Integrated pipeline filling and draining collaborative optimization model construction method and system thereof

Abstract

This invention discloses an integrated pipeline irrigation and drainage collaborative optimization model construction method and system, relating to the field of farmland water conservancy engineering technology. The method includes: acquiring basic data to determine irrigation and drainage design flow rates; constructing an initial pipeline network topology based on this, establishing a pipeline network life-cycle resource cost assessment function and hydraulic constraints; using a non-dominated sorting genetic algorithm to iteratively obtain a set of pipeline network configuration schemes; in drainage mode, acquiring actual water level receding time-series data; performing drainage simulation based on a hydrodynamic model to output simulated time-series data; retrieving the actual friction coefficient caused by sediment deposition within the pipeline with the goal of minimizing the error between the two; before switching to irrigation mode, feeding back the actual friction coefficient to the hydraulic constraints for re-optimization, and outputting a corrected water distribution strategy. This method can dynamically compensate for changes in pipe wall resistance caused by sediment, ensuring pump head matching and stable terminal water pressure. This invention also provides a collaborative optimization system for executing the above method.

Description

Technical Field

[0001] This invention relates to the field of farmland water conservancy engineering technology, and in particular to a method and system for constructing an integrated pipeline irrigation and drainage collaborative optimization model. Background Technology

[0002] In farmland irrigation projects, the water supply network and drainage network are typically physically separate. Operating two independent networks requires substantial construction investment and occupies a significant amount of land when laid in the fields. Integrating these two physical networks into a single underground pipeline topology is an effective and novel technological approach in this field.

[0003] Integrated irrigation and drainage networks require alternating two-way fluid control operations, altering between pressurized water supply and gravity drainage, during actual service. Surface runoff, under drainage conditions, carries sediment particles into underground pipe sections, and sediment settles due to gravity in areas where the flow rate slows. This physical phenomenon inevitably leads to sediment adhering to the pipe walls in the later stages of drainage, altering the internal surface condition of the pipeline.

[0004] The silt adhering to the pipe walls causes dynamic changes in the overall friction coefficient of the pipe network. If the altered pipe wall friction coefficient due to silt adhesion is not corrected after flood drainage, and the system is directly put into subsequent water supply and irrigation operations, the theoretically preset hydraulic parameters will deviate from the actual fluid resistance. This objective parameter mismatch leads to the problem that the pump output head cannot overcome the increased friction resistance, resulting in a decrease in water pressure at the end of the pipe network and affecting the rational water distribution at the terminal. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for constructing an integrated pipeline irrigation and drainage collaborative optimization model, in order to solve the problem mentioned in the background art that after flood drainage, the friction coefficient of the pipe wall changes due to silt deposition in the integrated pipeline network. If this change is not corrected and directly introduced into the water supply operation, it will cause the pump output head mismatch and the attenuation of the terminal water pressure.

[0006] To achieve the aforementioned objective, the technical solution of the present invention is as follows:

[0007] A method for constructing an integrated pipeline irrigation and drainage collaborative optimization model includes the following steps:

[0008] S10. Obtain basic data corresponding to the target farmland area, determine the irrigation design flow and drainage design flow based on the basic data, and form an irrigation mode and a drainage mode. The basic data includes meteorological and hydrological parameters, soil and crop parameters, as well as system configuration and energy consumption parameters.

[0009] S20. Based on the irrigation design flow and the drainage design flow, construct the initial pipeline network topology and establish the pipeline network life cycle resource cost assessment function and hydraulic constraints.

[0010] S30. Use a non-dominated sorting genetic algorithm to iteratively optimize the pipe parameters included in the initial pipeline network topology to obtain a set of pipeline network configuration schemes;

[0011] S40. Output the pipeline configuration parameters according to the target scheme in the pipeline configuration scheme set, and obtain the actual water level receding time series data of the physical entity pipeline network constructed based on the target scheme in the drainage mode. The actual water level receding time series data is the data of farmland water flowing into the physical entity pipeline network through irrigation and drainage well piles.

[0012] S50. Based on the non-steady flow dynamics model, the drainage process of the drainage mode is simulated, and the simulated water level receding time series data is output.

[0013] S60. Minimize the error between the actual water level receding time series data and the simulated water level receding time series data as the optimization objective, and calculate the actual friction coefficient inside the pipe of the physical entity pipe network caused by silt deposition.

[0014] S70. Switch between drainage mode and irrigation mode. Before switching to irrigation mode, feed back the actual friction coefficient into the hydraulic constraints and trigger the non-dominated sorting genetic algorithm to re-expand the optimization iteration in order to output the corrected pipeline water distribution strategy, thereby constructing an irrigation and drainage coordinated optimization model.

[0015] The solution further involves determining the irrigation design flow rate and drainage design flow rate based on the aforementioned basic data, including:

[0016] The irrigation modulus and drainage modulus are calculated based on the meteorological and hydrological parameters and the soil and crop parameters.

[0017] The larger of the flow rates corresponding to the irrigation modulus and the flow rates corresponding to the drainage modulus is determined as the pipeline design flow rate.

[0018] The solution further includes: In S20, the pipeline network full life cycle resource cost assessment function is obtained by weighted summation of the pipeline network construction infrastructure cost assessment value and the pipeline network operation cost assessment value;

[0019] The cost assessment value of the pipeline network construction infrastructure consists of pipeline material consumption index, water pump configuration equivalent, and irrigation and drainage well pile construction equivalent.

[0020] The estimated cost of pipeline operation consists of the power consumption of water pumps and the cost of maintenance and repair.

[0021] The hydraulic constraints include nodal pressure constraints, pipe flow velocity constraints, and pipe diameter reduction constraints.

[0022] The scheme further includes: In S30, the step of using a non-dominated sorting genetic algorithm to iteratively optimize the pipe parameters included in the initial pipeline network topology includes:

[0023] The number of main pipe sections, the number of branch pipe sections, the pipe diameter of each pipe section, and the water pressure distribution ratio of each intersection node are used as coding variables to generate an initial population composed of multiple decision variables.

[0024] Crossover and mutation operations are performed on each individual in the initial population to generate a progeny population.

[0025] A further step in the solution is to use the infeasibility method to address the hydraulic constraints during the optimization iteration process.

[0026] For a specific individual that does not meet the hydraulic constraints, calculate the total infeasibility value corresponding to the violation of each constraint by the specific individual.

[0027] When the total infeasibility value exceeds the dynamic decrease threshold, the specific individual is removed from the offspring population, and the removed individual is replaced by the individual with the highest feasibility in the current population.

[0028] The solution further includes: In S50, the simulation of the drainage process based on the non-steady flow dynamics model for the drainage mode includes:

[0029] The narrow-slit method is used to integrate and characterize the unpressurized and pressurized flows inside the pipe, and a non-conservative Saint-Venant equation system is established.

[0030] The non-conservative Saint-Venant equations are spatially discretized using spatial staggered cells and scalar dissipation finite volume method to establish a coupled algebraic equation system.

[0031] The coupled algebraic equations are discretized and solved using a fully implicit time scheme, and the simulated water level receding time series data at different time steps are output.

[0032] The solution further includes: In S40, obtaining the actual water level receding time series data of the physical entity pipe network constructed based on the target solution under the flood drainage mode includes:

[0033] The actual liquid level values ​​at multiple discrete time points during the water receding process are continuously collected by the liquid level sensor installed inside the irrigation and drainage well pile.

[0034] The collected actual liquid level values ​​are arranged in chronological order to generate the actual water level receding time series data.

[0035] The solution further includes: In S60, minimizing the error between the actual water level receding time series data and the simulated water level receding time series data is set as the optimization objective, and the actual friction coefficient inside the physical network pipes due to silt deposition is calculated by inversion, including:

[0036] Construct a target evaluation function, which is used to quantify the root mean square error of the actual water level receding time series data and the simulated water level receding time series data at corresponding time points;

[0037] A heuristic search algorithm is used to continuously adjust the preset friction coefficient input into the non-steady flow dynamics model until the root mean square error decreases below the convergence condition.

[0038] The preset friction coefficient corresponding to the convergence condition is locked as the actual friction coefficient.

[0039] The solution further includes: In S70, feeding back the actual friction coefficient to the hydraulic constraints and triggering the non-dominated sorting genetic algorithm to re-expand the optimization iteration to output the corrected pipeline water distribution strategy, including:

[0040] The actual friction coefficient is used to replace the preset friction coefficient originally set in the hydraulic adjustment calculation, and the formula for calculating head loss along the friction is reconstructed.

[0041] The water pressure demand values ​​of each node within the pipeline configuration scheme set are updated using the reconstructed head loss calculation formula.

[0042] The target water pump operating frequency that meets the updated water pressure requirement value is re-selected using the non-dominated sorting genetic algorithm, and the target water pump operating frequency is established as the corrected pipeline water distribution strategy.

[0043] An integrated pipeline irrigation and drainage collaborative optimization model construction system, used as a model construction system for the integrated pipeline irrigation and drainage collaborative optimization model construction method, includes:

[0044] The data acquisition module is used to acquire basic data corresponding to the target area of ​​farmland, and determine the irrigation design flow and drainage design flow based on the basic data;

[0045] The initial architecture establishment module is used to construct the initial pipeline network topology based on the irrigation design flow and the drainage design flow, and to establish the pipeline network full life cycle resource cost assessment function and hydraulic constraints.

[0046] The algorithm iteration module is used to perform optimization iteration on the pipeline parameters included in the initial pipeline topology using a non-dominated sorting genetic algorithm to obtain a set of pipeline configuration schemes;

[0047] The actual data acquisition module is used to output pipeline configuration parameters according to the target scheme in the pipeline configuration scheme set; and to acquire the actual water level receding time series data of the physical entity pipeline network constructed based on the target scheme in the drainage mode. The actual water level receding time series data is the data of farmland water flowing into the physical entity pipeline network through irrigation and drainage well piles.

[0048] The drainage simulation module is used to perform drainage process simulation on the drainage mode based on a non-steady flow dynamics model, and output simulated water level receding time series data.

[0049] The inversion calculation module is used to set minimizing the error between the actual water level receding time series data and the simulated water level receding time series data as the optimization objective, and to invert and calculate the actual friction coefficient inside the pipe of the physical entity pipe network caused by silt deposition.

[0050] The feedback correction module is used to switch between drainage mode and irrigation mode. Before switching to irrigation mode, the actual friction coefficient is fed back into the hydraulic constraints and the non-dominated sorting genetic algorithm is triggered to re-expand the optimization iteration to output the corrected pipeline water distribution strategy, thereby constructing the final optimization model.

[0051] The beneficial effects of this invention are:

[0052] The present invention has achieved the following beneficial effects:

[0053] This invention constructs an integrated underground pipe network topology that incorporates bidirectional scheduling of drainage and irrigation, controlling the capital investment and land occupation scale of infrastructure. Addressing the physical phenomenon of alternating drainage drainage water carrying sediment deposition leading to changes in pipe wall friction under alternating operating conditions, this scheme records actual water level receding time-series data in drainage mode. Combined with simulation results from a hydrodynamic model, the actual friction coefficient affected by sediment is derived with the goal of minimizing error. Before switching to irrigation mode, this scheme feeds back the extracted actual friction coefficient into the hydraulic constraints and triggers a re-optimization iteration, outputting a corrected water distribution strategy. This adaptive adjustment mechanism compensates for the increase in physical resistance caused by sediment adhesion, avoids hydraulic calculation deviations caused by using fixed theoretical parameters, ensures that the pump output head matches the actual resistance, and maintains stable water pressure delivery at the end of the pipe network.

[0054] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0055] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0056] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0057] Figure 1 A flowchart illustrating a method for constructing an integrated pipeline irrigation and drainage collaborative optimization model, as provided in an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram illustrating the interaction between the system hardware execution environment and the field sensor architecture provided in an embodiment of the present invention.

[0059] Figure 3 This is a structural breakdown diagram of the pipeline network lifecycle resource cost assessment function provided in this embodiment of the invention;

[0060] Figure 4 A flowchart illustrating the optimization iteration using a non-dominated sorting genetic algorithm provided in this embodiment of the invention;

[0061] Figure 5 A flowchart illustrating the inversion calculation of the actual friction coefficient and the secondary optimization correction provided in this embodiment of the invention;

[0062] Figure 6 The module composition structure diagram of the integrated pipeline irrigation and drainage collaborative optimization model construction system provided in the embodiments of the present invention is shown. Detailed Implementation

[0063] In traditional farmland irrigation projects, water supply networks and drainage networks are often physically separated. This results in substantial construction costs and a large land area required. Integrating these two networks into a single underground topology is a technological approach that aims to reduce costs and increase efficiency.

[0064] However, in a bidirectional fluid dispatching network that alternates between pressurized water supply and gravity drainage, the physical pipe walls experience sediment buildup during the later stages of drainage operations. This sediment buildup causes dynamic distortion in the pipe wall friction coefficient. If this distortion is not corrected before subsequent water supply operations, it will lead to pump head mismatch and pressure drop at the terminal. Therefore, this application provides a method for constructing an integrated pipeline irrigation and drainage collaborative optimization model, such as... Figure 1 As shown.

[0065] like Figure 2As shown, to execute the aforementioned model construction method, the system pre-configures a corresponding hardware execution environment. This environment includes a central processing unit (CPU), memory units, solid-state storage devices, and an industrial communication interface. The CPU calls the computer instruction sequence within the solid-state storage device and allocates multi-threaded computing space in the memory units to execute the non-dominated sorting genetic algorithm. The industrial communication interface continuously receives low-level physical environment status data transmitted from field sensors in the target farmland area. These field sensors include ultrasonic level transmitters in low-lying areas of the field, diffused silicon pressure sensors at the branch nodes of the main pipeline, and electromagnetic flowmeters at the throats of irrigation and drainage well piles. Through the data link between the hardware and the sensors, the underlying perception foundation of the physical environment is provided for the optimization model.

[0066] A method for constructing an integrated pipeline irrigation and drainage collaborative optimization model, such as Figure 1 As shown, it includes the following steps:

[0067] Step S10: Obtain basic data corresponding to the target farmland area, determine the irrigation design flow rate and drainage design flow rate based on the basic data, and form an irrigation mode and a drainage mode. The basic data includes meteorological and hydrological parameters, soil and crop parameters, and system configuration and energy consumption parameters.

[0068] Building the foundational data infrastructure involves the sensing, cleaning, and transformation of multi-dimensional physical environment parameters. Specifically, the control logic unit establishes a two-way encrypted link with IoT sensor nodes deployed in the target farmland area via an industrial communication interface.

[0069] The aforementioned sensing nodes include: ultrasonic level transmitters installed in low-lying catchment areas of the field; diffused silicon pressure sensors deployed at the branching points of the main pipeline; electromagnetic flowmeters embedded in the irrigation and drainage well channels; and frequency domain reflection moisture detection probes deeply buried in different soil profiles. In the sensing stage: the ultrasonic level transmitter uses a piezoelectric ceramic transducer to transmit and receive high-frequency acoustic pulses, and calculates the reference liquid level elevation by combining round-trip time difference and dynamic temperature compensation. The diffused silicon pressure sensor, based on the semiconductor piezoresistive effect, transmits the hydrostatic pressure of the fluid through a stainless steel corrugated diaphragm and silicone oil to a Wheatstone bridge, outputting a voltage signal proportional to the reference pressure, which is then amplified and converted from analog to digital to generate a standard digital message. The electromagnetic flowmeter, based on Faraday's law of electromagnetic induction, uses an alternating excitation magnetic field to induce an electromotive force by cutting magnetic lines of force with conductive fluid, thereby calculating the instantaneous flow rate. The frequency domain reflection moisture detection probe emits high-frequency electromagnetic waves into the soil medium to measure the dielectric constant, thus retrieving the soil volumetric water content.

[0070] In the data cleaning stage, the raw physical parameters collected by all sensor nodes are continuously transmitted to the cloud via a narrowband IoT communication protocol. To eliminate signal glitches and high-frequency noise generated by the sensors in complex farmland environments, the system invokes a Kalman filter-based data cleaning module in the cloud. This module, during the time-lapse process, alternately executes two steps: state prediction and observation update, ultimately generating smooth physical state time-series data.

[0071] In practical implementation, the system constructs a two-dimensional state vector with the actual liquid level value and its rate of change as variables. For this two-dimensional state vector, the system internally constructs a state transition matrix. Specifically defined as ,in: The actual sampling step size of the system is dynamically substituted; observation matrix Specifically defined as This is used to extract the actual observed liquid level. For the hyperparameter configurations required by the filtering algorithm, such as the state transition matrix, process noise covariance matrix, and measurement noise covariance matrix, the system does not use rigid fixed values, but rather dynamically initializes them based on the physical nominal parameters of the field sensors. Specifically, the time step parameter in the state transition matrix is ​​dynamically substituted according to the actual sampling frequency of the system (usually set to a value of...). The measured noise covariance parameter is directly mapped from the factory-specified variance of the selected ultrasonic level sensor (usually within 1000 ppm). The process noise covariance is set as a small diagonal matrix based on the empirical interference level of external wind and waves on water surface fluctuations. In this embodiment, the specific value selection logic of the small diagonal matrix includes: using the main diagonal elements to represent the inherent physical uncertainty of water level state transitions, and its numerical range is fixed within a range based on the sensor sampling frequency and the empirical wave amplitude. to Between these parameters, all off-diagonal elements within the matrix are set to zero. This calibration method ensures that the filter suppresses high-frequency wave clutter while preserving the true low-frequency trend of water level decline. Through adaptive configuration of the above parameters, the algorithm can adaptively filter abrupt clutter and output high-confidence water level receding time-series reference data.

[0072] To ensure the feasibility of the subsequent output of pipeline topology and pump frequency combination for engineering implementation, the aforementioned basic data was divided into three core categories: meteorological and hydrological parameters, soil and crop parameters, and system configuration and energy consumption parameters, and storage instructions were submitted to the relational database for each category.

[0073] Meteorological and hydrological parameters also include long-sequence rainfall duration and rainfall records. When processing the rainfall records and fitting the Pearson type III frequency curve, a least-squares numerical optimization iterative algorithm is used. Specifically, this includes: extracting statistical characteristic parameters from historical rainfall samples, including the mean and coefficient of variation; using the extracted coefficient of variation as a baseline value, setting a preset step size, and performing increment or decrement operations on the skewness coefficient within each iteration cycle; at each skewness coefficient update node, generating a corresponding candidate theoretical cumulative frequency distribution curve, and calculating the sum of squared deviations between the candidate theoretical cumulative frequency distribution curve and all historical rainfall discrete measured sample points; continuously performing iterative calculations until the sum of squared deviations reaches a preset convergence condition (i.e., drops to a minimum value), locking the skewness coefficient corresponding to the minimum value to complete the curve fitting. Finally, based on the set return period, the corresponding design rainfall amount is extracted from the fitted theoretical cumulative frequency distribution curve as the input boundary condition for the drainage system design.

[0074] For soil and crop parameters, including soil field water holding capacity, soil wilting coefficient, soil dry bulk density, planned wetting layer physical depth, and crop water consumption coefficients for each independent growth stage of the target crop (dynamically evolving throughout its life cycle), the system calculates the maximum effective water storage space of the farmland soil based on these parameters and multiplies the reference crop evapotranspiration by the corresponding growth stage water consumption coefficient to obtain the actual evapotranspiration of the target crop. To ensure that the output results match the absorption needs of plant roots, the system combines the maximum effective water storage space of farmland soil and calculates the net irrigation quota through the water balance equation. ):

[0075] ;

[0076] in, Net irrigation quota (unit: mm); The actual evapotranspiration of the target crop (unit: mm); Effective rainfall (unit: mm); Capillary recharge of deep groundwater (unit: mm); This represents surface runoff loss (unit: mm; the default value for the drip irrigation system in this scheme is 0). The amount of deep seepage loss (unit: mm); This represents the change in soil water storage (unit: mm; positive sign indicates an increase in soil water storage, and negative sign indicates a decrease).

[0077] For the remaining hydrological parameters in the above equations, the system's internal modules are configured with standardized derivation and calculation logic. Specifically, the effective rainfall is calculated using an empirical conversion factor method, that is, by multiplying the actual rainfall measured by the on-site rain gauge by an effective utilization factor (this factor is combined with the historical hydrological characteristics of the target farmland area, and the system's preferred value is between 0.70 and 0.85). The capillary recharge and deep seepage loss of deep groundwater are dynamically calculated by calling the frequency domain reflectance moisture detection probe data deployed on-site. Specifically, this includes: extracting the difference in dielectric constant fed back by the detection probes of two adjacent soil profiles at different depths in real time, and converting the difference into a vertical soil volumetric water content gradient; then, based on the real-time water content gradient and combined with the pre-initialized nominal value of the target soil texture pore distribution, solving for the capillary flux and deep seepage flux at the current moment through a water vertical transport algorithm model. The water vertical transport algorithm model adopts a one-dimensional discrete difference scheme based on Darcy's law, and its exact equation is: ;in, This refers to capillary upward flux or deep seepage flux (positive values ​​for upward migration and negative values ​​for downward migration). The unsaturated permeability coefficient is dynamically updated based on the extracted soil volumetric moisture content. This is the matrix potential gradient derived between adjacent probes based on the water content gradient. This calculation process replaces the empirical constant lookup table step, controlling the system's computing power overhead while meeting engineering accuracy requirements.

[0078] The system configuration and energy consumption parameters include the material assessment base per unit length and the nominal value of the physical roughness of the pipe wall for standardized PVC or high-density polyethylene pipes produced in the market. They also include the procurement cost of various rated power centrifugal pumps or submersible pumps, and the motor power factor and operating efficiency degradation curves. Furthermore, they include the comprehensive cost of civil engineering pouring and shell manufacturing for various types of intelligent irrigation and drainage well piles. Finally, they include the basic energy consumption conversion rate, tiered energy consumption weighting strategy, time-based cost reduction factor, and expected service life of the water conservancy project provided by the local power supply department. These system configuration and energy consumption parameters provide a pricing benchmark and a basis for time-based cost reduction calculations for the subsequent full life-cycle resource cost assessment function of the pipeline network.

[0079] Regarding the determination of irrigation and drainage design flow rates based on fundamental data, the logic for determining the irrigation design flow rate includes: dividing the calculated net irrigation quota by a preset field water resource transport and utilization efficiency coefficient to obtain the gross irrigation quota. The field water resource transport and utilization efficiency coefficient assesses the comprehensive losses during water flow transmission within pipelines, leakage at joints and gaps, and evaporation during spraying. Combining the total control area of ​​the target farmland region, the duration limit of a single irrigation cycle, and the system's average daily effective working hours, a basic irrigation modulus is calculated to meet the needs of the entire region's rotational irrigation grouping. Multiplying the basic irrigation modulus by the total control area of ​​the region yields the overall irrigation design flow rate.

[0080] The logic for establishing the drainage design flow includes: First, after rainfall falls into farmland and is reduced by canopy interception, depression filling, and infiltration losses, it converges into surface runoff. The system uses an empirical model of runoff coefficient to convert the derived design rainfall into the surface design runoff depth. ):

[0081] ;

[0082] in, Design runoff depth for the surface (unit: mm); Design rainfall amount (unit: mm); The comprehensive runoff coefficient (dimensionless, with a built-in correspondence table in the system: 0.45 for clay soil and bare land during the sowing season; 0.25 for loam soil and crop vegetation coverage greater than 80%).

[0083] Subsequently, the drainage modulus was derived by combining the maximum flood tolerance duration that a specific crop within the target farmland area could withstand:

[0084] ;

[0085] in, Drainage modulus (unit: ); Design runoff depth for the surface (unit: mm); To ensure that the dimensions of subsequent fluid dynamics models are consistent, a conversion constant (dimensionless) is used. The maximum flood tolerance duration parameter for a specific crop (unit: h, such as a fixed value of 24h for dryland soybeans and 72h for rice).

[0086] Finally, by multiplying the drainage modulus by the effective catchment area of ​​the entire region, the overall drainage design flow can be derived.

[0087] To prevent physical water conveyance bottlenecks in the pipeline network under complex conditions of alternating drought and flood, the system compares the overall irrigation design flow rate and the overall drainage design flow rate, locking the higher of the two values ​​as the baseline design flow rate for the pipeline network. This baseline value is used to limit the lower limit of the flow capacity of all subsequent main pipelines. An extreme value envelope design rule is adopted to ensure that the underground pipeline network has a flow cross-section that meets physical requirements, preventing fluid congestion and pressure buildup under both drought and flood conditions.

[0088] Step S20: Construct an initial pipeline network topology based on the irrigation design flow rate and the drainage design flow rate, and establish the pipeline network life-cycle resource cost assessment function and hydraulic constraints:

[0089] Understandably, when constructing the initial pipeline topology, the system uses a digital elevation model of the target farmland area, with water intake points or drainage catchment areas as the root nodes, and employs a shortest path algorithm based on spatial geometric constraints to generate spatial wiring connections to the coordinates of each field irrigation and drainage well. To balance the distribution of water pressure at the ends, redundant connections are added between the nodes of the main pipeline, resulting in a tree-like hybrid topology that includes local closed loops.

[0090] At the underlying implementation level, the shortest path algorithm based on spatial geometric constraints specifically calls Kruskal's minimum spanning tree algorithm. To construct a local closed loop network that prevents flow interruption while minimizing the total pipeline length, the system's embedded exact redundancy rule determination mechanism works as follows: During the execution of the Kruskal algorithm, it records all non-tree edges (i.e., candidate edge sets) that are discarded by the algorithm because they would form a closed loop; then, it calls a pseudo-random number generator, and based on a set topological redundancy constant (e.g., set to 10%), randomly selects 10% of the connecting edges from the discarded non-tree edge set and adds them to the final directed graph edge set, thereby establishing the hybrid topology.

[0091] Finally, all generated spatial connections are digitized into a directed graph edge set data structure, which specifically includes attribute fields such as start point number, end point number, pipe segment length, and pipeline spatial elevation.

[0092] When establishing the resource cost assessment function for the entire life cycle of the pipeline network, such as Figure 3 As shown, the system breaks down the total cost into the assessed value of the pipeline network construction infrastructure cost and the assessed value of the pipeline network operation cost, and then performs a weighted summation calculation on the two costs. The assessed value of the pipeline network construction infrastructure cost consists of three parts: pipeline material consumption indicators, pump configuration equivalent, and irrigation and drainage well pile construction equivalent.

[0093] Specifically, the calculation logic for each evaluation indicator and its equivalent is as follows:

[0094] Firstly, the pipeline material consumption index is reflected in the pipeline material cost. This cost is calculated by multiplying the length of all discrete pipe segments in the initial pipeline network topology by the sum of the material evaluation base per unit length currently allocated to the corresponding pipe segment by the algorithm.

[0095] Secondly, the pump configuration equivalent is reflected in the pump cost. This cost is obtained by traversing the pump performance database and matching the pump models and average purchase prices corresponding to the total head and total flow required by the current topology node.

[0096] Third, the cost of irrigation and drainage well piles is calculated by multiplying the total number of outlet nodes set in the topology network by the standard cost of a single set of well piles.

[0097] It is important to note that the aforementioned pipeline material costs include not only the material cost per unit length of the pipe itself, but also the equivalent of earthwork excavation costs, which are positively correlated with pipe diameter. When the algorithm attempts to increase the diameter of a specific pipe section during the optimization process, the unit price of the pipe increases accordingly. Simultaneously, the bottom width and depth of the trench excavation section are forced to expand accordingly, further increasing the mechanical workload for earthwork excavation, subgrade laying, pipe wall backfilling and compaction, and excess soil removal. Finally, the control logic unit integrates all the aforementioned derivative expenses into the pipeline material costs using a pre-input construction quota matrix.

[0098] The estimated cost of pipeline operation consists of pump power consumption and maintenance depreciation. Pump power consumption costs are calculated based on the pump shaft power, and the derivation logic includes: first, using the Darcy-Weisbach formula to calculate the head loss along each pipe section (…). ):

[0099] ;

[0100] in, Head loss along the friction (unit: m); The friction constant coefficient (dimensionless, its value is obtained from a built-in mapping table, which is pre-calculated and stored in the system memory based on the Colebrook-White formula. For example, for commonly used high-density polyethylene (HDPE) or polyvinyl chloride (PVC) pipes, the system directly retrieves the nominal value of its friction constant coefficient under absolutely smooth conditions, which is a constant between 0.009 and 0.011). The length of the physical pipe section (unit: m); Inner diameter (unit: m); Flow velocity (unit: m / s); Let gravitational acceleration be a scalar (take...) ).

[0101] By accumulating the resistance losses segment by segment and superimposing the terrain elevation difference and the minimum working head at the outlet, the actual total operating head of the water pump is obtained, and then the power of the pump shaft is further calculated:

[0102] ;

[0103] in, The power of the pump shaft (unit: kW). The density constant of the fluid medium (e.g., configured as) ); To calculate the total flow (unit: ); The transient operating efficiency of the water pump is dimensionless. The actual total operating head of the water pump (in meters); the 1000 in the denominator is a rigid physical adjustment dimension factor for converting watts to kilowatts; It is a scalar of gravitational acceleration.

[0104] The system calculates the average annual power consumption based on the pump's shaft power and average annual cumulative operating time, and then converts it into the present value of the total power cost over the operating period by combining this with the local basic energy consumption discount rate. This conversion process uses the present value annuity factor (PFA) based on the discount rate and the project's expected service life. )formula:

[0105] ;

[0106] in, The present value factor for annuities (dimensionless); This is the discount factor for the time decay of funds (dimensionless, with system values ​​ranging from 4.0% to 5.0%). The expected service life of a water conservancy project (unit: years, for example, set to 20 years).

[0107] In addition, maintenance and repair costs are extracted at a fixed percentage based on a specific constant of pipeline operating costs.

[0108] The specific constant used to extract maintenance and repair costs is quantified and set in the system's underlying logic as, for example, the total cost of pipeline operation. .

[0109] The infrastructure cost assessment value of the pipeline construction and the pipeline operation cost assessment value are added together to form the fitness function objective value required for the optimization solution stage.

[0110] The system establishes several hydraulic constraints, which are quantified into the following three sets of logical inequalities within the algorithm module:

[0111] (1) Node pressure constraints: The actual usable pressure head at any end node must meet the minimum rated outflow head of the micro-sprinkler, and the internal hydrostatic pressure of the pipe section, after being superimposed with the water hammer warning buffer pressure, should not exceed the nominal pressure bearing physical limit of the pipe material.

[0112] ;

[0113] in, For the network topology The actual available pressure head of each end node (unit: m); The physical number (dimensionless) of the end node in the network topology. The minimum rated outflow head of the micro-sprinkler (unit: m, fixed value is 10.0 m). This represents the nominal physical pressure limit of the pipe (unit: m, for example, a value of 102.0 m). The buffer pressure for water hammer early warning (unit: m, for example, a value of 15.0 m).

[0114] (2) Pipeline flow velocity constraints: The actual average operating velocity of water flow in any pipe section must be between the minimum non-silting velocity and the maximum non-scouring velocity to prevent sediment deposition or excessive friction on the pipe wall.

[0115] ;

[0116] in, For the initial pipeline topology, the first The actual cross-sectional average flow velocity within each pipe section (unit: m / s). This is an independent number (dimensionless) for each pipe segment in the initial pipeline network topology. Minimum non-silting flow velocity (unit: m / s, range: 0.6~0.8m / s); Maximum non-impact velocity (unit: m / s, range 2.5~3.0m / s).

[0117] (3) Pipe diameter decreasing constraint: In network topology traversal, the inner diameter of the pipe segment of the upstream parent node with direct physical connection must be greater than or equal to the inner diameter of the pipe segment of the downstream child node to avoid reverse diameter expansion configuration in the solution space:

[0118] ;

[0119] in, The inner diameter of the pipe segment of the upstream parent node with a direct physical connection (unit: m); The inner diameter of the downstream sub-node pipe section (unit: m).

[0120] If any individual in the optimization population exceeds any of the closed intervals of the above inequalities, the system immediately marks it as not meeting the conditions and performs penalty filtering.

[0121] Step S30: Use a non-dominated sorting genetic algorithm to iteratively optimize the pipe parameters included in the initial pipeline network topology to obtain a set of pipeline network configuration schemes.

[0122] like Figure 4 As shown, the specific optimization iteration process of the non-dominated sorting genetic algorithm includes:

[0123] The system uses the number of main pipe sections, the number of branch pipe sections, the pipe diameter specifications of each section, and the water pressure distribution ratio at each intersection node as encoding variables to generate an initial population composed of multiple decision variables. Considering that the inner diameter of commercial pipe materials exhibits a discrete, stepped distribution, the system employs an integer sequence encoding mechanism at the bottom layer: each gene position on the chromosome corresponds to a specific pipe section, and its internal integer value is mapped to the pipe diameter index in the commercial pipe diameter model database. Subsequently, the system uses a pseudo-random number generator to generate an initial chromosome sequence matrix of a preset size in batches, within the domain of the pipeline network's full life-cycle resource cost assessment function, thus completing the establishment of the initial population.

[0124] When generating the offspring population, the system performs crossover and mutation operations on each individual in the initial population.

[0125] In the crossover operation, the system employs a multi-point uniform fragment crossover strategy and a simulated binary crossover algorithm: two parent chromosomes are randomly selected, and according to a preset crossover probability, the pipe diameter index value sequences are exchanged at multiple random sites along the gene chain to recombine the local hydraulic characteristics retained by the parents in the offspring. For the selected parent chromosome pair, the system sets a distribution index to control the spatial span of the offspring genes deviating from the parent genes, and balances the computational power allocation between local development and global exploration by dynamically adjusting the distribution index.

[0126] In the mutation operation phase, the system employs a dynamic perturbation strategy and a multinomial mutation algorithm: based on small mutation probabilities, a portion of gene loci on a specific chromosome are randomly selected, a mutation distribution index is set, and gene mutations are triggered according to a specific probability. The mutation operator injects multinomial perturbations into the gene chain, forcing the value representing the pipe diameter index to randomly switch between multiple adjacent standard specification levels. The underlying function of this mutation operation is to effectively maintain the diversity of the genetic material matrix within the population. The multi-point crossover and multinomial mutation operators used in this embodiment can both adopt the basic operational logic of conventional non-dominated sorting genetic algorithms (such as NSGA-II) in this field. To achieve customized optimization of physical pipe networks, this solution is characterized by strict constraints on the physical boundaries of decision variables: all floating-point values ​​of pipe diameter indices generated by the mutation operation must be forcibly mapped to the system's preset library of commercially available standard pipe inner diameter specifications (such as the actual inner diameters corresponding to nominal outer diameters of DN50, DN63, DN75, DN90, etc.) through the nearest matching rule. This effectively eliminates non-standard pipe diameters that may be generated by the algorithm and cannot be manufactured in engineering, ensuring the engineering feasibility of the solution set.

[0127] For the offspring population generated by crossover and mutation operations, the evolutionary algorithm introduces a fast non-dominated sorting mechanism and a crowding calculation rule for selection. When evaluating the performance of each chromosome individual, the system calculates the full life cycle cost target and the node pressure deviation target. The node pressure deviation target measures the pressure uniformity at each irrigation outlet in the pipeline network to ensure consistent water supply to field crops. The specific extraction and calculation logic is as follows: after each hydraulic adjustment calculation, the actual measured pressure head of all end nodes (i.e., outlets) in the current pipeline topology is extracted, and the root mean square error (RMSE) of these discrete head values ​​relative to the average head of the entire region is calculated. In the evaluation dimension of non-dominated sorting, the system uses the minimization of this MMS error as the second optimization objective; the smaller the MMS error, the more stable the water pressure distribution of the corresponding individual. The system projects all individuals in the population into a multi-objective function space for comparison: if individual A is not inferior to individual B in all dimensions of objective function scores, and is at least strictly superior to individual B in independent objective scores, then individual A is logically determined to dominate individual B. Based on the dominance association mapping formed by matrix traversal comparison, the system divides the entire population into different non-dominated hierarchical echelons. Individuals in the primary non-dominated hierarchical echelon represent the frontier of the Pareto optimal solution set in the current population that is not suppressed by any remaining individuals.

[0128] To prevent the algorithm's convergence trajectory from concentrating on specific spatial coordinates, the system calculates the crowding distance for all individuals within the same non-dominated layer. Since there is a significant numerical difference between the total lifecycle cost and node pressure deviation, to prevent large-scale objectives from masking smaller-scale objectives and causing evaluation distortion, the system introduces a range normalization formula to scalarly sum the differences in the objective functions of adjacent individuals. Its underlying algebraic calculation formula is as follows:

[0129] ;

[0130] For individuals Crowding distance (dimensionless); The total number of target dimensions (dimensionless, equal to 2 here); The dimension index of the target dimension being traversed (dimensionless); and They were respectively in the second After sorting by fitness in ascending order for each objective, the individuals immediately adjacent to the currently evaluated individuals are listed next to each objective. The function values ​​(dimensionless) of the two individuals before and after; and The current group is in the th The maximum and minimum extrema on the target dimension (dimensionless). This is the dimensionless numerical index number of the individual sorted by fitness within the current non-dominated hierarchy.

[0131] When selecting individuals to be retained for the next mating pool iteration, the system follows a hierarchical priority logic: individuals at the top of the non-dominated level are given priority for advancement; if they are in the same non-dominated level, individuals with a larger crowding distance are given priority for advancement. This mechanism ensures that the evolved set of pipeline configuration schemes maintains extensibility in the target space distribution.

[0132] In the multi-dimensional optimization and iterative deduction process, the infeasibility method of the solution is used to handle hydraulic constraints. For specific out-of-bounds individuals that do not meet the hydraulic constraints, the system calculates the total infeasibility value corresponding to the violation of each constraint for that specific individual. The system establishes standardized deviation scalar functions for node pressure constraints, pipeline velocity constraints, and pipe diameter reduction constraints. If the actual measured pressure at the end node is lower than the minimum rated head, the system extracts the deviation scalar between the two and exponentially amplifies the deviation scalar as a single infeasibility penalty score; if a specific pipe segment violates the physical principle of downstream diameter reduction, a constant-level infeasibility penalty score is assigned. The specific quantification mechanism for the individual infeasibility penalty score is as follows: For node pressure constraints and pipeline flow velocity constraints, the difference scalar between the actual measured value and the boundary threshold (e.g., the difference between the actual head and the minimum rated head) is extracted and multiplied by a preset dimensionless amplification weighting constant (e.g., a value of 10.0~50.0) as the penalty score for that item; for pipe diameter reduction constraints that violate the physical bottom line (i.e., reverse diameter expansion), a veto-type maximum constant penalty value (e.g., a value of 10000) is directly assigned. The total infeasibility value can be obtained by linearly adding the above penalty scores. The system accumulates and integrates all individual infeasibility penalty scores triggered by a specific individual under all physical constraint clauses, and outputs the total infeasibility value. As the algorithm iterates through generations, the system pre-sets a dynamic decline threshold decay curve. In the early evolution stage, the dynamic decline threshold setting is broad. When the evolution algorithm enters the later convergence stage, the system lowers the dynamic decline threshold parameter according to an exponential decay law.

[0133] In the hyperparameter configuration of the underlying non-dominated sorting genetic algorithm, the system utilizes simulated binary crossover and polynomial mutation operators to maintain population diversity. To achieve a balance between the algorithm's global exploration capability and local convergence speed, the system configures a set of widely validated hyperparameter ranges in engineering: the fragment crossover probability is typically set at... Between (preferably 0.85 in this embodiment); the crossover index parameter and the variation distribution index, which control the deviation span of offspring genes, are usually set between (in this embodiment, 0.85 is preferred); Within the range; the probability of minor mutations is conventionally set as the reciprocal of the total length of the chromosome coding integer sequence.

[0134] Furthermore, the system lowers the dynamic descent threshold parameter according to an exponential decay law, with the dynamic descent penalty tolerance threshold decreasing exponentially from an initial broad threshold constant with the number of iterations. Regarding the termination condition for optimization, the system does not employ a single, absolute, constant number of iterations. The maximum total number of optimization iterations is set to a flexible threshold range (e.g., ...). During iterations, the system continuously monitors the convergence state of the Pareto front. If the optimal solution combination for several consecutive iterations no longer shows a significant shift, convergence termination is triggered prematurely. Regarding the decay control constant parameter that controls the kurtosis of the threshold decay curve, the system adaptively adjusts the decay rate based on the degree of violation in the early stages of iteration, typically limiting it to a certain value. Within the smooth interval. The specific iterative algebraic equation for lowering the dynamic descent threshold parameter according to the exponential decay law is: ;in, For the first Dynamic descent threshold during the second optimization iteration; This is the initial broad threshold constant set; The attenuation control constant parameter; This represents the current iteration step number; This represents the total number of iterations for maximizing the optimization.

[0135] During a single iteration of the evaluation process, when the calculated total infeasibility value exceeds the current dynamic reduction threshold, the system triggers a culling operation. This removes specific out-of-bounds individuals from the offspring population sequence and uses the retained individual with high feasibility and the highest fitness score for the entire lifecycle cost objective to replicate and replace the data space occupied by the culled individual. Through this dynamic infeasibility filtering mechanism, the algorithm engine can identify a set of pipeline configuration schemes that satisfy the underlying fluid dynamics equations and reduce the resource cost evaluation function throughout the pipeline's lifecycle.

[0136] After acquiring the set of pipeline configuration schemes (i.e., the Pareto optimal solution set), the system outputs the target scheme using a conditional filtering mechanism based on preset boundary conditions. Specifically, the system receives an externally input maximum tolerable threshold for the total lifecycle cost of pipeline construction (i.e., the upper limit of the project construction budget). Then, the system traverses the set of pipeline configuration schemes, automatically eliminating all candidate schemes whose total lifecycle cost assessment value exceeds the maximum tolerable threshold, forming a candidate subset. Finally, the system directly extracts the individual with the smallest node pressure deviation value (i.e., the most stable end-point water pressure distribution) from the candidate subset, locks it, and outputs it as the target scheme. If all individuals within the set of pipeline configuration schemes exceed the maximum tolerable threshold, the system triggers a backup rule, directly extracting the individual with the lowest total lifecycle cost assessment value as the target scheme.

[0137] Step S40: Output the pipeline configuration parameters according to the target scheme in the pipeline configuration scheme set, and obtain the actual water level receding time series data of the physical entity pipeline network constructed based on the target scheme in the drainage mode. The actual water level receding time series data is the data of farmland water flowing into the physical entity pipeline network through irrigation and drainage well piles, generated by farmland water flowing into the physical entity pipeline network through irrigation and drainage well piles.

[0138] After acquiring the pipeline network configuration scheme set, when the system is in drainage mode, surface runoff carries sediment particles into underground pipe sections. This sediment settles due to gravity in sections where the flow velocity slows, causing variations in the pipe wall roughness. To address this physical phenomenon, the system needs to acquire the actual water level receding time series data resulting from farmland water flowing into the corresponding pipes within the pipeline network configuration scheme set via irrigation and drainage well piles.

[0139] In the actual data recording phase, the liquid level sensor deployed inside the irrigation and drainage well piles employs high-frequency ultrasonic echo time difference ranging technology. In the fluid medium environment, the liquid level sensor outputs an electronic signal representing the current reference liquid level elevation at a set sampling time step. Subsequently, the system continuously captures multiple actual liquid level values ​​throughout the entire drainage and precipitation cycle, and structurally binds them together with a hardware clock reference timestamp. Based on the evolution sequence of the reference timestamp, the system stitches together all discrete data nodes, ultimately generating a time-series data sequence of actual water level receding that characterizes the dynamic trajectory of the drainage water level decline.

[0140] It should be noted that due to the increased resistance caused by sediment deposition, the actual flow capacity of accumulated water discharged through the pipe network is lower than that under clean pipe network conditions. This directly results in the slope of the actual water level receding time series data curve exhibiting a physical tailing characteristic.

[0141] Step S50: Perform a drainage process simulation on the drainage mode based on a non-steady flow dynamics model, and output simulated water level receding time series data:

[0142] It is understandable that the actual drainage process is accompanied by complex fluid phase transitions: in the initial stage of drainage, the peak flow inflow stage, the internal space of the pipe network is filled with fluid, presenting a full-pipe pressurized flow phase with pressure on all sides; while in the middle and later stages of drainage, the flow decay stage, the water level in the pipe drops and moves away from the top boundary of the pipe, and it transforms into an unpressurized open flow phase driven by gravity on the free surface.

[0143] To mathematically unify the handling of these two phases, the system introduces the virtual topology assumption of the narrow slit method. Specifically, in the spatial geometric modeling stage for a circular closed cross-section, the system extends a virtual narrow slit vertically upward at the top vertex of the pipe cross-section. When the physical water body submerges the top wall of the pipe and enters a pressurized full-flow state, the mathematical model assumes that the excess pressurized water body overflows upward into the virtual narrow slit.

[0144] Because the cross-sectional width of the virtual slit is set to an extremely small parameter, even a tiny increase in water volume will cause the water surface elevation inside the slit to rise vertically upwards. At this point, the virtual free water surface elevation inside the slit can effectively replace the scalar pressure head of the piezometer under pressurized full-flow conditions. Through this ingenious geometric domain transformation of the slit method, the pressure conduction of pressurized full-flow and the gravity conduction of unpressurized open flow are seamlessly integrated. The Saint-Venant equations and the scalar dissipation finite volume method used in this application to solve the unsteady flow hydrodynamic model are both based on conventional mathematical models in the field of fluid mechanics. To adapt to the bidirectional flow alternation scenario of the integrated pipe network in this scheme, the cross-sectional width of the virtual slit in the virtual topology assumption of the slit method is... The setting follows the equivalent water hammer wave velocity formula:

[0145] ;

[0146] in, The width of the virtual narrow slit (unit: m); Acceleration due to gravity (unit: ); The full cross-sectional area of ​​the pipe (unit: ); The water hammer wave velocity constant (unit: m / s, determined by the elastic modulus of the selected solid pipe material; for high-density polyethylene or polyvinyl chloride water supply pipes, it typically ranges from 200 m / s to 400 m / s). Using this specific width as the boundary condition ensures that the equations maintain the nonsingularity of the iterative matrix and numerical stability when encountering abrupt phase changes.

[0147] The continuous partial differential equations constructed based on the principle of mass conservation of fluid infinitesimal elements, and the momentum conservation partial differential equations constructed based on the momentum theorem, operate within the continuous mathematical framework. The precise mathematical sub-unit composition and internal data flow of the non-conservational Saint-Venant partial differential equation system include:

[0148] The continuous partial differential equation characterizing mass conservation data:

[0149] ;

[0150] in, Effective cross-sectional area of ​​water flow (unit: ); For time step variable (unit: s); Volumetric flow rate (unit: ); The spatial distance variable (unit: m) along the direction of water flow in the pipeline. Lateral inflow (unit: ).

[0151] The partial differential equation for momentum conservation that characterizes the transformation of momentum conservation is:

[0152] ;

[0153] in, Volumetric flow rate (unit: ); For time step variable (unit: s); The spatial distance variable (unit: m) along the direction of water flow in the pipeline. This is a momentum correction constant (dimensionless, for example, a value of 1.02 for the system). Effective cross-sectional area of ​​water flow (unit: ); Acceleration due to gravity (unit: ); This represents the absolute elevation of the water surface (unit: m). This is the frictional dissipation term (dimensionless).

[0154] The algebraic mapping formula for calculating the frictional dissipation term (i.e., calling Manning's formula):

[0155] ;

[0156] in, This is the frictional dissipation term (dimensionless); The physical pipe wall roughness (dimensionless); Volumetric flow rate (unit: ); Effective cross-sectional area of ​​water flow (unit: ); The hydraulic radius is in meters.

[0157] In the computational evolution, the computational accelerator module first uses Manning's formula to calculate the current frictional dissipation. Then, the scalar is substituted into the last term of the momentum conservation equation to achieve a coupled closed loop.

[0158] To transform the continuously distributed, non-conservative Saint-Venant equations into discrete algebraic data that can be processed by the central processing unit, the computational logic module uses spatial staggered cells and the scalar dissipative finite volume method to perform discrete operations on the spatial dimensions.

[0159] In terms of mesh generation and layout, the system adopts a spatial staggered cell strategy. Scalar variables of the fluid (water level and pipe cross-sectional area) are defined at the center points of discrete mesh cells; vector variables (flow velocity and cross-sectional volumetric flow rate) are defined at the interfaces between adjacent mesh cells. Through this staggered arrangement rule, the system can accurately calculate the pressure gradient force driving the fluid forward based on the difference between adjacent scalar nodes.

[0160] For handling convection flux, the system employs the scalar dissipative finite volume method for interface flux reconstruction. Given that the scalar dissipative finite volume method can determine the physical direction of fluid information propagation based on eigenvalue signs, the algorithm uses an upwind difference scheme and injects numerical dissipative terms when calculating flux data across interfaces to effectively suppress hydraulic jumps caused by abrupt changes in confluence. This rigorous spatial discretization process ensures the strict conservation of fluid dynamic mass and momentum during transfer between adjacent grids, successfully mapping a continuous system of partial differential equations into a discrete coupled algebraic equation system.

[0161] In the time-division step solution, the microprocessor employs a fully implicit time-difference scheme to stably solve the discretized coupled algebraic equations. The fully implicit scheme allows the system to span larger time steps while maintaining numerical stability. Within each time step, the microprocessor constructs a matrix of nonlinear equations containing unknowns about the full-grid water level and flow distribution, and uses a nonlinear iterative algorithm (such as the Newton-Raphson iteration method combined with a conjugate gradient solver) for successive approximation calculations.

[0162] To determine whether the numerical inversion within a single time step has converged, the system extracts the L2 norm of the incremental residuals of the grid unknowns. When the residual norm drops to a preset machine accuracy tolerance range (usually set to...), the convergence is determined. When the value reaches a certain level (e.g., a specific time step), the system is considered to have converged successfully. Simultaneously, to prevent infinite loops caused by numerical divergence, the system has a preset... The maximum number of iterations per step is capped. If convergence is not achieved after reaching this cap, the system will trigger an anti-loop interruption mechanism, automatically halving the current time step and performing a rollback and recalculation.

[0163] As the time step progresses, the system extracts and records transient water level elevation data at each irrigation and drainage well pile node location. The discrete water level elevations output from all time steps are concatenated and integrated to generate simulated water level receding time-series data describing farmland drainage efficiency.

[0164] Step S60: Minimize the error between the actual water level receding time series data and the simulated water level receding time series data as the optimization objective, and calculate the actual friction coefficient inside the physical pipeline network caused by silt deposition:

[0165] A target evaluation function is constructed, and the approximation between the simulated liquid level and the actual liquid level is quantified by calculating the root mean square error (RMSE). The specific calculation logic model is as follows:

[0166] ;

[0167] in, Root mean square error (unit: m); The total number of discrete sampling time points in the flood drainage time series data (dimensionless). The traversal index variable (dimensionless) is used for discrete sampling time points. For the first Each base timestamp corresponds to the extracted actual liquid level elevation (unit: m). For the simulation engine in the first Simulated liquid level elevation (unit: m) output at the same time step.

[0168] Simultaneously, the root mean square error decreasing to below the convergence condition has a quantified logical termination condition, including: the backend of the inversion calculation module establishing a depth of... The data sliding monitoring window of the generation, when it is continuously Within each optimization iteration, the amplitude variation of the globally optimal RMSE value (the difference between the highest and lowest fluctuation extremes) is continuously less than the lower bound constant. If the absolute value of the current RMSE calculated in a single measurement drops below the preset lower limit of 0.01m, any one of these conditions is met and the convergence condition is immediately determined, thus interrupting the subsequent exploration loop.

[0169] like Figure 5 As shown, the system employs a heuristic search algorithm to adjust the preset friction coefficient input into the unsteady flow dynamics model. In this embodiment, the heuristic search algorithm preferably adopts the particle swarm optimization algorithm. The system first initializes multiple search particles (each particle represents multiple sets of assumed values ​​for pipe segment friction coefficients), substitutes them into the unsteady flow dynamics model to perform flood drainage simulation, and calculates the root mean square error (RMSE) corresponding to the generated simulated water level receding time series data.

[0170] Based on the error feedback gradient, the search particle, guided by both its individual optimal value and the global optimal value of the population, continuously moves towards the parameter space with lower error within the inversion space. It updates its own velocity vector ( ) and position coordinates ( The mathematical iteration step size formula is:

[0171] Velocity vector update:

[0172] ;

[0173] in, For specific particles In the The velocity vector (dimensionless) during the next inversion iteration; For specific particles In the The velocity vector at the next iteration (dimensionless); This is the inertia weighting coefficient (dimensionless, typically decreasing smoothly from 0.8 to 0.9 initially to 0.4 to 0.5 later). The individual cognitive learning factor (dimensionless, ranging from 1.5 to 2.5, with an optimal configuration of around 2.0); and It is an independent floating-point random number (dimensionless) in the interval [0,1]. The optimal coordinates (dimensionless) of the individual in the particle's cognitive memory that minimizes RMSE. For specific particles In the Position coordinates at the next iteration (dimensionless); The group social learning factor (dimensionless, ranging from 1.5 to 2.5, with an optimal configuration of around 2.0). The globally optimal coordinates (dimensionless) locked for the collaborative memory of the entire group. Numerical index number (dimensionless) for a specific search particle within the multidimensional search space. This represents the current inversion iteration number (dimensionless).

[0174] Location coordinates updated:

[0175] ;

[0176] in, For specific particles In the The position coordinates (dimensionless) at the next iteration.

[0177] Regarding the parameter configuration of the optimization algorithm, the system sets the various learning-driving parameters of the particle swarm optimization algorithm within a reasonable heuristic range. Among these, the inertia weight coefficient... A dynamic decay strategy is typically employed, usually starting from the initial... Smoothly decreases to later stages To balance the overall expansion in the early stages with the detailed local search in the later stages; individual cognitive learning factors Social learning factors of groups Usually in The values ​​are set between 2 and 0 (in this embodiment, both are preferably configured to be around 2.0) to ensure that the particles maintain a balanced tension between individual experience and group cooperation.

[0178] Furthermore, to prevent the search particles from falling into a failure range that is not permissible under fluid dynamics principles during numerical iteration, the system employs a boundary rebound protection mechanism at the multidimensional spatial boundary. This mechanism limits the optimization closed range of the actual friction coefficient based on the characteristics of the engineering pipe material: the lower limit of the physical constant threshold is set to the roughness reference value of extremely clean, newly manufactured pipe material (usually set to...). The upper limit of the physical constant threshold is set as the empirical limit reference value for a surge in friction caused by severe backflow and siltation (usually set to...). ).

[0179] After the aforementioned forward hydrodynamic simulation and reverse parameter adjustment cycle, the root mean square error shows a decreasing trend. When it falls below the preset convergence condition, the system immediately stops exploring and locks the corresponding set of preset friction coefficients as the final inverted actual friction coefficients.

[0180] The boundary bounce mechanism includes a logical judgment step and a damping penalty rule. After coordinate displacement, if the fractal coordinates of the search particle exceed the set upper and lower thresholds, the system forcibly truncates the out-of-bounds distance and replaces its position with boundary coordinates. Simultaneously, the flight velocity in that dimension is extracted to perform boundary collision velocity decay calculation, which involves multiplying its out-of-bounds velocity component by an amplitude damping reduction coefficient (typically set to a certain value). The inversion is performed by reducing the excess velocity components of out-of-bounds particles, thus strictly constraining the inversion calculation within a solution space that conforms to the pipe material and the physical properties of the sediment, ensuring that the results converge to a truth point that is of guiding significance.

[0181] Step S70: Switch between drainage mode and irrigation mode. Before switching to irrigation mode, feed back the actual friction coefficient into the hydraulic constraints and trigger the non-dominated sorting genetic algorithm to re-expand the optimization iteration to output the corrected pipeline water distribution strategy, thereby constructing an irrigation and drainage coordinated optimization model.

[0182] Before feeding back the actual friction coefficient obtained from the inversion to the hydraulic constraints, a pre-set trans-state hydraulic parameter mapping model is invoked for parameter transformation. The parameter transformation process includes: receiving the actual pipe wall roughness parameters obtained from the drainage condition; using the pipe inner diameter specifications from the pipe network configuration scheme set and the set pipe network flow velocity range as input, and using the equivalent roughness transformation model, mapping the actual pipe wall roughness parameters under unpressurized open flow conditions to the constant coefficient of friction loss required by the Darcy-Weisbach calculation framework. The equivalent roughness transformation model is constructed based on the principle of shear force conservation in fluid mechanics, and its specific algebraic transformation formula for mapping the constant coefficient of friction loss is as follows: ;in, This is the constant coefficient of Darcy-Weisbach friction loss in the mapped output; It is a scalar of gravitational acceleration; These are the actual pipe wall roughness parameters derived under drainage conditions; This is the hydraulic radius of the corresponding pipe section under full-pipe irrigation pressurization conditions (i.e., one-quarter of the pipe section's inner diameter). Subsequently, the original preset value is replaced by the constant coefficient of friction loss output by the conversion, the formula for calculating friction head loss is reconstructed, and the water pressure demand values ​​of each node in the pipeline network are updated accordingly to compensate for the friction increase under alternating operating conditions.

[0183] When triggering the non-dominated sorting genetic algorithm for secondary optimization, given that the underground pipe network has been completed and solidified, the system implements a variable locking operation: freezing the decision variables controlling pipe diameter scaling and strictly limiting their crossover and mutation, thereby focusing the optimization computing power on resampling and iteration within the Hertz range of the water pump's variable frequency speed regulation. The Hertz range for resampling here is essentially a direct mapping of the water pressure allocation ratio variables of each crossover node in the initial population encoding to the physical equipment. The system linearly maps the floating-point decision variables originally used for water pressure allocation to the legal operating frequency band supported by the on-site variable frequency water pump (such as the 30Hz~50Hz closed range), and adjusts the water pump motor speed by changing the Hertz encoding value, accurately compensating for the increased actual resistance of the pipe network due to silt deposition.

[0184] Subsequently, the algorithm re-executes the screening process based on the hydraulic operating boundary reflecting actual resistance. Under the premise of meeting the pipeline pressure bearing limit, it outputs the optimal pump operating frequency that can overcome the additional resistance of silt and ensure that the terminal irrigation and drainage well piles reach the rated flow. This target frequency is established as the modified pipeline water distribution strategy to guide the variable frequency pump to smoothly deliver pressurized water to the field terminal, thus completing the final construction of the adaptive integrated collaborative optimization model.

[0185] The embodiment provides an integrated pipeline irrigation and drainage collaborative optimization model construction system, which is a model construction system used in the integrated pipeline irrigation and drainage collaborative optimization model construction method, such as... Figure 6 As shown, the system includes a data acquisition module 100, an initial architecture establishment module 200, an algorithm iteration module 300, an actual data acquisition module 400, a flood drainage simulation module 500, an inversion calculation module 600, and a feedback correction module 700.

[0186] The data acquisition module 100 is used to acquire basic data corresponding to the target farmland area, determine the irrigation design flow and drainage design flow based on the basic data, and form an irrigation mode and a drainage mode. The basic data includes meteorological and hydrological parameters, soil and crop parameters, as well as system configuration and energy consumption parameters.

[0187] The initial architecture establishment module 200 is used to construct the initial pipeline network topology based on the irrigation design flow and the drainage design flow, and to establish the pipeline network full life cycle resource cost assessment function and hydraulic constraints.

[0188] The algorithm iteration module 300 is used to perform optimization iteration on the pipeline parameters included in the initial pipeline network topology using a non-dominated sorting genetic algorithm to obtain a set of pipeline network configuration schemes.

[0189] The actual data acquisition module 400 is used to output pipeline configuration parameters according to the target scheme in the pipeline configuration scheme set; and to acquire the actual water level receding time series data of the physical entity pipeline network constructed based on the target scheme in the drainage mode. The actual water level receding time series data is the data of farmland water flowing into the physical entity pipeline network through irrigation and drainage well piles.

[0190] The drainage simulation module 500 is used to perform drainage process simulation based on a non-steady flow dynamics model and output simulated water level receding time series data.

[0191] The inversion calculation module 600 is used to set minimizing the error between the actual water level receding time series data and the simulated water level receding time series data as the optimization objective, and to invert and calculate the actual friction coefficient inside the pipe of the physical entity pipe network caused by silt deposition.

[0192] The feedback correction module 700 is used to switch between drainage mode and irrigation mode. Before switching to irrigation mode, the actual friction coefficient is fed back into the hydraulic constraints and the non-dominated sorting genetic algorithm is triggered to re-expand the optimization iteration in order to output the corrected pipeline water distribution strategy, thereby constructing the final optimized model.

[0193] At the system's underlying deployment level, each functional module is deployed based on the industrial control computer motherboard and the real-time operating system kernel. The industrial control computer motherboard is equipped with a multi-core central processing unit matrix and error-correcting memory modules. Each module runs as an independent thread within the central processing unit matrix, and the threads efficiently exchange tensor matrices and control signaling through a shared circular buffer queue within the error-correcting memory modules.

[0194] The specific details of the underlying hardware and software collaboration implementation for each functional module are as follows:

[0195] For the data acquisition module 100, its embedded communication protocol stack parsing firmware is responsible for stripping the transmission control protocol message headers from the influx of field IoT sensors and extracting meteorological, soil and hydrological physical payload data; the extracted payload data is then loaded into a relational structured database table after being formatted and mapped.

[0196] For the initial architecture establishment module 200, it calls the discrete graph theory operator library to project data structure entities representing water supply nodes and water transmission pipeline segments onto the two-dimensional plane of the farmland digital elevation model, thereby constructing a multi-adjacency matrix including vertex three-dimensional coordinates and edge resistance weights.

[0197] For the algorithm iteration module 300, a contiguous addressing memory pool is allocated in the dynamic random access memory to accommodate the chromosome sequence encoding matrix; at the same time, the floating-point instruction set parallel processing hardware architecture is invoked to synchronously and in parallel execute the solution of hydraulic adjustment algebraic equations.

[0198] For the actual data acquisition module 400, its hardware input pin is bound to a precision temperature-compensated real-time clock chip, which provides a reference timestamp for the acquired liquid level sensing signal; and calls a low-pass digital filter operator to remove high-frequency jitter noise in real time.

[0199] For the drainage simulation module 500, an accelerator for solving sparse matrix mathematics based on Newton-Raphson nonlinear iterative logic is encapsulated inside. The simulation solution of the drainage process is completed by geometric discretization of the non-conservative Saint-Venant partial differential dynamic equations with spatially staggered grid cells and fully implicit stable time step derivation.

[0200] The inversion calculation module 600 is responsible for driving the underlying multi-core concurrent computing process pool and activating the particle swarm intelligent tracking collaborative logic.

[0201] The feedback correction module 700 has a hardware interrupt intervention instruction issuance channel, which is used to replace the scalar constant of the smooth pipe wall resistance coefficient in the basic physical calculation equation library.

[0202] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for constructing an integrated pipeline irrigation and drainage collaborative optimization model, characterized in that, Includes the following steps: S10. Obtain basic data corresponding to the target farmland area, determine the irrigation design flow and drainage design flow based on the basic data, and form an irrigation mode and a drainage mode. The basic data includes meteorological and hydrological parameters, soil and crop parameters, as well as system configuration and energy consumption parameters. S20. Based on the irrigation design flow and the drainage design flow, construct the initial pipeline network topology and establish the pipeline network life cycle resource cost assessment function and hydraulic constraints. S30. Use a non-dominated sorting genetic algorithm to iteratively optimize the pipe parameters included in the initial pipeline network topology to obtain a set of pipeline network configuration schemes; S40. Output the pipeline configuration parameters according to the target scheme in the pipeline configuration scheme set, and obtain the actual water level receding time series data of the physical entity pipeline network constructed based on the target scheme in the drainage mode. The actual water level receding time series data is the data of farmland water flowing into the physical entity pipeline network through irrigation and drainage well piles. S50. Based on the non-steady flow dynamics model, the drainage process of the drainage mode is simulated, and the simulated water level receding time series data is output. S60. Minimize the error between the actual water level receding time series data and the simulated water level receding time series data as the optimization objective, and calculate the actual friction coefficient inside the pipe of the physical entity pipe network caused by silt deposition. S70. Switch between drainage mode and irrigation mode. Before switching to irrigation mode, feed back the actual friction coefficient into the hydraulic constraints and trigger the non-dominated sorting genetic algorithm to re-expand the optimization iteration in order to output the corrected pipeline water distribution strategy, thereby constructing an irrigation and drainage coordinated optimization model.

2. The model construction method according to claim 1, characterized in that, The irrigation design flow rate and drainage design flow rate are determined based on the aforementioned basic data, including: The irrigation modulus and drainage modulus are calculated based on the meteorological and hydrological parameters and the soil and crop parameters. The larger of the flow rates corresponding to the irrigation modulus and the flow rates corresponding to the drainage modulus is determined as the pipeline design flow rate.

3. The model construction method according to claim 1, characterized in that, In S20, the pipeline network full life cycle resource cost assessment function is obtained by weighted summation of the pipeline network construction infrastructure cost assessment value and the pipeline network operation cost assessment value; The cost assessment value of the pipeline network construction infrastructure consists of pipeline material consumption index, water pump configuration equivalent, and irrigation and drainage well pile construction equivalent. The estimated cost of pipeline operation consists of the power consumption of water pumps and the cost of maintenance and repair. The hydraulic constraints include nodal pressure constraints, pipe flow velocity constraints, and pipe diameter reduction constraints.

4. The model construction method according to claim 1, characterized in that, In S30, the step of using a non-dominated sorting genetic algorithm to iteratively optimize the pipe parameters included in the initial pipeline network topology includes: The number of main pipe sections, the number of branch pipe sections, the pipe diameter of each pipe section, and the water pressure distribution ratio of each intersection node are used as coding variables to generate an initial population composed of multiple decision variables. Crossover and mutation operations are performed on each individual in the initial population to generate a progeny population.

5. The model construction method according to claim 4, characterized in that, During the optimization iteration process, the infeasibility method of the solution is used to handle the hydraulic constraints: For a specific individual that does not meet the hydraulic constraints, calculate the total infeasibility value corresponding to the violation of each constraint by the specific individual. When the total infeasibility value exceeds the dynamic decrease threshold, the specific individual is removed from the offspring population, and the removed individual is replaced by the individual with the highest feasibility in the current population.

6. The model construction method according to claim 1, characterized in that, In S50, the simulation of the drainage process based on the non-steady flow dynamics model for the drainage mode includes: The narrow-slit method is used to integrate and characterize the unpressurized and pressurized flows inside the pipe, and a non-conservative Saint-Venant equation system is established. The non-conservative Saint-Venant equations are spatially discretized using spatial staggered cells and scalar dissipation finite volume method to establish a coupled algebraic equation system. The coupled algebraic equations are discretized and solved using a fully implicit time scheme, and the simulated water level receding time series data at different time steps are output.

7. The model construction method according to claim 1, characterized in that, In S40, obtaining the actual water level receding time-series data of the physical entity pipeline network constructed based on the target scheme under the flood drainage mode includes: The actual liquid level values ​​at multiple discrete time points during the water receding process are continuously collected by the liquid level sensor installed inside the irrigation and drainage well pile. The collected actual liquid level values ​​are arranged in chronological order to generate the actual water level receding time series data.

8. The model construction method according to claim 1, characterized in that, In S60, minimizing the error between the actual water level receding time series data and the simulated water level receding time series data is set as the optimization objective. The actual friction coefficient inside the physical pipe network due to silt deposition is calculated by inversion, including: Construct a target evaluation function, which is used to quantify the root mean square error of the actual water level receding time series data and the simulated water level receding time series data at corresponding time points; A heuristic search algorithm is used to continuously adjust the preset friction coefficient input into the non-steady flow dynamics model until the root mean square error decreases below the convergence condition. The preset friction coefficient corresponding to the convergence condition is locked as the actual friction coefficient.

9. The model construction method according to claim 1, characterized in that, In S70, feeding back the actual friction coefficient to the hydraulic constraints and triggering the non-dominated sorting genetic algorithm to re-expand the optimization iteration to output the corrected pipeline water distribution strategy includes: The actual friction coefficient is used to replace the preset friction coefficient originally set in the hydraulic adjustment calculation, and the formula for calculating head loss along the friction is reconstructed. The water pressure demand values ​​of each node within the pipeline configuration scheme set are updated using the reconstructed head loss calculation formula. The target water pump operating frequency that meets the updated water pressure requirement value is re-selected using the non-dominated sorting genetic algorithm, and the target water pump operating frequency is established as the corrected pipeline water distribution strategy.

10. An integrated pipeline irrigation and drainage collaborative optimization model construction system, which is a model construction system used in the integrated pipeline irrigation and drainage collaborative optimization model construction method of claim 1, characterized in that, include: The data acquisition module is used to acquire basic data corresponding to the target area of ​​farmland, and determine the irrigation design flow and drainage design flow based on the basic data; The initial architecture establishment module is used to construct the initial pipeline network topology based on the irrigation design flow and the drainage design flow, and to establish the pipeline network full life cycle resource cost assessment function and hydraulic constraints. The algorithm iteration module is used to perform optimization iteration on the pipeline parameters included in the initial pipeline topology using a non-dominated sorting genetic algorithm to obtain a set of pipeline configuration schemes; The actual data acquisition module is used to output pipeline configuration parameters according to the target scheme in the pipeline configuration scheme set; and to acquire the actual water level receding time series data of the physical entity pipeline network constructed based on the target scheme in the drainage mode. The actual water level receding time series data is the data of farmland water flowing into the physical entity pipeline network through irrigation and drainage well piles. The drainage simulation module is used to perform drainage process simulation on the drainage mode based on a non-steady flow dynamics model, and output simulated water level receding time series data. The inversion calculation module is used to set minimizing the error between the actual water level receding time series data and the simulated water level receding time series data as the optimization objective, and to invert and calculate the actual friction coefficient inside the pipe of the physical entity pipe network caused by silt deposition. The feedback correction module is used to switch between drainage mode and irrigation mode. Before switching to irrigation mode, the actual friction coefficient is fed back into the hydraulic constraints and the non-dominated sorting genetic algorithm is triggered to re-expand the optimization iteration to output the corrected pipeline water distribution strategy, thereby constructing the final optimization model.

Images