Intelligent regulation and control method and system for water transportation, distribution and back-supply in mining area

Abstract

The invention discloses an intelligent regulation and control method and system for water transportation, distribution and back supplement in a mining area. The method comprises the steps that pipe network operation data of a transportation and distribution pipe network and current multi-source geological and hydrological data are obtained; determining an optimal state estimation value of the pipe network by using a real-time hydraulic model and an ensemble Kalman filtering algorithm, and calculating a pressure residual error of each monitoring node; when the residual represents an abnormal working condition, determining a leakage result by using a leakage positioning algorithm; generating an emergency regulation and control decision by using an optimization scheduler based on the leakage result; when the residual represents a normal working condition, a scheduler is used for executing multi-time-scale predictive control through a real-time hydraulic model based on pipe network operation data to generate a transmission and distribution scheduling decision, and the pipe network is regulated and controlled to realize back-replenishing water transmission and distribution; performing multi-physics field coupling real-time simulation on the whole diffusion and migration process of the back-replenishing water by using the geologic structure digital twinborn model to obtain a back-replenishing prediction result; and evaluating the risk of the candidate supplementation scheme based on a reliability theory, and solving by using a multi-objective optimization algorithm to obtain a target supplementation scheme.

Description

A method and system for intelligent control of water transmission, distribution and replenishment in mining areas Technical Field

[0001] This application relates to the field of mine water treatment technology, and relates to, but is not limited to, an intelligent control method and system for mine water transmission, distribution and replenishment. Background Technology

[0002] Groundwater recharge is a key engineering technology for ensuring the balance of groundwater resources and restoring the ecological environment in mining areas and other regions. With the increasing demand for ecological restoration in mining areas, higher requirements are being placed on the operational efficiency, safety, reliability and energy efficiency management of groundwater recharge systems.

[0003] However, existing groundwater recharge systems generally adopt a traditional segmented management model, with each link operating independently and lacking coordination. This results in limited overall system efficiency, making it difficult to meet the precise control requirements under complex working conditions. It also suffers from drawbacks such as insufficient overall optimization efficiency, poor geological adaptability, weak recharge safety risk management, and delayed dynamic response. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method and system for intelligent control of water transmission, distribution and replenishment in mining areas, which at least solves the problems of insufficient overall optimization efficiency, poor geological adaptability, weak control of replenishment safety risks, and lagging dynamic response.

[0005] The technical solution of this application embodiment is implemented as follows: Firstly, this application embodiment provides a method for intelligent control of water transmission, distribution, and replenishment in mining areas. The method includes: acquiring pipeline operation data and current multi-source geological and hydrological data of the transmission and distribution network, wherein the pipeline operation data includes pipeline topology data, pipe material parameters, and real-time monitoring data; using a real-time hydraulic model in the dynamic control model of the transmission and distribution network, based on the pipeline operation data, and using an ensemble Kalman filter algorithm, determining the optimal state estimate of the transmission and distribution network; based on the optimal state estimate and the real-time monitoring data, determining the pressure residual of each monitoring node in the transmission and distribution network; when the pressure residual indicates that the transmission and distribution network is in an abnormal operating condition, using a leakage location algorithm in the dynamic control model of the transmission and distribution network, based on the pressure residual of each monitoring node, determining the target leakage result of the transmission and distribution network; based on the target leakage result, using the optimal hydraulic model in the dynamic control model of the transmission and distribution network... An optimized scheduler generates emergency control decisions. Under the condition that the pressure residual indicates the distribution network is in normal operating condition, the optimized scheduler, based on the network operation data, performs multi-timescale predictive control through the real-time hydraulic model, generating distribution scheduling decisions with the goal of minimizing total operating cost and ensuring safe operation. Based on the emergency control decisions or the distribution scheduling decisions, the distribution network is regulated to achieve the distribution of replenishment water. Using the geological structure digital twin model in the replenishment response prediction model, based on the current multi-source geological and hydrological data, a multi-physics field coupled real-time simulation of the entire diffusion and migration process of replenishment water under multiple candidate replenishment schemes is performed to obtain the replenishment prediction results corresponding to each candidate replenishment scheme. Based on reliability theory, a risk assessment is performed on each replenishment prediction result to obtain the evaluation results of multiple candidate replenishment schemes. Based on the evaluation results of the multiple candidate replenishment schemes, a multi-objective optimization algorithm is used to solve for the target replenishment scheme.

[0006] Secondly, embodiments of this application provide an intelligent control system for water transmission, distribution, and replenishment in coal mining areas. The system includes: a data acquisition module for acquiring network operation data and current multi-source geological and hydrological data of the transmission and distribution network, wherein the network operation data includes network topology data, pipe material parameters, and real-time monitoring data; a state estimation module for using a real-time hydraulic model in the dynamic control model of the transmission and distribution network, based on the network operation data, and employing an ensemble Kalman filter algorithm to determine the optimal state estimate of the transmission and distribution network; and based on the optimal state estimate and the real-time monitoring data, determining the pressure residuals of each monitoring node in the transmission and distribution network; and an emergency decision generation module for determining the target leakage result of the transmission and distribution network based on the pressure residuals of each monitoring node, using a leakage location algorithm in the dynamic control model of the transmission and distribution network when the pressure residuals indicate that the transmission and distribution network is in an abnormal operating condition; and based on the target leakage result, using the optimization scheduler of the dynamic control model of the transmission and distribution network... The system generates emergency control decisions; a transmission and distribution decision generation module, used to generate transmission and distribution scheduling decisions based on the network operation data and the real-time hydraulic model, using the optimized scheduler, to generate transmission and distribution scheduling decisions with the goal of minimizing total operating costs and ensuring safe operation, under the condition that the pressure residual characterizes the transmission and distribution network as being in normal operating condition; based on the emergency control decisions or the transmission and distribution scheduling decisions, the system performs operation control on the transmission and distribution network to achieve the transmission and distribution of replenishment water; a replenishment scheme generation module, used to perform multi-physics field coupling real-time simulation of the entire process of replenishment water diffusion and migration under multiple candidate replenishment schemes using the geological structure digital twin model in the replenishment response prediction model, based on the current multi-source geological and hydrological data, to obtain the replenishment prediction results corresponding to each candidate replenishment scheme; based on reliability theory, the system performs risk assessment on each replenishment prediction result to obtain the evaluation results of multiple candidate replenishment schemes; based on the evaluation results of the multiple candidate replenishment schemes, the system uses a multi-objective optimization algorithm to solve for the target replenishment scheme.

[0007] The beneficial effects of the technical solution provided in this application include at least the following: breaking the limitations of traditional segmented management, optimizing the operation strategy from a global perspective through collaborative decision-making between transmission and distribution scheduling and replenishment schemes, avoiding overall efficiency loss caused by local optima; accurately depicting the geological characteristics of the replenishment area based on the digital twin model of the geological structure, and realizing dynamic prediction of replenishment water migration by combining multi-physics field coupling simulation, thereby improving the adaptability to complex geological conditions; introducing reliability theory for risk assessment, providing quantitative risk basis for the replenishment scheme, and effectively reducing geological risks; realizing rapid identification and emergency control of abnormal pipeline conditions through dynamic correction of real-time hydraulic models and ensemble Kalman filtering, and multi-timescale predictive control, eliminating the time lag of parameter adjustment; and simultaneously realizing full-process coordination of transmission and distribution and replenishment, taking into account both operating costs and safety objectives, thereby improving the operating efficiency and reliability of the mining area water system. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments 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. Among them: Figure 1 is a flowchart of a method for intelligent control of water transmission and distribution and replenishment in a mining area provided by an embodiment of this application; Figure 2 is a framework diagram of a dynamic control model for a transmission and distribution network provided by an embodiment of this application; Figure 3 is a schematic diagram of a leak location algorithm based on the fusion of sparse signal reconstruction and machine learning provided by an embodiment of this application; Figure 4 is a framework diagram of a replenishment response prediction model provided by an embodiment of this application; Figure 5 is a framework diagram of a multi-objective adaptive replenishment optimization and evaluation model provided by an embodiment of this application; Figure 6 is a schematic diagram of the composition structure of an intelligent control system for water transmission and distribution and replenishment in a mining area provided by an embodiment of this application. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0010] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0011] It should be noted that the terms "first, second, and third" used in the embodiments of this application are merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0012] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments of this application pertain. It should also be understood that terms such as those defined in general dictionaries should be understood to have a meaning consistent with their meaning in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0013] This application provides a method for intelligent control of water transmission, distribution, and replenishment in mining areas, applied to electronic devices. These electronic devices include, but are not limited to, mobile phones, laptops, tablets, handheld internet devices, multimedia devices, streaming media devices, mobile internet devices, wearable devices, or other types of electronic devices. The functions implemented by this method can be achieved by a processor in the electronic device calling program code. The program code can be stored in a computer storage medium; therefore, the electronic device includes at least a processor and a storage medium. The processor can be used to process the intelligent control of water transmission, distribution, and replenishment in mining areas, and the memory can be used to store the data required and generated during the intelligent control of water transmission, distribution, and replenishment in mining areas.

[0014] Figure 1 is a flowchart illustrating an intelligent control method for water transmission, distribution, and replenishment in a mining area according to an embodiment of this application. As shown in Figure 1, the method includes at least the following steps: Step S110, acquiring pipeline operation data and current multi-source geological and hydrological data of the transmission and distribution network. The pipeline operation data includes pipeline topology data, pipe material parameters, and real-time monitoring data. The transmission and distribution network refers to the pipelines and ancillary facilities that undertake the tasks of transporting, regulating, and distributing replenishment water, production water, or domestic water in the mining area. In addition to the pipelines themselves, the transmission and distribution network also includes supporting monitoring equipment (such as pressure sensors, flow sensors, etc.), control equipment (such as pumps, valves, etc.), and data transmission equipment.

[0015] The pipeline network operation data refers to the operational status and physical attribute data of the transmission and distribution pipeline network itself. This includes static basic data such as pipeline topology data (e.g., pipeline segment direction, node distribution, pump and valve locations, etc.) and pipe material parameters (e.g., pipe diameter, pipe material, friction coefficient, etc., which determine hydraulic characteristics). This type of data is relatively fixed and forms the basic framework for building a real-time hydraulic model. The pipeline network operation data also includes dynamic real-time monitoring data during pipeline network operation (e.g., node pressure, pipeline segment flow, pump and valve opening, energy consumption, etc.). This type of data changes in real time and is a key basis for correcting the model and judging operating conditions. The multi-source geological and hydrological data refers to the geological and hydrological conditions of the external environment in which the pipeline network is located. This type of data affects the diffusion and migration patterns of replenishment water and also indirectly affects the operational status of the pipeline network.

[0016] Step S120: Using the real-time hydraulic model in the dynamic control model of the transmission and distribution network, and based on the network operation data, the optimal state estimate of the transmission and distribution network is determined using the ensemble Kalman filter algorithm; based on the optimal state estimate and the real-time monitoring data, the pressure residual of each monitoring node in the transmission and distribution network is determined; wherein, the dynamic control model of the transmission and distribution network is also known as the multi-source information synthesis and fusion based hydraulic dynamic model of the network. The Dynamic Model (MISF-HDM) is used to dynamically regulate the replenishment water distribution decision based on the operating conditions of the pipeline network, ensuring the safety and economy of the distribution process. The dynamic regulation model of the distribution pipeline network includes a built-in real-time hydraulic model, a leak location algorithm, and an optimization scheduler. The real-time hydraulic model is a digital simulation model of the distribution pipeline network, which can predict the theoretical operating state of the pipeline network based on relevant hydraulic equations. The Kalman filter algorithm can assimilate and fuse the theoretical operating state with real-time monitoring data, correct the deviation between the theoretical value obtained by the real-time hydraulic model and the actual operating state of the pipeline network, and finally obtain the optimal state estimate.

[0017] Using the optimal state estimate of each monitoring node corrected by the ensemble Kalman filter algorithm as a benchmark, the difference between the estimate and the actual monitoring data of the corresponding monitoring node collected by the sensor is calculated to obtain the pressure residual of the corresponding monitoring node.

[0018] Step S130: When the pressure residual indicates that the transmission and distribution network is in an abnormal operating condition, the leakage location algorithm in the dynamic control model of the transmission and distribution network is used to determine the target leakage result of the transmission and distribution network based on the pressure residual of each monitoring node; based on the target leakage result, the optimization scheduler of the dynamic control model of the transmission and distribution network is used to generate an emergency control decision; wherein, the pressure residual can be compared with a preset residual threshold. If the pressure residual is within the preset residual threshold, it indicates that the actual operating state of the network is consistent with the prediction of the real-time hydraulic model, which is a normal operating condition, indicating that the network has no leakage or blockage and is in a stable operating state. If the pressure residual exceeds the preset residual threshold, it indicates that the network may have abnormal operating conditions such as leakage or blockage.

[0019] Leakage location algorithms can determine the target leak by combining the residual distribution characteristics represented by pressure residuals with basic data such as pipeline topology and pipe material parameters, through numerical inversion or pattern recognition methods. Based on the target leak result, an optimization scheduler can generate emergency control decisions to quickly suppress the impact of the leak and ensure the overall safe operation of the pipeline network. The optimization scheduler can generate targeted emergency control decisions based on the severity of the leak. These emergency control decisions may include closing valves upstream and downstream of the leak point, cutting off the faulty pipe section to prevent the leak from expanding; adjusting the operating power of surrounding pumping stations to compensate for pressure loss in the faulty area; and switching to backup transmission and distribution paths to ensure normal water supply or replenishment needs in non-faulty areas.

[0020] Step S140: When the pressure residual indicates that the transmission and distribution network is in normal operating condition, the optimization scheduler, based on the network operation data, performs multi-time-scale predictive control through the real-time hydraulic model to generate transmission and distribution scheduling decisions with the goal of minimizing total operating cost and safe operation. Based on the emergency control decision or the transmission and distribution scheduling decision, the operation of the transmission and distribution network is regulated to achieve the transmission and distribution of replenishment water. The multi-time-scale predictive control is a time-segmented dynamic optimization strategy, including two levels: short-term (minute / hour) and medium-term (day / week). The short-term level can fine-tune pump and valve openings and pipe flow distribution based on real-time monitoring data to ensure that the pressure at network nodes remains stable within a safe range, avoiding network damage caused by excessively high or low instantaneous pressure. The medium-term level optimizes pump unit operation combinations and transmission and distribution paths based on network load change trends (such as fluctuations in replenishment water demand and peak / valley electricity prices) to achieve the goal of minimizing total operating cost. The transmission and distribution scheduling decisions may include pump and valve operating parameters, flow distribution schemes, and load adjustment plans.

[0021] Step S150: Using the geological structure digital twin model in the replenishment response prediction model, based on the current multi-source geological and hydrological data, perform multi-physics field coupling real-time simulation of the entire diffusion and migration process of replenishment water under multiple candidate replenishment schemes to obtain the replenishment prediction results corresponding to each candidate replenishment scheme; perform risk assessment on each replenishment prediction result based on reliability theory to obtain the evaluation results of multiple candidate replenishment schemes; based on the evaluation results of the multiple candidate replenishment schemes, use a multi-objective optimization algorithm to solve for the target replenishment scheme.

[0022] The recharge response prediction model is used to simulate and optimize the diffusion and transport process of recharge water based on multi-source geological and hydrological data, ensuring the scientific validity and safety of the recharge process. The model includes a built-in digital twin model of the geological structure, a reliability risk assessment module based on reliability theory, and an optimization decision module containing a multi-objective optimization algorithm. The digital twin model of the geological structure is a digital mapping of the underground geological structure and hydrological conditions of the mining area, including core parameters such as stratigraphic distribution, permeability, groundwater level, and lithology. It can simultaneously consider the interaction of multiple physical fields such as the seepage field, stress field, and temperature field of the recharge water underground, thus affecting the diffusion and transport of the recharge water. The entire relocation process is simulated in real time using multiphysics coupling to obtain recharge prediction results. These results include key indicators such as the diffusion range of recharge water, migration path, groundwater level rise, recharge efficiency, and impact on surrounding aquifers. The reliability theory is a quantitative analysis method for assessing the risk probability of different candidate recharge schemes. The optimization of recharge schemes usually involves multiple mutually constraining objectives, such as maximizing recharge efficiency, minimizing operation and maintenance costs, and minimizing environmental risk. Single-objective algorithms cannot simultaneously achieve all these objectives, while multi-objective optimization algorithms can find the optimal balance among multiple objectives, thereby obtaining a recharge scheme that balances multiple dimensions of requirements, including safety, efficiency, and cost.

[0023] In the above embodiments, the limitations of traditional segmented management are broken. By coordinating the decision-making of transmission and distribution scheduling and replenishment schemes, the operation strategy is optimized from a global perspective, avoiding the overall efficiency loss caused by local optima. Relying on the digital twin model of geological structure, the geological characteristics of the replenishment area are accurately depicted. Combined with multi-physics field coupling simulation, dynamic prediction of replenishment water movement is realized, improving the adaptability to complex geological conditions. Reliability theory is introduced for risk assessment, providing a quantitative basis for replenishment schemes and effectively reducing geological risks. Through real-time hydraulic models and dynamic correction of ensemble Kalman filtering, and multi-timescale predictive control, rapid identification and emergency control of abnormal pipeline conditions are realized, eliminating the time lag of parameter adjustment. At the same time, the entire process of transmission and distribution and replenishment is coordinated, taking into account both operating costs and safety objectives, and improving the operating efficiency and reliability of the mining area water system.

[0024] In some embodiments, step S120, "using the real-time hydraulic model in the dynamic control model of the transmission and distribution network, based on the network operation data, and using the ensemble Kalman filter algorithm, to determine the optimal state estimate of the transmission and distribution network," includes: step S1201, using the real-time hydraulic model in the dynamic control model of the transmission and distribution network, based on the network topology data, the pipe material parameters, the pre-established extended node flow continuity equation and the pipe section pressure drop equation, and solving using the extended node equation method to determine the initial state estimate of the transmission and distribution network; wherein, the real-time hydraulic model is based on the mass conservation and energy conservation laws describing the water flow motion within the network, and the mathematical expression of the mass conservation is the extended node flow continuity equation, as shown in formula (1): Formula (1); where, The flow vector for the pipe segment (the water velocity in all pipe segments, m) 3 / s); Let be the node head vector (the water level height of all nodes, in meters); The vector of node water demand (including leakage and the flow rate consumed / lost at each node) (m) 3 / s); , This is the correlation matrix that describes the pipeline network topology.

[0025] The mathematical expression of the energy conservation is the pipe section pressure drop equation, as shown in formula (2): Formula (2); where, The head loss vector of the pipe section (the water level loss caused by friction and resistance when the water flows in the pipe section, in m) is calculated by the Heize-Williams formula or the Darcy-Weisbach formula. This is the correlation matrix that describes the pipeline network topology.

[0026] As shown in Figure 2, formulas (1) and (2) can be solved simultaneously using the extended node equation method to determine the initial state estimate of the transmission and distribution network. The initial state estimate includes the pressure of each node and the flow rate of each pipe segment.

[0027] Step S1202: Based on the ensemble Kalman filter algorithm and the real-time monitoring data, the initial state estimate is optimized to obtain the optimal state estimate of the transmission and distribution network.

[0028] The Ensemble Kalman Filter (EnKF) algorithm enables dynamic self-correction of real-time hydraulic models. A model state vector *x* is defined, which is a comprehensive vector containing all core variables such as Q, H, D, and E, encompassing complete dynamic information of the system. During algorithm initialization, based on the initial state estimates of the distribution network and preset model uncertainties, an initial covariance matrix (i.e., the initial state estimation error covariance matrix in Figure 2) is assigned to each real-time hydraulic model member. This covariance matrix quantifies the uncertainty range of the initial state estimates. The Ensemble Kalman Filter algorithm continuously integrates real-time observation data into multiple real-time hydraulic model members. Through data assimilation calculations, it corrects the initial state estimates of each group of real-time hydraulic model members, obtaining the optimal state estimate. This makes the statistical characteristics (mean, variance) of the set of multiple real-time hydraulic model members more closely resemble actual operating conditions. In practical deployment, a set of 50 real-time hydraulic model members can be maintained, with each member representing a possible system state. The uncertainty is quantified through the statistical characteristics of the set.

[0029] As shown in Figure 2, the data assimilation (optimization) process includes a prediction step and an update step, as shown in Equation (3). The prediction step can predict the state at the current moment based on the state estimate of the previous moment and the real-time hydraulic model: = +w formula (3); where, It is the predicted value at time t. It is a hydraulic model operator. t is the optimal state estimate at time t-1, and w is the hydraulic model error.

[0030] In the update step, when new observation data (such as partial node pressure) is obtained, EnKF calculates the Kalman gain and optimally fuses the predicted value with the observed value to obtain the analysis value (i.e. the optimal state estimate), as shown in formulas (4) and (5): Formula (4); where, It is an observation operator. It is the forecast error covariance matrix. This is the observation error covariance matrix. This process enables the digital twin to continuously approximate the real pipeline network state, providing a reliable virtual experimental environment for subsequent precise positioning and optimization.

[0031] Formula (5); where, It is the optimal state estimate at time t. It is the predicted value at time t. It is the Kalman gain at time t. .

[0032] The dynamic control model for transmission and distribution networks serves as the intelligent central hub of the control system, solving problems such as inaccurate dynamic leak location, delayed scheduling response, and insufficient system resilience in traditional methods. The core of this model is a high-precision real-time hydraulic model based on the extended nodal equations method. It not only solves steady-state conditions but also introduces the simulation capability of unsteady hydraulic transient processes. Matrices D (node ​​water demand) and E (pipeline head loss) are time-varying. The real-time hydraulic model performs dynamic correction (data assimilation) using real-time monitoring data (pressure, flow rate). Employing an ensemble Kalman filter algorithm, it continuously assimilates the observed data, ensuring a high degree of consistency between the digital twin of the transmission and distribution network and the physical system. This provides a realistic and reliable virtual experimental environment for subsequent optimization and early warning. In practical applications, up to 50 ensemble members can be set, with data assimilation completed every 5 minutes, stabilizing the root mean square error between the simulated pressure and the measured pressure in the real-time hydraulic model to within 0.15 meters.

[0033] In the above embodiments, the accuracy of the state estimation of the transmission and distribution network is improved by using a two-step method of initial solution of the real-time hydraulic model and ensemble Kalman filter correction: the initial solution based on the basic hydraulic equations ensures the theoretical rationality of the state estimation; combined with EnKF correction of real-time monitoring data, the actual operating condition information is effectively integrated, the model calculation error is reduced, and the optimal state estimate value that is more in line with reality is output, providing an accurate data foundation for subsequent leak location and scheduling decisions.

[0034] In some embodiments, step S130, "using the leak location algorithm in the dynamic control model of the transmission and distribution network, and based on the pressure residuals of each monitoring node, to determine the target leak result of the transmission and distribution network," includes the following steps: Step S1301, based on the optimal state estimate, determining the real-time operating condition of the transmission and distribution network, and using the real-time hydraulic model to analyze the real-time operating condition to obtain the enhanced sensitivity matrix; wherein, as shown in Figure 3, the decision-making process of the leak location algorithm is a closed-loop system, consisting of a forward model and a reverse optimization problem. Its core logic follows a progressive structure of detection, modeling, solving, and verification, aiming to transform sparse monitoring data into reliable leak location results. The specific process includes leak detection and data preprocessing, physical enhancement and construction of the sensitivity matrix, leak signal reconstruction based on sparse optimization, and result verification and feedback optimization.

[0035] Step S1302: Based on the pressure residuals of each monitoring node, determine the pressure change vector; as shown in Figure 3, in the leak detection and data preprocessing step, real-time pressure monitoring data and steady-state reference pressure data can be obtained. First, calculate the pressure residual sequence of each monitoring point, and set a dynamic threshold based on historical statistical data or engineering experience; determine whether the pressure residual continuously and significantly exceeds the threshold through statistical testing to trigger a leak warning. Subsequently, filter and denoise the pressure changes confirmed as abnormal and perform data standardization processing to form a standardized pressure change vector for subsequent positioning. Output the pressure change vector ΔP representing the abnormal state of the pipeline network as shown in formula (6):

[0036] ΔP= Formula (6); As shown in Figure 3, in the physical enhancement and construction of the sensitivity matrix (i.e., the physical enhancement construction of the sensitivity matrix), the actual operating conditions of the current pipeline network (including node flow, pressure distribution, valve opening, etc.) can be obtained. Based on the real-time hydraulic model, linearization analysis is performed at the current operating point to calculate the basic pressure-leakage sensitivity matrix (Jacobi matrix). In order to improve the model's ability to characterize the actual physical constraints, a weighting factor is introduced as shown in Formula (7), which comprehensively considers factors such as node distance, pipe section impedance and hydraulic connectivity, and performs weighted correction on the basic sensitivity matrix to generate a physical information-enhanced sensitivity matrix as shown in Formula (8). This step aims to make the observation model more closely match the real hydraulic response characteristics of the pipeline network. The enhanced sensitivity matrix is ​​output as the observation matrix in compressed sensing.

[0037] Formula (7); where, It is the distance from node i to sensor j; It is the effective wave velocity, and e is the natural constant.

[0038] Formula (8); where, Weighting factors; It is the element in the i-th row and j-th column of the enhanced sensitivity matrix.

[0039] Step S1303: Based on the pressure change vector and the enhanced sensitivity matrix, construct and solve a sparse optimization problem to determine the initial leakage result of the transmission and distribution network; wherein, the initial leakage result may include the location of the leakage point and the leakage amount, as shown in Figure 3. In the leakage signal reconstruction (i.e., sparse signal reconstruction) step based on sparse optimization, the enhanced sensitivity matrix and the preprocessed pressure change vector can be obtained, and the leakage location problem is constructed as a norm-regularized convex optimization problem. Its objective function aims to minimize the pressure fitting error while forcing the solution vector to be sparse. The Iterative Shrinkage-Thresholding Algorithm (ISTA) is used for efficient solution. This algorithm explicitly promotes the sparsity of the solution through a soft threshold operator, and finally obtains a sparse leakage amount change vector (i.e., a sparse solution vector). Output the initial leakage location result, i.e., the node index (leakage location) and its magnitude (leakage amount estimate) corresponding to the non-zero elements in the leakage amount change vector. The physical relationship between the pressure change caused by the leakage and the leakage amount can be expressed as a forward observation model as shown in formula (9): Formula (9); where ΔP is the pressure change vector, ΔD is the leakage change vector; J is the enhanced sensitivity matrix (Jacobi matrix). It is the observation noise vector.

[0040] The leak localization problem can be transformed into a sparse optimization problem as shown in equation (10): Formula (10); where ΔD is the vector of change in node leakage to be determined (a sparse vector, where non-zero values ​​indicate leakage points); J is related to the state estimate, determined by the current operating conditions ( The hydraulic model is obtained by linearizing the hydraulic model (which simulates the flow state under the current operating conditions) to provide real-time input. It is the regularization coefficient; It is the L2 norm. It is the L1 norm. The introduction of this regularization term is the key innovation. It forces the solution vector to be sparse (most elements are 0, and only a few non-zero elements correspond to leakage points), thereby directly locating a few leakage points and breaking through the resolution limitation of traditional methods.

[0041] Step S1304: Based on the initial leakage result, perform forward pressure simulation using the real-time hydraulic model to obtain the simulated pressure change value; evaluate the reliability of the initial leakage result based on the degree of fit between the simulated pressure change value and the actual pressure change value; Step S1305: If the reliability does not reach the preset standard, adaptively adjust the parameters of the real-time hydraulic model; based on the adjusted real-time hydraulic model, re-execute the step of obtaining the enhanced sensitivity matrix, iteratively solve, until the reliability reaches the preset standard, and determine the corresponding leakage result as the target leakage result.

[0042] As shown in Figure 3, in the result verification and feedback optimization (i.e., verification evaluation feedback) step, the leak location and quantitative results obtained in step S1303 can be obtained. The estimated leakage amount is used as a known input and substituted into the real-time hydraulic model for forward pressure simulation to obtain the simulated pressure change value. During the fitting effect evaluation, the reliability of the leak location result is quantitatively evaluated by comparing the degree of fit between the simulated pressure change value and the actual pressure change value (e.g., calculating the residual norm or coefficient of determination). If the fitting effect evaluation result does not meet the predetermined standard, the verification fails, and the evaluation information is fed back to step S1301 or step S1303, triggering adaptive adjustment of model parameters or optimization thresholds for iterative re-optimization until the target leak result that meets the reliability requirements is obtained. If the fitting effect evaluation result meets the preset standard, the verification passes, a final leak location report with a reliability evaluation is output, and the execution of emergency control strategies and system feedback optimization is triggered.

[0043] To address the problems of sparse monitoring points, high noise interference, and low positioning accuracy in leak localization of water supply networks, this application transforms the leak localization problem into a sparse signal reconstruction problem within a compressed sensing framework. The basic principle is that leaks in actual pipeline networks typically occur at a few localized locations, resulting in a naturally sparse leakage vector, where most elements are zero, and non-zero elements correspond to the leak points. By utilizing this sparse prior information, even when the number of monitoring points is far less than the number of pipeline nodes (M≪N), the leakage localization problem can be solved. n Under the condition of limited pressure observation data, this method can accurately reconstruct the leakage signal from the limited pressure observation data. Based on compressed sensing theory, this method can recover the original signal from observations with a sampling rate far lower than the Nyquist rate, provided that the signal satisfies sparsity. In addition, it combines regularization methods, using L1 norm regularization to promote the sparsity of the solution, thereby significantly improving the accuracy of leakage location and noise resistance.

[0044] In the above embodiments, the leak location algorithm significantly improves the accuracy and reliability of leak location through an iterative process of sensitivity matrix enhancement, sparsity optimization, and credibility verification. Relying on the enhanced sensitivity matrix under real-time operating conditions, it solves the location problem in sparse monitoring point scenarios, improving the location accuracy from the traditional hundred-meter level to the meter level. Through the iterative mechanism of forward pressure simulation and credibility assessment, it avoids the deviation of single model calculation, ensures the accuracy of leak results, and provides a reliable basis for emergency control. It realizes the adaptive adjustment of model parameters, enhancing the algorithm's adaptability to complex pipeline network conditions.

[0045] In some embodiments, as shown in Figure 2, the real-time operating condition of the transmission and distribution network can be judged based on the optimal state estimate. Normal operating condition refers to the state in which the pressure residual of the transmission and distribution network is within the preset threshold range, there are no abnormal events such as leakage or pipe burst, and the system operates in accordance with the design expectations. Abnormal operating condition refers to the state in which the pressure residual of the transmission and distribution network exceeds the preset threshold, or other abnormal signals are detected (such as sudden changes in flow or sudden drops in pressure), indicating that there may be faults such as leakage or pipe burst in the transmission and distribution network.

[0046] When the pressure residual characterizes the distribution network under normal operating conditions, the system optimizes the scheduler module. The inputs of the scheduler include water demand forecast, electricity price signal, equipment status, and real-time status from the hydraulic kernel (i.e., real-time hydraulic model). The core algorithm utilizes multi-timescale model predictive control (MPC). The structure of the core algorithm includes a day-ahead scheduling layer and a real-time control layer. The day-ahead scheduling layer solves a mixed integer programming problem with a 24-hour cycle to determine the start-up and shutdown plan of the pump group. The real-time control layer solves a nonlinear programming problem with a 15-minute rolling optimization window to finely adjust the pump speed and valve opening. In this application, a toughness constraint is introduced, such as requiring that the system can still guarantee water supply to key nodes under any single pipe section failure condition. The objective function of the core algorithm is shown in formulas (11) and (12): Formula (11); where t is the time index, the time step within the optimization period, and t=1,2,…,24 represents 24 hours in a day; The change in electricity price at time t over time is a key input parameter for the optimization model; is the total power consumption of the system at time t, and the total power consumption of all operating devices at time t, which is the optimization variable; It is the cost coefficient for a single start-up and shutdown, which penalizes the cost of frequent start-ups and shutdowns of the pump set, including the costs of electrical shock and mechanical wear; It is a binary start / stop status variable. =1 indicates that a state transition occurs at time t, otherwise it is 0.

[0047] Formula (12); where k is the prediction step index, k=0,1,…,N−1, representing the discrete time step in the prediction time domain; N is the prediction time domain length, the number of steps the MPC algorithm predicts forward, usually corresponding to 15 minutes; It is the nodal head vector at time k; It is the head setpoint vector at time k; It is the control input vector at time k; It is the weighted norm of the state tracking error; It is the weighted norm for controlling energy consumption; R is the state weight matrix; R is the control weight matrix.

[0048] The MISF-HDM model, through the integration of physical models and data in principle, the coupling of leakage localization and optimization scheduling algorithms in structure, and the multi-mode closed-loop feedback mechanism in decision-making, achieves a leap from passive response to proactive early warning, and from local optimization to global resilient scheduling, forming the core of system intelligence. The decision logic of this model is a state-based closed-loop feedback control process, including economical and safe transmission and distribution scheduling decisions under normal operating conditions and emergency control decisions under abnormal operating conditions.

[0049] As shown in Figure 2, the economic and safety scheduling decision under normal operating conditions is a state-based closed-loop feedback control process. The MPC multi-timescale optimization forms a closed-loop control process through setpoints, MPC controllers, passive objects, and feedback correction, including the following steps: state perception; the real-time hydraulic model outputs the current network state through EnKF, and the setpoints of the MPC controller are dynamically generated by the day-ahead scheduling layer based on water supply safety standards, water demand forecasts, resilience constraints, and equipment operation limitations.

[0050] Prediction; Based on water usage prediction, the MPC optimizer predicts system behavior over a future period in forward simulation.

[0051] Optimization solution: With the goal of minimizing total operating cost and ensuring operational safety, the optimal sequence of pump station and valve control commands is obtained, but only the first step is executed (to avoid decision-making bias caused by uncertainty of future operating conditions).

[0052] Decision execution; issuing control commands to the actuators (i.e., controlled objects such as pumps and valves).

[0053] Rolling optimization: In the next cycle, the above steps are repeated based on new real-time monitoring data, and the system is re-optimized according to the new state measurement values ​​to form a closed-loop feedback, which effectively overcomes model errors and external disturbances and ensures that the decision is always optimal.

[0054] Ultimately, the system outputs an economic and safe scheduling decision, achieving the dual goals of minimizing total operating costs and ensuring operational safety.

[0055] As shown in Figure 2, when the pressure residual indicates that the transmission and distribution network is in an abnormal operating condition (such as leakage), the system enters the leakage location algorithm module. This leakage location algorithm module includes sub-steps such as information preprocessing, feature extraction, fusion calculation, and leakage location, and outputs the initial leakage point location and leakage amount estimate.

[0056] Subsequently, the sparse optimization solution module is used to quickly analyze and accurately solve the pipeline network status under the leakage scenario, and finally complete the location and quantification of the leakage point.

[0057] Based on the location results, emergency control decisions under abnormal operating conditions include the following steps: Detection and location: When the pressure residual exceeds the threshold, the leak location algorithm is triggered, and the location of the leak point and the estimated amount of leakage are output.

[0058] Impact assessment: The hydraulic kernel rapidly simulates the pressure distribution and flow allocation of the pipeline network under leakage scenarios to identify the affected areas.

[0059] Emergency plan generation: The optimized scheduler aims to minimize the scope of water supply interruption and maintain pressure at critical nodes. It simulates different valve closure and isolation schemes to generate the optimal emergency dispatch strategy.

[0060] Decision-making and execution: Recommend emergency plans that include closing specific valves and adjusting relevant pump station operating parameters to operators for confirmation and execution, or execute them automatically.

[0061] Continuous monitoring: Continue to monitor the effectiveness of emergency measures until the leak is repaired and the system returns to normal operation.

[0062] Ultimately, the system outputs emergency control decisions, enabling intelligent operation and maintenance and risk management under abnormal operating conditions.

[0063] As shown in Figure 2, the actuator control module transforms economic and safety scheduling decisions or emergency control decisions into pump and valve actions, driving the pipeline system to change its operating state. The effect monitoring module collects the actual operating data of the pipeline system in real time, evaluates the control effect, and the data assimilation feedback module transmits the monitoring data back to the real-time hydraulic model, providing the latest observations for the ensemble Kalman filter, realizing the dynamic update of the optimal state estimate, thus forming a complete closed-loop control system.

[0064] In some embodiments, step S150, "using the geological structure digital twin model in the replenishment response prediction model, based on the current multi-source geological and hydrological data, to perform multi-physics field coupling real-time simulation of the entire process of replenishment water diffusion and migration, and obtain replenishment prediction results," includes the following steps: Step S1501, acquiring historical multi-source geological and hydrological data, fusing the historical multi-source geological and hydrological data to obtain historical fused geological and hydrological data; as shown in Figure 4, the historical multi-source geological and hydrological data may include geological exploration data, geophysical data, remote sensing data, historical hydrological data, and field test data, etc.

[0065] Step S1502: Based on the historical fused geological and hydrological data, a geological structure digital twin model containing three-dimensional stratigraphic structure and aquifer spatial distribution is constructed using geostatistics methods; as shown in Figure 4, this step aims to create a three-dimensional geological structure digital twin that reflects the heterogeneity of the actual site, including data fusion and three-dimensional geological modeling, parameter field spatial variability analysis, hydrogeological parameter field construction, and initial boundary condition setting. First, in the data fusion and 3D geological modeling stage, the data acquisition and fusion phase integrates multi-source heterogeneous data, including geological exploration borehole data, geophysical exploration profiles, remote sensing images, and historical hydrogeological data. Subsequently, based on geostatistical methods such as Kriging interpolation and sequential indicator simulation, a 3D stratigraphic structure and aquifer spatial distribution model is constructed. Finally, in the parameter field spatial variability analysis stage, through the parameter determination process, the spatial distribution patterns of hydrogeological parameters such as permeability coefficient, porosity, and specific yield are analyzed by combining key point hydraulic parameters obtained from field double-ring permeability tests and Guelph permeameter tests. In the hydrogeological parameter field construction stage, accurate hydrogeological parameter fields (such as permeability coefficient, porosity, and specific yield) are assigned to the 3D stratigraphic structure and aquifer spatial distribution model using geostatistical inversion or spatial interpolation techniques. In the initial boundary condition setting stage, initial water level, boundary flow, and other conditions are set for subsequent numerical simulations, thereby completing a parameterized 3D geological structure digital twin that can be used for numerical simulation.

[0066] Step S1503: Based on the current multi-source geological and hydrological data, initial and boundary conditions are set. In the digital twin model of the geological structure, the finite element method or finite difference method is used to couple and solve Darcy's law describing groundwater flow and the convection-dispersion equation describing solute migration, so as to realize the real-time simulation of the entire process of recharge water diffusion and transport through multi-physics field coupling, and obtain the recharge prediction results.

[0067] As shown in Figure 4, the core governing equations form the theoretical basis for multiphysics coupled simulation, describing physical processes such as groundwater flow, solute transport, and heat transfer. Based on this digital twin model of the geological structure, a dynamic high-fidelity simulation of the recharge process is initiated. Before the simulation, initial conditions (such as the current groundwater flow field, water level, and background water quality concentration field) and boundary conditions (including planned recharge intensity, spatial location, duration, and source water quality) need to be set according to the monitored current multi-source geological and hydrological data. The numerical solution process uses the finite element method or finite difference method to couple and solve Darcy's law describing groundwater flow with the convection-dispersion equation describing solute migration. This simulation can predict in real time the spatiotemporal evolution of groundwater level rise, the advance front of the recharge water body, and the transport paths and concentration distribution of potential pollutants under different recharge scenarios, providing dynamic and refined recharge prediction results for risk assessment.

[0068] In the real-time simulation of the entire diffusion and transport process of replenished water using multiphysics coupling, this application adopts the modified Richards equation as shown in formula (13) to describe the transport process of water in the vadose zone and saturation zone: Formula (13); where θ is the volumetric water content, which is the volume percentage of water in a unit volume of soil; and ψ is the pressure head, which is the potential energy state of water in the soil. is the target moisture content, t is time; K(ψ) is the unsaturated hydraulic conductivity, the soil's ability to conduct water; z is the elevation head; S is the source-sink term for water replenishment or discharge.

[0069] The soil moisture characteristic curves are described using the van Genuchten model as shown in equations (14) to (16): Formula (14); where, It is the residual moisture content; It is the saturated moisture content; It is the reciprocal of the air intake value, n is the soil pore size distribution parameter, and m is the pore diameter.

[0070] Formula (15); Formula (16); where, It is the effective saturation. It is saturated hydraulic conductivity. It is the actual hydraulic conductivity.

[0071] Considering the combined effects of convection, dispersion, and chemical reactions, the evolution of replenishment water quality is predicted using formulas (17) and (18): Where C is the solute concentration; D d It is the hydrodynamic dispersion tensor; q is the Darcy velocity; It is the reaction rate constant. It is the rate of change of solute mass over time.

[0072] Formula (18); In the above embodiments, the accuracy and geological adaptability of recharge prediction are improved by constructing a digital twin model of geological structure and simulating multi-physics fields; the three-dimensional geological digital twin model constructed based on geostatistics accurately depicts the stratigraphic structure and aquifer distribution, solving the problem of poor adaptability of traditional simplified geological models; the multi-physics field simulation coupled with Darcy's law and convection-dispersion equation realizes the dynamic prediction of the entire process of recharge water "water flow-solute migration", providing a more comprehensive geological response basis for the design of recharge schemes.

[0073] In some embodiments, the assessment result includes a risk level. Step S150, "based on reliability theory, performs a risk assessment on the refill prediction result to obtain the assessment results of multiple candidate refill schemes; based on the assessment results of the multiple candidate refill schemes, a multi-objective optimization algorithm is used to solve for the target refill scheme," includes the following steps: Step S1504, determining the probability distribution characteristics of the current multi-source geological and hydrological data; based on the probability distribution characteristics, using reliability theory, through a first-order second-moment method or Monte Carlo simulation, performing a risk assessment on the refill prediction result for a specified risk event to obtain the failure probability of multiple candidate refill schemes corresponding to the specified risk event; Step S1505, based on the failure probability of the multiple candidate refill schemes, determining the risk level of the corresponding candidate refill scheme; As shown in Figure 4, to address the inherent uncertainties of geological parameters and models, this step can introduce reliability theory to perform a probabilistic quantitative assessment of the risk of candidate refill schemes. First, parameter uncertainty analysis is performed to identify and determine the probability distribution characteristics of key hydrogeological parameters (such as permeability coefficient). Subsequently, a reliability calculation method is used to calculate the failure probability of a specified risk event (such as excessive pollutants or rapid water level rise leading to submersion). Finally, based on the calculated failure probability values, risk classification and early warning are performed, forming risk levels such as "low, medium, high, and extremely high," and corresponding early warning information is automatically triggered, providing intuitive and quantitative risk signals for management decisions.

[0074] This model transforms traditional deterministic safety assessment into probabilistic risk assessment by introducing reliability theory, and establishes a limit state function as shown in formula (19). The failure probability calculation framework shown in Equation (20) or Equation (21) provides a quantitative safety index for candidate replenishment schemes.

[0075] Formula (19); where X=[K,n, [Qrecharge, ...] represents a vector of basic random variables, such as the permeability coefficient K (the rate of soil water infiltration), the soil pore size distribution parameter n, and the reciprocal of the air inlet value. , recharge flow rate Qrecharge, etc.; R is the system resistance (maximum allowable backwater level, etc.); S(X) is the load effect (actual backwater level, etc.).

[0076] Formula (20); Formula (21); where, Let X be the joint probability density function of the random variable X; and These are the mean and standard deviation of the limit state function, respectively.

[0077] The failure probability of the recovery system is The second-order moment method (FORM) or Monte Carlo simulation can be used to calculate... The Risk Response Index (RRI) can be defined as RRI = Or the related reliability index β. This method provides a probabilistic and scientific basis for replenishment safety, rather than the traditional deterministic safety factor method. In practical applications, we use field test and monitoring data to determine the probability distribution function (usually assumed to be log-normal) of key parameters (such as the permeability coefficient K), making the risk assessment results more reliable. For example, in a coal mine in North China, the probability that the water level in the surrounding residents' wells would exceed the safety threshold under the predetermined replenishment plan was calculated to be... =0.003, which meets the safety standard of less than 0.01, providing a quantitative basis for the implementation of the plan.

[0078] Step S1506: Based on the risk level and core risk indicators of each candidate replenishment scheme, compare and select candidate replenishment schemes with different replenishment location, intensity, and timing parameter combinations to screen out potential schemes with controllable risks; Step S1507: Use a multi-objective optimization algorithm to iteratively adjust the replenishment engineering parameters of the potential schemes to seek the optimal parameter combination with the lowest risk while meeting the replenishment objectives; Based on the optimal parameter combination, determine the target replenishment scheme.

[0079] This step constitutes the model's decision optimization closed loop. Based on the aforementioned risk levels, the system automatically performs multi-candidate replenishment scheme comparison and analysis, comparing and analyzing core risk indicators (such as maximum failure probability and impact range) under different replenishment locations, intensities, and timing sequences. It then enters the parameter optimization and adjustment and optimal replenishment scheme generation stage. Controllable replenishment engineering parameters can be adjusted using optimization algorithms (such as genetic algorithms and particle swarm optimization) to seek the optimal replenishment scheme (i.e., the target replenishment scheme) with the lowest risk while meeting the replenishment objectives. Finally, the model outputs a comprehensive decision support report, covering the recommended optimal replenishment scheme, the expected risk level distribution map, and explanations of key uncertainty factors, providing direct scientific basis for engineering design and risk management.

[0080] Once the target replenishment plan is determined, the system enters the full-process implementation and closed-loop optimization stage of the execution monitoring layer shown in Figure 4: First, the execution monitoring of the replenishment plan is initiated, and the optimal replenishment plan in the comprehensive decision support report is broken down into specific engineering instructions to guide the real-time monitoring of the replenishment project. The key engineering parameters, groundwater flow field changes, and solute transport patterns during the replenishment construction process are dynamically monitored throughout the entire process. Then, through the effect evaluation module, the actual data obtained from the monitoring is compared with the prediction results of the multi-physics field coupling simulation to quantitatively evaluate the actual implementation effect of the target replenishment plan, focusing on verifying whether the replenishment efficiency and risk control level have achieved the expected goals.

[0081] After the evaluation is completed, the real-time monitoring data of the backfilling project and the effect evaluation results are synchronously transmitted back to the multiphysics coupling simulation module and optimization decision module of the model through the monitoring data feedback module: if the backfilling effect meets expectations, the current target backfilling plan is continuously executed and the closed loop is completed; if the expected effect is not met, dynamic feedback and backfilling plan optimization decision are triggered, and the backfilling project parameters are readjusted and the backfilling plan is optimized based on the feedback data, and the iterative optimization process is entered again.

[0082] In the above embodiments, by combining reliability theory with multi-objective optimization, the risk quantification and precise optimization of the replenishment scheme are realized; based on the risk assessment of first-order second-moment / Monte Carlo simulation, the failure probability and risk level of the replenishment scheme are quantified, solving the problem of the lack of quantitative basis in traditional safety evaluation; through the two-step method of risk screening and parameter optimization, multi-objective synergistic optimization of replenishment efficiency, cost and risk is achieved under the premise of ensuring risk controllability, thereby improving the scientificity and practicality of the scheme.

[0083] In some embodiments, the multi-objective optimization algorithm includes a multi-objective evolutionary algorithm and a proximal policy optimization algorithm. Step S1507, "using the multi-objective optimization algorithm to iteratively adjust the backfilling engineering parameters of the potential solution, and seeking the optimal parameter combination with the lowest risk under the premise of satisfying the backfilling objective," includes the following steps: Step S15071, using the multi-objective evolutionary algorithm to decompose the multi-objective optimization problem into single-objective sub-problems, wherein the multi-objective optimization problem is a three-objective collaborative optimization that simultaneously achieves the highest backfilling efficiency, the lowest operating cost, and the lowest risk. In some embodiments, step S15071, "using the multi-objective evolutionary algorithm to decompose the multi-objective optimization problem into single-objective sub-problems," includes the following steps: Step S21, based on a dynamic weight adjustment mechanism, the multi-objective evolutionary algorithm is improved to obtain an improved multi-objective evolutionary algorithm; Step S22, based on the improved multi-objective evolutionary algorithm, the Chebyshev scalarization method is used to decompose the multi-objective optimization problem into single-objective sub-problems.

[0084] Improvements in dynamic weight adjustment and Chebyshev scalarization enhance the flexibility and optimization efficiency of the multi-objective evolutionary algorithm. The dynamic weight adjustment mechanism adapts to changes in target priority under different backfilling scenarios, solving the problem that fixed weights cannot match differences in working conditions. The Chebyshev scalarization method achieves efficient decomposition of multi-objective problems into single-objective subproblems, improving the convergence speed and solution set diversity of the algorithm, and ensuring the global optimality of the optimization results.

[0085] Step S15072: Utilizing information co-evolution between adjacent single-objective sub-problems, iteratively optimize the backup position, backup strength, and backup timing of the potential solutions to generate an initial Pareto optimal solution set; Step S15073: Using a reinforcement learning agent based on the proximal policy optimization algorithm, dynamically verify and iteratively update the parameter combinations in the initial Pareto optimal solution set to obtain an updated Pareto optimal solution set; Step S15074: Based on the backup objective, select the optimal parameter combination from the updated Pareto optimal solution set.

[0086] Among them, the Multi-objective Adaptive Recharge Optimization and Evaluation Model (MAREEM) can optimize the recharge scheme, as shown in formula (22). The multi-objective optimization module can be established based on the decomposition multi-objective evolutionary algorithm (MOEA / D) framework: Where x is the decision variable (parameters of the backup plan). =(λ1,λ2,λ3) is the weight vector, corresponding to the three optimization objectives; f1(x) is the ideal point vector; f2(x) is the recovery efficiency objective function (actual value); f3(x) is the operating cost objective function; f4(x) is the risk indicator objective function. It is the Chebyshev scalarized objective function; the improved weight adjustment mechanism is shown in formula (23): Formula (23); where It is an adaptive learning rate that is dynamically adjusted based on historical optimization results. These are the initial weights for the i-th optimization objective. ) is the weight of the i-th optimization objective at time t. ) is the weight of the j-th optimization objective at time t.

[0087] As shown in Figure 5, the reinforcement learning training module is the core of the MAREEM model to achieve adaptive optimization. By modeling the backfilling process as a Markov decision process, the system can learn the optimal decision strategy from historical experience.

[0088] The design of the state vector in the reinforcement learning training module embodies the idea of ​​multi-source information fusion, which includes direct monitoring data (pressure, flow, water level, water quality), model output results (risk index), and historical decision information, to ensure that the agent can fully perceive the system state.

[0089] As shown in formula (24), the state space fully describes the system's operation at time t, providing an information basis for agent decision-making: ;in, It is the vector of the recovery scheme at time t-1. It is a vector of pipeline pressure monitoring data; It is a node traffic monitoring data vector; It is a vector of groundwater level monitoring data; It is a vector of water quality index concentrations; It is a scalar of the risk index; It is the previous time step's backup plan vector.

[0090] As shown in formula (25), the action space adopts a continuous-discrete hybrid design, which supports both fine-tuning of flow rate and switching of replenishment positions, reflecting the control requirements of actual engineering. The action space defines the control operations that the agent can perform: Formula (25); where, It is the amount of flow adjustment to replenish; It is the amount of adjustment for the recovery time; It is the position adjustment vector for backfilling.

[0091] As shown in formula (26), the reward function is the key to guiding the agent's learning, and the design principle is to balance multiple optimization objectives: ;in It is the reward function value at time t. It is the replenishment efficiency at time t, calculated as the ratio of effective replenishment water volume to total replenishment water volume; It is an efficiency weighting coefficient, reflecting the degree of importance attached to water resource utilization. It is the operating cost per unit of replenishment, including electricity consumption, medicine consumption, etc. It is a risk index for recovery, calculated by the URRTM model; It is an operation frequency penalty to prevent excessively frequent equipment adjustments; , , These are the corresponding penalty weight coefficients. The weight coefficients are determined using the Analytic Hierarchy Process (AHP) and are dynamically adjusted based on decision preferences during the learning process.

[0092] As shown in formula (27), the PPO algorithm ensures the stability of the training process by limiting the magnitude of policy updates: Formula (27); where, It is the probability ratio between the old and new strategies; It is the advantage function estimator, calculated using the GAE method; These are the trimming parameters, typically set to 0.1-0.3; These are policy network parameters. The value of the pruning loss function.

[0093] In the above embodiments, the efficiency and robustness of the backfill scheme parameter optimization are improved by fusing multi-objective evolutionary algorithms and reinforcement learning algorithms; the decomposition strategy of the multi-objective evolutionary algorithm efficiently solves the three-objective optimization problem of "efficiency-cost-risk" and generates a Pareto solution set covering multiple trade-off dimensions; the dynamic verification of the proximal policy optimization algorithm utilizes the autonomous learning capability of reinforcement learning to iteratively update the solution set, improving the adaptability of parameter combinations to actual working conditions and ensuring the executability of the optimal solution.

[0094] As shown in Figure 5, the replenishment optimization decision system of the application embodiment consists of three parts: multi-objective optimization core, reinforcement learning decision core, and reinforcement learning training module, forming a complete process of data input, multi-objective optimization, intelligent decision-making, training evolution, and closed-loop re-optimization.

[0095] In the core part of multi-objective optimization, the input data includes multi-source monitoring data and upstream model prediction information. Multi-source monitoring data includes real-time monitoring data such as pipeline pressure, node flow, groundwater level, and water quality concentration. Upstream model prediction information includes the remediation risk index and historical decision schemes output by the risk assessment module. The MOEA / D algorithm is adopted, based on the Chebyshev scalar objective function as shown in formula (22), to decompose the three-objective optimization problem of "highest remediation efficiency, lowest operating cost, and lowest risk" into single-objective sub-problems. Through objective function decomposition and co-evolutionary decomposition, the information co-evolution between adjacent single-objective sub-problems is used to iteratively optimize the remediation position, remediation intensity, and remediation timing, generate the Pareto front PF covering multiple trade-off dimensions, and output the Pareto optimal solution set. When selecting the decision mode, automatic decision-making or manual decision-making is supported. The weights shown in formula (23) can be adjusted by the decision-maker's preference input, and finally the reinforcement learning intelligent decision is output, which transmits the Pareto optimal solution set to the reinforcement learning decision core.

[0096] In the core of reinforcement learning decision-making, during the fusion decision generation phase, system state awareness can be achieved. Action decisions are made based on the policy network through the PPO policy network, and policy optimization is guided by the value function evaluation and reward function. Based on the Pareto optimal solution set, a pre-reinforcement learning policy is used to generate an executable recovery control scheme (i.e., a fusion decision scheme). During the recovery control execution phase, the recovery control scheme is distributed to the execution mechanism to drive the recovery project implementation. During the state re-evaluation phase, the system state data after recovery is collected to evaluate the control effect. During the performance evaluation phase, if the performance meets the target, the experience data is stored to complete the current round of decision-making. If the performance does not meet the target, rolling optimization and adjustment are triggered, and a new round of multi-objective optimization is initiated. The experience data is then transferred to the experience playback pool of the reinforcement learning training module.

[0097] In the reinforcement learning training module, experience data is obtained from the experience replay pool, the training process is started, the PPO training algorithm is executed, and network parameters are updated and policies are improved according to the algorithm gradient. Through multiple rounds of training iterations, the policy is continuously optimized, and the improved policy is deployed to the reinforcement learning decision core to realize the adaptive learning loop of model autonomous evolution and policy network update. By triggering the multi-objective optimization core, the Pareto optimal solution set is resolved based on the new policy to complete the entire closed-loop iteration.

[0098] In this embodiment, MAREEM's decision-making is a closed-loop adaptive process of "prediction-optimization-learning-execution-evolution," including the following steps: Step S31, multi-objective optimization solution and solution set generation; the system receives prediction information (such as future inflows and risk fields) and the current real-time state from upstream models (such as MISF-HDM and URRTM). The MOEA / D optimizer is started, and under the premise of satisfying all engineering constraints, it quickly solves and outputs a set of Pareto optimal replenishment solutions covering different trade-off points.

[0099] Step S32: Intelligent decision-making based on learning experience; the decision-maker can initially select a baseline solution from the Pareto solution set based on current management objectives (such as focusing on efficiency or cost in this period). Simultaneously, the built-in PPO agent is activated, which evaluates the baseline solution based on a deep understanding of historical data and system dynamics, and may propose forward-looking, sequential fine-tuning suggestions. For example, the agent might suggest: moderately increasing replenishment volume for "energy storage" during periods of low electricity prices to reduce overall energy costs; or fine-tuning the dosage of chemicals in the water treatment unit several hours in advance based on water quality change predictions. The final decision is a fusion of the static optimal solution given by the optimizer and the dynamic adjustment strategies suggested by the agent.

[0100] Step S33: Robust Execution and Rolling Re-optimization; the fused decision scheme is distributed to the field execution agency. The system enters Model Predictive Control (MPC), looping: the system state data after actual execution is re-collected at a fixed frequency (e.g., every 15 minutes) and compared with the predicted values. Once the deviation exceeds a set threshold, the system immediately uses the latest state as the starting point and re-triggers steps S31 and S32 to perform rolling optimization and decision adjustment, ensuring that the system always dynamically tracks the globally optimal trajectory and has strong robustness to external disturbances.

[0101] Step S34: Experience Accumulation and Autonomous Model Evolution. Data (state, action, reward, new state) generated from each complete "decision-execution-feedback" cycle is stored in the experience replay pool. The PPO agent periodically samples data from the pool to update its policy network parameters. This means that every operation makes the system smarter. For example, during a six-month trial run in a mining area, the system, through continuous adaptive learning, gradually improved its replenishment efficiency from an initial 68% to a stable level above 82%, achieving a leap from optimization to autonomous intelligent decision-making. In actual deployment, the agent is first extensively pre-trained in a simulation environment built from high-fidelity digital twins and historical data, and then connected to the real system for safe and controlled online fine-tuning, ensuring the reliability and security of engineering applications.

[0102] MAREEM's innovation lies in transforming traditional offline optimization into an adaptive optimization system with online learning and self-evolution capabilities, achieving a theoretical leap from static optimization to dynamic autonomous decision-making. Its principles are rooted in the deep integration of two levels. The first level is multi-objective Pareto optimization, using a decomposition-based multi-objective evolutionary algorithm (MOEA / D) as its framework to address the three conflicting core objectives of maximizing backfilling efficiency, minimizing operating costs, and minimizing hydrogeological risks. MOEA / D decomposes the complex multi-objective problem into a set of single-objective sub-problems (using Chebyshev scaling) and leverages the information between adjacent sub-problems for co-evolution, efficiently generating a uniformly distributed set of Pareto optimal solutions. This model improves upon traditional MOEA / D by introducing a dynamic weight adjustment mechanism, enabling the algorithm to adaptively explore the complex and non-convex regions of the Pareto front, ensuring the completeness and diversity of solutions. The second level, sequential decision-making and online learning, models the entire backfilling system's operation as a sequential decision-making process and constructs the MAREEM model itself as a reinforcement learning agent. The agent learns through continuous interaction with its environment (the real-world feedback system and its digital twin), with the Proximal Policy Optimization (PPO) algorithm at its core. By limiting the magnitude of policy updates, PPO strikes a balance between exploration and exploitation, learning stability and convergence speed, enabling the agent to autonomously extract optimized policies from historical operational data and real-time feedback, gradually gaining forward-looking decision-making capabilities.

[0103] This model upgrades traditional groundwater numerical simulation from an offline analysis tool to an online prediction and decision support system, enabling real-time simulation and risk assessment of the water diversion channels during groundwater recharge. The model's decision-making logic is a closed-loop intelligent decision-making process, from basic construction to dynamic simulation, risk assessment, and scheme optimization. Based on multiphysics coupling theory, the model comprehensively considers multiple factors such as hydrogeological conditions, soil characteristics, and recharge water quality. By establishing a high-precision digital twin of the geological structure, it achieves visualized prediction of the entire recharge response process. The model's principle is based on the coupled simulation of water flow and solute transport in the saturated-unsaturated zones. It uses the Richards equation to describe the water transport process (water flow transport) in the vadose zone and saturated zone, combined with convection-dispersion equations to simulate solute transport. By introducing reliability theory, it transforms traditional deterministic safety assessment into probabilistic risk assessment, providing a quantitative basis for scientific decision-making on recharge schemes, and ultimately achieving dynamic optimization and decision support for recharge schemes.

[0104] This application constructs an intelligent decision-making system for end-to-end collaborative optimization. A closed-loop intelligent control architecture of "data perception - model prediction - optimization decision - execution control" is established, achieving deep collaboration among all stages through a unified data bus. Specifically, geological digital twin technology, multi-objective optimization algorithms, and reinforcement learning mechanisms are organically combined to form an intelligent decision-making system with autonomous learning and continuous improvement capabilities. This architectural innovation enables the system to optimize operational strategies from a global perspective, rather than relying on the localized optimization of traditional technologies.

[0105] Regarding leak location, this application introduces compressed sensing theory into pipeline leak detection and proposes a sparse optimization algorithm based on L1 regularization, solving the problem of accurate location under sparse monitoring point conditions. By introducing a physically enhanced sensitivity matrix, the leak location accuracy is improved from the traditional hundred-meter level to the meter level. Regarding remediation decision-making, this application creates a risk assessment model based on reliability theory, transforming traditional deterministic safety judgments into probabilistic risk assessments, defining a remediation risk index, and providing a quantitative basis for remediation safety.

[0106] This application achieves multi-timescale optimization decision fusion, including day-ahead economic dispatch, real-time safety control, and online adaptive learning. Specifically, it combines the reinforcement learning PPO algorithm with the multi-objective evolutionary algorithm MOEA / D, enabling the system to learn optimization strategies from historical data. In practical applications, the system can automatically adjust its operating strategy based on electricity price signals, increasing replenishment for energy storage during off-peak periods and reducing energy consumption during peak periods, thereby maximizing economic benefits.

[0107] In terms of system performance, this application significantly improves water resource utilization efficiency through end-to-end optimization. Practical applications demonstrate that the system effectively coordinates the operational status of each stage, achieving optimized energy consumption control while ensuring replenishment effectiveness. Compared to traditional methods, the system exhibits significant advantages in replenishment efficiency and operating costs. Regarding reliability, this application greatly enhances system security through intelligent early warning and rapid response mechanisms. Advanced leak detection algorithms can promptly identify anomalies, and risk assessment models can accurately predict operational risks, providing an effective means for prevention. The system's adaptive learning capability ensures performance stability during long-term operation. In terms of economy, this application generates considerable economic benefits through optimized operating strategies. The intelligent scheduling system can adjust operating plans based on real-time conditions, reducing operating costs while ensuring service quality. The system's modular design also reduces maintenance complexity and lowers long-term operating expenses.

[0108] In a comprehensive water resource utilization project in a mining area, the system proposed in this application was successfully applied to the reuse treatment of high-salinity mine water. By establishing a geological digital twin model, the characteristics of groundwater flow were accurately depicted; a scientific recharge plan was formulated using a multi-objective optimization algorithm; and real-time monitoring of operational risks was achieved with the help of an intelligent early warning system. The system demonstrated good adaptability and reliability.

[0109] After adopting the method described in this application, a groundwater recharge project in a certain region achieved refined and dynamic management. By collecting operational data in real time through an intelligent monitoring network, automatically adjusting recharge parameters using optimization algorithms, and providing timely warnings of potential problems based on a risk assessment model, this method effectively improved recharge efficiency and ensured the safe operation of the project.

[0110] In recharge projects under complex geological conditions, this application establishes a high-precision groundwater model by integrating multi-source monitoring data. Combined with real-time optimization algorithms, precise control of the recharge process is achieved. The system exhibits stable performance under various operating conditions, providing a reliable solution for water resource management under similar conditions.

[0111] Compared to traditional groundwater management systems, this application is the first to deeply integrate modern intelligent algorithms with hydrogeological models, realizing a complete technology chain from monitoring and early warning to optimization decision-making. At the algorithmic level, the hybrid optimization strategy proposed in this application overcomes the limitations of traditional methods. By combining physical models with data-driven methods, it ensures both the physical rationality of the results and improves the system's adaptability. The effectiveness of this method has been verified in multiple practical engineering projects.

[0112] This application constructs a full-process intelligent decision-making system with dynamic regulation of the transmission and distribution process as its core, changing the traditional fragmented operation mode. The technical solution of this system aims to achieve coordinated intelligent control of the entire chain of water treatment, transmission and distribution optimization, and geological replenishment through real-time data-driven approaches. The core of its method lies in establishing a central nervous system with real-time perception, intelligent decision-making, and dynamic execution capabilities.

[0113] Specifically, this solution first utilizes sensor arrays (pressure, acoustic, and online water quality monitors) deployed at key nodes of the transmission and distribution network to conduct real-time risk perception of the network's operational status. This enables early identification and precise location of leaks, while simultaneously dynamically tracking water quality changes to prevent secondary pollution. Building upon this foundation, the system employs a high-precision hydraulic model (digital twin) as its intelligent decision-making engine. This model not only simulates the water flow, pressure, and water quality within the network in real time but also dynamically optimizes pump station operation strategies based on real-time replenishment needs and time-of-use electricity pricing signals. This stabilizes the network pressure within a safe and efficient range and significantly reduces transmission and distribution energy consumption.

[0114] In response to abnormal operating conditions, the system incorporates a multi-level safety early warning and emergency decision support mechanism. Once the monitoring data triggers the early warning threshold, the system immediately alarms and rapidly simulates various emergency control schemes (such as valve operation and pump station adjustment) based on the current state, providing operators with the optimal handling decision and minimizing the impact of accidents. To ensure the safety of the distribution endpoint, this application also introduces a geological structure digital twin cognitive module. This module accurately characterizes the geological features of the replenishment zone, providing safety constraints on replenishment flow and pressure for the distribution process, preventing geological risks induced by improper replenishment. Ultimately, all links form a closed loop through the replenishment scheme optimization and full-process effect evaluation module. By continuously learning key performance indicators in the distribution process, the system continuously optimizes control strategies, ultimately forming a safe dynamic control closed loop of monitoring-diagnosis-decision-execution-evaluation-optimization, thereby ensuring the stable operation of the system in a safe, efficient, and adaptive state.

[0115] This application constructs a closed-loop intelligent control system with prediction, optimization, decision-making, and learning capabilities. It surpasses traditional rule-based or local optimization-based control methods, achieving autonomous optimization throughout the entire process—from real-time perception to intelligent decision-making and adaptive learning—through a series of interconnected advanced mathematical models.

[0116] Based on the foregoing embodiments, this application further provides an intelligent control system for water transmission, distribution and replenishment in mining areas. The system includes various modules and sub-modules, and each unit of each sub-module can be implemented by a processor in an electronic device; of course, it can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0117] Figure 6 is a schematic diagram of the composition structure of a smart control system for water transmission, distribution, and replenishment in a mining area provided in an embodiment of this application. As shown in Figure 6, the system 600 includes: a data acquisition module 61, used to acquire pipeline operation data and current multi-source geological and hydrological data of the transmission and distribution network, wherein the pipeline operation data includes pipeline topology data, pipe material parameters, and real-time monitoring data; a state estimation module 62, used to determine the optimal state estimate of the transmission and distribution network based on the pipeline operation data and using an ensemble Kalman filter algorithm, using the real-time hydraulic model in the dynamic control model of the transmission and distribution network; and to determine the pressure residual of each monitoring node in the transmission and distribution network based on the optimal state estimate and the real-time monitoring data; and an emergency decision generation module 63, used to determine the target leakage result of the transmission and distribution network based on the pressure residual of each monitoring node, using the leakage location algorithm in the dynamic control model of the transmission and distribution network when the pressure residual indicates that the transmission and distribution network is in an abnormal operating condition; and to determine the target leakage result based on the pressure residual of each monitoring node, using the dynamic control model of the transmission and distribution network. The system employs an optimized scheduler to generate emergency control decisions. A transmission and distribution decision generation module 64, under the condition that the pressure residual indicates the transmission and distribution network is in normal operating condition, utilizes the optimized scheduler and, based on the network operation data, executes multi-timescale predictive control through the real-time hydraulic model to generate transmission and distribution scheduling decisions with the goal of minimizing total operating costs and ensuring safe operation. Based on the emergency control decisions or the transmission and distribution scheduling decisions, the system performs operational control of the transmission and distribution network to achieve the transmission and distribution of replenishment water. A replenishment scheme generation module 65 utilizes the geological structure digital twin model in the replenishment response prediction model and, based on the current multi-source geological and hydrological data, performs multi-physics field coupling real-time simulation of the entire diffusion and migration process of replenishment water under multiple candidate replenishment schemes to obtain the replenishment prediction results corresponding to each candidate replenishment scheme. Based on reliability theory, it performs risk assessment on each replenishment prediction result to obtain the evaluation results of multiple candidate replenishment schemes. Based on the evaluation results of the multiple candidate replenishment schemes, it uses a multi-objective optimization algorithm to solve for the target replenishment scheme.

[0118] In some possible embodiments, the state estimation module 62 includes: a fourth determination submodule, used to determine the initial state estimate of the transmission and distribution network by using the real-time hydraulic model in the dynamic control model of the transmission and distribution network, based on the network topology data, the pipe parameters, the pre-established extended node flow continuity equation and the pipe segment pressure drop equation, and solving the equation using the extended node equation method; and an optimization submodule, used to optimize the initial state estimate based on the ensemble Kalman filter algorithm and the real-time monitoring data to obtain the optimal state estimate of the transmission and distribution network.

[0119] In some possible embodiments, the emergency decision generation module 63 includes: a first determination submodule, used to determine the real-time operating condition of the transmission and distribution network based on the optimal state estimate, and to analyze the real-time operating condition using the real-time hydraulic model to obtain an enhanced sensitivity matrix; a second determination submodule, used to determine the pressure change vector based on the pressure residuals of each monitoring node; a third determination submodule, used to construct and solve a sparse optimization problem based on the pressure change vector and the enhanced sensitivity matrix to determine the initial leakage result of the transmission and distribution network; a credibility assessment submodule, used to perform forward pressure simulation using the real-time hydraulic model based on the initial leakage result to obtain simulated pressure change values; to assess the credibility of the initial leakage result based on the degree of fit between the simulated pressure change values ​​and the actual pressure change values; and an update submodule, used to adaptively adjust the parameters of the real-time hydraulic model if the credibility does not reach a preset standard, and to re-execute the step of obtaining the enhanced sensitivity matrix based on the adjusted real-time hydraulic model, iteratively solving until the credibility reaches the preset standard, and determining the corresponding leakage result as the target leakage result.

[0120] In some possible embodiments, the recharge scheme generation module 65 includes: an acquisition submodule, used to acquire historical multi-source geological and hydrological data, and fuse the historical multi-source geological and hydrological data to obtain historical fused geological and hydrological data; a construction submodule, used to construct a geological structure digital twin model containing three-dimensional stratigraphic structure and aquifer spatial distribution based on the historical fused geological and hydrological data and using geostatistical methods; and a simulation submodule, used to set initial conditions and boundary conditions based on the current multi-source geological and hydrological data, and in the geological structure digital twin model, use the finite element method or finite difference method to couple and solve Darcy's law describing groundwater flow and the convection-dispersion equation describing solute migration, to realize real-time simulation of the entire diffusion and transport process of recharge water through multi-physics field coupling, and obtain recharge prediction results.

[0121] In some possible embodiments, the assessment result includes a risk level. The remediation scheme generation module 65 includes: a fifth determination submodule, used to determine the probability distribution characteristics of the current multi-source geological and hydrological data; a risk assessment submodule, used to assess the risk of the remediation prediction result for a specified risk event based on the probability distribution characteristics, using reliability theory, through a first second-order moment method or Monte Carlo simulation, to obtain the failure probability of multiple candidate remediation schemes corresponding to the specified risk event; a sixth determination submodule, used to determine the risk level of the corresponding candidate remediation scheme based on the failure probability of the multiple candidate remediation schemes; a screening submodule, used to compare candidate remediation schemes with different combinations of remediation location, intensity, and time series parameters based on the risk level and core risk indicators of each candidate remediation scheme, and screen out potential schemes with controllable risk; an iterative adjustment submodule, used to iteratively adjust the remediation engineering parameters of the potential schemes using a multi-objective optimization algorithm, and seek the optimal parameter combination with the lowest risk under the premise of meeting the remediation target; and a seventh determination submodule, used to determine the target remediation scheme based on the optimal parameter combination.

[0122] In some possible embodiments, the multi-objective optimization algorithm includes a multi-objective evolutionary algorithm and a proximal policy optimization algorithm. The iterative adjustment submodule includes: a decomposition unit, used to decompose the multi-objective optimization problem into single-objective sub-problems using the multi-objective evolutionary algorithm, wherein the multi-objective optimization problem is a three-objective collaborative optimization that simultaneously achieves the highest backfill efficiency, lowest operating cost, and lowest risk; an iterative optimization unit, used to utilize information co-evolution between adjacent single-objective sub-problems to iteratively optimize the backfill position, backfill strength, and backfill timing of the potential solutions to generate an initial Pareto optimal solution set; an iterative update unit, used to use a reinforcement learning agent based on the proximal policy optimization algorithm to dynamically verify and iteratively update the parameter combinations in the initial Pareto optimal solution set to obtain an updated Pareto optimal solution set; and a filtering unit, used to filter the optimal parameter combination from the updated Pareto optimal solution set based on the backfill objective.

[0123] In some possible embodiments, the decomposition unit includes: an improvement subunit, used to improve the multi-objective evolutionary algorithm based on a dynamic weight adjustment mechanism to obtain an improved multi-objective evolutionary algorithm; and a decomposition subunit, used to decompose the multi-objective optimization problem into single-objective sub-problems based on the improved multi-objective evolutionary algorithm using the Chebyshev scalarization method.

[0124] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0125] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0126] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0127] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected to achieve the purpose of the embodiments of this application according to actual needs. In addition, each functional unit in the embodiments of this application may be fully integrated into one processing unit, or each unit may be a separate unit, or two or more units may be integrated into one unit; the integrated unit may be implemented in hardware or in the form of hardware plus software functional units.

[0128] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause the device automatic test line to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0129] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined to obtain new method embodiments without conflict. The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined to obtain new method embodiments or device embodiments without conflict.

[0130] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent control of water transmission, distribution, and replenishment in mining areas, the method comprising: Acquire pipeline operation data and current multi-source geological and hydrological data of the transmission and distribution pipeline network. The pipeline operation data includes pipeline topology data, pipe material parameters, and real-time monitoring data. Using the real-time hydraulic model in the dynamic control model of the transmission and distribution pipeline network, and based on the network operation data, the optimal state estimate of the transmission and distribution pipeline network is determined using the ensemble Kalman filter algorithm; based on the optimal state estimate and the real-time monitoring data, the pressure residual of each monitoring node in the transmission and distribution pipeline network is determined. When the pressure residual indicates that the transmission and distribution network is in an abnormal operating condition, the leakage location algorithm in the dynamic control model of the transmission and distribution network is used to determine the target leakage result of the transmission and distribution network based on the pressure residual of each monitoring node. Based on the target leakage results, an emergency control decision is generated using the optimization scheduler of the dynamic control model of the transmission and distribution network. Under the condition that the pressure residual indicates that the transmission and distribution network is in normal operating condition, the optimization scheduler is used to perform multi-time-scale predictive control based on the network operation data and the real-time hydraulic model to generate transmission and distribution scheduling decisions with the goal of minimizing total operating cost and safe operation. Based on the emergency control decision or the transmission and distribution scheduling decision, the transmission and distribution pipeline network is operated and controlled to achieve the transmission and distribution of replenishment water; using the geological structure digital twin model in the replenishment response prediction model, based on the current multi-source geological and hydrological data, a multi-physics field coupled real-time simulation of the entire diffusion and migration process of replenishment water under multiple candidate replenishment schemes is performed to obtain the replenishment prediction results corresponding to each candidate replenishment scheme; based on reliability theory, a risk assessment is performed on each replenishment prediction result to obtain the evaluation results of multiple candidate replenishment schemes; based on the evaluation results of the multiple candidate replenishment schemes, a multi-objective optimization algorithm is used to solve the problem to obtain the target replenishment scheme.

2. The method according to claim 1, characterized in that, The method of using the real-time hydraulic model in the dynamic control model of the transmission and distribution network, based on the network operation data, and using the ensemble Kalman filter algorithm to determine the optimal state estimate of the transmission and distribution network includes: using the real-time hydraulic model in the dynamic control model of the transmission and distribution network, based on the network topology data, the pipe material parameters, the pre-established extended node flow continuity equation and the pipe section pressure drop equation, solving the extended node equation method to determine the initial state estimate of the transmission and distribution network; and optimizing the initial state estimate based on the ensemble Kalman filter algorithm and the real-time monitoring data to obtain the optimal state estimate of the transmission and distribution network.

3. The method according to claim 1, characterized in that, The method of using the leakage location algorithm in the dynamic control model of the transmission and distribution network to determine the target leakage result of the transmission and distribution network based on the pressure residuals of each monitoring node includes: determining the real-time operating condition of the transmission and distribution network based on the optimal state estimate; analyzing the real-time operating condition using the real-time hydraulic model to obtain an enhanced sensitivity matrix; determining the pressure change vector based on the pressure residuals of each monitoring node; constructing and solving a sparse optimization problem based on the pressure change vector and the enhanced sensitivity matrix to determine the initial leakage result of the transmission and distribution network; performing forward pressure simulation using the real-time hydraulic model based on the initial leakage result to obtain simulated pressure change values; evaluating the reliability of the initial leakage result based on the degree of fit between the simulated pressure change values ​​and the actual pressure change values; and adaptively adjusting the parameters of the real-time hydraulic model if the reliability does not reach the preset standard, re-executing the step of obtaining the enhanced sensitivity matrix based on the adjusted real-time hydraulic model, iteratively solving until the reliability reaches the preset standard, and determining the corresponding leakage result as the target leakage result.

4. The method according to claim 1, characterized in that, The method utilizes a geological structure digital twin model within the recharge response prediction model. Based on the current multi-source geological and hydrological data, it performs real-time multi-physics field coupling simulation of the entire process of recharge water diffusion and migration to obtain recharge prediction results. This includes: acquiring historical multi-source geological and hydrological data; fusing the historical multi-source geological and hydrological data to obtain historical fused geological and hydrological data; constructing a geological structure digital twin model containing three-dimensional stratigraphic structure and aquifer spatial distribution using geostatistical methods based on the historical fused geological and hydrological data; setting initial and boundary conditions based on the current multi-source geological and hydrological data; and using the finite element method or finite difference method in the geological structure digital twin model to couple and solve Darcy's law describing groundwater flow and the convection-dispersion equation describing solute migration, thereby achieving real-time multi-physics field coupling simulation of the entire process of recharge water diffusion and migration to obtain recharge prediction results.

5. The method according to claim 1, characterized in that, The assessment results include risk levels. The risk assessment of the replenishment prediction results based on reliability theory yields assessment results for multiple candidate replenishment schemes. Based on the assessment results of these multiple candidate replenishment schemes, a multi-objective optimization algorithm is used to solve for and obtain the target replenishment scheme. This includes: determining the probability distribution characteristics of the current multi-source geological and hydrological data; based on the probability distribution characteristics, using reliability theory and through first-order second-moment method or Monte Carlo simulation, assessing the risk of the replenishment prediction results for a specified risk event, obtaining the failure probability of multiple candidate replenishment schemes corresponding to the specified risk event; determining the risk level of the corresponding candidate replenishment scheme based on the failure probability of the multiple candidate replenishment schemes; comparing candidate replenishment schemes with different replenishment location, intensity, and time series parameter combinations based on the risk level and core risk indicators of each candidate replenishment scheme, and selecting potential schemes with controllable risk; using a multi-objective optimization algorithm to iteratively adjust the replenishment engineering parameters of the potential schemes, seeking the optimal parameter combination with the lowest risk while meeting the replenishment objective; and determining the target replenishment scheme based on the optimal parameter combination.

6. The method according to claim 5, characterized in that, The multi-objective optimization algorithm includes a multi-objective evolutionary algorithm and a proximal policy optimization algorithm. The multi-objective optimization algorithm iteratively adjusts the backfilling engineering parameters of the potential solution to find the optimal parameter combination with the lowest risk while satisfying the backfilling objective. This includes: using the multi-objective evolutionary algorithm to decompose the multi-objective optimization problem into single-objective sub-problems, where the multi-objective optimization problem is a three-objective collaborative optimization that simultaneously achieves the highest backfilling efficiency, lowest operating cost, and lowest risk; utilizing information co-evolution between adjacent single-objective sub-problems to iteratively optimize the backfilling position, backfilling strength, and backfilling timing of the potential solution, generating an initial Pareto optimal solution set; using a reinforcement learning agent based on the proximal policy optimization algorithm to dynamically verify and iteratively update the parameter combinations in the initial Pareto optimal solution set, obtaining an updated Pareto optimal solution set; and selecting the optimal parameter combination from the updated Pareto optimal solution set based on the backfilling objective.

7. The method according to claim 6, characterized in that, The step of using the multi-objective evolutionary algorithm to decompose the multi-objective optimization problem into single-objective sub-problems includes: improving the multi-objective evolutionary algorithm based on a dynamic weight adjustment mechanism to obtain an improved multi-objective evolutionary algorithm; and using the Chebyshev scalarization method based on the improved multi-objective evolutionary algorithm to decompose the multi-objective optimization problem into single-objective sub-problems.

8. A smart control system for water transmission, distribution, and replenishment in a mining area, the system comprising: The data acquisition module is used to acquire pipeline operation data and current multi-source geological and hydrological data of the transmission and distribution pipeline network. The pipeline operation data includes pipeline topology data, pipe material parameters and real-time monitoring data. The state estimation module is used to determine the optimal state estimate of the transmission and distribution pipeline network based on the network operation data and the real-time hydraulic model in the dynamic control model of the transmission and distribution pipeline network, using an ensemble Kalman filter algorithm; and to determine the pressure residual of each monitoring node in the transmission and distribution pipeline network based on the optimal state estimate and the real-time monitoring data. The emergency decision generation module is used to determine the target leakage result of the transmission and distribution network based on the pressure residual of each monitoring node when the pressure residual indicates that the transmission and distribution network is in an abnormal operating condition. Based on the target leakage results, an emergency control decision is generated using the optimization scheduler of the dynamic control model of the transmission and distribution network. The transmission and distribution decision generation module is used to generate transmission and distribution scheduling decisions by using the optimization scheduler and the real-time hydraulic model to perform multi-time-scale predictive control based on the pipeline operation data, under the condition that the pressure residual characterizes the transmission and distribution network as being in normal operating condition, with the goal of minimizing total operating cost and safe operation. Based on the emergency control decision or the transmission and distribution scheduling decision, the transmission and distribution pipeline network is operated and controlled to realize the transmission and distribution of replenishment water; the replenishment scheme generation module is used to use the geological structure digital twin model in the replenishment response prediction model, based on the current multi-source geological and hydrological data, to perform multi-physics field coupling real-time simulation of the entire process of replenishment water diffusion and migration, and obtain replenishment prediction results; based on reliability theory, the replenishment prediction results are risk-assessed to obtain the evaluation results of multiple candidate replenishment schemes, and based on the evaluation results of the multiple candidate replenishment schemes, a multi-objective optimization algorithm is used to solve for the target replenishment scheme.

9. The system according to claim 8, characterized in that, The emergency decision generation module includes: a first determination submodule, used to determine the real-time operating condition of the transmission and distribution network based on the optimal state estimate, and to analyze the real-time operating condition using the real-time hydraulic model to obtain an enhanced sensitivity matrix; a second determination submodule, used to determine the pressure change vector based on the pressure residuals of each monitoring node; a third determination submodule, used to construct and solve a sparse optimization problem based on the pressure change vector and the enhanced sensitivity matrix to determine the initial leakage result of the transmission and distribution network; a credibility assessment submodule, used to perform forward pressure simulation using the real-time hydraulic model based on the initial leakage result to obtain simulated pressure change values; to assess the credibility of the initial leakage result based on the degree of fit between the simulated pressure change values ​​and the actual pressure change values; and an update submodule, used to adaptively adjust the parameters of the real-time hydraulic model if the credibility does not reach a preset standard, and to re-execute the step of obtaining the enhanced sensitivity matrix based on the adjusted real-time hydraulic model, iteratively solving until the credibility reaches the preset standard, and determining the corresponding leakage result as the target leakage result.

10. The system according to claim 8, characterized in that, The recharge scheme generation module includes: an acquisition submodule, used to acquire historical multi-source geological and hydrological data, and fuse the historical multi-source geological and hydrological data to obtain historical fused geological and hydrological data; a construction submodule, used to construct a geological structure digital twin model containing three-dimensional stratigraphic structure and aquifer spatial distribution based on the historical fused geological and hydrological data and using geostatistical methods; and a simulation submodule, used to set initial and boundary conditions based on the current multi-source geological and hydrological data, and in the geological structure digital twin model, use the finite element method or finite difference method to couple and solve Darcy's law describing groundwater flow and the convection-dispersion equation describing solute migration, to realize real-time simulation of the entire diffusion and transport process of recharge water through multi-physics field coupling, and obtain recharge prediction results.

Images