Reservoir group multi-target game scheduling model based on space-time collaborative dimensionality reduction and solving method thereof

Abstract

The invention discloses a reservoir group multi-target game scheduling model based on space-time collaborative dimensionality reduction and a solving method thereof, and the method comprises the steps: taking the water level of each reservoir as a decision variable, calculating each interest related party and constraint condition of a reservoir scheduling system for reservoir group scheduling, and constructing a reservoir group multi-target game scheduling model; a long-medium-short term layered nesting mechanism is constructed, and month, day and hour scale reservoir group target water levels are obtained based on a reservoir group multi-target game scheduling model to generate month, day and hour scale dynamic constraint intervals; returning a daily scale based on a deviation feedback mechanism at an hour scale to carry out iterative optimization on the model; reservoir key indexes serve as the basis for judging the importance of reservoirs, high-importance reservoirs are reserved, and low-importance reservoirs are clustered into equivalent virtual reservoirs. The method has the advantages that the problem of dimensionality disaster caused by space-time dual-dimension coupling in large-scale reservoir group multi-target scheduling of a traditional scheduling model is solved, and the calculation complexity is reduced on the premise that the scheduling precision is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir group operation, and particularly to a multi-objective game operation model for reservoir groups based on spatio-temporal collaborative dimensionality reduction and a solution method therefor. Background Art

[0002] Reservoir group operation plays a key role in multiple tasks such as flood control, water supply, irrigation, and power generation. A scientific and reasonable reservoir group operation strategy can not only effectively meet the diversified needs of social and economic development, but also achieve the efficient utilization of resources on the basis of ensuring ecological safety, realize the dynamic balance among multiple objectives, and maximize the comprehensive benefits.

[0003] However, there are still some deficiencies in the existing reservoir group operation methods. First, the current reservoir group operation models mostly adopt a single time scale and are based on traditional mathematical optimization methods such as linear programming and integer programming. When facing complex game situations involving multiple decision-making subjects, diverse decision variables, and complex constraint conditions, with the extension of the operation period and the increase in the number of reservoirs, there are situations where the decision variables increase exponentially and the computational complexity exceeds the scope of conventional computing power, that is, the "curse of dimensionality" problem. Second, insufficient consideration of the cross-time scale collaborative mechanism leads to a lack of dynamic connection between long-term objectives (such as annual power generation plans) and short-term responses (such as real-time flood operation), making it difficult to effectively respond to complex environmental changes; at the same time, the independent operation of models at different time scales causes error accumulation in the transmission of key parameters such as water levels and flows, further exacerbating the problem of model decoupling distortion.

[0004] In summary, the existing methods cannot achieve global optimization of multiple scales, multiple objectives, and multiple constraints while ensuring computational efficiency. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-objective game operation model for reservoir groups based on spatio-temporal collaborative dimensionality reduction and a solution method therefor, so as to solve the foregoing problems existing in the prior art.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0007] A multi-objective game operation model for reservoir groups based on spatio-temporal collaborative dimensionality reduction and a solution method therefor include the following steps.

[0008] S1. Construction of the game operation model: Determine the stakeholders of the reservoir operation system, use the water levels of each reservoir as decision variables, calculate each stakeholder and constraint conditions of the reservoir group operation, so as to construct a multi-objective game operation model for reservoir groups.

[0009] S2. Time-scale decomposition and nesting: Construct a hierarchical and nested mechanism for long-term, medium-term, and short-term. Based on the multi-objective game scheduling model of the reservoir group, obtain the target water levels of the reservoir group at monthly, daily, and hourly scales to generate dynamic constraint intervals at monthly, daily, and hourly scales respectively; and at the hourly scale, return to the daily scale level based on the deviation feedback mechanism to iteratively optimize the model.

[0010] S3. Spatial-scale clustering: Use the key indicators of the reservoir as the basis for judging the importance of the reservoir, retain the reservoirs with high importance, and cluster the reservoirs with low importance into equivalent virtual reservoirs.

[0011] Preferably, step S2 specifically includes the following content.

[0012] S21. Take the monthly-scale water level as the decision variable, and use the multi-objective game scheduling model of the reservoir group to calculate the target water level of the reservoir group at the monthly scale, realize target decomposition, and generate a dynamic constraint interval at the monthly scale.

[0013] S22. Take the daily-scale water level as the decision variable, and use the monthly-scale target water level of the reservoir group as the constraint, and use the multi-objective game scheduling model of the reservoir group to calculate the target water level of the reservoir group at the daily scale, realize target decomposition, and generate a dynamic constraint interval at the daily scale.

[0014] S23. Take the hourly-scale water level as the decision variable, and use the daily-scale target water level of the reservoir group as the constraint, and use the multi-objective game scheduling model of the reservoir group to calculate the target water level of the reservoir group at the hourly scale; based on the deviation feedback mechanism, calculate the deviation between the hourly-scale water level and the daily-scale target water level of the reservoir group. When the deviation is greater than the preset threshold, return to the daily scale level for iterative optimization to achieve fine adjustment within the day.

[0015] Preferably, the iterative optimization at the daily scale level in step S23 is specifically to perform multi-objective game equilibrium solution on the multi-objective game scheduling model, and take the minimum deviation between the daily-scale average water level and the monthly-scale target water level of the reservoir group as the goal. According to the Pareto front, obtain the optimal solution set at the daily scale to achieve fine adjustment within the day.

[0016] Preferably, introduce a real-time penalty mechanism at the hourly scale level, and optimize the multi-objective game scheduling model in real time every 24 hours to achieve fine adjustment within the day based on the daily scale level.

[0017] Preferably, the calculation formula for the dynamic constraint interval at the monthly scale is

[0018]

[0019] Where is the target water level of the reservoir group at the monthly scale; is the water level constraint interval at the monthly scale; ΔZ m is the relaxation threshold at the monthly scale; ΔZ baseis the base slack; are the standard deviation and mean of the predicted error of the inflow for the current month, respectively;

[0020] The calculation formula for the daily-scale dynamic constraint interval is

[0021]

[0022] where is the daily-scale target water level; is the daily-scale water level constraint interval; ΔZ d is the daily-scale relaxation threshold; ΔZ base is the base slack; are the standard deviation and mean of the predicted error of the inflow for the current day, respectively;

[0023] The calculation formula for the hourly-scale dynamic constraint interval is

[0024]

[0025] where is the hourly-scale target water level; is the hourly-scale water level constraint interval; ΔZ t is the daily-scale relaxation threshold; ΔZ base is the base slack; are the standard deviation and mean of the predicted error of the inflow for the current hour, respectively.

[0026] Preferably, step S3 specifically includes the following content

[0027] S31. Set corresponding thresholds for the two key reservoir indicators, the regulation capacity index or ecological sensitivity;

[0028] S32. Calculate the regulation capacity index or ecological sensitivity of each reservoir, and determine whether the regulation capacity index or ecological sensitivity of each reservoir is greater than or equal to the corresponding threshold. If so, determine that reservoir as a high-importance reservoir; otherwise, determine that reservoir as a low-importance reservoir;

[0029] S33. Use spectral clustering or the K-means algorithm to construct a spatial feature matrix for each low-importance reservoir with reservoir location, storage capacity, maximum discharge, and design function as elements, and cluster the spatial feature matrix to cluster the low-importance reservoirs into equivalent virtual reservoirs.

[0030] Preferably, the calculation formulas for the regulation capacity index and ecological sensitivity of the reservoir are

[0031]

[0032] E s = ∑w k × fk (ΔQ)

[0033] Among them, I r is the regulation capacity index of the reservoir; V eff is the effective regulation storage capacity; Q avg is the average runoff of the reservoir within the scheduling period; E s is the ecological sensitivity of the reservoir; w k is the weight of the k-th ecological target; f k (ΔQ) is the ecological response function that quantifies the negative impact degree of the flow change ΔQ on the k-th ecological target.

[0034] Preferably, the constructed spatial feature matrix is

[0035]

[0036] Among them, F i is the spatial feature matrix of the i-th reservoir; Lat and Lon are the longitude and latitude coordinates respectively; is the maximum storage capacity of the i-th reservoir; is the maximum discharge of the i-th reservoir; w ik is the scheduling target weight, representing the weight of the i-th reservoir on the k-th functional target;

[0037] Clustering the spatial feature matrix, it is

[0038]

[0039] Among them, W ij is the similarity between reservoir i and j; F j is the spatial feature matrix of the j-th reservoir; σ takes the median of the feature spacing.

[0040] Preferably, step S3 further includes

[0041] S34. Establish an equivalent storage capacity curve for the equivalent virtual reservoir,

[0042]

[0043] Among them, V agg (h) is the equivalent storage capacity of the equivalent virtual reservoir at the water level height of h; V i (h) is the original storage capacity of the i-th reservoir at the water level height of h; a i is the storage capacity weight of the i-th reservoir allocated according to the regulation capacity; b is the standard deviation correction coefficient for compensating the clustering error; is the average storage capacity of the reservoir; C is the total number of reservoirs.

[0044] Preferably, in step S1, the reservoir operation system includes five stakeholders, namely flood control objective, power generation objective, shipping objective, water supply objective, and ecological objective;

[0045] The constraint conditions include water balance constraint, water level upper and lower limit constraint, reservoir discharge capacity constraint, out - flow discharge constraint, hourly water balance constraint, and flow mutation limit.

[0046] The beneficial effects of the present invention are as follows: 1. The present invention proposes a multi - scale dimensionality reduction method, significantly reducing the computational intensity when facing complex game systems. 2. The present invention dynamically couples the rolling - horizon optimization and hierarchical dimensionality reduction methods to achieve cross - scale multi - objective coordination, providing theoretical support and practical guidance for the efficient allocation of water resources. 3. The present invention clusters reservoirs to reduce the dimension of scheduling calculations, thereby improving the computational efficiency of the entire scheduling process. Description of the Drawings

[0047] Figure 1 is the flowchart of the method in the embodiment of the present invention. Detailed Embodiments

[0048] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0049] Embodiment 1

[0050] In this embodiment, aiming at the defects such as low computational efficiency of the existing model when facing complex game situations, the present invention proposes a three - layer time - nested optimization method at monthly, daily, and hourly scales and a spatial clustering method based on "equivalent virtual reservoirs" for the multi - objective requirements of flood control, power generation, shipping, water supply, and ecological protection. A multi - objective game scheduling model and solution method for reservoir groups based on spatio - temporal coordination dimensionality reduction are proposed, which helps to achieve cross - scale multi - objective coordination and efficient resource allocation of reservoir groups. This method introduces a rolling optimization mechanism on the time scale. Through hierarchical modeling and cross - scale parameter transfer mechanism, long - term problems are decomposed into short - term problems, and gradually optimized by rolling, improving the accuracy and real - time performance of system optimization, and realizing the coordinated optimization of long - term strategic objectives and short - term dynamic responses. In the spatial dimension, multiple reservoirs are classified according to factors such as their functions, geographical locations, and scheduling requirements, and reservoirs belonging to the same class are merged into an "equivalent virtual reservoir". This spatio - temporal coordination dimensionality reduction method can solve the "curse of dimensionality" problem caused by the coupling of spatio - temporal dual dimensions in the multi - objective scheduling of large - scale reservoir groups, and reduce the computational complexity by two orders of magnitude while ensuring the scheduling accuracy. The method specifically includes the following parts.

[0051] I. Construction of the Game Scheduling Model

[0052] Identify the stakeholders of the reservoir operation system, use the water levels of each reservoir as decision variables, calculate the various stakeholders and constraints of the reservoir group operation, and construct a multi-objective game operation model for the reservoir group. Specifically:

[0053] 1.1 Definition of game players: The reservoir operation system includes five stakeholders, namely flood control objective, power generation objective, shipping objective, water supply objective, and ecological objective.

[0054] 1.2 Calculation of game objectives and constraints: Use the water levels of each reservoir as decision variables, calculate the flood control objective, power generation objective, shipping objective, water supply objective, ecological objective and their constraints of the reservoir group operation according to the following formulas, and construct a multi-objective game operation model for the reservoir group.

[0055] The calculation formulas for each objective are as follows:

[0056] ① Power generation objective, represented by the maximum total power generation of the cascade reservoirs in the calculation period,

[0057]

[0058] N i (t) = K i Q i (t)H i (t)

[0059] Among them, T is the total number of calculation periods; n is the number of reservoirs in the cascade group; N i (t) is the output of the i-th reservoir in the t-th period; K i is the output coefficient of the i-th reservoir; Q i (t) is the power generation flow of the i-th reservoir in the t-th period; H i (t) is the average power generation head of the i-th reservoir in the t-th period; Δt is the calculation period step size.

[0060] ② Water supply objective: Represented by the minimum water shortage,

[0061]

[0062] Among them, Q out,i (t) is the outflow of the i-th reservoir in the t-th period; Q eco,i (t) is the required water shipping objective flow of the i-th reservoir in the t-th period; n is the number of cascade reservoir groups; Δt is the calculation period step size.

[0063] ③ Shipping objective: Represented by the maximum navigation guarantee rate,

[0064]

[0065] Among them, Qout,i (t) is the outflow of the i-th reservoir at the t-th time period; Q ship,i (t) is the minimum navigation flow of the i-th reservoir at the t-th time period; #() is an indicator function, which is 1 if the condition in the parentheses is satisfied, otherwise it is 0.

[0066] ④ Flood control objective: expressed by the maximum flood control safety guarantee rate,

[0067]

[0068] Among them, Q out,i (t) is the outflow of the i-th reservoir at the t-th time period; Q saf,i (t) is the safe flood discharge of the i-th reservoir at the t-th time period; #() is an indicator function, which is 1 if the condition in the parentheses is satisfied, otherwise it is 0.

[0069] ⑤ Ecological objective: expressed by the minimum ecological water shortage,

[0070]

[0071] Among them, Q out,i (t) is the outflow of the i-th reservoir at the t-th time period; Q eco,i is the minimum ecological flow of the i-th reservoir.

[0072] The calculation formulas for each constraint condition are as follows:

[0073] ① Water balance constraint:

[0074]

[0075] Among them, and are the storage capacity, inflow and outflow of the i-th reservoir at the t-th time period during the scheduling period, respectively.

[0076] ② Water level (storage capacity) upper and lower limit constraints:

[0077]

[0078] Among them, is the water level in front of the dam of the i-th reservoir at the t-th time period; is the water level constraint interval of the reservoir.

[0079] ③ Reservoir outflow capacity constraint:

[0080]

[0081] Among them, is the maximum outflow capacity at the water level in front of the dam of the i-th reservoir at the t-th time period under the water level.

[0082] ④ Outflow rate constraint:

[0083]

[0084] In the formula, is the lower limit value of the outflow rate of the i-th reservoir; is the outflow rate of the i-th reservoir in the (t-1)-th period during the scheduling period; ΔQ i is the upper limit value of the daily outflow rate variation of the i-th reservoir.

[0085] ⑤ Hourly water balance constraint:

[0086]

[0087] Among them, V h+1 and V h are the reservoir storage capacities at the h-th hour and the (h-1)-th hour respectively; and are the inflow rate and the outflow rate of the reservoir at the h-th hour respectively.

[0088] ⑥ Flow mutation limit:

[0089]

[0090] Among them, and are the outflow rates of the reservoir at the h-th hour and the (h-1)-th hour respectively, ΔQ max is the upper limit of the adjacent hourly flow change rate.

[0091] II. Time scale decomposition and nesting

[0092] Establish a hierarchical nesting mechanism of long (month)-medium (day)-short (hour) periods, obtain the target water levels of the reservoir group at the monthly, daily, and hourly scales based on the multi-objective game scheduling model of the reservoir group, so as to generate the dynamic constraint intervals at the monthly, daily, and hourly scales respectively; and return to the daily scale level based on the deviation feedback mechanism at the hourly scale level to iteratively optimize the model.

[0093] 2.1. Monthly scale: Taking the monthly scale water level as the decision variable, use the multi-objective game scheduling model of the reservoir group to calculate the target water level of the reservoir group at the monthly scale, realize the target decomposition and generate the dynamic constraint interval at the monthly scale. The calculation formula of the dynamic constraint interval at the monthly scale is,

[0094]

[0095] Among them, is the target water level of the reservoir group at the monthly scale; is the water level constraint interval at the monthly scale; ΔZ m is the relaxation threshold at the monthly scale; ΔZbase is the basic slack; are respectively the standard deviation and the mean of the predicted error of the monthly inflow into the reservoir.

[0096] 2.2 Daily scale: Taking the daily scale water level as the decision variable and the monthly scale target water levels of the reservoir group as the constraints, the multi-objective game scheduling model of the reservoir group is used to calculate the daily scale target water levels of the reservoir group, realizing the objective decomposition and generating the daily scale dynamic constraint interval. The calculation formula of the daily scale dynamic constraint interval is,

[0097]

[0098] where, is the daily scale target water level; is the daily scale water level constraint interval; ΔZ d is the daily scale relaxation threshold; ΔZ base is the basic slack; are respectively the standard deviation and the mean of the predicted error of the daily inflow into the reservoir.

[0099] 2.3 Hourly scale: Taking the hourly scale water level as the decision variable and the daily scale target water levels of the reservoir group as the constraints, the multi-objective game scheduling model of the reservoir group is used to calculate the hourly scale target water levels of the reservoir group. The calculation formula of the hourly scale dynamic constraint interval is,

[0100]

[0101]

[0102] where, is the hourly scale target water level; is the hourly scale water level constraint interval; ΔZ t is the daily scale relaxation threshold; ΔZ base is the basic slack; are respectively the standard deviation and the mean of the predicted error of the current hourly inflow into the reservoir.

[0103] At the hourly scale level, based on the deviation feedback mechanism, the deviation between the hourly scale water level and the daily scale target water levels of the reservoir group is calculated. When the deviation is greater than the preset threshold, it returns to the daily scale level for iterative optimization to achieve refined adjustment within the day.

[0104] The specific process is as follows: Perform multi-objective game equilibrium solution on the multi-objective game scheduling model (the objective functions include flood control objective, power generation objective, shipping objective, water supply objective, ecological objective), and take the minimum deviation between the daily scale average water level and the monthly scale target water levels of the reservoir group as the objective. According to the Pareto front, obtain the optimal solution set at the daily scale to achieve refined adjustment within the day.

[0105] The formula aiming at minimizing the deviation between the daily-scale average water level and the monthly-scale target water level of the reservoir group is expressed as

[0106]

[0107] where is the water level of the reservoir group on the t-th day at the daily scale, is the target water level of the reservoir group at the monthly scale.

[0108] At the hourly scale, a real-time penalty mechanism also needs to be introduced. The real-time optimization model is updated every 24h, and fine-grained intraday adjustment is carried out based on the medium-term plan to optimize the objective function of the real-time optimization model.

[0109] III. Spatial Scale Clustering

[0110] Take the key indicators of the reservoir as the basis for judging the importance of the reservoir, retain the reservoirs with high importance, and cluster the reservoirs with low importance into equivalent virtual reservoirs.

[0111] 3.1. Set corresponding thresholds for the two key indicators of the reservoir, namely the regulation capacity index or the ecological sensitivity.

[0112] 3.2. Calculate the regulation capacity index or ecological sensitivity of each reservoir, and judge whether the regulation capacity index or ecological sensitivity of each reservoir is greater than or equal to the corresponding threshold. If so, the reservoir is judged as a reservoir with high importance and retained; otherwise, the reservoir is judged as a reservoir with low importance. The calculation formulas for the regulation capacity index and ecological sensitivity are as follows

[0113]

[0114] E s =∑w k ×f k (ΔQ)

[0115] where I r is the regulation capacity index of the reservoir; V eff is the effective regulation storage capacity; Q avg is the average runoff of the reservoir during the scheduling period; E s is the ecological sensitivity of the reservoir; w k is the weight of the k-th ecological target; f k (ΔQ) is the ecological response function that quantifies the negative impact degree of the flow change ΔQ on the k-th ecological target.

[0116] 3.3. Adopt spectral clustering or K-means algorithm to construct a spatial feature matrix for each low-importance reservoir with elements such as reservoir location, storage capacity, maximum discharge flow, and design function, and cluster the spatial feature matrix to cluster the low-importance reservoirs into equivalent virtual reservoirs.

[0117] The spatial feature matrix is expressed as

[0118]

[0119] where F i is the spatial feature matrix of the i-th reservoir; Lat and Lon are the longitude and latitude coordinates respectively; is the maximum storage capacity of the i-th reservoir; is the maximum discharge of the i-th reservoir; w ik is the scheduling objective weight, representing the weight of the i-th reservoir on the k-th functional objective.

[0120] The formula for clustering the spatial feature matrix is

[0121]

[0122] where W ij is the similarity between reservoir i and j; F j is the spatial feature matrix of the j-th reservoir; σ takes the median of the feature spacing.

[0123] 3.4. Establish an equivalent storage capacity curve for the equivalent virtual reservoir; the formula is

[0124]

[0125] where V agg (h) is the equivalent storage capacity of the equivalent virtual reservoir at the water level height of h; V i (h) is the original storage capacity of the i-th reservoir at the water level height of h; a i is the storage capacity weight of the i-th reservoir allocated according to the regulation capacity; b is the standard deviation correction coefficient for compensating the clustering error; is the average storage capacity of the reservoir; C is the total number of reservoirs.

[0126] Embodiment 2

[0127] A series of comprehensive water conservancy projects with good regulation capacity and large storage capacity have been planned in the upper reaches of the Yangtze River, including the middle reaches of the Jinsha River reservoir group composed of six cascade reservoirs, the lower reaches of the Jinsha River reservoir group, the Yalong River reservoir group, the Min River reservoir group, the Jialing River reservoir group, the Wu River reservoir group, and the main stream Three Gorges - Gezhouba cascade reservoir group, with a regulated storage capacity of 110 billion m 3。In this embodiment, taking the cascade reservoir group in the upper reaches of the Yangtze River as an example, it includes a cascade system composed of 12 reservoirs, namely Liyuan (R1), Ahai (R2), Jin'anqiao (R3), Longkaikou (R4), Ludila (R5), Guanyinyan (R6), Wudongde (R7), Baihetan (R8), Xiluodu (R9), Xiangjiaba (R10), Tingzikou (R11), and Three Gorges (R12). In this embodiment, the method provided in Embodiment 1 is adopted to optimize the dispatching strategies of the above-mentioned reservoir group during the period from June to August 2023, specifically including the following contents,

[0128] I. Construction of the game dispatching model

[0129] 1.1 Definition of game players: The dispatching objectives of the cascade reservoir group in the upper reaches of the Yangtze River include 5 game objectives, namely flood control objective, power generation objective, shipping objective, water supply objective, and ecological objective.

[0130] 1.2 Calculation of game objectives and constraint conditions: Taking the water levels of 12 reservoirs as decision variables, calculate the flood control objective, power generation objective, shipping objective, water supply objective, ecological objective and their constraint conditions of the reservoir group dispatching according to the following formulas, and construct the overall framework of the multi-objective game dispatching model of the reservoir group. The calculation formulas of relevant objectives and constraint conditions can be found in Embodiment 1.

[0131] II. Time scale decomposition and nesting

[0132] On the long-term (monthly) scale, goal decomposition and constraint generation need to be realized. The monthly-scale target water levels are output by the model, and dynamic constraint intervals are generated.

[0133] Then, taking the long-term output target water level as a constraint, the daily-scale target water levels are output by the model, and dynamic constraint intervals are generated.

[0134] Finally, taking the medium-term (daily) output target water level as a constraint, the hourly-scale target water levels are output by the model, and dynamic constraint intervals are generated

[0135] On the short-term (hourly scale) level, a deviation feedback mechanism is introduced to calculate the deviation between the hourly-scale water level and the daily-scale target water level of the reservoir group. When the deviation is greater than the preset threshold, return to the daily-scale level for iterative optimization to achieve refined intra-day adjustment. The specific optimization process is as follows: Through multi-objective game equilibrium solution (the multi-objectives include power generation objective, water supply objective, shipping objective, and flood control objective), with the minimum deviation between the daily-scale average water level and the monthly-scale target water level of the reservoir group as the goal, according to the Pareto front, the optimal solution set on the medium-term scale is obtained.

[0136] A real-time penalty mechanism is introduced at the short-term level. The real-time optimization model is updated every 24h, and refined intra-day adjustment is carried out based on the medium-term plan to optimize the objective function of the real-time optimization model.

[0137] III. Spatial Scale Clustering

[0138] Calculate the regulation capacity index or ecological sensitivity of the key reservoir indicators, and determine the reservoirs with high regulation capacity index or high ecological sensitivity as high-importance reservoirs and retain them. In this embodiment, taking the regulation capacity judgment as an example, according to experience, the reservoirs with high regulation capacity (I r > 0.8) are high-importance reservoirs and need to be optimized independently, while the low-importance reservoirs can be simplified through clustering.

[0139] Taking the Three Gorges Reservoir as an example, its I r value is 1.05, and the I r values of the other reservoirs are 0.78, 0.69, 0.91, 0.86, 0.8, 0.72, 1.23, 1.76, 1.45, 0.92, 0.78, 1.05 respectively. Then there are a total of 8 high-importance reservoirs and 4 low-importance reservoirs.

[0140] Cluster the 4 low-importance reservoirs. First, construct the spatial feature vector. In actual calculation, the parameters are uniformly standardized. Specifically, the longitude and latitude are normalized to the interval [0, 1], the reservoir capacity is divided by 109m 3 , and the discharge is divided by 5000m 3 / s, and the standardized vector is obtained: F1 = [0.1, 0.34, 0.8, 1.2, 0.5], F2 = [0.13, 0.35, 0.6, 0.9, 0.7], F5 = [0.12, 0.34, 0.7, 1.3], F6 = [0.14, 0.36, 0.6, 1.2]; the clustering results are C1 = {R1, R2, R3}, C2 = {R4, R5, R6}, and a total of 2 equivalent virtual reservoirs are clustered. Then establish the equivalent reservoir capacity curve for the equivalent virtual reservoirs.

[0141] In this embodiment, the traditional method and the method of the present invention are respectively used for comparative evaluation. From the variable dimension, the number of variables required for calculating 12 reservoirs by the traditional method is 12×92×24 = 26496, and after the time-space collaborative dimensionality reduction by the method of the present invention, the number of variables is (5×30 + 3×5 + 4×1)×24 = 4056, and the variables are reduced by about 85%. In terms of calculation time, it takes 147 minutes for the traditional method to run the model once, and the method of the present invention only takes 24 minutes, and the efficiency is increased by 84%. In terms of overall benefits, the benefits of each scheduling objective of the method of the present invention are the same as those of the traditional method.

[0142] By adopting the above technical solutions disclosed by the present invention, the following beneficial effects are obtained:

[0143] The present invention provides a multi-objective game scheduling model for reservoir groups based on spatio-temporal collaborative dimensionality reduction and a solution method thereof. The present invention proposes a dimensionality reduction method at different scales, significantly reducing the computational intensity when facing a complex game system. The present invention dynamically couples the rolling horizon optimization and the hierarchical dimensionality reduction method to achieve cross-scale multi-objective collaboration, providing theoretical support and practical guidance for the efficient allocation of water resources. The present invention reduces the dimension of the scheduling calculation by clustering the reservoirs, thereby improving the computational efficiency of the entire scheduling process.

[0144] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art of this technology, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A multi-objective game scheduling model and its solution method for reservoir groups based on spatio-temporal collaborative dimensionality reduction, characterized in that: It includes the following steps: S1. Construction of the game scheduling model: Determine the stakeholders of the reservoir scheduling system, use the water levels of each reservoir as decision variables, calculate each stakeholder and constraint conditions of the reservoir group scheduling, so as to construct a multi-objective game scheduling model for the reservoir group; S2. Time-scale decomposition and nesting: Construct a hierarchical nesting mechanism of long-term, medium-term and short-term. Based on the multi-objective game scheduling model of the reservoir group, obtain the target water levels of the reservoir group at monthly, daily and hourly scales, so as to generate dynamic constraint intervals at monthly, daily and hourly scales respectively; and at the hourly scale level, return to the daily scale level based on the deviation feedback mechanism to iteratively optimize the model; S3. Spatial-scale clustering: Use the key indicators of the reservoir as the basis for judging the importance of the reservoir, retain the reservoirs with high importance, and cluster the reservoirs with low importance into equivalent virtual reservoirs.

2. The multi-objective game scheduling model and its solution method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 1, characterized in that: Step S2 specifically includes the following content: S21. Use the monthly-scale water level as the decision variable, and call the multi-objective game scheduling model of the reservoir group to calculate the target water level of the reservoir group at the monthly scale, realize target decomposition and generate a dynamic constraint interval at the monthly scale; S22. Use the daily-scale water level as the decision variable, and use the target water level of the reservoir group at the monthly scale as the constraint, and call the multi-objective game scheduling model of the reservoir group to calculate the target water level of the reservoir group at the daily scale, realize target decomposition and generate a dynamic constraint interval at the daily scale; S23. Use the hourly-scale water level as the decision variable, and use the target water level of the reservoir group at the daily scale as the constraint, and call the multi-objective game scheduling model of the reservoir group to calculate the target water level of the reservoir group at the hourly scale; Based on the deviation feedback mechanism, calculate the deviation between the hourly-scale water level and the target water level of the reservoir group at the daily scale. When the deviation is greater than the preset threshold, return to the daily scale level for iterative optimization to achieve fine adjustment within the day.

3. The multi-objective game scheduling model and its solving method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 2, characterized in that: The iterative optimization at the daily scale level in step S23 is specifically to perform multi-objective game equilibrium solution on the multi-objective game scheduling model, and take the minimum deviation between the daily-scale average water level and the target water level of the reservoir group at the monthly scale as the goal, and obtain the optimal solution set at the daily scale according to the Pareto front to achieve fine adjustment within the day.

4. The reservoir group multi-objective game scheduling model and its solution method based on spatio-temporal collaborative dimensionality reduction according to claim 3, characterized in that: Introduce a real-time penalty mechanism at the hourly scale level, optimize the multi-objective game scheduling model in real time every 24 hours, and achieve fine adjustment within the day based on the daily scale level.

5. The multi-objective game scheduling model and its solution method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 2, characterized in that: The calculation formula for the dynamic constraint interval at the monthly scale is Among them, is the target water level of the reservoir group at the monthly scale; is the water level constraint interval at the monthly scale; ΔZ m is the relaxation threshold at the monthly scale; ΔZ base is the basic relaxation amount; are the standard deviation and mean of the prediction error of the monthly inflow respectively; The calculation formula for the dynamic constraint interval at the daily scale is Among them, is the daily-scale target water level; is the daily-scale water level constraint interval; ΔZ d is the daily-scale relaxation threshold; ΔZ base is the basic relaxation amount; are respectively the standard deviation and mean of the predicted error of the daily inflow; The calculation formula for the dynamic constraint interval at the hourly scale is Among them, is the target water level at the hourly scale; is the water level constraint interval at the hourly scale; ΔZ t is the daily scale relaxation threshold; ΔZ base is the basic relaxation amount; are the standard deviation and mean of the current hourly inflow prediction error, respectively.

6. The multi-objective game scheduling model and its solution method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 1, characterized in that: Step S3 specifically includes the following content: S31. Set corresponding thresholds for the two key indicators of the regulation capacity index or ecological sensitivity of the reservoir; S32. Calculate the regulation capacity index or ecological sensitivity of each reservoir, and judge whether the regulation capacity index or ecological sensitivity of each reservoir is greater than or equal to the corresponding threshold. If so, determine the reservoir as a reservoir with high importance, otherwise, determine the reservoir as a reservoir with low importance; S33. Use the spectral clustering or K-means algorithm to construct a spatial feature matrix for each low-importance reservoir with the reservoir location, storage capacity, maximum discharge flow, and design function as elements, and cluster the spatial feature matrix to cluster the low-importance reservoirs into equivalent virtual reservoirs.

7. The multi-objective game scheduling model and its solving method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 1, characterized in that: The calculation formulas for the regulation capacity index and ecological sensitivity of the reservoir are E s = ∑w k × f k (ΔQ) Among them, I r is the regulation capacity index of the reservoir; V eff is the effective regulation storage capacity; Q avg is the average runoff of the reservoir within the scheduling period; E s is the ecological sensitivity of the reservoir; w k is the weight of the k-th ecological target; f k (ΔQ) is the ecological response function that quantifies the negative impact degree of the flow change ΔQ on the k-th ecological target.

8. The multi-objective game scheduling model and its solution method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 1, characterized in that: The constructed spatial feature matrix is Among them, F i is the spatial feature matrix of the i-th reservoir; Lat and Lon are the longitude and latitude coordinates respectively; is the maximum storage capacity of the i-th reservoir; is the maximum discharge of the i-th reservoir; w ik is the scheduling objective weight, representing the weight of the i-th reservoir on the k-th functional objective; Clustering the spatial feature matrix, which is Among them, W ij is the similarity between reservoir i and j; F j is the spatial feature matrix of the j-th reservoir; σ takes the median of the characteristic spacing.

9. The multi-objective game scheduling model and its solving method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 1, characterized in that: Step S3 also includes S34. Establish an equivalent storage capacity curve for the equivalent virtual reservoir Among them, V agg (h) is the equivalent storage capacity of the equivalent virtual reservoir at the water level height of h; V i (h) is the original storage capacity of the i-th reservoir at the water level height of h; a i is the storage capacity weight of the i-th reservoir allocated according to the regulation capacity; b is the standard deviation correction coefficient for compensating the clustering error; is the average storage capacity of the reservoir; C is the total number of reservoirs.

10. The multi-objective game scheduling model and its solving method for reservoir groups based on spatio-temporal collaborative dimensionality reduction according to claim 1, characterized in that: In step S1, the reservoir operation system includes five stakeholders, namely flood control objective, power generation objective, shipping objective, water supply objective, and ecological objective The constraint conditions include water balance constraint, upper and lower water level constraints, reservoir discharge capacity constraint, outflow discharge constraint, hourly water balance constraint, and flow mutation limit

Images