Reservoir Group Scheduling Decision-making Method Considering Multi-source Uncertainty Propagation and Evolution Tracking

Abstract

The present invention discloses a reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking, which relates to the technical field of reservoir scheduling and is used to solve the problem of multi-objective risk decision-making for reservoir scheduling in a stochastic environment. By constructing a spatio-temporal two-dimensional risk deduction model, refined risk diagnosis and dynamic deduction are carried out on the reservoir group scheduling process to obtain the uncertainty propagation and evolution tracking mechanism, effectively improving the robustness and reliability of reservoir scheduling decisions in an uncertain environment. And in the acquisition of group preference information, a preference mediator FAHP-OWA that comprehensively considers subjective weights and preference conflicts is used to obtain a high-dimensional weight vector. At the same time, it combines multi-round interactive optimization and weight allocation mechanisms, which can gradually introduce the preference information of decision-makers, making the scheduling process more flexible and the coordination of the decision-making group better, improving the decision-making efficiency and accuracy of the decision-making group in an uncertain environment, and realizing the accurate discrimination of the advantages and disadvantages of the schemes.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir operation, and more specifically, to a reservoir group operation decision-making method considering the propagation and evolution tracking of multi-source uncertainties. Background Art

[0002] In the context of population growth, climate change, and increasing energy demand, reservoirs, as important water supply sources, control the hydrological situation of the basin by formulating operation plans to achieve water resource utilization and allocation. Real-time reservoir operation is based on meteorological and runoff forecasting processes, comprehensively considering operation objectives such as flood control, power generation, and ecology, and seeking the optimal water storage or discharge strategy for the reservoir. The effectiveness of reservoir operation not only depends on the design of the operation plan but also on the accuracy and robustness of the forecasting system. In particular, uncertainty factors such as meteorological data, flood forecasting, and decision-makers' preference information make the decision-making process in reservoir operation highly complex. These uncertainties not only directly interfere with the operation plan but also are transmitted layer by layer through the "meteorology-forecasting-operation-decision-making" chain, further amplifying the potential risks in the implementation process.

[0003] Reservoir operation is a complex task involving optimization decisions in multiple aspects such as flood control, water supply, and power generation. However, it is often difficult to balance the above-mentioned objectives, which leads to considerable risks and uncertainties. To solve such complex problems, it is necessary to develop a multi-objective optimization model that can take into account different objectives simultaneously. Existing reservoir operation methods do not fully consider the impact of uncertainties on operation decisions. Usually, the operation model relies on single meteorological forecasting data or historical hydrological data and fails to comprehensively consider the propagation effect of multi-source uncertainty factors. The accumulation of systematic forecasting errors and uncertainties may lead to a deviation between the operation result and the actual situation, thereby increasing the possibility of decision failure and the risk of the reservoir group system.

[0004] In addition, in the process of reservoir operation decision-making, the multi-attribute decision-making (MADM) model also faces limitations. Traditional MADM methods rely on a definite decision matrix and index weights and are difficult to cope with the uncertainty of index values in the decision matrix caused by forecasting uncertainties in the actual operation process and the weight uncertainty caused by the fuzziness of decision-makers' subjective preferences and the conflict of multi-agent interests. It is necessary to establish a stochastic multi-attribute decision-making model that can handle double uncertainties.

[0005] With the increasing uncertainty of the watershed hydrological system, the scheduling decision not only requires the optimization model to provide reasonable solutions during the scheduling stage, but also must conduct risk assessment and dynamic adjustment after the implementation of the solutions. The existing reservoir scheduling technologies lack systematic spatio-temporal dimension deduction tools and fail to conduct refined diagnosis and dynamic deduction of the risks at each hydrological characteristic node in the reservoir group. This deficiency limits the application of scheduling decisions under complex spatio-temporal coupling conditions, especially the actual impacts of the interactions between upstream and downstream reservoir systems and the uncertainties of variables such as flood flow and rainfall intensity on the scheduling scheme, which are difficult to be effectively quantified and evaluated. The existing technologies mostly rely on empirical evaluation methods and lack tools for comprehensively quantitatively evaluating the uncertainties and risks after scheduling. Therefore, developing a spatio-temporal two-dimensional deduction model for reservoir scheduling risks is of great promising value.

[0006] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0007] To overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a reservoir group scheduling decision method considering multi-source uncertainty propagation and evolution tracking to solve the problems raised in the above background art.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A reservoir group scheduling decision method considering multi-source uncertainty propagation and evolution tracking includes the following steps:

[0010] In the hydrological forecasting stage, a hydrological forecasting uncertainty evolution model based on a temporal fusion transformer (TFT) is constructed to simulate the dynamic characteristics of hydrological variables and generate the probability distribution of hydrological forecasts through bias calibration;

[0011] In the scheduling stage, a multi-objective distributionally robust optimization model considering meteorological variable uncertainty, runoff stochastic scenario uncertainty, and scheduling objective conflict is constructed, and a moss growth optimization algorithm is used to solve the non-dominated solution set;

[0012] In the scheduling risk deduction stage, a spatio-temporal two-dimensional deduction of the flood control scheduling risk of the reservoir group is carried out based on the spatio-temporal topological relationship constructed by a directed acyclic graph, and the risk level of the reservoir scheduling scheme is calculated;

[0013] In the decision-making stage, a preference mediator is used to coordinate the preferences of the decision-making group, a random decision matrix and weight samples are generated based on a hypercube simulation engine, a random average solution distance model is constructed through a weighted matrix fusion module, and the comprehensive evaluation value is calculated and the uncertainty scheme is optimized.

[0014] In a preferred embodiment, the uncertainty of the hydrological forecast is quantified through the following steps;

[0015] Simulate and correct meteorological forecast errors based on the TFT model to generate the probability distribution of meteorological variables;

[0016] Use a composite probability distribution model to describe runoff uncertainty, including a mixed form of normal distribution and skewed distribution:

[0017] ;

[0018] In the formula, is the standard normal distribution function, is the skewed distribution function, μ is the expected value of runoff, σ is the standard deviation of runoff, γ is the skewness coefficient, and α is the weight parameter;

[0019] Use Monte Carlo simulation for the conditional probability distribution of runoff output disturbed by meteorological errors:

[0020] ;

[0021] In the formula, is the conditional probability of the runoff result given the error E and the parameter , is the posterior probability distribution given the error E and the parameter , is the parameter of the meteorological variable.

[0022] In a preferred embodiment, the objective function of the multi-objective distribution robust optimization model includes:

[0023] Minimization of the flood control storage occupancy rate: ;

[0024] In the formula, is the storage capacity of the i-th reservoir at time t under the l-th random scenario; is the storage capacity corresponding to the designed flood limit water level of the i-th reservoir; is the maximum flood control storage capacity of the i-th reservoir;

[0025] Maximization of the safety degree of downstream control nodes: ;

[0026] In the formula, is the outflow discharge of the i-th reservoir at time t under the l-th random scenario; is the maximum allowable outflow discharge of the i-th reservoir.

[0027] In a preferred embodiment, the optimization process of the moss growth optimization algorithm includes:

[0028] Generate an initial solution through random perturbation of meteorological variables and form an initial population of solutions in combination with hydrological uncertainty;

[0029] Dynamically adjust the expansion path of the solution according to the wind direction mechanism: ;

[0030] In the formula, represents the similarity of individual i, is the direction weight of the wind direction;

[0031] Simulate the dual reproduction mechanism of moss, and expand along the best direction or adjust the local search range in the solution space.

[0032] In a preferred embodiment, the spatio-temporal two-dimensional deduction is realized through the following steps:

[0033] Construct a directed acyclic graph based on the spatial relationships of upstream and downstream, main and tributary streams, and left and right banks to represent the spatio-temporal topology of reservoir operation;

[0034] Use the graph attention network to perform feature learning on each node and dynamically adjust the information weights between nodes: ; ;

[0035] In the formula, and respectively represent the linear transformation of the feature vectors of hydrological variable nodes i and j passing through the weight matrix W; a is a learnable parameter used to learn the interaction between the features of hydrological variable nodes; LeakyReLU is a non-linear activation function used to introduce non-linearity and prevent the problem of gradient disappearance; is the set of neighbor nodes of hydrological variable node i;

[0036] Deduce the spatio-temporal propagation path, generate a risk distribution map, and characterize the evolution of scheduling risks at each node and time step.

[0037] In a preferred embodiment, the decision-making stage specifically includes the following steps:

[0038] Extract the scheme eigenvalue from the non-dominated solution set and construct a random decision matrix;

[0039] Adopt a preference mediator that comprehensively considers subjective weights and preference conflicts to obtain a high-dimensional weight vector, and calculate the comprehensive evaluation value by weighted calculation of the forward distance and the reverse distance;

[0040] The random average solution distance model ranks the priorities of each scheduling scheme in different decision-making scenarios through the comprehensive evaluation value.

[0041] In a preferred embodiment, the risk-based multi-attribute decision-making model simultaneously considers the following uncertainties:

[0042] The uncertainty of decision-making index values, including the highest water level, reservoir overtopping risk, maximum discharge flow, downstream flood risk, and final water level;

[0043] The uncertainty of weights coordinates the group preference conflicts through the fuzzy analytic hierarchy process combined with the ordered weighted averaging operator, and quantifies the randomness of weight distribution through the feasible weight space.

[0044] The technical effects and advantages of the reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking in the present invention:

[0045] The present invention comprehensively and systematically considers the multi-source uncertainties in the meteorology, forecasting, scheduling, and decision-making chain. By constructing a spatio-temporal two-dimensional risk deduction model, it conducts refined risk diagnosis and dynamic deduction on the reservoir group scheduling process, obtains the uncertainty propagation and evolution tracking mechanism, and effectively improves the robustness and reliability of reservoir scheduling decisions in an uncertain environment;

[0046] The stochastic average solution distance method created in the present invention, which is applicable to multiple uncertainties, provides a new method for quantitatively evaluating the advantages and disadvantages of scheduling schemes. In the acquisition of group preference information, a preference mediator FAHP-OWA that comprehensively considers subjective weights and preference conflicts is used to obtain a high-dimensional weight vector. At the same time, it combines multi-round interactive optimization and weight allocation mechanisms, can gradually introduce the preference information of decision-makers, makes the scheduling process more flexible and the coordination of the decision-making group better, improves the decision-making efficiency and accuracy of the decision-making group in an uncertain environment, and realizes the accurate discrimination of the advantages and disadvantages of the schemes. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic flow chart of the reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking in the present invention;

[0048] Figure 2 It is a flow chart of the numerical solution algorithm of the MC-EDAS model in the present invention;

[0049] Figure 3 It is the two-stage decision-making process of risk-based group decision-making in the present invention;

[0050] Figure 4 It is a three-dimensional column chart of the scheme ranking probability of the MC-EDAS model in the present invention under four different types of weights;

[0051] Figure 5 It is the central weight vector of each flood control scheduling scheme in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] By integrating runoff stochastic simulation, multi-objective stochastic optimization scheduling, spatio-temporal two-dimensional deduction of reservoir operation risks, and risk-based multi-attribute decision-making modules, the present invention effectively improves the robustness and reliability of reservoir operation decision-making in an uncertain environment.

[0054] Embodiment The present invention provides a multi-objective scheduling decision-making method for a reservoir group considering the propagation and evolution tracking of multi-source uncertainties, including the following steps. Refer to Figure 1 :

[0055] S1: Collect data such as meteorology, engineering, basin, and underlying surface required for modeling, and perform data cleaning and preprocessing operations on the data set to obtain the data set of the target basin.

[0056] S2: Construct a meteorological statistical post-processing module, and use a temporal fusion transformer (TFT) to simulate and quantify the uncertainties of various meteorological variables such as precipitation.

[0057] S3: Construct a general basin rainfall-runoff model, use the statistically post-processed meteorological forecast results as input, and fuse the general probability distribution line types to generate a runoff probability forecast that can comprehensively consider normal and skewed (positive skew / negative skew) conditions.

[0058] S4: Establish a real-time reservoir operation distributed robust optimization model that couples the minimum occupancy rate of the flood control storage capacity and the highest safety level of the downstream control nodes.

[0059] S5: Use the moss growth optimization algorithm (MGO) to solve the real-time reservoir operation distributed robust optimization model that couples the minimum occupancy rate of the flood control storage capacity and the highest safety level of the downstream control nodes to obtain a set of stochastic non-dominated solutions.

[0060] S6: Based on the obtained set of stochastic non-dominated solution samples, use the cumulative rainfall forecast error, reservoir and reach peak flow forecast error, reservoir and reach flood volume forecast error, cross-section combined peak flow, and flood regulation highest water level as network nodes, and draw a directed acyclic graph neural network according to the spatial topological relationships of upstream and downstream, main and tributary, left and right banks, following the principle of "cause nodes first, result nodes later", and use a graph attention network (GAT) to perform spatio-temporal two-dimensional deduction of reservoir operation risks.

[0061] S7: Based on the average solution distance method, a stochastic average solution distance method applicable to multiple uncertain environments was created. The stochastic decision matrix and the high-dimensional weight vector aggregated from the preferences of numerous decision-makers were extracted and input into the stochastic average solution distance model to quantitatively evaluate the solutions on the set of stochastic non-dominated solutions, thereby obtaining a scheduling plan that conforms to the decision-makers' preferences.

[0062] Further, the specific steps of step S1 are as follows:

[0063] S11: For each hydrological node, relevant variable data were obtained and integrated, including key elements such as precipitation intensity field, aerodynamic temperature field, and fluid motion velocity field. Physical constraint data such as reservoir storage capacity constraints and spillway design thresholds were extracted, and engineering parameters were calculated and determined. The geographical spatial information of the basin was obtained by analyzing the complex topological structure of the river channel, extracting terrain elevation data, and determining the spatio-temporal distribution of hydrological observation stations. The underlying surface characteristics were quantified using indicators such as soil permeability coefficient and vegetation coverage rate. All data were collected, cleaned, and normalized relying on a multi-source monitoring technology platform and integrated into a high-dimensional spatio-temporal dataset to provide accurate basic support for the subsequent modeling process.

[0064] S12: The construction of the dataset needs to perform a time series rearrangement operation according to the spatio-temporal distribution characteristics of the multi-source data and perform spatial cropping according to the range of each target hydrological node to ensure the consistency of the data time series.

[0065] S13: After the dataset is cleaned, it is stored in a data management system. This system dynamically provides data access support through API interfaces to ensure that the subsequent model modules can directly obtain the spatio-temporal data of the target basin required, realizing the transfer of cross-module data streams.

[0066] Further, the specific formulas for the time series rearrangement and spatial cropping in step S12 are as follows:

[0067] (1) Rearrange the time series of the data of hydrological variable node i to ensure data consistency:

[0068] ;

[0069] In the formula, is the time offset calculated according to the specific requirements of the hydrological node.

[0070] (2) Perform spatial cropping according to the range of the target hydrological node to obtain the dataset of the cropped target basin:

[0071] ;

[0072] In the formula, is the dataset after spatial cropping. It represents the data value at a specific time t and spatial position i.

[0073] Further, the specific formula for the data set cleaning and storage in step S13 is as follows:

[0074] (1) Clean the cropped data set, remove outliers, and obtain the cleaned data set

[0075] ;

[0076] In the formula, is the median calculated for the data set at spatial position i within the time window , which is not affected by extreme values (outliers).

[0077] (2) Store the cleaned data set in the data management system and provide data access support through the API interface:

[0078] ;

[0079] In the formula, parameters include parameters such as time series and spatial range.

[0080] Further, the specific steps of step S2 are as follows:

[0081] S21: Construct a temporal fusion transformer (TFT) to describe the spatio-temporal variation characteristics of complex hydrological variables, especially for key hydrological elements such as rainfall intensity, river flow, and wind speed field, and provide data support for the reservoir runoff forecasting process.

[0082] S22: Perform error correction and bias calibration on the meteorological prediction data, compare the historical hydrological observation data with the forecast data, generate the meteorological forecast uncertainty assessment results according to the TFT model, and output the probability distribution of the prediction error, the corrected meteorological data, and the confidence interval.

[0083] S23: Conduct sensitivity analysis on different meteorological variables (such as rainfall, wind speed, etc.) to evaluate the impact of their errors on the runoff and water level prediction results. Use Monte Carlo simulation to perturb the meteorological variable errors and observe how these errors propagate and affect the final runoff output. The specific formula is as follows:

[0084] ;

[0085] In the formula, is the conditional probability of the runoff result given the error E and the parameter , is the posterior probability distribution given the error E and the parameter , Parameters for meteorological variables, such as precipitation, wind speed, etc., and E is the error term of meteorological prediction.

[0086] Furthermore, the initialized temporal fusion transformer (TFT) architecture adopted in step S21 includes the following modules: GRN, VSN, SCE, TFD, and the specific operations are as follows:

[0087] (1) Perform GRN processing on the time series data (X_t) to ensure the effective flow of meteorological and hydrological information in the multi-layer network:

[0088] ;

[0089] (2) Use the variable selection network (VSN) to assign weights and select the temporal inputs of multiple hydrological variables. VSN is used to capture the most relevant features in the input variables and generate an adaptive variable selection weight matrix:

[0090] ;

[0091] In the formula, Wv and bv are respectively the trainable weight and bias parameters.

[0092] (3) The temporal coding module SCE maps the time information into a low-dimensional space representation through a periodic function:

[0093] ;

[0094] In the formula, t is the time step, T is the period length of the time series, and the temporal coding result et will be input into the model together with the input hydrological variables.

[0095] (4) The temporal fusion decoder module TFD aggregates the temporal features of meteorological variables through the attention mechanism to generate the final meteorological prediction data, and the calculation formula is as follows:

[0096] ;

[0097] In the formula, Q, K, and V are respectively the query, key, and value matrices, and dk is the dimension of the key vector.

[0098] Furthermore, the specific steps for error correction and bias calibration of the meteorological prediction data in step S22 are as follows:

[0099] (1) Generate the meteorological forecast error: ;

[0100] In the formula, is the residual between the meteorological forecast data and the actual value at time t, is the actual observed value at time t, is the actual observed value at time t.

[0101] (2) Correct the predicted value of meteorological data through the residual, and calculate the corrected predicted value as:

[0102] ;

[0103] In the formula, is the corrected predicted value at time t + 1, is the preliminary predicted value (uncorrected) at time t + 1, is the expected value of the residual at time t.

[0104] (3) Calculate the confidence interval as: ;

[0105] In the formula, is the corrected predicted value at time t + 1, is the preliminary predicted value (uncorrected) at time t + 1, is the expected value of the residual at time t, is the standard deviation of the residual.

[0106] Furthermore, the specific steps of step S3 are as follows:

[0107] S31: Construct a dynamic representation framework for the general rainfall-runoff model. The model covers the non-linear dynamic response mechanisms of rainfall, surface runoff, soil infiltration, and groundwater recharge, and through the multi-scale and multi-factor coupling mechanism under complex basin conditions, the hydrological variables are modeled as a whole. The model input is provided by the time series forecast of meteorological data, and multi-dimensional simulations are carried out for key variables such as rainfall intensity and wind speed field. The formula framework can be expressed as: ;

[0108] In the formula, Q represents the total runoff volume per unit time of the basin, P is the time series rainfall input volume, A is the area of the basin geographical unit, C represents the complex dynamic coefficient in the hydrological cycle, which depends on the underlying surface characteristics, vegetation coverage rate, and soil saturation, and Φ represents the dynamic response function of the runoff coefficient, which is adjusted based on the basin characteristics.

[0109] S32: Integration and standardization of multi-dimensional time series input data. The meteorological data set includes multi-dimensional elements such as rainfall intensity, wind speed field, and temperature change. Adopting standardization and data preprocessing strategies, the statistically post-processed meteorological data generated by the time series fusion converter (TFT) is integrated at a high level to ensure the spatio-temporal consistency and continuity of the input data.

[0110] S33: Use a compound distribution model to simulate runoff uncertainty. The model characterizes the asymmetry and discreteness of runoff through a mixed form of two probability distributions (normal distribution and skewed distribution), and is especially suitable for the runoff process under non-linear rainfall events. The formula is as follows:

[0111] ;

[0112] In the formula, is the standard normal distribution function, is the skewed distribution function, μ is the expected value of runoff, σ is the runoff standard deviation, γ is the skewness coefficient, and α is the weight parameter that controls the contributions of the normal and skewed distributions.

[0113] S34: Calibrate the model with high-precision parameters. Use the Bayesian inference method to optimize the key parameters μ, σ, and γ in the composite distribution model to maximize the model fitting degree and prediction accuracy. Dynamically adjust the model parameters through the inversion calculation of historical runoff observation data. The Bayesian inference formula is as follows: ;

[0114] In the formula, is the posterior distribution of the model parameter under the given observation data D, is the likelihood function, is the prior distribution, representing the initial estimate of the parameter.

[0115] S35: Generate high-confidence runoff probability forecast results. Through the calibrated model, output the runoff probability distribution of multiple time steps, and at the same time combine the confidence interval to characterize the uncertainty of runoff. The forecast results can quantitatively reflect the distribution of runoff under future rainfall scenarios, especially the basin response under extreme meteorological conditions. ;

[0116] In the formula, represents the runoff prediction at the (t + 1)-th moment, is the prediction interval, and α is the confidence level, which characterizes the uncertainty.

[0117] Furthermore, the specific steps of step S4 are as follows:

[0118] S41: Multidimensional modeling of the objective function. The objective function is set with the minimization of the flood control storage occupancy rate and the maximization of the safety of downstream nodes as the core to ensure that the scheduling strategy takes into account both flood control and safety. The model transforms the storage capacity limit, the outflow limit, and the safety of downstream nodes into an optimizable objective function, which is specifically expressed as:

[0119] The first objective function: Minimize the flood control storage occupancy rate of each reservoir under different random scenarios. The objective formula is as follows:

[0120] ;

[0121] In the formula, is the storage capacity of the i-th reservoir at time t under the l-th random scenario; is the storage capacity corresponding to the designed flood limit water level of the \(i\)-th reservoir; is the maximum flood control storage capacity of the \(i\)-th reservoir.

[0122] The second objective function: maximize the safety degree of the downstream control node, and the objective formula is as follows:

[0123] ;

[0124] In the formula, is the discharge flow of the \(i\)-th reservoir at time \(t\) under the \(l\)-th random scenario; is the maximum allowable discharge flow of the \(i\)-th reservoir.

[0125] By introducing the weight parameter \(\lambda\), the multi-objective optimization is transformed into a unified objective function:

[0126] ;

[0127] S42: The constraint conditions of the multi-dimensional expansion model of the constraint conditions cover the water balance limit of the reservoir, the discharge flow limit, and the safety flow limit of the downstream control node, etc. All constraint conditions depend on the dynamically monitored meteorological and hydrological data to ensure that the real-time scheduling operates within the physical boundaries:

[0128] (1) Storage capacity balance constraint: ;

[0129] In the formula, and are the storage capacities of the \(i\)-th reservoir at \(t + 1\) and \(t\) under the \(l\)-th random scenario, respectively. The same applies to other variables under the \(l\)-th random scenario.

[0130] (2) Water level constraint: ;

[0131] (3) Discharge flow constraint: ;

[0132] (4) Discharge flow amplitude constraint: ;

[0133] (5) Maximum discharge capacity constraint: ;

[0134] (6) Initial and boundary conditions: ;

[0135] In the formula, is the initial water level of the reservoir under the \(l\)-th random scenario; is the target end water level of the reservoir under the \(l\)-th random scenario.

[0136] S43: The model processes multi-source random perturbations (such as meteorological forecast errors, rainfall intensity deviations, etc.) to form a robust scheduling strategy. The model uses distributionally robust optimization to handle uncertainties, quantifies the potential impact of uncertainties on the reservoir group scheduling, and thus obtains the optimal solution under multi-source uncertainties. Assume that the perturbations come from an uncertainty set and the influence of these perturbations is simulated through a distribution set. The key formula in the optimization process is as follows: ;

[0137] In the formula, u is the scheduling decision variable, representing the inflow and outflow decisions of the reservoir; p is the perturbation parameter, representing meteorological and hydrological uncertainties; is the uncertainty set, which defines the possible distribution range of meteorological variables (such as precipitation, wind speed, etc.).

[0138] S44: Build an adaptive scheduling mechanism with multiple time scales, and dynamically update the scheduling strategy according to different time steps. In the short term, based on the timeliness of hydrological forecasts, and in the medium and long term, combined with multi-source data, to achieve robust real-time adjustment of the system. Make dynamic adjustments for different time steps: within a short time step (such as the hourly scale), the model is frequently updated according to real-time hydrological forecasts; within a long time step (such as the daily or weekly scale), the model makes global adjustments by combining meteorological forecast trends and historical data. The specific formula is as follows:

[0139] (1) Short-term step adjustment ;

[0140] (2) Long-term step adjustment ;

[0141] In the formula, N is the length of the scheduling time step (such as the number of days or hours); is the expected value of the uncertainty parameter.

[0142] Furthermore, the specific steps of step S5 are as follows:

[0143] S51: Combine the uncertainty factors of meteorological and flood forecasts, and generate the initial solution of the moss growth algorithm through random perturbations and heuristic strategies: ;

[0144] In the formula, lb and ub are the upper and lower bounds of the variable respectively, and rand(0,1) generates a random number between 0 and 1. After the initial solution is generated, the system applies it to the multi-objective flood control scheduling model and randomly initializes it according to meteorological uncertainties such as flood forecasts and rainfall intensity, so as to ensure that the generated solution can cope with multi-source uncertainties.

[0145] S52: Determine the wind direction and expand the growth direction of the solutions. In the moss growth model, the wind direction is one of the key factors for individuals to be affected by external factors (such as meteorological disturbances, flood forecasts). The introduction of the wind direction mechanism enables most solutions to develop towards the optimal solution direction, thereby reducing the model search space. The wind direction calculation formula in this process is as follows: ;

[0146] In the formula, represents the similarity of individual i, is the direction weight of the wind direction. The wind direction enables the solutions to flexibly adjust their evolution paths according to the current meteorological and hydrological information.

[0147] S53: Simulate the growth and expansion process of moss. Some solutions are used as "seeds" to spread along the wind direction, while other solutions readjust their paths according to environmental disturbances. The propagation path of the seeds can be described by the following formula:

[0148] ;

[0149] S54: Implement a dual reproduction mechanism to simulate the growth strategies of seed reproduction and moss reproduction. The reproduction formula is:

[0150] ;

[0151] ;

[0152] S55: When the moss is in a long-term ineffective or locally optimal situation, simulate the moss entering the metabolic dormancy state through the cryptobiotic strategy, and resume and continue the search when the conditions are appropriate. The cryptobiotic mechanism ensures that in extreme meteorological and hydrological environments, the algorithm can dynamically resume its search ability, thereby enhancing the robustness of the global search.

[0153] Furthermore, the specific steps of step S6 are as follows:

[0154] S61: Process the set of random non-dominated solution samples: These samples contain the multi-objective optimization results, covering different reservoir operation schemes. Each sample constructs its probability distribution function according to different meteorological variables (such as cumulative rainfall forecast error, flood peak flow forecast error, etc.). The formula is as follows:

[0155] ;

[0156] In the formula, represents the uncertainty probability distribution of the i-th reservoir at time t; Eprecip, Eflow, and Estorage represent the forecast errors of cumulative rainfall, flood peak flow, and reservoir storage capacity respectively. Set each reservoir as a network node to construct the core node of spatio-temporal risk propagation, and describe the dynamic evolution process of the whole system by using the interaction relationship between meteorological variables and reservoir operation.

[0157] S62: Construct a directed acyclic graph (DAG) according to the spatial topological relationships such as upstream and downstream, main and tributary rivers, and left and right banks of the reservoir and its basin. Represent the physical connections during the reservoir and its operation through this structure. The edges between nodes represent the water flow transfer paths and time delays, ensuring clear causal relationships, that is, the principle of "the cause node comes before and the result node comes after".

[0158] S63: Use the graph attention network (GAT) for spatio-temporal two-dimensional deduction. GAT deduces the changes in the states of each node at different time steps. Through the superposition of time steps, the propagation and evolution process of reservoir operation risks in space and time are shown. Specifically, it includes:

[0159] (1) Learn the weight relationships between nodes, identify the nodes that have the greatest impact on the decision-making risks of the reservoir system (such as the flow forecast error of the upstream reservoir), and dynamically adjust the weights according to the influence magnitude.

[0160] ;

[0161] ;

[0162] In the formula, and respectively represent the linear transformation of the feature vectors of hydrological variable nodes i and j passing through the weight matrix W; a is a learnable parameter used to learn the interaction between the features of hydrological variable nodes; LeakyReLU is a non-linear activation function used to introduce non-linearity and prevent the problem of gradient disappearance. is the set of neighbor nodes of hydrological variable node i.

[0163] (2) Dynamically adjust the information transfer weights between nodes. Based on the results of the dynamic weight adjustment, determine the optimal reservoir operation path under specific spatio-temporal conditions, ensure that both the flood control storage capacity can be fully utilized and the safety of the downstream area can be maximally protected, and accurately capture the spatio-temporal evolution of reservoir operation risks;

[0164] ;

[0165] In the formula, is the updated feature of each reservoir node i, is the activation function.

[0166] (3) Generate a reservoir operation risk distribution map, which includes the risk levels of each reservoir node and its adjacent nodes at different time steps, identify key nodes and high-risk areas, and then perform prediction and risk inference on any node in the network to improve the accuracy of reservoir operation decision-making.

[0167] ;

[0168] In the formula, represents the risk level of the hydrological variable node i at the time step t.

[0169] Furthermore, the specific steps of the step S7 are as follows:

[0170] S71: Extract the variable feature information from the set of random non-inferior solutions, obtain the distribution of each index value by using statistical analysis, and then construct a decision matrix under the uncertain environment.

[0171] S72: Aggregate the preferences of reservoir operation experts, and use the preference mediator FAHP-OWA that comprehensively considers subjective weights and preference conflicts to obtain a high-dimensional weight vector. The preference mediator includes three components: a basic preference expressor, an information aggregator, and a feature extractor. The basic preference expressor is used to initially obtain the high-dimensional weight information of each operation expert, and this component supports the information expression in the form of triangular fuzzy numbers; the information aggregator uses the OWA operator to fuse and aggregate the high-dimensional weight information; the feature extractor fits the distribution characteristics according to the high-dimensional index weight set and conducts spatial expansion.

[0172] Furthermore, in the step S72, the information aggregator aggregates the high-dimensional weight vector into a coordinated weight vector, and the output of the basic preference expressor is a high-dimensional weight information vector , k = 1, 2, …, L, where L is the number of operation experts participating in the group decision-making, Calculated from triangular fuzzy numbers, the weight fusion formula of the information aggregator coupled with the OWA operator is as follows:

[0173] ;

[0174] ;

[0175] ;

[0176] The coordinated weight vector that aggregates all experts' preferences is as follows:

[0177] ;

[0178] Furthermore, in the step S72, since the decision needs to consider as much preference information as possible, the feature extractor conducts spatial expansion on the coordinated weight vector according to a certain probability distribution, and the specific formula is as follows:

[0179] ;

[0180] In the formula, is the preference conflict level. When the preference conflict of the decision-making group is small and the satisfaction with the calculated coordinated weight vector is high, A smaller value can be taken, and vice versa A larger value is taken to make the weight space cover a larger range.

[0181] S73: The random average solution distance model can handle high-dimensional weight vectors and random decision matrices, and this model ranks the priority of the uncertain solutions in the set of random non-dominated solutions.

[0182] Furthermore, the random version of the average solution distance model in step S73 consists of the following key components, as Figure 2 shown:

[0183] (1) Hypercube simulation engine: The Latin hypercube sampling technique is used to generate a sample set of the decision matrix and high-dimensional weight vectors. The value range of each random variable is divided into several equally spaced intervals. A sample point is randomly selected within each interval to ensure the uniform distribution of the samples of each variable.

[0184] (2) Weighted matrix fusion module: The sampled decision matrix and high-dimensional weight vectors are combined, and corresponding index weights are assigned to each sample to ensure the representativeness of all combinations of variables and weights.

[0185] (3) Running simulation component: Model operations are performed on each generated weighted decision matrix sample to evaluate the distance between each sample and the average solution, and different decision scenarios are simulated. This step can be calculated using the EDAS model.

[0186] (4) Probability decision component: The output results of all samples are collected, and distribution characteristics are extracted, including calculating the mean, variance, confidence interval, etc., and statistical indicators such as the probability of scheme ranking and weighted comprehensive evaluation index are output.

[0187] Furthermore, the main calculation steps of the average solution distance (EDAS) model in step S73 are as follows:

[0188] (1) Select the evaluation index of the reservoir operation scheme.

[0189] (2) The n-dimensional index values of m schemes form the decision matrix X.

[0190] ;

[0191] In the formula, Xij represents the j-th index value of the i-th scheme.

[0192] (3) Calculate the average scheme. ;

[0193] (4) According to the different types of indicators, calculate the distance between each alternative scheme and the average scheme to determine the positive distance matrix and the reverse distance matrix Among them and The calculation formulas are as follows:

[0194] ;

[0195] ;

[0196] (5)Determine the weighted sum of positive distances and the weighted sum of negative distances .

[0197] (6)Normalize the weighted sum of positive distances and the weighted sum of negative distances .

[0198] (7)Calculate the comprehensive evaluation value ASi of each scheme, and sort the schemes accordingly. The larger the comprehensive evaluation value, the better the scheme. ;

[0199] Furthermore, three statistical indicators, namely the scheme sorting probability, the weighted comprehensive evaluation index, and the central weight vector, are defined in step S73 as follows:

[0200] (1)The scheme sorting probability ( ) represents the proportion of the number of times the reservoir operation scheme Ai obtains the sorting priority r in the Latin hypercube sampling. The calculation formula is: ;

[0201] In the formula, refers to the weight space corresponding to the scheme Ai obtaining the sorting r ( ).

[0202] (2)The weighted comprehensive evaluation index ( ) represents the probability that the reservoir operation scheme Ai obtains all sorts. The calculation formula is: ;

[0203] In the formula, is the secondary weight, reflecting the contribution degree of all of the scheme Ai to the weighted comprehensive evaluation index. is the comprehensive evaluation of the scheme Ai.

[0204] (3)The central weight vector ( ) is defined as the center of the weight expansion space when the scheme Ai obtains the optimal sorting: ;

[0205] In the formula, is the weight when the sorting is optimal. Characterizes the corresponding relationship between different weight combinations and the optimal reservoir operation decision.

[0206] Taking the reservoir system in the Dadu River Basin as an example, including the Pubugou Reservoir and the downstream protected object, Leshan City, the effectiveness and rationality of the method are illustrated.

[0207] The Pubugou Reservoir plays an important role in flood control, power generation, water supply, etc. The normal storage level of the reservoir is 850 m, and the total reservoir capacity is 5.39 billion cubic meters, of which the flood regulation capacity is 1.056 billion cubic meters. Leshan City is an important flood control control point downstream. In this numerical experiment, 40 runoff scenarios are randomly generated using a general-purpose watershed rainfall-runoff model. The generated runoff scenario set is used as the input of the reservoir real-time operation distributionally robust optimization model, and the moss growth optimization algorithm (MGO) is used to solve the model to obtain a set of stochastic non-dominated solutions under uncertain conditions. The stochastic decision matrix of the multi-objective optimal operation of the reservoir obtained therefrom consists of 30 random variables and 20 constants, as shown in Table 1.

[0208] Table 1 Stochastic decision matrix of the real-time multi-objective optimal operation of the reservoir

[0209]

[0210] During the actual operation period, the decision-making is carried out in two stages. The specific steps are as Figure 3 shown. In the first stage, it is not required that the decision maker express any preference, and the MC-EDAS model is run under the condition that the index weights are completely unknown. Figure 4 shows the three-dimensional bar chart of the probability of the scheme ranking under the condition that the weight information is completely unknown. It can be seen from the figure that the probabilities of the schemes A1, A2, A9, and A10 ranking at the back are relatively large. Among them, the probabilities of A1 and A2 obtaining a poor ranking are very large (the probabilities of ranking in the top five are both 0), and the sum of the probabilities of A9 - A10 ranking in the top five is less than 10%. The above results indicate that these four schemes can be judged as significantly inferior schemes and eliminated from the scheme set. The schemes A3 - A5 all have relatively large probabilities of ranking at the front or in the middle, and can be identified as compromise schemes. In addition, the schemes A6 - A8 all have a certain probability of obtaining the optimal ranking. Among them, the probability of scheme A7 is 47.0%, the probability of scheme A6 is 24.6%, and the probability of scheme A8 is 15.3%. These three schemes can be identified as significantly superior schemes and should be focused on in the second stage.

[0211] Figure 5 shows the central weight vector when each flood control operation scheme obtains the optimal ranking. It can be seen from the figure that the central weight vector of scheme A3 output by the MC-EDAS model assigns relatively small weights to Qmax and PQ, which are 0.1774 and 0.0499 respectively; while relatively large weights are assigned to Zmax, PZ, and Ze, which are 0.3317, 0.246, and 0.195 respectively.

[0212] In the second stage of the decision-making process, based on the preliminary optimal selection results of the first stage and the central weight vector, four reservoir operation experts are invited to express their preference information in the form of triangular fuzzy numbers. The fuzzy judgment matrix given by the expert group is shown in Table 2.

[0213] Table 2 Fuzzy judgment matrix given by four reservoir operation experts

[0214]

[0215] The FAHP-OWA coordinated weighting method is used to weight each index to obtain the coordinated weights. As shown in Table 3, first, the fuzzy judgment matrix of the expert group is transformed into quantitative weight values according to the FAHP method. Secondly, the OWA operator is used to perform an ordered weighted aggregation on the conflicts of the preferences of the four experts to obtain the coordinated weights. On the basis of the coordinated weights, the uncertainty of the weight information is further quantified by using the feasible weight space, as shown in Table 4.

[0216] Table 3 Weight quantification based on the fuzzy analytic hierarchy process and the results of weight assignment by the OWA operator

[0217]

[0218] Table 4 Three different forms of weight information

[0219]

[0220] Taking the random decision matrix and three types of weights as inputs, the MC-EDAS model is run again. Comparing (b), (c), and (d) in Figure 4 with (a) in Figure 4 , after adding the preference information of the decision-making group, the output results of the MC-EDAS model become more explicit and can better distinguish significantly superior and inferior schemes.

[0221] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0222] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0223] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application of the technical solution and the invention constraints. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0224] In addition, in each embodiment of this application, the functional modules can be integrated into a processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0225] As mentioned above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0226] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking, characterized in that: The steps include: In the hydrological forecasting stage, a hydrological forecast uncertainty evolution model based on a time series fusion converter is constructed to simulate the dynamic characteristics of hydrological variables and generate the probability distribution of hydrological forecasts through deviation calibration; In the dispatching stage, a multi-objective distributed robust optimization model is constructed that comprehensively considers the uncertainty of meteorological variables, the uncertainty of runoff random scenarios, and the conflict of dispatching objectives, and the moss growth optimization algorithm is used to solve the non-inferior solution set; In the operation risk deduction stage, the spatiotemporal two-dimensional deduction of the flood control operation risk of the reservoir group is carried out based on the spatiotemporal topological relationship constructed by the directed acyclic graph, and the risk level of the reservoir operation plan is calculated; In the decision-making stage, the preference mediator is used to coordinate the preferences of the decision-making group, and the random decision matrix and weight samples are generated based on the hypercube simulation engine. The random average solution distance model is constructed through the weighted matrix fusion module, and the comprehensive evaluation value is calculated and the priority of each scheduling scheme in different decision-making scenarios is ranked; The objective function of the multi-objective distributed robust optimization model includes: Minimize flood control storage capacity occupancy rate: ; In the formula, is the storage capacity of the i-th reservoir at time t under the l-th random scenario; is the storage capacity corresponding to the design flood limit water level of the i-th reservoir; is the maximum flood control storage capacity of the i-th reservoir; Maximize the safety of downstream control nodes: ; In the formula, is the outflow of the ith reservoir at time t under the lth random scenario; is the maximum outflow allowed by the i-th reservoir; The two-dimensional space-time deduction is achieved by the following steps: Construct a directed acyclic graph based on the spatial relationship between upstream and downstream, main and tributary rivers, and left and right banks to represent the spatiotemporal topology of reservoir operation; Use the graph attention network to learn the features of each node and dynamically adjust the information weights between nodes: ; ; In the formula, and They represent the linear transformation of the feature vectors of hydrological variable nodes i and j after the weight matrix W; a is a learnable parameter used to learn the interaction between the features of hydrological variable nodes; LeakyReLU is a nonlinear activation function used to introduce nonlinearity and prevent the gradient vanishing problem; is the set of neighbor nodes of hydrological variable node i; The spatiotemporal propagation path is deduced to generate a risk distribution map, which characterizes the evolution of scheduling risk at each node and time step.

2. The reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking according to claim 1 is characterized in that: The uncertainty of hydrological forecasts is quantified through the following steps; Based on the TFT model, the meteorological forecast errors are simulated and corrected to generate the probability distribution of meteorological variables; A composite probability distribution model is used to describe runoff uncertainty, including a mixture of normal and skewed distributions: ; In the formula, is the standard normal distribution function, is the skewed distribution function, μ is the expected value of runoff, σ is the standard deviation of runoff, γ is the skewness coefficient, and α is the weight parameter; The conditional probability distribution of runoff output disturbed by meteorological errors is obtained using Monte Carlo simulation: ； In the formula, Given the error E and parameters The conditional probability of runoff outcome when Given the error E and parameters The posterior probability distribution when , is the parameter of meteorological variables.

3. The reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking according to claim 1 is characterized in that: The optimization process of the moss growth optimization algorithm includes: The initial solution is generated by random perturbations of meteorological variables and the initial population of solutions is formed by combining hydrological uncertainties; Dynamically adjust the expansion path of the solution according to the wind direction mechanism: ; In the formula, represents the similarity of individual i, is the directional weight of wind direction; The dual reproduction mechanism of moss is simulated to expand or adjust the local search range along the optimal direction in the solution space.

4. The reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking according to claim 1 is characterized in that: The decision-making stage includes the following steps: Extract the solution eigenvalues ​​from the non-inferior solution set and construct a random decision matrix; A preference mediator that comprehensively considers subjective weights and preference conflicts is used to obtain a high-dimensional weight vector, and a comprehensive evaluation value is calculated based on the weighted forward distance and reverse distance. The random average solution distance model ranks the priorities of various scheduling schemes in different decision-making scenarios through comprehensive evaluation values.

5. The reservoir group scheduling decision-making method considering multi-source uncertainty propagation and evolution tracking according to claim 4 is characterized in that: The risk-based multi-attribute decision-making model considers the following uncertainties simultaneously: Uncertainty in the values ​​of decision indicators, including maximum water level, reservoir overtopping risk, maximum outflow, downstream flood risk, and final water level; The uncertainty of weights is addressed by using the fuzzy analytic hierarchy process combined with the ordered weighted average operator to coordinate group preference conflicts, and the randomness of weight distribution is quantified through the feasible weight space.

Images