for the next time step, the method further includes: The updated water state vector at the next moment is used as the current water state vector at the next moment, and the updated regulation behavior vector at the next moment is used as the current regulation behavior vector at the next moment, forming a closed-loop self-evolution of the self-evolutionary water quality simulation model.

[0011] Secondly, the present invention provides a self-evolutionary water quality regulation device for implementing the aforementioned self-evolutionary water quality regulation method, the device comprising: The structured monitoring task description information acquisition module is used to construct a self-evolving water quality simulation model, acquire the current water state vector and current regulation behavior vector generated by the self-evolving water quality simulation model at the current moment, as well as the current comprehensive environmental data of the target water area at the current moment; and perform semantic understanding and task parsing based on the current water state vector and current comprehensive environmental data to obtain the structured monitoring task description information. The water quality analysis tool set acquisition module is used to automatically construct the corresponding water quality analysis tools based on the structured monitoring task description information, and obtain the water quality analysis tool set; The current water quality numerical result vector acquisition module is used to acquire the current comprehensive monitoring data of the target water area at the current moment, and to perform numerical processing on the current comprehensive monitoring data through a set of water quality analysis tools to obtain the current water quality numerical result vector. The updated water state vector acquisition module is used to update the current water state vector using the current water quality numerical result vector and the current regulation behavior vector, so as to obtain the updated water state vector at the next moment. The water quality regulation task execution module is used to dynamically evolve the self-evolving water quality simulation model based on the updated water state vector at the next moment, obtain the updated regulation behavior vector at the next moment, and perform the water quality regulation task at the next moment based on the updated regulation behavior vector at the next moment.

[0012] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the self-evolution-based water quality regulation method.

[0013] Fourthly, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the self-evolution-based water quality regulation method described above.

[0014] The present invention has the following beneficial effects: 1. By constructing a self-evolving water quality simulation model, and then sequentially executing semantic understanding and task parsing, dynamic generation of water quality analysis tools, numerical processing, state updates, and dynamic evolution, a complete closed-loop iteration from the current water state vector to the next moment's regulatory behavior vector is formed. This process completely decouples semantic reasoning from numerical computation, avoiding the unreliability of water quality analysis caused by existing large language models directly participating in numerical calculations. By integrating multi-source heterogeneous information such as comprehensive environmental data and comprehensive monitoring data, water quality analysis tools are dynamically generated on demand according to monitoring tasks, thereby overcoming the shortcomings of insufficient adaptability of static predefined tools. By feeding back the updated water state vector to the model and continuously evolving it, bidirectional interaction and closed-loop self-optimization between water quality regulation decisions and changes in the water environment state are realized, significantly improving the accuracy and robustness of water quality regulation.

[0015] 2. In the water quality analysis tool generation stage, a second set of constraints, consisting of syntactic validity, causal correctness, numerical stability, and interface compliance, is introduced to perform multiple verifications on the generated water quality analysis tools. This ensures that the tool code produced by the large language model is syntactically correct, avoids look-ahead bias in terms of timing, and can safely handle boundary conditions numerically. Furthermore, it standardizes input and output interface specifications, significantly improving the reliability and execution stability of the generated tools. It also frees the large language model from unreliable numerical computation, allowing it to focus on code generation and task planning, thus achieving deep decoupling between semantic reasoning and numerical computation. The constraint verification mechanism allows for the selection of stable tools for reuse, reducing the overhead of repeated generation. In dynamic scenarios such as sudden pollution and seasonal hydrological changes, it can quickly and reliably adapt to analysis tasks, significantly enhancing the generalization ability and automation level of water quality regulation.

[0016] 3. By mapping the regulatory behavior vector to a regulatory influence change vector through a behavior influence function, and using a result mapping function to convert the numerical water quality result vector into a water quality change vector, the first response coefficient, the second response coefficient, and a random disturbance term are introduced. The water state change vector is then accurately calculated using a weighted linear combination method, ultimately updating the current water state vector. This quantitatively decouples the water quality regulation behavior from the tool calculation results during the state update. The two response coefficients can be independently adjusted according to actual needs, thus flexibly balancing the contribution of human regulation measures and analysis results to environmental evolution. The addition of the random disturbance term simulates natural uncontrollable factors such as rainfall runoff and hydrological fluctuations, elevating the water quality simulation environment from a traditional deterministic model to an evolutionary system with real randomness, significantly enhancing the robustness of the self-evolving water quality simulation model to noise interference from actual water areas. The state changes in each iteration are clearly traceable, effectively ensuring the stability and accuracy of the self-evolving water quality simulation model in long-term evolution. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a structural diagram of the device according to an embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0020] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0021] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0022] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0023] Reference Figure 1 The present invention provides a water quality regulation method based on self-evolution, including steps S100-S500.

[0024] The S100 constructs a self-evolving water quality simulation model, obtains the current water state vector and current regulation behavior vector generated by the self-evolving water quality simulation model at the current moment, as well as the current comprehensive environmental data of the target water area at the current moment; based on the current water state vector and the current comprehensive environmental data, semantic understanding and task parsing are performed to obtain structured monitoring task description information.

[0025] In some embodiments, the self-evolving water quality simulation model includes a water state update unit, a large language model, and multiple water quality monitoring decision-making entities; The water state update unit includes a water quality semantic analysis function, a behavior influence function, and a result mapping function; The water state update unit is used to dynamically update the water state vector. The water state vector includes water state data for each water zone, and the water state data includes water quality data, flow data, water level data, and sewage discharge data. The water quality monitoring decision-making body is used to generate control behaviors based on the water state vector, and the control behaviors of all water quality monitoring decision-making bodies are combined into a control behavior vector.

[0026] In some embodiments, the water state update unit is the core computational module of the self-evolving water quality simulation model, which contains three key functions: water quality semantic analysis function. Behavioral Influence Function and result mapping function The water state update unit is responsible for dynamically updating the water state vector, that is, calculating the water state vector at the next moment based on the current water quality status, regulation behavior, and environmental input.

[0027] Water state vector It describes the target water area at a given time. A vector representing the overall environmental state, including water state data for each water area partition. Assume the target water area has a total of... Each water area is divided into sections (e.g., upstream section, midstream section, downstream section), the first... The water status data for each water zone includes water quality data, flow rate data, water level data, and wastewater discharge data. Specifically, water quality data includes dissolved oxygen (DO) (mg / L), ammonia nitrogen (NH3-N) (mg / L), total phosphorus (TP) (mg / L), chemical oxygen demand (COD) (mg / L), and pH (dimensionless), etc.; flow rate data... (Unit: m) / s); water level data (Unit: m); Wastewater discharge data This includes the discharge volume from each sewage outlet. Therefore, the current water state vector at time t... It can be represented as:

[0028] in, Indicates the first Water status data for each water zone.

[0029] Using a specific watershed as the simulation object, the watershed is approximately 120 kilometers long and has three water quality monitoring sections, two sewage outlets, and one sewage treatment plant distributed along the river. The initial water state vector is set based on the statistical characteristics of the watershed's historical water quality monitoring data over the past three years, and the initial water state vector is defined as follows. for:

[0030] in, This represents the total number of state variables in the water area. Indicates the first The initial values ​​of the water area status data.

[0031] The water quality monitoring decision-making entity is an intelligent agent responsible for generating regulatory behaviors within a self-evolving water quality simulation model. Each entity is driven by a large language model instance and possesses independent capabilities for aquatic environmental state perception, water quality semantic analysis, regulatory decision generation, and tool invocation. Each water quality monitoring decision-making entity operates at any given time... Receive the current water state vector It uses an internal large language model for semantic understanding and decision-making reasoning to generate targeted regulatory behaviors. .

[0032] The regulatory behaviors of all water quality monitoring decision-making entities are arranged in ascending order of entity number, forming the current regulatory behavior vector. : in, The number of decision-making bodies for water quality monitoring; Indicates the first The initial regulatory behavior of each entity, such as water diversion flow rate and wastewater treatment reduction rate, is used as one of the inputs to the water state update unit, participating in the calculation of the water state at the next time step.

[0033] The initial control behavior vector consists of the initial control behaviors of multiple water quality monitoring decision-making entities. Recorded as: in, Indicates the first The initial regulatory behaviors of individual entities, such as water diversion flow rate and sewage treatment reduction rate.

[0034] In this embodiment, five water quality monitoring decision-making bodies are configured, each corresponding to a different management function: Entity 1: The main body responsible for water quality supervision in the basin, including macro-control such as water diversion and regional total pollutant discharge control; Entity 2: The wastewater treatment plant operator, responsible for adjusting treatment processes and improving pollutant reduction rates; Entity 3: The polluting enterprise, responsible for controlling the concentration and flow rate of its own discharge outlets; Entity 4: The main body responsible for ecological restoration, including wetland purification, ecological water replenishment, and other natural restoration measures; Entity 5: Emergency response entity, responsible for emergency interception, adsorption and treatment in the event of a sudden pollution incident.

[0035] In some embodiments, semantic understanding and task parsing are performed based on the current water state vector and current integrated environmental data to obtain structured monitoring task description information, including: The current water state vector and current comprehensive environmental data are input into the water quality semantic analysis function for semantic understanding and task parsing to obtain structured monitoring task description information; The structured monitoring task description information includes monitoring objectives, a set of first constraints, and an execution path. The monitoring objectives include water quality improvement targets and risk warning indicators. The set of first constraints includes time constraints, data access constraints, and execution security constraints. The execution path includes the execution order and dependencies of water quality analysis steps.

[0036] In some embodiments, the system obtains the current water state vector generated by the self-evolving water quality simulation model at the current time t. and current regulatory behavior vector Simultaneously, acquire the current comprehensive environmental data of the target water area at the current time t. The current comprehensive environmental data includes meteorological forecast data (such as rainfall and temperature in the next 7 days), pollution source emission data (such as real-time concentration and flow rate of sewage outlets), and water diversion and dispatch instructions (such as the adjustable water volume of upstream reservoirs).

[0037] Input the current water state vector and current comprehensive environmental data into the water quality semantic analysis function. : The water quality semantic analysis function first performs multi-level semantic understanding and task parsing on the water quality data, flow rate and water level data, and sewage discharge data of each water area in the current water state vector. This includes watershed-level water quality trend analysis (calculating the 30-day moving average and trend direction of each water quality parameter), cross-sectional water quality fluctuation analysis (identifying short-term abnormal water quality fluctuations and key parameter mutation points), and water quality parameter correlation analysis (mining the correlation between different water quality parameters). Then, the semantic understanding and task parsing results are fused with the current comprehensive environmental data to generate structured monitoring task description information. .

[0038] Structured monitoring task description information The framework includes monitoring objectives, a set of primary constraints, and an implementation path. Monitoring objectives include water quality improvement targets and risk warning indicators, such as "reducing ammonia nitrogen concentration at the midstream section to below Class III standards" or "predicting the dissolved oxygen trend at the downstream section over the next 24 hours." The set of primary constraints includes time constraints (e.g., analysis results must be returned within 5 minutes), data access constraints (using only current and historical monitoring data to avoid forward bias), and implementation safety constraints (e.g., control measures must not cause water quality degradation at other sections). The implementation path includes the execution sequence and dependencies of water quality analysis steps, such as first performing data preprocessing, then calculating water quality indicators, then performing pollution source contribution analysis, and finally generating control recommendations.

[0039] S200 automatically constructs corresponding water quality analysis tools based on the structured monitoring task description information, thus obtaining a set of water quality analysis tools.

[0040] In some embodiments, a corresponding water quality analysis tool is automatically constructed based on the structured monitoring task description information to obtain a set of water quality analysis tools, including: The second set of constraints is determined based on the monitoring objectives, the first set of constraints, and the execution path described in the structured monitoring task description information. The second set of constraints includes syntax validity constraints, causal correctness constraints, numerical stability constraints, and interface compliance constraints. The structured monitoring task description information is converted into generation prompts. The large language model generates water quality analysis tools that satisfy the second set of constraints based on the generation prompts. The types of water quality analysis tools include comprehensive water quality indicator index calculation tools, data preprocessing tools, dissolved oxygen saturation calculation tools, pollution source contribution assessment tools, and water quality trend prediction tools. All of these water quality analysis tools constitute the water quality analysis tool set.

[0041] In some embodiments, the second set of constraints includes: syntactic validity constraints: the generated tool code must conform to syntactic specifications and be correctly parsed by the interpreter; causal correctness constraints: the tool must not use predicted data from future moments during the calculation process to avoid look-ahead bias and ensure the temporal rationality of the analysis results; numerical stability constraints: the tool must handle boundary conditions during execution to avoid numerical overflow, division by zero errors, and logarithmic calculation anomalies in water quality parameter calculations; and interface compliance constraints: the tool must follow a unified input / output interface specification to ensure composability and scalability.

[0042] The structured monitoring task description information is converted into a generation prompt, which includes the task description, constraints, available data types, and output requirements.

[0043] A collection of water quality analysis tools generated from large language models : in, Indicates the generated first A water quality analysis tool, This represents the total number of tools. The types of water quality analysis tools include: Comprehensive Water Quality Index (WQI) calculation tools: used to calculate water quality evaluation indicators such as the WQI, single-factor evaluation index, and eutrophication index; Data preprocessing tools: used to preprocess raw monitoring data (outlier removal, missing value imputation, data normalization) or extract features; Dissolved oxygen saturation calculation tools: used to calculate the current dissolved oxygen saturation of the water body; Pollution source contribution assessment tools: used to assess the contribution of different pollution sources to the water quality of the monitoring section and the expected effects of control measures; Water quality trend prediction tools: used to predict future water quality change trends based on historical water quality data and hydrological conditions.

[0044] The S300 acquires the current comprehensive monitoring data of the target water area at the current moment, and performs numerical processing on the current comprehensive monitoring data through a set of water quality analysis tools to obtain the current water quality numerical result vector.

[0045] In some embodiments, the current comprehensive monitoring data is numerically processed using a set of water quality analysis tools to obtain a vector of current water quality numerical results, including: The current comprehensive monitoring data is input into each water quality analysis tool in the set of water quality analysis tools for numerical mapping, and the current water quality numerical results corresponding to each water quality analysis tool are obtained. All current water quality numerical results are arranged into a vector according to the sequence number of the water quality analysis tool.

[0046] In some embodiments, current integrated monitoring data of the target water area at the current moment is obtained. The current comprehensive monitoring data includes water quality monitoring data (real-time values ​​of water quality parameters at each cross-section), hydrological data (flow rate, water level, and flow velocity), and meteorological data (temperature and rainfall). This is achieved through a set of water quality analysis tools. Numerical processing is performed on the current comprehensive monitoring data to obtain the current water quality numerical result vector. :

[0047] In practice, the current comprehensive monitoring data is input into each water quality analysis tool in the set of water quality analysis tools for numerical mapping, obtaining the current water quality numerical result corresponding to each water quality analysis tool. All current water quality numerical results are then arranged according to the sequence number of the water quality analysis tool to form a current water quality numerical result vector. All numerical calculations are performed by the generated water quality analysis tools. The large language model is only responsible for water quality monitoring task planning, interpretation of analysis results, and generation of control suggestions; it does not directly participate in numerical calculations.

[0048] Taking this embodiment as an example, the data preprocessing tool first imputes missing values ​​and replaces outliers in the original monitoring data, outputting cleaned data. The comprehensive water quality index calculation tool, based on the cleaned data, calculates the WQI value of the midstream section to be 108.63, classifying it as moderate or higher pollution; the dissolved oxygen saturation calculation tool calculates the dissolved oxygen saturation to be 66.8%, indicating an oxygen deficit; the pollution source contribution assessment tool calculates that the suspected underground discharge contributes 45.3% of the total increase in pollution discharge; and the water quality trend prediction tool predicts that the dissolved oxygen in the downstream section will continue to decline over the next three days. These calculation results collectively form the current water quality numerical result vector. .

[0049] S400 updates the current water state vector by using the current water quality numerical result vector and the current control behavior vector, and obtains the updated water state vector for the next moment.

[0050] In some embodiments, the current water state vector is updated using the current water quality numerical result vector and the current regulation behavior vector to obtain the updated water state vector, including: The current regulatory behavior vector is mapped to the regulatory influence change vector through the behavior influence function, and the current water quality numerical result vector is mapped to the water quality change vector through the result mapping function. Set the first response coefficient and the second response coefficient, and obtain the random perturbation term at the current time. The water state change vector is calculated based on the first response coefficient, the change vector of the control effect, the second response coefficient, the water quality change vector, and the random disturbance term. Add the current water state vector to the water state change vector to obtain the updated water state vector.

[0051] In some embodiments, each water quality monitoring decision-making body bases its decisions on the current water state vector. Independent generation of regulatory behavior ,in , The number of decision-making entities for water quality monitoring, and all regulatory actions constitute the current regulatory action vector. .

[0052] Through behavioral influence function For each water quality monitoring decision-making entity, the current control behavior is used to calculate the pollutant load change, with the first... Taking a single water area as an example, the pollutant load change resulting from the current regulation behavior vector is calculated using a centralized water quality regulation and coordination engine. : in, Indicates the first The water quality monitoring decision-making body in the first The current control measures will be implemented in each water area zone. Subsequently, the pollutant flux introduced (unit: kg / d). Indicates the first The water quality monitoring decision-making body in the first The current control measures will be implemented in each water area zone. Then, the amount of pollutant flux removed.

[0053] Changes in pollutant load Normalization to the change in the effect of regulation : in, The water quality impact coefficient is determined based on historical watershed data; in this embodiment, it is taken as... ; For the first Each water zone at time Average daily flow (unit: ); The function compresses the normalized load change to an interval. This is to avoid excessive influence from a single entity that could lead to sudden state changes.

[0054] By combining the changes in the regulatory impact across all water zones, we obtain the vector of changes in the regulatory impact. : in This represents the total number of water area zones in this embodiment. This corresponds to the upstream, midstream, and downstream sections.

[0055] Result mapping function Based on the deviation between the water quality analysis results and the water quality targets, water quality adjustment amounts are generated for each water area zone, for example: If the Comprehensive Water Quality Index (WQI) exceeds the Class III standard threshold (WQI>80), it indicates that the water quality exceeds the standard. The mapping function outputs a negative change, meaning that the pollutant concentration needs to be reduced. If the dissolved oxygen saturation is below 80%, it indicates that the water body is hypoxic, and the resulting mapping function outputs a negative change, meaning that dissolved oxygen needs to be increased. If the pollution source contribution analysis finds that the contribution of a certain pollution source is too high, the mapping function will output the corresponding reduction amount. If water quality forecasts indicate future deterioration, the resulting mapping function outputs a preventative adjustment amount in advance.

[0056] Specifically, for the first The first water zone For a given water quality parameter (such as ammonia nitrogen concentration), the resulting mapping function can be expressed as: in, For the current number The first partition Measured concentrations of several water quality parameters; The target concentration for this parameter, such as the Class III standard limit; To adjust the coefficients, this embodiment uses values ​​between 0.05 and 0.2, depending on the parameter sensitivity; combining the changes in all zones and all parameters yields the water quality change vector. .

[0057] First response coefficient Used to regulate the intensity of the impact of control behaviors on water quality. The larger the value, the more significant the immediate effect of the main regulatory behavior. In this embodiment, the calibration is based on historical regulation data of the Qingjiang River Basin. .

[0058] Second response coefficient Used to adjust the intensity of the influence of tool calculation results on water quality status. The higher the value, the greater the contribution of the water quality analysis tool's recommendations to the status update. In this embodiment, we take... This reflects the design principle that large language models do not directly participate in numerical calculations, but are only indirectly affected by tool results.

[0059] random disturbance term This is used to simulate the impact of uncontrollable factors such as natural hydrological fluctuations, rainfall runoff, and water temperature changes on water quality. In this embodiment, the random disturbance term follows a normal distribution with a mean of 0:

[0060] Standard deviation Determined based on natural fluctuation statistics of historical monitoring data, taking... Independent sampling at each time step The value.

[0061] Substituting the above quantities into the linear combination formula yields the vector of water state changes. : In this formula, each term is in vector form, corresponding to the same dimension of the water state variable.

[0062] Add the corresponding elements of the current water state vector to the water state change vector to obtain the updated water state vector for the next time step. : For example, the ammonia nitrogen concentration at the midstream section is calculated as follows: Corresponding item: Corresponding item: Sampled values: therefore: The updated results for ammonia nitrogen concentration at the midstream section are as follows: In this way, the system organically combines the calculation results of water quality analysis tools with the regulatory behavior of multiple agents, and realizes the adaptive evolution of water state under random natural disturbance, providing accurate state input for the closed-loop self-evolution in the next moment.

[0063] The S500 dynamically evolves the self-evolving water quality simulation model based on the updated water state vector at the next moment, obtains the updated regulation behavior vector at the next moment, and performs the water quality regulation task at the next moment based on the updated regulation behavior vector at the next moment.

[0064] In some embodiments, after performing the water quality regulation task for the next time step based on the updated regulation behavior vector for the next time step, the method further includes: The updated water state vector at the next moment is used as the current water state vector at the next moment, and the updated regulation behavior vector at the next moment is used as the current regulation behavior vector at the next moment, forming a closed-loop self-evolution of the self-evolutionary water quality simulation model.

[0065] In some embodiments, the system repeats the above process continuously. After each round of water quality regulation task—that is, a complete state update and regulation behavior generation—the system automatically uses the latest updated water state vector as the current water state vector for the next round, and the latest updated regulation behavior vector as the current regulation behavior vector for the next round. Simultaneously, it acquires new comprehensive environmental data and comprehensive monitoring data, and re-executes semantic understanding and task parsing, tool generation, numerical processing, and state updates. This mechanism enables the self-evolving water quality simulation model to autonomously adjust its monitoring strategies and regulation schemes under the drive of changes in water quality state and environmental conditions, without human intervention.

[0066] After each iteration, the system determines whether preset convergence conditions are met. These conditions include: all key water quality parameters at monitoring sections meet the Class III limits of the Surface Water Environmental Quality Standard; and the magnitude of the water state change vector over multiple consecutive steps is less than a preset threshold. When the convergence conditions are met, the system stops iterating, outputs the current water quality regulation strategy and analysis toolset as the final solution, and records the entire evolution trajectory in a historical database for strategy reuse in similar scenarios. If the convergence conditions are not met, the system automatically enters the next iteration and continues the complete process from semantic understanding and task parsing to state updates.

[0067] Reference Figure 2 This invention provides a self-evolutionary water quality regulation device 20 for implementing a self-evolutionary water quality regulation method. The device includes: The structured monitoring task description information acquisition module 21 is used to construct a self-evolving water quality simulation model, acquire the current water state vector and current regulation behavior vector generated by the self-evolving water quality simulation model at the current moment, as well as the current comprehensive environmental data of the target water area at the current moment; and perform semantic understanding and task parsing based on the current water state vector and current comprehensive environmental data to obtain the structured monitoring task description information. The water quality analysis tool set acquisition module 22 is used to automatically construct the corresponding water quality analysis tools based on the structured monitoring task description information and obtain the water quality analysis tool set. The current water quality numerical result vector acquisition module 23 is used to acquire the current comprehensive monitoring data of the target water area at the current moment, and to perform numerical processing on the current comprehensive monitoring data through a set of water quality analysis tools to obtain the current water quality numerical result vector. The updated water state vector acquisition module 24 is used to update the current water state vector by using the current water quality numerical result vector and the current regulation behavior vector, so as to obtain the updated water state vector at the next moment. The water quality regulation task execution module 25 is used to dynamically evolve the self-evolving water quality simulation model according to the updated water state vector at the next moment, obtain the updated regulation behavior vector at the next moment, and perform the water quality regulation task at the next moment according to the updated regulation behavior vector at the next moment.

[0068] This application provides an electronic device, including a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements any of the above-described schemes of a self-evolutionary water quality regulation method.

[0069] Specifically, the processor may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may be a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of this application.

[0070] Memory can be any medium capable of containing, storing, transmitting, propagating, or transmitting instructions. For example, memory can include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, instruments, or propagation media. Specific examples of memory include: magnetic storage devices such as magnetic tape or hard disk drives (HDDs); optical storage devices such as optical discs (CD-ROMs); and also random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0071] This application also provides a computer-readable medium storing a computer program thereon, which, when executed by a processor, implements the self-evolutionary water quality regulation method of any of the above-described schemes. This computer-readable medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The aforementioned computer-readable medium carries one or more programs, which, when executed, implement the method as described in the embodiments of this application.

[0072] According to embodiments of this application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wired, optical fiber, radio frequency signals, etc., or any suitable combination thereof.

[0073] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments and / or claims of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application. Therefore, the scope of this application should not be limited to the above embodiments, but should be defined not only by the appended claims, but also by their equivalents. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A water quality regulation method based on self-evolution, characterized in that, include: Construct a self-evolving water quality simulation model, obtain the current water state vector and current regulation behavior vector generated by the self-evolving water quality simulation model at the current moment, as well as the current comprehensive environmental data of the target water area at the current moment; perform semantic understanding and task parsing based on the current water state vector and current comprehensive environmental data to obtain structured monitoring task description information; Automatically construct corresponding water quality analysis tools based on the structured monitoring task description information to obtain a set of water quality analysis tools; Obtain the current comprehensive monitoring data of the target water area at the current moment, and perform numerical processing on the current comprehensive monitoring data through a set of water quality analysis tools to obtain the current water quality numerical result vector; The current water state vector is updated by using the current water quality numerical result vector and the current regulation behavior vector to obtain the updated water state vector for the next moment. The self-evolving water quality simulation model is dynamically evolved based on the updated water state vector at the next moment to obtain the updated regulation behavior vector at the next moment, and the water quality regulation task at the next moment is carried out based on the updated regulation behavior vector at the next moment.

2. The water quality regulation method based on self-evolution according to claim 1, characterized in that, The self-evolving water quality simulation model includes a water state update unit, a large language model, and multiple water quality monitoring decision-making entities; The water state update unit includes a water quality semantic analysis function, a behavior influence function, and a result mapping function; The water state update unit is used to dynamically update the water state vector. The water state vector includes water state data for each water zone, and the water state data includes water quality data, flow data, water level data, and sewage discharge data. The water quality monitoring decision-making body is used to generate control behaviors based on the water state vector, and the control behaviors of all water quality monitoring decision-making bodies are combined into a control behavior vector.

3. The water quality regulation method based on self-evolution according to claim 2, characterized in that, Based on the current water state vector and current comprehensive environmental data, semantic understanding and task parsing are performed to obtain structured monitoring task description information, including: The current water state vector and current comprehensive environmental data are input into the water quality semantic analysis function for semantic understanding and task parsing to obtain structured monitoring task description information; The structured monitoring task description information includes monitoring objectives, a set of first constraints, and an execution path. The monitoring objectives include water quality improvement targets and risk warning indicators. The set of first constraints includes time constraints, data access constraints, and execution security constraints. The execution path includes the execution order and dependencies of water quality analysis steps.

4. The water quality regulation method based on self-evolution according to claim 2, characterized in that, Based on the structured monitoring task description information, corresponding water quality analysis tools are automatically constructed to obtain a set of water quality analysis tools, including: The second set of constraints is determined based on the monitoring objectives, the first set of constraints, and the execution path described in the structured monitoring task description information. The second set of constraints includes syntax validity constraints, causal correctness constraints, numerical stability constraints, and interface compliance constraints. The structured monitoring task description information is converted into generation prompts. The large language model generates water quality analysis tools that satisfy the second set of constraints based on the generation prompts. The types of water quality analysis tools include comprehensive water quality indicator index calculation tools, data preprocessing tools, dissolved oxygen saturation calculation tools, pollution source contribution assessment tools, and water quality trend prediction tools. All of these water quality analysis tools constitute the water quality analysis tool set.

5. The water quality regulation method based on self-evolution according to claim 2, characterized in that, By using a suite of water quality analysis tools to numerically process the current comprehensive monitoring data, a vector of current water quality numerical results is obtained, including: The current comprehensive monitoring data is input into each water quality analysis tool in the set of water quality analysis tools for numerical mapping, and the current water quality numerical results corresponding to each water quality analysis tool are obtained. All current water quality numerical results are arranged into a vector according to the sequence number of the water quality analysis tool.

6. The water quality regulation method based on self-evolution according to claim 2, characterized in that, The current water state vector is updated by combining the current water quality numerical result vector and the current regulation behavior vector, resulting in an updated water state vector, including: The current regulatory behavior vector is mapped to the regulatory influence change vector through the behavior influence function, and the current water quality numerical result vector is mapped to the water quality change vector through the result mapping function. Set the first response coefficient and the second response coefficient, and obtain the random perturbation term at the current time. The water state change vector is calculated based on the first response coefficient, the change vector of the control effect, the second response coefficient, the water quality change vector, and the random disturbance term. Add the current water state vector to the water state change vector to obtain the updated water state vector.

7. The water quality regulation method based on self-evolution according to claim 2, characterized in that, After performing the water quality regulation task for the next time step based on the updated regulatory behavior vector for the next time step, it also includes: The updated water state vector at the next moment is used as the current water state vector at the next moment, and the updated regulation behavior vector at the next moment is used as the current regulation behavior vector at the next moment, forming a closed-loop self-evolution of the self-evolutionary water quality simulation model.

8. A self-evolutionary water quality regulation device, used to implement the self-evolutionary water quality regulation method as described in any one of claims 1 to 7, characterized in that, The device includes: The structured monitoring task description information acquisition module is used to construct a self-evolving water quality simulation model, acquire the current water state vector and current regulation behavior vector generated by the self-evolving water quality simulation model at the current moment, as well as the current comprehensive environmental data of the target water area at the current moment; and perform semantic understanding and task parsing based on the current water state vector and current comprehensive environmental data to obtain the structured monitoring task description information. The water quality analysis tool set acquisition module is used to automatically construct the corresponding water quality analysis tools based on the structured monitoring task description information, and obtain the water quality analysis tool set; The current water quality numerical result vector acquisition module is used to acquire the current comprehensive monitoring data of the target water area at the current moment, and to perform numerical processing on the current comprehensive monitoring data through a set of water quality analysis tools to obtain the current water quality numerical result vector. The updated water state vector acquisition module is used to update the current water state vector using the current water quality numerical result vector and the current regulation behavior vector, so as to obtain the updated water state vector at the next moment. The water quality regulation task execution module is used to dynamically evolve the self-evolving water quality simulation model based on the updated water state vector at the next moment, obtain the updated regulation behavior vector at the next moment, and perform the water quality regulation task at the next moment based on the updated regulation behavior vector at the next moment.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the self-evolution-based water quality regulation method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the self-evolution-based water quality regulation method as described in any one of claims 1 to 7.

Images