Search

CN-122022159-A - Small-river-basin non-point source pollution load dynamic evaluation and regulation method

CN122022159ACN 122022159 ACN122022159 ACN 122022159ACN-122022159-A

Abstract

The invention discloses a dynamic evaluation and regulation method for pollution load of a small watershed non-point source, which relates to the technical field of data processing and aims to solve the technical problem that evaluation distortion is caused by space-time variation of key environmental behavior parameters of a model so as to cause the lack of accuracy and effectiveness of a regulation scheme based on the model; S2, dynamically evaluating a non-point source pollution process model based on a digital twin body, wherein the operation mechanism and the data are driven by two, and S3, generating an adaptive regulation and control scheme in the digital twin body through a layered decision framework based on the corrected model state and a pollution load control target. According to the invention, by constructing a physical constraint transfer learning model, dynamic and high-precision inversion of the environmental behavior parameters of the key pollutants is realized, and the problem of evaluation distortion caused by space-time variation of model parameters is fundamentally solved.

Inventors

  • MA HAIXIA

Assignees

  • 定西市水土保持科学研究所

Dates

Publication Date
20260512
Application Date
20260128

Claims (10)

  1. 1. The small-river basin non-point source pollution load dynamic evaluation and regulation method is characterized by comprising the following steps of: S1, constructing a digital twin body of a small river basin, wherein the digital twin body is used for integrating geospatial data, real-time monitoring data and controllable water conservancy facility states; S2, dynamically evaluating an operation mechanism and data double-driven non-point source pollution process model based on the digital twin body, wherein the method comprises the following steps of: dynamically inverting environmental behavior parameters of the key pollutants through a physical constraint transfer learning model based on-line monitoring time sequence data and laboratory simulation measurement data; assimilating real-time monitoring data into the non-point source pollution process model by utilizing a data assimilation technology so as to continuously correct the model state; S3, generating an adaptive regulation and control scheme in the digital twin body through a layered decision architecture based on the corrected model state and the pollution load control target, wherein the layered decision architecture comprises a strategy layer based on scene simulation for long-term strategy selection, a tactical layer based on rolling optimization for medium-term instruction generation and an execution layer based on real-time feedback for instantaneous adjustment; And S4, issuing the self-adaptive regulation and control scheme to a corresponding controllable water conservancy facility for execution, and feeding back execution effect data to the digital twin body to form an evaluation and regulation closed loop.
  2. 2. The method for dynamically evaluating and regulating the pollution load of a small-river basin non-point source according to claim 1, wherein in step S1, the construction of the digital twin body specifically comprises: Accessing multi-source heterogeneous data comprising geospatial data, meteorological data, on-line monitoring sensing data and facility running state data; Carrying out standardization, fusion and space-time alignment treatment on the multi-source heterogeneous data to form a unified space-time data set for driving the digital twin; the on-line monitoring sensing data at least comprises flow, water level, turbidity, pH and specific pollutant concentration indexes from a key section.
  3. 3. The method for dynamically evaluating and regulating the pollution load of a small-river basin non-point source according to claim 1, wherein the environmental behavior parameters of the key pollutants are dynamically inverted, specifically: taking laboratory simulation measurement data obtained regularly as a target tag, and taking on-line monitoring time sequence data in a corresponding time period as an input characteristic; Training the physical constraint transfer learning model, and establishing a dynamic mapping relation from on-line monitoring data to environmental behavior parameters by minimizing a joint loss function, wherein the joint loss function is as follows: ; In the formula, For task loss function, model prediction results are measured With real labels Differences between; Punishment functions for physical constraints; Is an input feature vector; And Is a weight super parameter; And calculating and outputting dynamically updated environmental behavior parameters according to the real-time monitoring data by using the trained model.
  4. 4. The method for dynamically evaluating and regulating the pollution load of a small-basin non-point source according to claim 3, wherein the physical constraint penalty term Based on mass conservation law construction, the calculation formula is as follows: ; In the formula, A scalar field of concentration for the contaminant; Time is; is a water flow velocity vector field; Is a divergence operator; For parameters of predicted environmental behaviour And other input features The constraint forces inversion parameters to meet the framework of the convection-diffusion reaction equation, ensuring the physical consistency thereof.
  5. 5. The method for dynamically evaluating and regulating the pollution load of a small-river basin non-point source according to claim 1, wherein in step S2, the model of the non-point source pollution process driven by the mechanism and the data is a multi-scale coupling model, comprising: a microscale model for simulating a pollutant microscale migration and transformation process; the plot and channel scale model is used for simulating plot runoff, soil erosion, and a process of transporting pollutants along with sediment and growing in a channel; the small-river-basin-scale distributed hydrological water quality model is used for simulating the transfer and collection of pollution loads from a source region to a receiving water body space; Wherein the different scale models are coupled with feedback through the transmission of key state variables.
  6. 6. The method for dynamically evaluating and regulating the pollution load of a small-basin non-point source according to claim 1, wherein the model state is continuously corrected by using a data assimilation technology, specifically by using a set kalman filter algorithm, and the model state update formula is as follows: ; In the formula, Representing a state vector of the model forecast; representing an optimal estimated state vector obtained after data assimilation analysis; Representing an observation vector; is an observation operator; the method is a Kalman gain matrix, and the short-term prediction accuracy is improved by fusing real-time observation data and dynamically calibrating the model track.
  7. 7. The method for dynamically evaluating and regulating the pollution load of a small-river basin non-point source according to claim 1, wherein in step S3, the tactical layer is generated based on the instructions of rolling optimization, and a model predictive control algorithm is adopted, which solves the following optimization problem in each control period: ; constraint conditions: 、 、 ; In the formula, Representing an objective function that needs to be minimized; indexing for discrete time steps; the method comprises the steps of (1) setting a control instruction sequence to be optimized; To control the time domain; Future predicted for model state corrected according to the data assimilation technique State variables of the steps; Is the target of the expected pollution load or water quality state; And Is a diagonal weight matrix; representing a state transfer function of a mechanism and data double-drive non-point source pollution process model after data assimilation and correction; Input for a predicted future disturbance; And Respectively representing upper and lower limits of the amplitude of the control instruction; And The algorithm rolls and solves the optimal control instruction in a future period of time.
  8. 8. The method for dynamically evaluating and controlling pollution load of small basin area source according to claim 7, wherein in step S3, the fine tuning of the execution layer based on real-time feedback is implemented by reinforcement learning agent, which learns fine tuning strategy, action value function by maximizing cumulative rewards Is following the bellman optimal equation: ; In the formula, For the moment of time Environmental state of (2) Including the model predictive control algorithm at time Is of the predicted state of (1) And actual observation state Deviation of the controllable facility, and real-time status of the controllable facility; for the moment of time The action taken by the agent is the reference control instruction output by the model predictive control algorithm Fine tuning amount of (2); is shown in the state Execute action downwards The instant reward signal obtained later; Is a discount factor; And the intelligent agent on-line self-adaptively compensates model errors and unknown disturbance.
  9. 9. The method for dynamically evaluating and regulating the pollution load of the small-river basin non-point source according to claim 1 is characterized in that in the step S4, time sequence control instructions in the self-adaptive regulation scheme are specific parameters aiming at the opening and closing time, opening degree or running water level of a pond dam, a gate or an adjustable interception dam, and when a regulating object is a plurality of interception dams distributed, the self-adaptive regulation scheme is a cooperative regulation instruction set which is generated according to the space-time distribution of the pollution load and is matched with each other on water storage and drainage time.
  10. 10. A small-river basin non-point source pollution load dynamic evaluation and control system for implementing the small-river basin non-point source pollution load dynamic evaluation and control method as defined in any one of claims 1-9, comprising: The data sensing and collecting module is used for acquiring multi-source data required by constructing and driving the digital twin; The digital twin body constructing and maintaining module is used for integrating the multi-source data and establishing and maintaining a virtual mapping model of the small drainage basin; the dynamic evaluation model engine is internally provided with a mechanism and data double-driven non-point source pollution process model, a transfer learning parameter inversion unit and a data assimilation unit; the intelligent regulation and control decision module is internally provided with the layered decision architecture and is used for generating an adaptive regulation and control scheme; the control instruction issuing and executing feedback module is used for converting the regulation and control scheme into a control instruction and issuing the control instruction to the field facility, and collecting and executing state data and feeding back the data to the system; The dynamic evaluation model engine and the intelligent regulation and control decision module operate in the digital twin environment to form a closed loop.

Description

Small-river-basin non-point source pollution load dynamic evaluation and regulation method Technical Field The invention relates to the technical field of data processing, in particular to a dynamic evaluation and regulation method for small-basin non-point source pollution load. Background Non-point source pollution, especially nitrogen, phosphorus, pesticides and other pollutants caused by agricultural activities diffuse along with runoff, and become a main cause of environmental quality deterioration of small watershed water bodies. The occurrence of the method has randomness, universality and hysteresis, and the load assessment and effective regulation are all the problems in the field of environmental management. At present, the technical development in the field mainly depends on a physical mechanism-based distributed hydrologic water quality model (such as SWAT, HSPF and the like) for simulation evaluation, and combines engineering and non-engineering measures for regulation planning. However, in practical applications, especially for specific pesticides (such as triazoles, sulfonylureas, etc.) with strong adsorptivity and complex environmental behaviors, the prior art systems expose a long-standing core bottleneck that the key environmental behavior parameters of the model (such as adsorption-desorption coefficients, degradation rates, etc. of pollutants) have high space-time variability. The fixed parameters measured under laboratory conditions are difficult to accurately reflect the dynamic changes of the fixed parameters in different fields, different soil humidity, different temperature, different pH and different microbial environments. The fundamental contradiction between the static parameter and the dynamic environment leads to the inherent deviation of the model in simulating the pollutant migration and transformation process, so that the reliability of the load evaluation result is insufficient. The design of the regulation scheme based on the method is similar to planning a path on a misalignment map, and the scientificity and the effectiveness of the regulation scheme are greatly reduced. Therefore, how to realize dynamic and self-adaptive identification of key parameters of a model in a real environment becomes a scientific and technical problem that the precise evaluation and high-efficiency regulation and control capability of small-river basin non-point source pollution must be overcome. The prior art either relies on frequent and costly field large-scale sampling and laboratory analysis to update parameters without timeliness, or relies on historical data to perform data-driven modeling completely, lacks physical mechanism constraint, has weak extrapolation capability and poor interpretability, and cannot fundamentally solve the problems. In view of the above, we propose a dynamic evaluation and control method for small-river basin non-point source pollution load. Disclosure of Invention The invention aims to provide a dynamic evaluation and regulation method for small-river basin non-point source pollution load, which aims to solve the technical problems that evaluation distortion is caused by space-time variation of key environmental behavior parameters of a model, and further, a regulation scheme based on the model lacks accuracy and effectiveness. In order to solve the technical problems, the invention provides the following technical scheme that the method for dynamically evaluating and regulating the pollution load of the small-river basin non-point source comprises the following steps: S1, constructing a digital twin body of a small river basin, wherein the digital twin body is used for integrating geospatial data, real-time monitoring data and controllable water conservancy facility states; S2, dynamically evaluating an operation mechanism and data double-driven non-point source pollution process model based on the digital twin body, wherein the method comprises the following steps of: dynamically inverting environmental behavior parameters of the key pollutants through a physical constraint transfer learning model based on-line monitoring time sequence data and laboratory simulation measurement data; assimilating real-time monitoring data into the non-point source pollution process model by utilizing a data assimilation technology so as to continuously correct the model state; S3, generating an adaptive regulation and control scheme in the digital twin body through a layered decision architecture based on the corrected model state and the pollution load control target, wherein the layered decision architecture comprises a strategy layer based on scene simulation for long-term strategy selection, a tactical layer based on rolling optimization for medium-term instruction generation and an execution layer based on real-time feedback for instantaneous adjustment; And S4, issuing the self-adaptive regulation and control scheme to a corresponding controllable water conservancy fa