Reservoir multi-objective optimization scheduling decision-making method based on digital twinning and AI simulation

Abstract

The invention discloses a reservoir multi-target optimization scheduling decision-making method based on digital twinning and AI simulation, and relates to the technical field of reservoir scheduling, the method effectively solves the problems of data fragmentation and information redundancy in traditional scheduling by constructing a multi-source data fused reservoir digital twinning body, realizes accurate virtual-real mapping of reservoir working conditions, and improves the scheduling efficiency. A high-reliability simulation basis is provided for subsequent optimization scheduling, and decision errors caused by data errors or model deviation are avoided; four core targets of flood control, power generation, ecology and water supply are considered, an engineering safety bottom line and rigidity requirements are defined through a layered constraint system, constraint satisfaction degree screening is enhanced by means of an improved multi-target optimization algorithm, and target weights can be dynamically adjusted according to a real-time scene. And after the non-dominated scheme set is optimized and output, candidate schemes are screened in combination with real-time demand quantitative scores.

Description

Technical Field

[0001] This invention relates to the field of reservoir scheduling technology, and more specifically, to a multi-objective optimization scheduling decision-making method for reservoirs based on digital twins and AI simulation. Background Technology

[0002] Traditional reservoir operation techniques suffer from significant limitations in integrating and utilizing multi-source data. Hydrological, meteorological, engineering, and ecological data related to reservoir operation are often collected and stored in a fragmented manner, lacking a unified and efficient fusion mechanism. This leads to a break in the correlation between data, easily resulting in information redundancy. Furthermore, data errors or differences in units of measurement may hinder a comprehensive and accurate understanding of the overall reservoir condition, thus providing insufficient reliability of the basic data support for subsequent operation decisions and increasing the risk of decision-making bias.

[0003] Traditional multi-objective scheduling of reservoirs has shortcomings in objective coordination and constraint control. On the one hand, core objectives such as flood control, power generation, ecology, and water supply are often considered separately or simply superimposed, without forming a scientific synergistic optimization mechanism. This can easily lead to situations where one objective is optimized while other objectives deteriorate, making it difficult to balance multi-dimensional needs. On the other hand, the design of constraint conditions lacks systematicity, focusing more on single safety indicators and failing to construct a hierarchical constraint system covering engineering safety, physical mechanisms, and rigid requirements. This may result in scheduling schemes exceeding the engineering safety threshold or failing to meet rigid requirements such as ecology and water supply.

[0004] The generation and verification stages of traditional reservoir scheduling schemes are disconnected, resulting in poor feasibility for implementation. Existing optimization methods primarily output schemes based on theoretical calculations, failing to fully integrate real-time scheduling scenarios such as flood and dry seasons to dynamically adjust target priorities, and lacking effective simulation verification methods. Since the generated schemes are not subjected to virtual simulations matched to actual operating conditions, they may exhibit constraint violations or poor target stability, leading to frequent adjustments during actual implementation. This not only affects scheduling efficiency but may also introduce operational risks, making it difficult to guarantee the stability and practicality of scheduling decisions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a multi-objective optimization scheduling decision-making method for reservoirs based on digital twins and AI simulation.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A multi-objective optimization scheduling decision-making method for reservoirs based on digital twins and AI simulation, including Step 1: Construct a digital twin of the reservoir by fusing multi-source data; Step 2: Determine the flood control target of the reservoir Power generation target Ecological goals and water supply targets And set water level constraints, outflow constraints, water balance constraints and engineering equipment constraints; Step 3: Perform multi-objective simulation optimization and solve the problem, output a set of scheduling decision optimization schemes, adjust the target weights according to the real-time scheduling scenario, calculate the comprehensive score of each scheduling decision optimization scheme, input the scheme with the highest score into the digital twin, simulate future working conditions, and verify the scheme. If the verification is satisfied, the final scheduling scheme is output. If it is not satisfied, the scheme with the second highest score is input into the digital twin, and so on.

[0007] Furthermore, a digital twin of the reservoir is constructed by fusing multi-source data, as follows: multi-source data on the reservoir is collected, and the multi-source data is weighted and fused using the entropy weight method to obtain comprehensive reservoir status data. A digital twin of the reservoir is then constructed based on the comprehensive reservoir status data.

[0008] Furthermore, the reservoir's multi-source data includes hydrological data, meteorological data, engineering data, and ecological data; Hydrological data: Reservoir inflow Outbound flow Water level ; Meteorological data: Basin rainfall Evaporation Short-term rainfall forecast ; Engineering data: Gate opening Generator output ; Ecological data: Minimum ecological flow in downstream river channels River water quality indicators .

[0009] Furthermore, flood control objectives ; This represents a time variable, indicating any point in time within the scheduling period; This refers to the preset time range for the reservoir's flood season. This refers to the flood control limit water level of the reservoir, which is the highest safe water level allowed during the flood season. This is the weighting coefficient, with a value ranging from 0 to 1; The "multi-year average rainfall during the flood season" for the reservoir-controlled watershed.

[0010] Furthermore, power generation targets T represents the total duration of the scheduling cycle; The duration of a single time step is the time interval between two adjacent computing nodes.

[0011] Furthermore, ecological goals ; For a complete scheduling cycle; This represents the minimum ecological flow for the downstream river channel. This is the weighting coefficient, with a value ranging from 0 to 1; The COD standard limit for river water quality.

[0012] Furthermore, water supply targets ; This is the preset water supply period time range; Let t be the total water demand at time t; Let be the water loss at time t.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The method of this invention effectively solves the problems of data fragmentation and information redundancy in traditional scheduling by constructing a reservoir digital twin with multi-source data fusion. It integrates hydrological, meteorological, engineering and ecological data with entropy weight method, combines physical model and LSTM network to correct nonlinear factors, and can also dynamically calibrate twin parameters based on real-time monitoring values ​​to achieve accurate virtual-real mapping of reservoir operating conditions. This provides a highly reliable simulation basis for subsequent optimized scheduling and avoids decision-making errors caused by data errors or model deviations. Taking into account four core objectives—flood control, power generation, ecology, and water supply—the design clarifies the engineering safety baseline and rigid requirements through a hierarchical constraint system. It also strengthens the constraint satisfaction screening through an improved multi-objective optimization algorithm and can dynamically adjust objective weights based on real-time scenarios. This design avoids the deterioration of other objectives caused by optimizing a single objective, while ensuring that the optimization direction aligns with the scheduling priorities of different periods such as flood season and dry season, making the optimization scheme more adaptable to various scenarios. By employing a closed-loop process of "optimization solution - scoring and screening - digital twin verification," the feasibility of the scheduling scheme is significantly improved. After optimizing the output set of non-dominated schemes, candidate schemes are selected by quantitative scoring based on real-time demand. Then, the constraint satisfaction and target stability of the schemes are verified through digital twin simulation. By iteratively eliminating schemes that violate regulations or deteriorate the target, the final output scheduling scheme not only possesses theoretical optimality but also effectively conforms to the actual operation of reservoir projects. Attached Figure Description

[0014] Figure 1 This is a flowchart of the present invention; Figure 2 A flowchart illustrating the process of constructing a digital twin of a reservoir based on comprehensive reservoir status data. Detailed Implementation

[0015] Reference Figures 1-2 A multi-objective optimization scheduling decision-making method for reservoirs based on digital twins and AI simulation, including Step 1: Construct a digital twin of the reservoir by fusing multi-source data.

[0016] The construction of a digital twin of a reservoir by fusing multi-source data is as follows: collect multi-source data on the reservoir, use the entropy weight method to weight and fuse the multi-source data to obtain comprehensive reservoir status data, and construct a digital twin of the reservoir based on the comprehensive reservoir status data.

[0017] The reservoir's multi-source data includes hydrological data, meteorological data, engineering data, and ecological data; Hydrological data: Reservoir inflow Outbound flow Water level (Data collected in real time via a water level gauge); Meteorological data: Basin rainfall (Using the precipitation observation channels of meteorological satellites to obtain watershed-scale rainfall data), evaporation (Evaporation from the water surface is measured using an evaporation pan, and the data is collected and uploaded in real time via sensors), short-term rainfall forecast. (Based on atmospheric dynamics and thermodynamic equations, real-time meteorological data (temperature, air pressure, humidity, etc.) are input, and calculations are performed using global (such as ECMWF, GFS) or regional (such as WRF) models to output a short-term (maximum 72 hours) rainfall forecast for the watershed.) Engineering data: Gate opening Generator output (Uploaded in real time via SCADA system); Ecological data: Minimum ecological flow in downstream river channels (Based on historical runoff data of the watershed, such as the Tennant method, 10%-30% of the multi-year average runoff is taken as the minimum ecological flow (lower limit during the dry season and upper limit during the wet season), and river water quality indicators. (The concentration of chemical oxygen demand (COD) in river water samples was rapidly obtained on-site using a portable COD analyzer (spectrophotometry).) Entropy weighting is used to weighted fuse multi-source data, specifically as follows: Sample values ​​of each data type are standardized to a uniform scale (mapped to the 0-1 range) to eliminate dimensional differences. The formula is: Standardized value = (Original value - Minimum value of that data type) / (Maximum value of that data type - Minimum value). For each data type, its information entropy is calculated based on the standardized value. The smaller the entropy value, the richer the effective information carried by the data (lower redundancy). The weights of each data type are then derived from the information entropy; the smaller the entropy value, the larger the weight. The total weights are 1, calculated as: Weight of a data type = (1 - Entropy value of that data type) / (Sum of (1 - Entropy values) of all data types). Weighted fusion: The weights of each data type are multiplied by their standardized values, and then summed to obtain the fused comprehensive reservoir status data, achieving redundancy elimination and information enhancement from the multi-source data.

[0018] A digital twin of the reservoir is constructed based on comprehensive reservoir status data, as detailed below: Using the integrated reservoir status data as input, a core physical model of the reservoir is established, including the water balance equation (to calculate the dynamic changes in reservoir capacity), the water flow equation (to simulate the evolution of water flow in the reservoir area and downstream river channel), and the power generation output equation (corresponding to water level and unit output). ), while embedding the gate opening The correlation between the discharge flow rate and the actual discharge flow rate ensures that the twin model can accurately represent the physical process of the reservoir from "inflow to storage to discharge to power generation." This addresses nonlinear factors that the core physical model of the reservoir cannot cover (such as the impact of siltation on reservoir capacity and water quality). (Change patterns), using historical monitoring values ​​from the reservoir's comprehensive state data and the output error of the reservoir's core physical model as samples, a Long Short-Term Memory (LSTM) network is trained: real-time parameters from the reservoir's comprehensive state data (such as watershed rainfall) are processed. River water quality indicators The input is a Long Short-Term Memory (LSTM) network, which outputs error compensation values ​​for the core physical model of the reservoir. This corrects the results of the core physical model and improves the adaptability of the digital twin to complex operating conditions. Based on the actual monitoring values ​​in the reservoir's comprehensive state data, the deviation between the digital twin's output value and the actual value is calculated. If the deviation exceeds a preset threshold (e.g., water level deviation > 0.1m), the key parameters of the digital twin (e.g., reservoir capacity curve coefficient, discharge coefficient) are corrected using real-time parameters from the reservoir's comprehensive state data to ensure that the digital twin's output is synchronized with the actual operating conditions of the reservoir in real time, ultimately forming a digital twin of the reservoir.

[0019] Step 2: Determine the flood control target of the reservoir Power generation target Ecological goals and water supply targets Furthermore, water level constraints, outflow constraints, water balance constraints, and engineering equipment constraints are set.

[0020] Flood control target ; This represents a time variable, indicating any point in time within the scheduling period; This refers to the preset time range for the reservoir's flood season. This refers to the flood control limit water level of the reservoir, which is the highest safe water level allowed during the flood season. This is a weighting coefficient used to balance the contributions of "current water level overshoot risk" and "short-term rainfall forecast deviation risk," with a value ranging from 0 to 1. It is applicable to rainy watersheds. It can be set to 0.4-0.5, and for arid watersheds it can be set to 0.1-0.2; The "multi-year average rainfall during the flood season" for the reservoir-controlled watershed serves as a benchmark for judging whether short-term rainfall forecasts are abnormal (obtained statistically from historical meteorological data); power generation target. T represents the total duration of the scheduling cycle; The duration of a single time step, i.e., the time interval between two adjacent computing nodes; ecological goals. ; For a complete scheduling cycle; The minimum ecological flow in the downstream river channel is the minimum discharge flow required to maintain the basic ecological functions of the river channel (such as fish survival and vegetation irrigation). This is a weighting coefficient used to balance the contributions of "ecological flow deficit risk" and "COD exceedance risk," with a value range of 0-1 (example value: flow deficit leads to a higher risk of fish mortality). The value can range from 0.2 to 0.3). The COD standard limit for river water quality refers to the maximum permissible concentration of COD required to meet the water body's function (such as drinking water source or fishery water use); water supply target ; This refers to the pre-set water supply period, which is the time period during which water demand needs to be prioritized. Let t be the total water demand at time t, that is, the total amount of water that urban and rural residents, industry, agriculture, etc. need to obtain from the reservoir during this period; The water loss at time t is the water loss caused by leakage, evaporation, and water use along the way (such as irrigation along the canal) during the process of water leaving the reservoir being transported to the water-using end through channels / pipelines. The following constraints are set: water level constraint, outflow constraint, water balance constraint, and engineering equipment constraint: Water level constraints: ; Dead water level To verify the flood level; Outbound flow constraints: ; The minimum discharge capacity is determined by the gate characteristic curve. Maximum discharge flow; Water balance constraints: ; The total reservoir capacity at time t refers to the total amount of water stored in the reservoir at the "t-th time step" within the scheduling cycle. The total reservoir capacity at time t+1; The seepage rate of the reservoir dam at time t: refers to the "instantaneous rate" of water loss from the reservoir due to seepage from the dam body and foundation at time t; Engineering equipment constraints: Unit output: ; Minimize the unit's technical output, such as 30% of its rated power, to prevent low-frequency vibrations. This is the rated power of the unit, taken from the equipment nameplate.

[0021] Step 3: Perform multi-objective simulation optimization to solve the problem and output a set of scheduling decision optimization schemes. Adjust the objective weights according to the real-time scheduling scenario (such as prioritizing flood control during the flood season and prioritizing water supply during the dry season), calculate the comprehensive score of each scheduling decision optimization scheme, input the scheme with the highest score into the digital twin, simulate future working conditions, and verify the scheme (verify whether it meets the requirements of "no constraint violation and no significant deterioration of objectives"). If the verification is met, the final scheduling scheme is output. If it is not met, the scheme with the second highest score is input into the digital twin, and so on.

[0022] Perform multi-objective simulation optimization to solve the problem, and output a set of scheduling decision optimization schemes, as follows: Decision variable definition: Each individual corresponds to the "outbound flow sequence for the next 72 hours" [ , … The range of values ​​is limited by the outbound flow rate constraint. Population size and parameters: The population size is set to 150 (to balance iteration efficiency and solution diversity), and the maximum number of iterations is 80 generations (verified by preliminary experiments: the objective function value converges after 80 generations, and there is no significant optimization in subsequent iterations). Initialization method: The initial population is generated by random perturbation based on the historical best scheduling decision scheme (such as the scheme that meets the constraints in the same period of the past 3 years) to avoid the population getting trapped in local optima.

[0023] For each individual (scheduling decision scheme), its Input a digital twin sequence to simulate reservoir conditions over the next 72 hours: The digital twin calculates the real-time water level based on the "water balance equation" in step one. Storage capacity ; Based on the objective function formula in step two, calculate f1 (flood risk), f2 (total power generation), f3 (ecological risk), and f4 (water supply deficit) for this individual. Simultaneously check whether the individual meets the constraints of step two (water level, flow rate, unit output, etc.).

[0024] Selection: A combination of "elite retention + roulette wheel selection" strategy is adopted - first retain the top 20% of individuals with high overall constraint satisfaction, and select the remaining 80% through roulette wheel selection (the higher the fitness, the greater the probability of being selected). Crossover and mutation: "Single-point crossover" is used (randomly selecting one time step and exchanging two individuals). The mutation probability is set to 0.05 for the latter half of the sequence (randomly fine-tuning a certain time step). (The amplitude should be ≤10% to avoid deviating from the constraint range). Convergence criterion: When the mean change of the objective function value in the non-dominated solution set is ≤1% for 10 consecutive generations, the iteration stops and the final set of non-dominated optimization schemes is output (usually containing 20-30 schemes, each with its own advantages in different objectives, such as Scheme A being superior in flood control and Scheme B being superior in power generation).

[0025] The target weights are adjusted according to the real-time scheduling scenario (such as prioritizing flood control during the flood season and prioritizing water supply during the dry season), and the comprehensive score of each scheduling decision optimization scheme is calculated as follows: For example, during the flood season, when prioritizing flood control while taking into account the ecology, the target weights can be set as follows (f1 corresponds to weight r1, f2 corresponds to weight r2, f3 corresponds to weight r3, and f4 corresponds to weight r4): r1 = 0.4, r2 = 0.2, r3 = 0.25, and r4 = 0.15; during the dry season, when prioritizing water supply while taking into account power generation, the target weights can be set as follows: r1 = 0.1, r2 = 0.3, r3 = 0.2, and r4 = 0.4. First, each scheduling decision optimization scheme is "min-max normalized" (to eliminate dimensional differences), and then the comprehensive score is calculated according to the adjusted weights. The higher the score, the more suitable the scheme is for the current scenario.

[0026] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0027] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0028] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0029] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0030] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0031] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0032] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0033] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A reservoir multi-objective optimal scheduling decision-making method based on digital twinning and AI simulation, characterized in that, include Step 1: Construct a digital twin of the reservoir by fusing multi-source data; Step two: determining the flood control target of the reservoir , the power generation target , the ecological target , and the water supply target , and setting the water level constraint condition, the outflow constraint condition, the water balance constraint condition, and the engineering equipment constraint condition; Step 3: Perform multi-objective simulation optimization and solve the problem, output a set of scheduling decision optimization schemes, adjust the target weights according to the real-time scheduling scenario, calculate the comprehensive score of each scheduling decision optimization scheme, input the scheme with the highest score into the digital twin, simulate future working conditions, and verify the scheme. If the verification is satisfied, the final scheduling scheme is output. If it is not satisfied, the scheme with the second highest score is input into the digital twin, and so on.

2. The reservoir multi-objective optimization scheduling decision-making method based on digital twinning and AI simulation according to claim 1, characterized in that, The construction of a digital twin of a reservoir by fusing multi-source data is as follows: collect multi-source data on the reservoir, use the entropy weight method to weight and fuse the multi-source data to obtain comprehensive reservoir status data, and construct a digital twin of the reservoir based on the comprehensive reservoir status data.

3. The reservoir multi-objective optimal scheduling decision-making method based on digital twinning and AI simulation according to claim 2, characterized in that, The reservoir's multi-source data includes hydrological data, meteorological data, engineering data, and ecological data; Hydrological data: inflow discharge , outflow discharge , water level ; Meteorological data: basin rainfall , evaporation , short-term forecast rainfall ; Engineering data: gate opening , generator set output ; Ecological data: minimum ecological flow of downstream river , river water quality index .

4. The reservoir multi-objective optimization scheduling decision-making method based on digital twin and AI simulation according to claim 1, characterized in that, Flood control target ; denotes a time variable, representing an arbitrary time within the scheduling period; is a preset flood season time interval of the reservoir; is the flood control water level of the reservoir, i.e., the highest safe water level allowed during the flood season; is a weight coefficient, with a value range of 0-1; is the "flood season average annual rainfall" of the reservoir control basin.

5. The reservoir multi-objective optimization scheduling decision-making method based on digital twin and AI simulation according to claim 1, characterized in that, Generation target ; T represents the total duration of the scheduling period; is the duration of a single time step, i.e. the time interval between two adjacent computation nodes.

6. The reservoir multi-objective optimization scheduling decision-making method based on digital twin and AI simulation according to claim 1, characterized in that, Ecological target ; is the complete scheduling period; is the minimum ecological flow of the downstream river channel; is a weight coefficient, with a value range of 0-1; is the COD standard limit of the river water quality.

7. The reservoir multi-objective optimization scheduling decision-making method based on digital twin and AI simulation according to claim 1, characterized in that, water supply target ; is a preset water supply period time interval; is a total water demand amount at time t; is a water delivery loss amount at time t.

Images