Method for selecting reconfiguration working condition of water power of tidal river network based on sensitive analysis of gate pump scheduling response

Abstract

The application discloses a kind of based on gate pump scheduling response sensitivity analysis's tidal river network hydrodynamic reconstruction working condition optimization method, comprising: obtaining the candidate gate pump scheduling working condition set and corresponding multidimensional response index matrix for target tidal river network, the feasibility screening of the candidate gate pump scheduling working condition set, obtain feasible working condition set;Based on the information dispersion of each response index and the response sensitivity of scheduling parameter, construct fusion weight, and the comprehensive performance score is obtained by using the fusion weight to comprehensively evaluate feasible working condition;And based on the key scheduling parameter and the preference constraint of preconfiguration, the recommended score is obtained by modifying the comprehensive performance score;Based on the spatial heterogeneity of index, each working condition is locally short board checked and eliminated, and the target hydrodynamic reconstruction scheduling scheme is output.The application realizes funnel type layer-by-layer optimization of mass scheduling working condition, while ensuring overall optimization efficiency, effectively avoids local stagnant water risk, and improves the reliability of engineering operation.

Description

Technical Field

[0001] This invention relates to the field of water environment management and water conservancy project scheduling technology, and in particular to a method for optimizing the hydrodynamic reconfiguration conditions of tidal river networks based on gate pump scheduling response sensitivity analysis. Background Technology

[0002] Under the impact of heavy rainfall or overflow pollution, complex tidal river networks often exhibit extremely uneven hydrodynamic spatial and temporal distribution due to the combined effects of tidal backwater and internal channel resistance, making it easy for pollutant plumes to linger in localized areas for extended periods. Utilizing the coordinated operation of external and internal gate and pump hubs to force water exchange and reshape the hydrodynamic flow field within the river network is a key engineering intervention to accelerate the recovery of water physical and chemical indicators. A rational and efficient gate and pump coordination scheme can maximize the use of limited pumping capacity and available tidal windows to accelerate the convective diffusion and degradation of pollutants.

[0003] Currently, scheduling decisions for improving the water environment of tidal river networks mainly rely on scenario simulations and multi-criteria evaluations using coupled hydrodynamic and water quality models. The conventional approach typically involves pre-setting several fixed combinations of gate / sluice gate opening / closing or pumping station flow rates, calculating the overall water quality improvement under typical hydrological cycles using numerical models, and then using entropy weighting or expert scoring for linear weighted ranking to derive a recommended scheme. However, with the increasing topological complexity of urban river networks, the weight allocation in the multi-objective trade-off process often falls into a rigid, static combination, making it difficult to dynamically adjust according to changes in the system's physical response intensity, thus limiting the objectivity of the optimization results. More importantly, existing comprehensive evaluation systems are generally based on the global average performance of various physical indicators at the river section or network-wide scale. This highly aggregated calculation logic can easily mask the risk of extreme water quality deterioration at specific cross-sections along the route, meaning that water diversion schemes with high evaluation scores may still cause severe waterlogging disasters at specific spatial nodes during actual engineering implementation.

[0004] Therefore, in complex water transfer projects with multivariate coupling, constructing a decision-making system that combines robustness in objective physical evaluation with the ability to prevent local secondary disasters has become a common technical challenge in this field. It is necessary to study a scheduling optimization method that can effectively balance global multi-objective performance and spatial extreme risks, and fully adapt to complex hydraulic conditions. Summary of the Invention

[0005] The purpose of this invention is to provide a method for optimizing the hydrodynamic reconfiguration conditions of tidal river networks based on gate pump scheduling response sensitivity analysis, in order to solve the aforementioned problems in the prior art.

[0006] Technical solution: A method for optimizing hydrodynamic reconfiguration conditions in tidal river networks based on gate pump scheduling response sensitivity analysis, comprising:

[0007] Obtain a set of candidate gate pump scheduling conditions for the target tidal river network, and a multi-dimensional response index matrix corresponding to the set of candidate gate pump scheduling conditions;

[0008] The feasibility of the candidate gate pump scheduling conditions is screened to obtain a feasible set of conditions.

[0009] A fusion weight is constructed based on the information dispersion of each response indicator and the response sensitivity to scheduling parameters. The fusion weight is then used to comprehensively evaluate each feasible working condition and obtain a comprehensive performance score.

[0010] Identify key scheduling parameters, and based on these key scheduling parameters and pre-configured preference constraints, correct the overall performance score to obtain a recommended score;

[0011] Based on the spatial heterogeneity of the indicators, local shortcomings are checked and the last-place elimination is performed for each working condition, and the target hydrodynamic reconfiguration scheduling scheme is output.

[0012] Beneficial effects: Addressing the challenge that static weights in multi-objective decision-making can easily lead to evaluation failure, this invention introduces an adaptive evaluation mechanism based on information discrimination, cleverly mapping the distribution characteristics of objective data with the underlying physical response sensitivity of the system through a dynamic convex combination. This approach effectively overcomes the rigid constraints of fixed weights, enabling the evaluation system to automatically suppress the interference of extreme noise indicators and improve the robustness and physical interpretability of multi-dimensional comprehensive hydrodynamic evaluation.

[0013] In adapting to complex hydraulic conditions, this invention innovatively constructs a sensitivity-weighted preference soft-constraint model. By transforming on-site management experience into a piecewise penalty function based on key parameters, it effectively reduces the recommended priority of extreme solutions that involve extremely high energy consumption or secondary pollution risks without compromising the absolute red line of engineering feasibility. This ensures that the final dispatch instructions highly match the economic and safe operation requirements of actual engineering projects.

[0014] Furthermore, to thoroughly prevent secondary disaster risks caused by global average effects, this invention delves into the evaluation granularity of variance analysis down to a single cross-section, accurately capturing the spatial heterogeneity differences of response indicators spreading along the river channel. By identifying the most vulnerable cross-section with the most severe utility deficit and setting a dynamic elimination threshold, a robust "barrel principle" defense is established, effectively reducing the extreme risks of localized waterlogging and black and odorous outbreaks, and providing a highly reliable decision-making foundation for the emergency management of impact pollution in plain tidal river networks. Attached Figure Description

[0015] Figure 1 This is a diagram illustrating the optimal method for hydrodynamic reconfiguration of tidal river networks based on the sensitivity analysis of gate pump scheduling response in this invention.

[0016] Figure 2This is the fusion weight extraction diagram described in this invention.

[0017] Figure 3 This is the recommended score acquisition diagram described in this invention.

[0018] Figure 4 This is a diagram of the spatial heterogeneity index extraction method described in this invention. Detailed Implementation

[0019] Combination Figure 1 Description of Example 1: A method for optimizing the hydrodynamic reconfiguration conditions of tidal river networks based on gate pump scheduling response sensitivity analysis is provided, including the following steps:

[0020] Step 101: Obtain a set of candidate gate pump scheduling conditions for the target tidal river network, and a multi-dimensional response index matrix corresponding to the set of candidate gate pump scheduling conditions.

[0021] This method is primarily applied to shock pollution scenarios affected by rainfall or pipeline overflows. The step of obtaining a candidate gate pump scheduling condition set for the target tidal river network is specifically triggered when a shock pollution event with a sudden deterioration in water quality indicators is identified by online monitoring equipment deployed in the target tidal river network.

[0022] Online monitoring equipment collects dissolved oxygen, ammonia nitrogen, or total phosphorus concentrations at key river sections in real time. When the monitored core indicators, such as ammonia nitrogen concentration, suddenly exceed the preset safety threshold, the system is determined to be subjected to severe impact pollution input, and the current working condition optimization and hydrodynamic reconfiguration process is automatically activated.

[0023] The candidate gate-pump scheduling case set is represented as a 2D parameter matrix. Each row represents an independent gate-pump joint scheduling case, and each column represents the specific values ​​of a type of gate-pump scheduling parameters. These scheduling parameters include at least the control water level, pump station start-up time, pump station operating time, and pump station inflow. The multidimensional response index matrix is ​​represented as a mapping data set between candidate cases and spatial cross-sections, containing the spatiotemporal response characteristic values ​​of each candidate case at different representative cross-sections and spatial scales. The response indices in the multidimensional response index matrix are divided into cross-sectional scale indices and channel scale indices.

[0024] The response indicators encompass the hydrodynamic, water environment, and water quality dimensions, which characterize the physical system's state, as well as the operational scheduling dimension, which characterizes the engineering control costs. By combining these four dimensions, the complex evolution patterns of the tidal river network as a whole and in its local areas under different scheduling combinations can be fully depicted.

[0025] In some alternative implementations, the process of obtaining the candidate gate pump scheduling condition set and the corresponding index matrix can also be performed by periodically loading a pre-stored local historical condition database, or by using an online prediction model to generate it in real time.

[0026] Step 102: Perform feasibility screening on the candidate gate pump scheduling condition set to obtain a feasible condition set; the feasibility screening process refers to introducing specific external red line rigid constraints to perform preliminary filtering of the physical feasibility of massive scheduling combinations.

[0027] Due to the dense network of sluice gates and pumping stations and the gently undulating terrain of the tidal plain river network, their operation is strictly constrained by flood control safety water level boundaries. By comparing the extreme maximum water level value during the simulation process with a pre-configured set of hard constraint indicators, it is possible to identify online which operating conditions might trigger breach risks or violate the engineering structural design red lines during operation.

[0028] By performing a feasibility screening, infeasible operating conditions that violate any hard constraints can be eliminated, and the remaining safe operating conditions constitute the set of feasible operating conditions. The coarse screening mechanism forms the first-level funnel of the multi-level optimization framework, ensuring that all subsequent calculations and sorting are carried out within safe engineering boundaries, thus saving the computational overhead of advanced optimization algorithms.

[0029] Step 103: Construct fusion weights based on the information dispersion of each response index and the response sensitivity to scheduling parameters, and obtain the response sensitivity contribution of each scheduling parameter. Use the fusion weights to comprehensively evaluate each feasible working condition and obtain a comprehensive performance score. Specifically, calculate the Euclidean distance between the weighted index vector of each feasible working condition and the positive ideal point, and the Euclidean distance between the weighted index vector and the negative ideal point. Use the ratio of the negative ideal distance to the sum of the positive and negative ideal distances as the comprehensive performance score of the working condition.

[0030] Specifically, after obtaining the set of feasible operating conditions, it is necessary to solve the multi-objective trade-off problem of conflicts and constraints among different physical indicators. By extracting the data discrimination features of the response indicators in the set of feasible operating conditions and the sensitivity information of the system, the combined weights of the two are used to conduct a comprehensive performance evaluation of each feasible operating condition.

[0031] Among them, the information dispersion of each response index within the feasible operating condition set is used to determine the basic weights and characterize the discrimination ability of each index at the data level; the relative sensitivity coefficient of each response index to changes in scheduling parameters is extracted through system main effect variance analysis to characterize the response intensity of each index under human control.

[0032] By deeply integrating information dispersion and response sensitivity, the final fusion weights can encompass both the objective statistical characteristics of the dataset itself and deeply embed the controllable characteristics of the physical system itself, thus avoiding the one-sidedness of traditional decision-making methods that are purely data-driven or purely subjective in assigning weights.

[0033] Based on the calculated fusion weights, the weighted Euclidean distance between each feasible operating condition and the theoretical positive and negative ideal solutions is calculated using a multi-criteria decision model, thereby obtaining a comprehensive performance score that characterizes the relative degree to which each operating condition approaches the ideal operating state.

[0034] Step 104: Identify key scheduling parameters based on the response sensitivity contribution; and correct the comprehensive performance score based on the key scheduling parameters and pre-configured preference constraints to obtain a recommended score.

[0035] Specifically, a comprehensive performance score generated purely based on response indicators may present problems such as optimal operating conditions being extremely uneconomical in terms of engineering costs, or contradicting local operation and management experience. For example, some high-scoring operating conditions require drainage gates to maintain high-speed drainage at extremely low water levels, which is physically highly likely to cause disturbance and release of heavy metals from riverbed sediments, resulting in secondary pollution. Therefore, this step introduces a soft constraint mechanism for scheduling preferences based on sensitivity information.

[0036] By evaluating the contribution of each scheduling parameter to the overall variance of the multidimensional system response, key scheduling parameters that play a dominant controlling role in hydrodynamic reconfiguration are identified. Subsequently, in conjunction with pre-configured parameter management constraints, segmented preference penalty terms are established for key parameters such as control water level or pump station operating time. These preference penalty terms are used to deduct and correct the overall performance score, thereby outputting a recommended score that aligns with local management preferences. This correction process adjusts the relative recommended ranking among feasible operating conditions in the form of soft constraints, prioritizing the reduction of the ranking of uneconomical or high-risk operating conditions.

[0037] Step 105: Based on the spatial heterogeneity of the indicators in the multidimensional response index matrix, perform local bottleneck verification and last-place elimination for each working condition, and select the working condition with the highest recommended score from the remaining working conditions that have passed the last-place elimination as the target hydrodynamic reconfiguration scheduling scheme for output.

[0038] Specifically, after preference adjustments, while the top-ranked operating conditions may demonstrate superior performance at the overall river scale or in terms of average comprehensive attributes, they may still harbor local weaknesses, such as complete flow stagnation and high pollutant concentration at certain cross-sections. To ensure the project's bottom line of preventing localized black and odorous water disasters, local weaknesses at the cross-sectional scale are checked for each operating condition based on the spatial heterogeneity of the indicators. By analyzing the fluctuations in response sensitivity along the longitudinal direction of the river, cross-section-dependent indicators with significant spatial heterogeneity are extracted as retained indicators.

[0039] Calculate the local bottleneck index for each operating condition under the retention index. This index is determined by the section with the worst utility deficit. Based on the statistical distribution of the bottleneck indices for all feasible operating conditions, set a dynamic screening threshold. Perform bottom-ranking elimination on operating conditions with the recommended score, removing schemes with unacceptable local bottlenecks. From the remaining operating conditions that simultaneously meet the safety red line, comprehensive performance, management preferences, and local reliability requirements, output the operating condition with the highest recommended score as the target hydrodynamic reconfiguration scheduling scheme.

[0040] After outputting the target hydrodynamic reconfiguration and scheduling scheme, the decision-making information will be extended to the underlying execution mechanism to achieve fully automatic closed-loop control of the physical system.

[0041] The target hydrodynamic reconfiguration scheduling scheme is analyzed, and the target control water level, pump station start-stop timing control signals, and introduced flow control commands corresponding to each hydraulic facility in the target tidal river network are extracted.

[0042] The target water level, the pump station start-stop timing control signal, and the inlet flow control command are sent to the corresponding gate actuators and pump station control systems. Through underlying physical actions such as frequency conversion drive or speed regulation, the inlet flow and drainage control water level of the pump station are precisely adjusted.

[0043] By altering the boundary hydraulic thrust, the spatial distribution pattern of hydrodynamics is reshaped within the real physical river network system, accelerating the dilution and water replacement of pollution plumes, thereby achieving efficient emergency treatment of shock pollution.

[0044] Through the implementation of the above-mentioned joint scheduling of gates and pumps, the target tidal river network can achieve a reduction in the number of days of water exchange and an accelerated attenuation of pollutant concentration peaks after hydrodynamic reconstruction. The specific degree of improvement depends on the initial intensity of the shock pollution load and the control margin of available hydraulic facilities.

[0045] Example 2: In one possible implementation, the feasibility of the candidate gate pump scheduling condition set is screened to obtain a feasible condition set, including the following steps:

[0046] Step 201: Obtain the pre-configured engineering operation boundary as hard constraint conditions;

[0047] Obtaining the pre-configured engineering operation boundaries is achieved by reading the design parameter specifications of the water conservancy facilities and the flood control scheduling procedures corresponding to the target tidal river network from the storage device. In the physical entity of the plain tidal river network, the scheduling of gates and pumps is constrained by physical and operational boundaries. The pre-configured engineering operation boundaries include the upper water level threshold, the lower water level threshold, and the upper limit threshold for continuous operation time of pumping stations.

[0048] For example, during the flood season when shock pollution occurs, the representative section of the target tidal river network has the highest safe water level for flood control. If the hydrodynamic reconstruction process causes the simulated water level of a specific section to exceed the highest safe water level for flood control, it will trigger the risk of overflow.

[0049] For example, drainage gates have minimum safe water level control requirements when opening for drainage. If the simulated water level at a specific cross-section is lower than the minimum safe water level, drainage will cause the bottom sediment to be disturbed by high-speed water flow shear stress, resulting in a secondary release of pollutants. The aforementioned physical boundary constraints and equipment protection constraints together constitute the hard constraints for the execution of scheduling assessments.

[0050] Step 202: Remove the operating conditions that violate any of the hard constraints from the candidate gate pump scheduling operating condition set, and form the feasible operating condition set with the remaining operating conditions.

[0051] The massive set of candidate gate pump scheduling conditions is traversed to extract the extreme state parameters generated by each candidate condition during simulation. These extreme state parameters are then compared item by item with the hard constraints.

[0052] If the simulated water level at a specific cross-section under a candidate gate pump scheduling condition is greater than the maximum safe flood control water level, or less than the minimum safe water level, the condition is determined to be physically infeasible or high-risk. The system will veto any condition that violates any hard constraint, removing it from the candidate queue.

[0053] After a thorough comparison of all candidate operating conditions, those without any violations or breaches of limits are retained and collectively constitute the set of feasible operating conditions for subsequent multi-objective comprehensive evaluation. By establishing hard constraint barriers, the computational domain of multi-dimensional indicator evaluation is confined to the physically feasible boundary, avoiding the waste of subsequent fusion weight calculation resources on ineffective operating states.

[0054] In some optional implementations, the pre-configured engineering operation boundaries can be dynamically configured according to different hydrological boundaries. For non-flood season periods, the maximum safe flood control water level setting is increased; for flood season periods, the maximum safe flood control water level setting is decreased. The system automatically loads matching hard constraints to perform elimination operations by determining the current hydrological time period.

[0055] Example 3: Further explanation of the construction of fusion weights based on the information dispersion of each response index and the response sensitivity to scheduling parameters.

[0056] In one possible implementation, the following steps are included:

[0057] Step 301: The response indicators participating in the comprehensive evaluation and the response indicators participating in the extraction of total sensitivity belong to different indicator sets; wherein, when calculating the total sensitivity, the response indicators of all dimensions are extracted for global evaluation; when constructing the fusion weight for the comprehensive evaluation, the engineering cost dimension indicators obtained by explicit combination of various scheduling parameters are removed from the response indicators to prevent the influence of the scheduling parameters from being repeatedly represented in the comprehensive evaluation and preference correction.

[0058] The multidimensional response indicator matrix contains seven response indicators belonging to four dimensions, specifically including:

[0059] The average flow velocity to reverse flux ratio in the hydrodynamic dimension is the ratio of the cumulative reverse flow volume through the representative cross section during the high tide period to the total flow volume during the simulated period.

[0060] Used to characterize the relative intensity of tidal backflow;

[0061] Ecological flow velocity compliance rate and water exchange days in the water environment dimension;

[0062] Mean ammonia nitrogen concentration in water quality;

[0063] The ratio of gate runtime to pump station energy consumption in the operation scheduling dimension.

[0064] Among them, the ecological flow velocity compliance rate represents the proportion of the cumulative time during which the absolute value of the flow velocity at the cross section reaches the ecological threshold to the total simulation time.

[0065] Water exchange days characterize the time required for a complete water exchange in a river channel;

[0066] The gate running time ratio reflects the gate's utilization efficiency of the available drainage tidal window;

[0067] Pump station energy consumption characterizes the physical energy consumed by the pump station during the simulation period.

[0068] The specific indicator for engineering cost is pump station energy consumption. Pump station energy consumption is calculated directly through time integration from two scheduling parameters: pump station inflow rate and pump station operating time. When extracting the overall sensitivity, all seven response indicators are used to establish a main effect assessment mapping covering the overall water flow, water quality, and energy consumption. When constructing the comprehensive evaluation matrix, pump station energy consumption is excluded, retaining only the remaining six calculated indicators to form the comprehensive evaluation indicator set. Including pump station energy consumption in the multi-criteria comprehensive ranking simultaneously would cause the control effects of pump station inflow rate and pump station operating time on system performance to be calculated multiple times during the evaluation phase and subsequent preference correction phase, resulting in logical bias in parameter weights.

[0069] Step 302: Calculate the information entropy of each of the response indicators within the feasible operating condition set to extract the basic weights.

[0070] Before calculating the information entropy, it is necessary to normalize and map each response index. Obtain the value distribution of each response index within the feasible working condition set remaining after hard constraint screening, and map it to a dimensionless interval of 0 to 1 using the range normalization method (maximum-minimum value normalization) based on the benefit-type attribute or cost-type attribute.

[0071] Since the normalization operation is only performed within a subset of feasible operating conditions, the upper and lower bounds of the normalized data matrix are determined by the feasible operating conditions. In the subsequent calculation of the multi-criteria decision distance, the positive and negative ideal points can be directly set to 1 and 0, respectively.

[0072] Based on the normalized values, the proportion of each working condition under the corresponding index is calculated, and the information entropy is calculated based on the proportion.

[0073] After calculating the information entropy of each indicator, the corresponding information entropy is subtracted from the constant 1, and then divided by the sum of the differences of all indicators to extract the basic weights representing the discrete attributes of the data.

[0074] It should be noted that the entropy weight method and the TOPSIS method are existing technologies and are knowledge known to those skilled in the art, and will not be elaborated upon here.

[0075] Step 303: For each of the response indices, calculate the main effect variance contribution of the response indices to the changes in each scheduling parameter to determine the total sensitivity, and convert the total sensitivity into a relative sensitivity coefficient.

[0076] Specifically, a two-level variance analysis logic is used to process the operating condition data, including:

[0077] For all seven response indicators, the inter-group mean variance of each indicator under different operating levels of a certain scheduling parameter is calculated, and the ratio of this ratio to the total variance of the indicator under all feasible operating conditions is calculated to obtain the main effect variance contribution. This ratio measures the control strength of a single scheduling parameter change on the fluctuation of a specific indicator.

[0078] Extract the main effect variance contribution of all four scheduling parameters corresponding to a certain response index, and sum the extracted contribution values ​​to obtain the total sensitivity of the response index.

[0079] A high overall sensitivity value indicates that the hydrodynamic or water quality indicator is more susceptible to changes in artificial scheduling boundaries. The overall sensitivity of all seven response indicators is extracted for global evaluation in subsequent identification of key scheduling parameters.

[0080] When constructing the fusion weights, only the total sensitivity of the six response indicators participating in the comprehensive evaluation is selected. The ratio of the total sensitivity of a single indicator to the sum of the total sensitivity of the six indicators is calculated to obtain the relative sensitivity coefficient limited to the range of 0 to 1.

[0081] Step 304: Combine the basic weights and the relative sensitivity coefficients to reweight, and obtain the fusion weights of each response index.

[0082] The base weights reflect the information discrimination provided by the differences in the values ​​of various indicators within the feasible working condition dataset, while the relative sensitivity coefficients reflect the response attributes of various indicators within the physical system under the control of external engineering interventions. Reweighting is performed on these two types of parameters, and the output is a fusion weight that simultaneously encompasses the information distribution characteristics of the dataset and the dynamic control characteristics of the system.

[0083] In some alternative implementations, the extraction of the total sensitivity for a specific response index is not limited to using a single main effect variance contribution. In river network topologies where scheduling parameters have coupled influences, the interaction effect variance contribution generated by the coordinated changes of the two scheduling parameters is obtained. This interaction effect variance contribution is then superimposed on the main effect variance contribution to generate an interaction effect enhanced sensitivity index. This enhanced sensitivity index is used to replace the original relative sensitivity coefficient in subsequent reweighting to incorporate the nonlinear synergistic effect between scheduling parameters.

[0084] When a certain response indicator takes almost no change in value among feasible operating conditions, such as when the information entropy is close to 1 and the basic weight is close to 0, even if the sensitivity coefficient of the indicator is high, its final weight is still effectively suppressed after fusion, thus avoiding interference from the ranking results by indicators without discriminative power.

[0085] Example 4: In one possible implementation, the fusion weights of each response index are obtained by reweighting the base weights and the relative sensitivity coefficients, including the following steps:

[0086] Step 401: Multiply the basic weight of each response index by the corresponding relative sensitivity coefficient to obtain an initial product result. The basic weight characterizes the data dispersion of each response index within the feasible operating condition set, reflecting the discriminative ability of static information. The relative sensitivity coefficient characterizes the response intensity of each response index driven by changes in gate pump scheduling parameters, reflecting the controlled attributes of the dynamic physical system.

[0087] The two types of positive attribute parameters are coupled and mapped through multiplication. The extracted initial product result maintains a high value if and only if a response index simultaneously possesses high information dispersion and high scheduling sensitivity. If either attribute value is close to zero, the role of that response index in multi-criteria decision-making is suppressed.

[0088] Wherein, the initial product result w of response index j _init_j =w j 0 *S_hat_j ;

[0089] w j 0 S is the base weight for response indicator j. _hat_j is the relative sensitivity coefficient of response index j.

[0090] Step 402: Normalize the initial product results of all response indicators to obtain the fusion weights.

[0091] The initial product set generated by the multiplication mapping no longer satisfies the mathematical constraint that the total weight sum is 1. To support the subsequent calculation of multi-criteria decision models such as weighted Euclidean distance, this sequence needs to be normalized to first-order linearity. The sum of the initial product results of all response indicators participating in the evaluation is calculated, and the target weights are updated using the ratio of the initial product result of each response indicator to this sum.

[0092] The fusion weight w of response index j j =w _init_j / W _sum ;

[0093] Among them, w _init_j In response to the initial product result of index j, W _sum This is the sum of the initial product results of all response indicators involved in the evaluation.

[0094] Through the above normalization process, the resulting fusion weights for each response index are distributed within a dimensionless range of 0 to 1, and the sum of the fusion weights for all response indices equals 1. This multiplicative reweighting fusion scheme can provide evaluation support with low computational complexity and clear physical meaning in scenarios where the basic weights and relative sensitivity coefficients are relatively uniformly distributed.

[0095] In some optional implementations, if there is noisy data close to the lower limit extreme value in the obtained basic weights or relative sensitivity coefficients, the input parameters smaller than the preset lower limit truncation threshold are replaced with small positive numbers by comparing with the preset lower limit truncation threshold before the multiplication mapping operation is performed, so as to prevent the product result from collapsing due to the extreme small value of a single information source.

[0096] like Figure 2 As shown in Example 5: To overcome the amplification and collapse effects that may be caused by multiplication reweighting when processing extremely small values ​​of a single information source, this example provides an adaptive convex combination fusion scheme based on information discriminability. The remaining steps are the same as in Example 3, and can be referred to the foregoing description, and will not be repeated here. In one possible implementation, it is specifically implemented in the following way:

[0097] Step 501: Calculate the ratio of the sample standard deviation to the sample mean of the basic weight set and the relative sensitivity coefficient set for all response indicators, respectively, to obtain two types of coefficients of variation that characterize the discriminative power of the information.

[0098] In a multi-index evaluation system, the internal dispersion of the weight sequence directly reflects the ability of the corresponding information source to distinguish between different working conditions. The coefficient of variation, as a dimensionless statistic, can eliminate the interference of the data benchmark mean on the evaluation of dispersion.

[0099] Specifically, the basic weight sequence corresponding to the response indicators of all evaluation items is obtained. The sample standard deviation and sample mean of the basic weight sequence are calculated, and the ratio of the sample standard deviation to the sample mean is calculated to obtain the coefficient of variation of the basic weights. Using the same processing logic, the relative sensitivity coefficient sequence corresponding to the response indicators of all evaluation items is obtained. The sample standard deviation and sample mean of the relative sensitivity coefficient sequence are calculated, and the ratio of the two is calculated to obtain the coefficient of variation of the relative sensitivity coefficients. Both types of coefficients of variation obtained are dimensionless positive real numbers.

[0100] Wherein, the coefficient of variation CV of the target set is σ / μ; σ is the sample standard deviation of the target set, and μ is the sample mean of the target set.

[0101] Step 502: Determine the adaptive fusion adjustment parameter based on the proportion of the coefficient of variation corresponding to the relative sensitivity coefficient in the sum of the two types of coefficients of variation.

[0102] Obtain the coefficients of variation of the basic weights and the coefficients of variation of the relative sensitivity coefficients calculated in the previous steps. Add these two types of coefficients of variation to obtain the total coefficient of variation, which characterizes the overall information dispersion. Calculate the ratio of the coefficient of variation of the relative sensitivity coefficient to the total coefficient of variation, and extract this ratio as an adaptive fusion adjustment parameter to control the relative contributions of the two information sources.

[0103] The logic behind this parameter construction is that information sources with stronger discriminative capabilities should be allocated a larger weight in the final fusion decision. When the dispersion of the relative sensitivity coefficient sequence is higher than that of the basic weight sequence, the system determines that the physical response sensitivity can provide stronger condition-discriminating features, and the extracted adaptive fusion adjustment parameter approaches 1. When the dispersion of the basic weight sequence is higher than that of the relative sensitivity coefficient sequence, the system determines that the objective information entropy is more discriminative, and the extracted adaptive fusion adjustment parameter approaches 0.

[0104] Adaptive fusion adjustment parameter β=CV _S / (CV _w +CV _S );

[0105] Among them, CV_S The coefficient of variation (CV) corresponds to the relative sensitivity coefficient. _w This refers to the coefficient of variation corresponding to the base weights. When the sum of the two coefficients of variation is 0, the adaptive fusion adjustment parameter is set to a preset value, such as 0.5.

[0106] In one example, the coefficient of variation (COP) of the relative sensitivity coefficient extracted by the system is 0.8, and the COP of the basic weights is 0.2. Substituting these values ​​into the above calculation relationship, the total COP is calculated to be 1.0. Further calculation of the proportion of the COP of the relative sensitivity coefficient to the total COP yields an adaptive fusion adjustment parameter of 0.8. This result will drive the subsequent linear allocation logic to tilt towards the sensitivity side.

[0107] Step 503: Perform a linear convex combination of the basic weights and the relative sensitivity coefficients according to the adaptive fusion adjustment parameters to obtain the fusion weights that automatically satisfy the normalization constraints.

[0108] The basic adjustment parameter is obtained by subtracting the constant 1 from the adaptive fusion adjustment parameter.

[0109] The first weighting term is obtained by multiplying the basic weight of each response index by the basic adjustment parameter.

[0110] The second weighting term is obtained by multiplying the relative sensitivity coefficient of each response index with the adaptive fusion adjustment parameter.

[0111] The first weighted term and the second weighted term of the corresponding response indicator are added together to obtain the final fusion weight of the corresponding indicator.

[0112] Among them, the relative sensitivity coefficient Perform and normalize.

[0113] ; γ and m are natural numbers;

[0114] Since the sum of the two adjustment parameters involved in the convex combination mapping is always 1, and the two types of original weight parameters have each satisfied the mathematical constraint that the sum of the elements is 1 before being input into the current model, the sum of the fusion weights of the indicators output by this linear convex combination operation is automatically equivalent to 1.

[0115] The fusion weight w of response index j j =(1-β)*w j 0 +β* ;

[0116] Where β is the adaptive fusion adjustment parameter, w j 0 The basic weights for responding to index j, The normalized relative sensitivity coefficient of response index j.

[0117] In some alternative implementations, if the system considers the coupling effect between gate pump scheduling parameters when extracting sensitivity information and calculates the interaction-enhanced sensitivity coefficient, then the relative sensitivity coefficient input sequence in the above computing architecture is replaced as a sequence of interaction-enhanced sensitivity coefficients.

[0118] Accordingly, the corresponding coefficient of variation is recalculated based on the interaction-enhanced sensitivity sequence and the adaptive fusion adjustment parameters are determined, thereby introducing the dynamic distinguishing features under the synergistic influence of multiple parameters into the linear convex combinational logic.

[0119] Example 6: In one possible implementation, key scheduling parameters are identified, and the overall performance score is corrected based on the key scheduling parameters and pre-configured preference constraints to obtain a recommended score, including the following steps:

[0120] Step 601: Based on the global sensitivity contribution of each scheduling parameter to the overall system response in the multidimensional response index matrix, evaluate the overall importance of each scheduling parameter.

[0121] Obtain the main effect variance contribution data of all participating scheduling parameters for each response indicator. Calculate the average main effect variance contribution of a single scheduling parameter for all seven response indicators, and use this average as the overall importance indicator for that scheduling parameter.

[0122] This calculation process transforms the comprehensive control intensity of different scheduling parameters on the entire fluid dynamics system and the water quality evolution environment into quantifiable and comparable numerical objects. The higher the overall importance value of a scheduling parameter, the stronger its global driving force in the hydrodynamic reconfiguration and pollution dispersion process.

[0123] Step 602: The scheduling parameters whose overall importance meets the preset ranking conditions are used as the key scheduling parameters, and normalized preference weights are assigned according to the magnitude of the overall importance; the preset conditions include the following meanings: the overall importance ranking is in the first preset number of positions.

[0124] All scheduling parameters are sorted in descending order of overall importance. Parameters with the highest overall importance or those with an overall importance value greater than a preset extraction threshold are identified as key scheduling parameters. For example, in a plain tidal river network, pump station operating time and changes in control water level typically contribute significantly to the system response variance, and these are extracted as key scheduling parameters. For the extracted set of key scheduling parameters, the proportion of each key scheduling parameter's overall importance value in the total overall importance value of the set is calculated, generating a corresponding normalized preference weight.

[0125] Among them, the normalized preference weight ω of the key scheduling parameter k k =I k / I sum ;

[0126] I k For the overall importance of the critical scheduling parameter k, I sum This is the sum of the overall importance of all critical scheduling parameters.

[0127] Step 603: Use the pre-configured segmented preference penalty function as the pre-configured preference constraint, and input the specific values ​​of the key scheduling parameters under each working condition into the corresponding segmented preference penalty function to obtain an independent penalty value;

[0128] In order to transform implicit management experience in the field of water conservancy project scheduling into explicit digital constraints, the system introduces the pre-configured segmented preference penalty function.

[0129] The segmented preference penalty function is internally divided into an exemption interval and a monotonic penalty interval. The system maps the specific values ​​of the acquired key scheduling parameters to different execution paths.

[0130] In the first execution path, when the specific value of the key scheduling parameter falls within the penalty-free range corresponding to conventional scheduling experience, the output independent penalty value is zero.

[0131] Specifically, for example, if the normal operating time of the target pumping station is set to 8 hours, and hydrodynamic reconfiguration is implemented to cope with sudden pollution, appropriately extending the pumping station's operating time is an acceptable scheduling operation. When the specific value of the pumping station's operating time is ≤10 hours, the system determines that the parameter value falls within the penalty-free range, directly outputs the value 0 as the independent penalty value for the parameter, and does not apply any negative intervention.

[0132] In the second execution path, when the specific value of the key scheduling parameter deviates from the conventional scheduling experience and falls into the monotonic penalty interval, the output independent penalty value increases monotonically with the degree of deviation.

[0133] Specifically, when the pump station's operating time exceeds 10 hours, the system determines that prolonged operation of the pump station will increase system energy consumption and result in uneconomical operation. Within this monotonic penalty interval, the farther the parameter value is from the boundary of the penalty-free interval, the larger the calculated independent penalty value.

[0134] Furthermore, the piecewise preference penalty function satisfies a continuity constraint at the connection points of adjacent intervals, meaning that the left-hand limit value is equal to the right-hand limit value at the boundary node. This structural design prevents the penalty value from experiencing numerical jumps at the piecewise critical points, thus avoiding oscillations in the optimization model.

[0135] Step 604: Use the normalized preference weights to perform a weighted summation of the independent penalty values ​​of each key scheduling parameter to generate a total preference penalty term;

[0136] Because different key scheduling parameters have varying degrees of impact on the hydrodynamic system, the system does not employ equal-weight aggregation. Instead, it uses the normalized preference weights obtained in step 602 for differentiated weighting. The independent penalty value of each key scheduling parameter under a specific operating condition is calculated as the product of its corresponding normalized preference weight. The products of all key scheduling parameters are then summed to obtain the total preference penalty term for that operating condition. The corresponding linear calculation relationship is as follows:

[0137] P i =∑(ω k *φ ik );

[0138] Among them, P i Let ω be the total preference penalty term for condition i, ∑ denotes the summation operation, and ω be the total preference penalty term for condition i. k φ is the normalized preference weight for the key scheduling parameter k. ik This is the independent penalty value corresponding to the key scheduling parameter k under this working condition i.

[0139] like Figure 3 As shown, in step 605, the overall performance score is deducted using the total preference penalty item to obtain the recommendation score.

[0140] Obtain the overall performance score output by the basic computing module.

[0141] A pre-configured correction coefficient is introduced to adjust the intensity of intervention by management preferences in the final decision.

[0142] Calculate the product of the total preference penalty term and the correction coefficient, and subtract the product from the comprehensive performance score to update the final recommendation score for this work condition.

[0143] The value of the correction coefficient can be adaptively configured by those skilled in the art based on the strength of management preferences for the specific river network. In this embodiment, the correction coefficient can be set to 0.1, so that the reduction in the overall performance score by the total preference penalty is controlled within a reasonable range.

[0144] R i =C i -λ*P i ;

[0145] Among them, R i C is the recommended score for operating condition i. i Let λ be the overall performance score for operating condition i, and P be the pre-configured correction coefficient. iThis is the total preference penalty term for condition i.

[0146] The correction coefficient λ must satisfy the following constraint: λ max (P i )≤C max -C min ; where C max and C min These are the maximum and minimum values ​​of the overall performance score, respectively, to ensure that preference adjustments do not reverse the original ranking.

[0147] Through the above computational logic, the system, without altering the initial physical feasibility boundaries, uses soft constraint penalties to lower the ranking of operating conditions with excessively high energy consumption or extreme scheduling behavior. This ensures that the system's final recommended output not only performs well in multi-dimensional water quality and hydrodynamic response indicators but also complies with the daily operating procedures of the field equipment.

[0148] Combination Figure 4 Description of Example 7: In one possible implementation, the local bottleneck verification and bottom-ranking elimination for each working condition based on the spatial heterogeneity of the indicators is specifically divided into a spatial heterogeneity indicator extraction stage and a local bottleneck elimination stage, including the following steps:

[0149] Step 701: Calculate the contribution of each of the key scheduling parameters to the cross-sectional scale variance of the response index on each representative cross-section in the multidimensional response index matrix.

[0150] In the physical entity of tidal river networks, the intensity of the boundary control effect of gates and pumps varies spatially across different river sections. To quantify this spatial distribution characteristic, the evaluation scale of the main effects analysis of variance needs to be scaled down to the cross-sectional level.

[0151] Obtain the spatial distribution sequence of specific response indicators across representative cross sections. For selected representative cross sections, use an analysis of variance algorithm to calculate the ratio of the variance variation of the indicator caused by different operational levels of a single key scheduling parameter to the total variance of the indicator at that cross section. This ratio output is the cross section-scale variance contribution, characterizing the statistical dependence of the local cross section's hydrodynamics and water quality environment on the control of this scheduling parameter.

[0152] Step 702: Evaluate the dispersion of the cross-sectional scale variance contribution along each representative cross-section to obtain the spatial variation coefficient;

[0153] Obtain the set of cross-sectional scale variance contributions from each representative cross-section. Calculate the sample standard deviation and sample mean of this set. Divide the sample standard deviation by the sample mean to perform dimensionless processing. To prevent system division-by-zero errors caused by the mean of variance contributions of some parameters approaching 0, a preset small positive number is added to the sample mean as the denominator, and finally, the spatial variation coefficient is calculated and output.

[0154] CV _space_jk =σ _s / (μ _s +ε);

[0155] Among them, CV _space_jk σ is the spatial variation coefficient of the key scheduling parameter k with respect to the response index j. _s The sample standard deviation, μ, of the contribution of cross-sectional scale variance to each representative cross-section. _s The sample mean of the contribution of cross-sectional scale variance to each representative cross-section, where ε is a preset small positive number to prevent division by zero.

[0156] Step 703: Extract the maximum value of the spatial variation coefficient among all the key scheduling parameters, and use it as the spatial heterogeneity index of the corresponding response index.

[0157] Iterate through all key scheduling parameters and extract the set of spatial variation coefficients generated for specific response indicators. Obtain the maximum value in the set of spatial variation coefficients and assign this maximum value as the spatial heterogeneity index of the corresponding response indicator.

[0158] Among them, the spatial heterogeneity index H of response index j _j =max(CV _space_jk );

[0159] `max` is the operator for extracting the maximum value. (CV) _space_jk Let be the spatial variation coefficient of the key scheduling parameter k with respect to the response index j.

[0160] Step 704: The response index that exceeds the preset variation threshold of the spatial heterogeneity index is retained as the retention index.

[0161] Obtain the pre-configured mutation threshold, which can be set to 1.0. Compare the spatial heterogeneity indices of the calculated response metrics with this mutation threshold.

[0162] When the spatial heterogeneity index > 1.0, it indicates that at least one key scheduling parameter causes the standard deviation of the variance contribution of the response index along the course to exceed its mean. The system determines that the response index exhibits spatial differences on a multiple order of magnitude along the main river channel, possessing the physical risk of evolving into a locally deteriorating node. The system selects and extracts response indices that meet this determination condition as retention indices for subsequent elimination assessment.

[0163] Step 705: Within the set of feasible working conditions retained after the feasibility screening, normalize each of the retained indices to obtain the cross-sectional utility value of each working condition on each of the representative cross sections.

[0164] Obtain the set of operating conditions remaining after prior hard constraint filtering and preference correction. Extract the specific values ​​of these operating conditions for each retained index on all representative cross sections.

[0165] For a single retained index, the extreme value distribution of all working conditions across all representative cross-sections is used as the reference domain, and a benefit-based or cost-based extreme value normalization mapping operation is performed. After processing, the absolute value of the index is transformed into a dimensionless utility characteristic distributed in the interval between 0 and 1, thus obtaining the cross-sectional utility value of each working condition on each of the aforementioned representative cross-sections.

[0166] Step 706: For each of the aforementioned retention indices, identify the minimum cross-sectional utility value of each working condition in all representative cross-sections, and calculate the corresponding maximum utility deficit accordingly; use the maximum utility deficit as a local weakness index to assess the local weakness of the corresponding working condition.

[0167] To prevent the deterioration of local hydraulic conditions from being masked by the high average performance of the overall river network, the system introduces a quantitative verification mechanism for single-point risks.

[0168] For a specific retention index, the system iterates through the sequence of cross-sectional utility values ​​for a given working condition across all representative cross-sections, extracting the minimum value. This minimum value reflects the worst spatial performance node for that working condition within that index dimension. The maximum utility deficit is calculated by subtracting this minimum cross-sectional utility value from a constant of 1. The system defines the extracted maximum utility deficit as the local bottleneck index. The larger the index value, the more unacceptable the risk of stagnant water or high pollution exists in a local river section of the physical system.

[0169] B _i q =1-min(y _isq );

[0170] Among them, B _i q Let y be the local bottleneck index of working condition i under the retention index q, min is the minimum value extraction operator, and y is the local bottleneck index of working condition i under the retention index q. _isq Let be the cross-sectional utility value of working condition i on the representative cross-section s.

[0171] Step 707: Set a dynamic screening threshold based on the statistical distribution of the local short board index for all working conditions; remove working conditions whose local short board index is greater than the dynamic screening threshold, and use the remaining working conditions that have not been removed as the set of remaining working conditions that have passed the local short board verification, thus completing the last-place elimination.

[0172] The system obtains a set of local bottleneck indices calculated for all operating conditions under specific retention indicators. This set of local bottleneck indices is then sorted in ascending order of value, and the statistical distribution points of this sequence are extracted as dynamic screening thresholds. These statistical distribution points can be specifically selected as the upper quartiles. The system compares the local bottleneck index of each operating condition with the dynamic screening threshold one by one. When the local bottleneck index of a certain operating condition exceeds the dynamic screening threshold, a data removal operation is performed. Operating conditions that are not removed under all retention indicators are ultimately verified and eliminated, thus ensuring that the output hydrodynamic reconfiguration scheme meets the minimum safety performance at all locations along the route.

[0173] Example 8: In one possible implementation, the data source for the global initial input is pre-constructed using offline simulation and full factorial design methods, specifically including the following steps:

[0174] Step 801: Determine various gate and pump scheduling parameters that affect the hydrodynamic distribution of the river network. The gate and pump scheduling parameters shall at least cover the control water level, pump station start-up time, pump station running time and pump station inlet flow.

[0175] To acquire topological data of hydraulic engineering projects corresponding to the target tidal river network, physical nodes that can be artificially controlled and have a direct driving effect on hydrodynamic distribution are identified. In plain tidal river networks, the control water level of drainage gates and the operating parameters of water diversion pumping stations are used as core scheduling variables.

[0176] Among them, the control water level determines the available tidal window for drainage between the outer river tidal level and the inner river tidal level; the pump station start-up time controls the phase coordination relationship between the introduced water flow and the tidal cycle; the pump station running time determines the total volume of water introduced into the river network in a single day; and the pump station introduced flow corresponds to the instantaneous input power driving water body renewal.

[0177] The above four types of parameters constitute the set of scheduling parameters that determine the boundary conditions for the spatial reconfiguration of the river network hydrodynamics.

[0178] Step 802: For each type of gate pump scheduling parameter, set discrete parameter levels to characterize the gradient of engineering operation.

[0179] Based on the design and operation specifications of the target water conservancy facilities and daily scheduling experience, physically feasible upper and lower limits for the scheduling parameters of various gate pumps are defined, and discrete parameter levels representing the operational gradient are set within the range of these upper and lower limits.

[0180] For example, taking a representative river system of a plain tidal river network as an example, the controlled water level distribution is set within a preset safe water level range, divided into 6 discrete parameter levels according to an operational gradient of 0.1 meters; the pump station start-up time is set to the morning period every day, divided into 3 discrete parameter levels according to an operational gradient of 1 hour; the pump station operating time is set according to emergency water diversion needs, divided into 15 discrete parameter levels according to an operational gradient of 0.5 hours; and the pump station inflow rate is set according to the number of pumps operating at a time, divided into 3 discrete parameter levels. The specific parameter boundaries and the number of levels are configured independently according to the engineering constraints of different areas.

[0181] Step 803: Using a full factorial experimental design, the discrete parameter levels of all the gate pump scheduling parameters are traversed and cross-combined to generate a pre-constructed set of candidate gate pump scheduling conditions that covers the space of coordinated changes of various parameters.

[0182] The discrete parameter levels of the various gate pump scheduling parameters set above are permuted and combined for calculation. To fully cover the coordinated variation space and nonlinear interaction effects of various scheduling parameters in the physical system, a full factorial experimental design method is adopted to perform cross-combination mapping of all levels without omission. The dataset containing all combination features generated by this operation is the candidate gate pump scheduling condition set.

[0183] The total number of operating conditions in the candidate gate pump scheduling operating condition set is N=L CW *L ST *L OD *L PD ;

[0184] Among them, L CW To control the water level, L ST L is the level number representing the pump station start-up time. OD L is the level number representing the operating time of the pumping station. PD The level number for introducing flow into the pumping station.

[0185] Step 804: Construct a one-dimensional hydrodynamic and water quality coupling model for the target tidal river network. The one-dimensional hydrodynamic and water quality coupling model includes a fluid dynamics control equation module for describing water flow motion and a convection and diffusion module for describing pollutant transport.

[0186] In order to simulate the physical driving effect of various scheduling conditions on the river network state in an offline state, a deterministic numerical simulation base based on a system of partial differential equations is established.

[0187] Obtain the channel topology data, cross-sectional geometric feature dimensions, and hydrological boundary conditions of the target tidal river network, and construct a one-dimensional hydrodynamic-water quality coupling model.

[0188] In the fluid dynamics governing equations module, the one-dimensional Saint-Venant equations (including the continuity equation and the momentum equation) are used to simulate the unsteady flow process in the tidal river network. In the convection-diffusion module, the water flow velocity and cross-sectional area calculated and output by the fluid dynamics governing equations module are used as inputs, and combined with the first-level comprehensive degradation coefficient and diffusion coefficient of water quality, the convection-diffusion process of pollutant concentration in the spatiotemporal dimensions is simulated.

[0189] The deterministic coefficient, Nash efficiency coefficient, and mean relative error are used as quantitative technical indicators for model parameter calibration and accuracy evaluation.

[0190] Step 805: The scheduling parameter combinations corresponding to each condition in the pre-constructed candidate gate pump scheduling condition set are sequentially used as boundary conditions and input into the one-dimensional hydrodynamic and water quality coupling model for batch dynamic simulation, and the spatiotemporal evolution state data of each river channel and representative section during the simulation period are extracted.

[0191] The simulation period is set to cover typical tidal cycles and rainfall-induced pollution events. A cyclic scheduling script is used to sequentially read specific parameter combinations from the candidate gate pump scheduling condition set. The extracted control water level, pump station start-up time, pump station runtime, and pump station inflow are converted into hydraulic boundary drive conditions for this specific simulation batch.

[0192] The one-dimensional hydrodynamic-water quality coupling model is run for numerical integration to solve the problem, and the state variable sequence of the physical system under this condition is output during the simulation period. The time series distributions of instantaneous flow velocity, water level, flow rate, and characteristic pollutant concentration are extracted along each river channel and at key monitoring sections, and stored as the spatiotemporal evolution state data.

[0193] Step 806: Based on the spatiotemporal evolution state data, calculate the evaluation response characteristics of each working condition under the corresponding physical dimension, and aggregate them to generate a pre-acquired multidimensional response index matrix.

[0194] Retrieve the spatiotemporal evolution state data output from batch dynamic simulation. Using a preset numerical integration algorithm, average value statistical algorithm, or duration cumulative comparison algorithm, transform the instantaneous state sequence with high temporal resolution into a single quantitative indicator characterizing the overall performance of the working condition.

[0195] The ecological flow velocity compliance rate is obtained by calculating the cumulative time it takes for the absolute value of the flow velocity to reach a preset benchmark threshold. The total water storage capacity of the river channel is approximated using a trapezoidal formula, and the number of days for water exchange is obtained by comparing it with the average daily net flow volume. Based on this conversion rule, evaluation values ​​for each working condition are generated in the dimensions of hydrodynamics, water quality, water environment, and engineering costs. The data structure containing all feasible working conditions and multi-dimensional response index features is merged to construct a multi-dimensional response index matrix to support online decision-making logic.

[0196] Compared to the traditional single entropy weight method for sorting and selection, this method shows a more balanced performance in terms of the absolute value of the minimum flow velocity at each representative section along the main river channel, thus avoiding extreme water stagnation at specific sections.

[0197] Example 9: Unlike Example 3 above, this example provides another method for determining the overall sensitivity for each response index. The specific difference lies in the introduction of the variance contribution of the interaction effects between scheduling parameters. The remaining steps are the same as in Example 3, and can be referred to the foregoing description; they will not be repeated here.

[0198] In the physical operation of the tidal river network, there is a nonlinear physical coupling effect between different gate and pump scheduling parameters.

[0199] Specifically, for example, an increase in the flow rate introduced by a pumping station will alter the hydrodynamic gradient of the inland river, thereby interactively affecting the actual drainage efficiency of the controlled water level at the Taopu River sluice gate. If only the main effect variance contribution of a single scheduling parameter is extracted, the control characteristics of the coordinated changes of parameters on the hydrodynamic and water quality responses will be lost.

[0200] To quantify such synergistic effects, the system obtains a multi-parameter cross-mapping matrix for the candidate operating condition set. For each response index, the inter-group interaction variance caused by the coordinated changes of multiple scheduling parameters is calculated. The ratio of the inter-group interaction variance to the total variance of the index is calculated to obtain the interaction effect variance contribution. The main effect variance contributions of each scheduling parameter are summed with the interaction effect variance contribution to generate the overall sensitivity of the enhanced interaction effect.

[0201] S j *=S j main +S j inter ;

[0202] Among them, S j *To enhance the overall sensitivity to the interaction effect of response index j, S j main S is the sum of the main effect variance contributions of each scheduling parameter to the response index j. j inter This is the sum of the variance contributions of the pairwise interaction effects generated by the pairwise combinations of each scheduling parameter, excluding third-order and higher-order interaction effects.

[0203] The interaction effects of all evaluation response indices are used to enhance the overall sensitivity. The ratio of the interaction effects of a single index enhancing the overall sensitivity to the sum of the corresponding values ​​of all indices is calculated to obtain the interaction enhancement relative sensitivity coefficient. This interaction enhancement relative sensitivity coefficient reflects the response tendency of the physical system under the combined effects of single-parameter driving and multi-parameter coupled driving.

[0204] Based on this, when using the adaptive convex combination fusion scheme described in Example 5, the original relative sensitivity coefficient sequence is replaced by the interactively enhanced relative sensitivity coefficient sequence. The ratio of the sample standard deviation to the sample mean of the interactively enhanced relative sensitivity coefficient sequence is calculated to obtain the updated sensitivity variation coefficient.

[0205] The updated sensitivity coefficient of variation and the coefficient of variation of the basic weights are used to redetermine the adaptive fusion adjustment parameters, and the linear convex combination operation is performed using the adaptive fusion adjustment parameters.

[0206] This alternative expands the evaluation dimension of fusion weights in multivariable nonlinear hydrodynamic environments by supplementing the sensitive physical quantities hidden in the parameter coupling.

Claims

1. A method for optimizing hydrodynamic reconfiguration conditions in tidal river networks based on gate pump scheduling response sensitivity analysis, characterized in that, include: Obtain a set of candidate gate pump scheduling conditions for the target tidal river network, and a multi-dimensional response index matrix corresponding to the set of candidate gate pump scheduling conditions; The feasibility of the candidate gate pump scheduling conditions is screened to obtain a feasible set of conditions. A fusion weight is constructed based on the information dispersion of each response indicator and the response sensitivity to scheduling parameters. The fusion weight is then used to comprehensively evaluate each feasible working condition and obtain a comprehensive performance score. Identify key scheduling parameters, and based on these key scheduling parameters and pre-configured preference constraints, correct the overall performance score to obtain a recommended score; Based on the spatial heterogeneity of the indicators, local shortcomings are checked and the last-place elimination is performed for each working condition, and the target hydrodynamic reconfiguration scheduling scheme is output.

2. The method according to claim 1, characterized in that, The feasibility screening of the candidate gate pump scheduling condition set to obtain a feasible condition set includes: Obtain the pre-configured engineering operation boundary as a hard constraint condition; The conditions that violate any of the hard constraints in the candidate gate pump scheduling condition set are removed, and the remaining conditions constitute the feasible condition set.

3. The method according to claim 1, characterized in that, The response indicators in the multidimensional response indicator matrix are divided into cross-sectional scale indicators and channel scale indicators. The response indicators include hydrodynamic, water environment, and water quality dimensions that characterize the state of the physical system, as well as the operation scheduling dimension that characterizes the cost of engineering control.

4. The method according to claim 1, characterized in that, The construction of fusion weights based on the information dispersion of each response index and the response sensitivity to scheduling parameters includes: Calculate the information entropy of each of the aforementioned response indicators within the feasible operating condition set to extract the basic weights; For each of the aforementioned response indicators, the main effect variance contribution of the response indicator to the changes in each scheduling parameter is calculated to determine the total sensitivity, and the total sensitivity is converted into a relative sensitivity coefficient; The fusion weights of each response index are obtained by reweighting the base weights and the relative sensitivity coefficients.

5. The method according to claim 4, characterized in that, The reweighting process, combining the base weights and the relative sensitivity coefficients, to obtain the fusion weights for each response index, specifically includes: The initial product result is obtained by multiplying the basic weight of each response index with the corresponding relative sensitivity coefficient. The initial product results of all response metrics are normalized to obtain the fusion weights.

6. The method according to claim 4, characterized in that, The reweighting of each response index by combining the basic weights and the relative sensitivity coefficients is specifically achieved through the following method: The ratios of the sample standard deviation to the sample mean of the basic weight set and the relative sensitivity coefficient set for all response indicators are calculated respectively to obtain two types of coefficients of variation that characterize the discriminative power of the information. The adaptive fusion adjustment parameter is determined by the proportion of the coefficient of variation corresponding to the relative sensitivity coefficient in the sum of the two types of coefficients of variation. The basic weights and the relative sensitivity coefficients are linearly convexly combined according to the adaptive fusion adjustment parameters to obtain the fusion weights that automatically satisfy the normalization constraints.

7. The method according to claim 1, characterized in that, The process of identifying key scheduling parameters and, based on these parameters and pre-configured preference constraints, correcting the overall performance score to obtain a recommended score includes: The overall importance of each scheduling parameter is evaluated based on its global sensitivity contribution to the overall system response. The scheduling parameters whose overall importance satisfies the preset ranking conditions are used as the key scheduling parameters, and normalized preference weights are assigned according to the magnitude of the overall importance. The pre-configured segmented preference penalty function is used as the pre-configured preference constraint. The specific values ​​of the key scheduling parameters under each working condition are input into the corresponding segmented preference penalty function to obtain an independent penalty value. The independent penalty values ​​of each key scheduling parameter are weighted and summed using the normalized preference weights to generate a total preference penalty term; The overall performance score is deducted using the total preference penalty item to obtain the recommendation score.

8. The method according to claim 7, characterized in that, The pre-configured segmented preference penalty function is internally divided into an exemption interval and a monotonic penalty interval; Wherein, when the specific value of the key scheduling parameter falls into the penalty-free range corresponding to conventional scheduling experience, the output independent penalty value is zero; When the specific value of the key scheduling parameter deviates from the conventional scheduling experience and falls into the monotonic penalty interval, the output independent penalty value increases monotonically with the degree of deviation.

9. The method according to claim 1, characterized in that, The method of checking local shortcomings and eliminating the bottom performers based on the spatial heterogeneity of indicators for each working condition is specifically divided into the spatial heterogeneity indicator extraction stage and the local shortcomings elimination stage. The spatial heterogeneity index extraction stage includes: Calculate the contribution of each of the key scheduling parameters to the cross-sectional scale variance of the response index on each representative cross-section in the multidimensional response index matrix; The spatial variation coefficient is obtained by evaluating the degree of dispersion of the cross-sectional scale variance contribution along each representative cross-section; Extract the maximum value of the spatial variation coefficient among all the key scheduling parameters, and use it as the spatial heterogeneity index of the corresponding response index; The response index that exceeds the preset variation threshold is used as the retention index.

10. The method according to claim 9, characterized in that, The localized weakness elimination phase includes: Within a preset set of working conditions, the retained indices are normalized to obtain the cross-sectional utility value of each working condition on each representative cross-section. For each of the aforementioned retention indices, the minimum cross-sectional utility value for each working condition in all representative cross-sections is identified, and the corresponding maximum utility deficit is calculated accordingly. The maximum utility deficit is used as a local weakness index to assess the local vulnerability of the corresponding working condition; A dynamic screening threshold is set based on the statistical distribution of the local short-board index for all operating conditions. The conditions where the local weakness index is greater than the dynamic screening threshold are eliminated, thus completing the last-place elimination.

Images