Hydropower enterprise joint replenishment optimization method and system based on collaborative decision

Abstract

This invention discloses a method and system for optimizing joint replenishment for hydropower enterprises based on collaborative decision-making, relating to the field of data processing technology. The method includes: constructing a joint operation data matrix for hydropower enterprises and extracting collaborative state vectors; establishing a set of joint heterogeneous prediction models, analyzing the collaborative state vectors, and obtaining the joint replenishment strategy generated by each model; evaluating the joint replenishment strategies and obtaining the decision-making dominance score for each model; determining the pattern based on the decision-making dominance score, and optimizing the set of joint replenishment strategies according to the pattern determination results to obtain an optimized joint replenishment strategy. This invention addresses the technical problem in existing technologies where multiple hydropower enterprises lack an effective collaborative decision-making mechanism during joint replenishment, leading to low accuracy in replenishment decisions and low overall collaborative efficiency. It achieves the technical effect of improving the collaborativeness of joint replenishment decisions and optimizing overall replenishment efficiency for hydropower enterprises.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for optimizing joint replenishment of hydropower enterprises based on collaborative decision-making. Background Technology

[0002] During the operation of hydropower enterprises, each enterprise typically needs to replenish materials based on equipment operating status, maintenance requirements, and inventory levels to ensure the stable operation of power generation equipment. When multiple hydropower enterprises have joint replenishment needs, replenishment decisions often rely on each enterprise's individual demand forecasting and replenishment plan formulation, lacking a unified collaborative decision-making mechanism. This results in insufficient information utilization among different enterprises, making it difficult to achieve overall coordination of replenishment strategies. In this situation, replenishment decisions are easily influenced by a single data source or a single forecasting method, failing to comprehensively reflect the operating status and material demand changes of each enterprise. Consequently, the accuracy of joint replenishment decisions is insufficient, and the overall efficiency of replenishment coordination needs to be improved. Summary of the Invention

[0003] This application provides a method and system for optimizing joint replenishment of hydropower enterprises based on collaborative decision-making, which is used to address the technical problem that the lack of an effective collaborative decision-making mechanism in the joint replenishment process of multiple hydropower enterprises in the prior art leads to low accuracy of replenishment decisions and low overall collaborative efficiency.

[0004] In view of the above problems, this application provides a method and system for optimizing joint replenishment of hydropower enterprises based on collaborative decision-making.

[0005] The first aspect of this application provides a joint replenishment optimization method for hydropower enterprises based on collaborative decision-making, the method comprising: Operational data from multiple hydropower enterprises are collected to construct a joint operation data matrix. A collaborative state vector is extracted from this matrix. A set of joint heterogeneous prediction models is established, and the collaborative state vector is analyzed to obtain the joint replenishment strategy generated by each model. The joint replenishment strategy is evaluated to obtain a decision-making dominance score for each model, including a risk mitigation contribution index, a collaborative consistency stability index, and an operational state adaptability index. A pattern determination is performed based on the decision-making dominance score, and the set of joint replenishment strategies output by the set of joint heterogeneous prediction models is optimized according to the pattern determination results to obtain an optimized joint replenishment strategy.

[0006] A second aspect of this application provides a joint replenishment optimization system for hydropower enterprises based on collaborative decision-making, the system comprising: The system comprises the following modules: a data matrix construction module for collecting operational data from multiple hydropower enterprises and constructing a joint operation data matrix; a vector extraction module for extracting the joint operation data matrix and extracting collaborative state vectors; a strategy generation module for establishing a set of joint heterogeneous prediction models, analyzing the collaborative state vectors for the set of joint heterogeneous prediction models, and obtaining a joint replenishment strategy generated by each model; a strategy evaluation module for evaluating the joint replenishment strategies and obtaining a decision-making dominance score for each model, which includes a risk mitigation contribution index, a collaborative consistency stability index, and an operational state adaptability index; and a strategy optimization module for performing pattern determination based on the decision-making dominance score, optimizing the set of joint replenishment strategies output by the set of joint heterogeneous prediction models according to the pattern determination results, and obtaining a joint replenishment optimization strategy.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application collects operational data from multiple hydropower enterprises to construct a joint operation data matrix. It then extracts a collaborative state vector from this joint operation data matrix. A set of joint heterogeneous prediction models is established, and the collaborative state vector is analyzed to obtain a joint replenishment strategy generated by each model. The joint replenishment strategy is evaluated to obtain a decision-making dominance score for each model, including a risk mitigation contribution index, a collaborative consistency stability index, and an operational state adaptability index. Based on the decision-making dominance score, a pattern determination is performed, and the set of joint replenishment strategies output by the set of joint heterogeneous prediction models is optimized according to the pattern determination results to obtain an optimized joint replenishment strategy. This invention addresses the technical problem in existing technologies where multiple hydropower companies lack an effective collaborative decision-making mechanism during joint replenishment, resulting in low accuracy of replenishment decisions and low overall collaborative efficiency. By constructing a joint operation data matrix of hydropower companies and extracting collaborative state vectors, and using a set of joint heterogeneous prediction models to generate multiple joint replenishment strategies, these strategies are evaluated and their patterns determined. This collaborative optimization of the joint replenishment strategies improves the collaborativeness of joint replenishment decisions among hydropower companies and optimizes overall replenishment efficiency. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1A schematic diagram of the process for optimizing joint replenishment of hydropower enterprises based on collaborative decision-making, provided in an embodiment of this application; Figure 2 A schematic diagram of the structure of the hydropower enterprise joint replenishment optimization system based on collaborative decision-making provided in this application embodiment.

[0010] Figure labeling: Data matrix construction module 11, vector extraction module 12, policy generation module 13, policy evaluation module 14, policy optimization module 15. Detailed Implementation

[0011] This application provides a method and system for optimizing joint replenishment of hydropower enterprises based on collaborative decision-making. It addresses the technical problem in existing technologies where multiple hydropower enterprises lack an effective collaborative decision-making mechanism during joint replenishment, leading to low accuracy in replenishment decisions and low overall collaborative efficiency. The method constructs a joint operation data matrix of hydropower enterprises and extracts collaborative state vectors. It then uses a set of joint heterogeneous prediction models to generate multiple joint replenishment strategies, evaluates and determines patterns, and collaboratively optimizes these strategies. This achieves the technical effect of improving the collaborativeness of joint replenishment decisions and optimizing overall replenishment efficiency among hydropower enterprises.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, this application provides a joint replenishment optimization method for hydropower enterprises based on collaborative decision-making, the method comprising: Step S100: Collect operational data from multiple hydropower companies and construct a joint operation data matrix for hydropower companies.

[0015] In this embodiment of the application, during the data acquisition phase, the operation data of multiple hydropower companies are uniformly acquired through the hydrological monitoring system deployed by each hydropower company. The operation data includes hydrological data, power generation load data, equipment operation data, maintenance plan data, inventory data, and inter-company material sharing accessibility data. Specifically, the hydrological data is generated by acquiring real-time inflow values, reservoir water levels, and inflow change sequences of the river basins where each hydropower company is located through a hydrological monitoring system; power generation load data is generated by acquiring power output values, unit load rates, and load change sequences of each hydropower unit within the corresponding time period through the enterprise production and operation management system; equipment operation data is generated by acquiring unit operating status identifiers, equipment operating efficiency parameters, and equipment fault records through an equipment operation and maintenance management system; maintenance plan data is generated by acquiring equipment maintenance time intervals, maintenance equipment numbers, and maintenance duration through an equipment operation and maintenance management system; inventory data is generated by acquiring the current inventory quantity, historical inbound and outbound quantity, and material consumption per unit time of various key materials through a materials management system; and inter-enterprise material allocation records are used to acquire transportation distances, transportation times, and available allocation path relationships between different hydropower companies to generate inter-enterprise material sharing accessibility data. This results in a data set that reflects the operating status and material supply status of multiple hydropower companies.

[0016] During the data processing phase, the collected operational data is organized into a unified data structure, enabling the operational data from multiple hydropower companies to be expressed correspondingly within the same time period. Specifically, the inflow rate, water level, and inflow sequence in the hydrological data are recorded at uniform time intervals; the generator output and load rate in the power generation load data are arranged according to the same time period; the equipment operation status identifiers and equipment efficiency parameters in the equipment operation data are grouped according to equipment number; the maintenance time intervals and maintenance equipment numbers in the maintenance plan data are matched with the corresponding time periods; the inventory quantity and material consumption in the inventory data are recorded according to material category; and the transportation distance, transportation time, and dispatch path relationships in the inter-enterprise material sharing accessibility data are organized according to enterprise node pairs, thereby forming a consistent data structure for the operational data of different hydropower companies in both the time and indicator dimensions.

[0017] In the data organization phase, the processed operational data is uniformly organized according to enterprise, operational indicator, and time dimensions. A matrix structure is used to structurally represent the operational data of multiple enterprises, thereby constructing a joint operation data matrix for hydropower enterprises. Specifically, hydropower enterprise nodes are used as the row dimensions of the matrix, operational indicator types as the column dimensions, and corresponding operational data values ​​within each time period as matrix elements. Data such as inflow rate, water level, generator output, unit load rate, equipment operating status indicators, equipment operating efficiency parameters, maintenance time intervals, material inventory quantity, material consumption, and inter-enterprise transportation time and distance are uniformly integrated across multiple hydropower enterprises within the same time period. This forms a joint operation data matrix for hydropower enterprises, representing the overall operational status and material supply status of multiple hydropower enterprises.

[0018] Step S200: Extract the joint operation data matrix of the hydropower enterprises and extract the collaborative state vector.

[0019] In this embodiment, after obtaining the joint operation data matrix of hydropower enterprises, the operation data in the joint operation data matrix is ​​first processed by index reading. The joint operation data matrix of hydropower enterprises records the operation status of multiple hydropower enterprises within a unified time window. Its matrix elements include data such as inflow rate, water level, unit power output, unit load rate, equipment operation status identifier, equipment operation efficiency parameters, maintenance time interval, material inventory quantity, material consumption, and inter-enterprise transportation time and distance. Specifically, according to the enterprise node dimension, the hydropower enterprise joint operation data, power generation load data, equipment operation data, maintenance plan data, inventory data, and inter-enterprise material sharing accessibility data of each hydropower enterprise within the current time window are read one by one from the joint operation data matrix, thereby obtaining the corresponding enterprise's operation index data set.

[0020] After obtaining the operational indicator data sets of each hydropower enterprise, the operational indicator data sets are processed by indicator combination. The data such as inflow rate, water level, generator output, unit load rate, equipment operating status, equipment operating efficiency parameters, maintenance time interval, material inventory quantity, material consumption, and inter-enterprise transportation time and distance within the same time window of the same hydropower enterprise are arranged according to the preset indicator order, thereby forming the corresponding hydropower enterprise's operational status data sequence, so that the operational status of each hydropower enterprise can be expressed through a unified data sequence.

[0021] After obtaining the operational status data sequences of each hydropower enterprise, the operational status data sequences of multiple hydropower enterprises are jointly combined and processed. The operational status data sequences of different hydropower enterprises within the same time window are sequentially spliced ​​according to the enterprise node order to form a data vector that describes the joint operational status of multiple hydropower enterprises. The data vector is used to comprehensively characterize the overall operational relationship between the water situation status, power generation load status, equipment operation status, and material inventory status of multiple hydropower enterprises within the current time window.

[0022] After completing the above data combination processing, the data vector is defined as a collaborative state vector. The collaborative state vector is used to express the unified characteristics of the operating status of multiple hydropower enterprises within the current time window, and to reflect the collaborative operating relationship between different hydropower enterprises in terms of water resource supply, power generation load demand, equipment operating status, and material inventory guarantee.

[0023] Step S300: Establish a set of joint heterogeneous prediction models, analyze the collaborative state vector for the set of joint heterogeneous prediction models, and obtain the joint replenishment strategy generated by each joint heterogeneous prediction model.

[0024] In this embodiment, a set of joint heterogeneous prediction models for joint replenishment decision analysis is first established. This set of models analyzes the joint operation status of multiple hydropower enterprises using prediction models with different decision logics, predicting material demand trends from the perspective of different influencing factors. Specifically, at least three types of prediction models are constructed in the set: operation-driven prediction models, health-driven prediction models, and pattern-driven prediction models. The operation-driven prediction model predicts material demand trends based on the physical operating status of hydropower enterprises; the health-driven prediction model predicts material demand trends based on equipment health indices; and the pattern-driven prediction model predicts material demand trends based on inventory fluctuation patterns.

[0025] After establishing the operation-driven forecasting model, the hydrological data and power generation load data in the collaborative state vector are analyzed as input data for the operation state. The hydrological data includes inflow and water level values, while the power generation load data includes unit output and unit load rate. Based on the relationship between changes in water supply and power generation load, the operation-driven forecasting model predicts trends in unit operating intensity and material consumption demand over future time periods. This yields material demand forecasts based on changes in operation state, and the corresponding replenishment quantities and timing for enterprises are determined based on these forecasts, thus generating a joint replenishment strategy corresponding to the operation-driven forecasting model.

[0026] After establishing a health-driven predictive model, equipment operation data and maintenance plan data from the collaborative state vector are analyzed as equipment status input data. Equipment operation data includes equipment operation status identifiers and equipment operation efficiency parameters, while maintenance plan data includes maintenance time intervals and maintenance equipment numbers. The health-driven predictive model analyzes changes in equipment operation status and maintenance cycles to calculate the trend of equipment health index changes. Based on these changes, it predicts potential equipment maintenance and material consumption needs within future time periods, thus obtaining material demand prediction results based on changes in equipment health status. Based on these prediction results, the corresponding replenishment quantity and replenishment time schedule are determined, thereby generating a joint replenishment strategy corresponding to the health-driven predictive model.

[0027] After establishing the pattern-driven forecasting model, inventory data and material consumption data from the collaborative state vector are used as input data for inventory status analysis. Inventory data includes current inventory levels and historical inventory changes, while material consumption data includes the amount of material consumed per unit time. By analyzing the trends in inventory quantity and material consumption, the pattern-driven forecasting model predicts the inventory decline trend of each hydropower enterprise over a future time period. Based on this, it determines the replenishment demand scale for the future time period, thus obtaining a material demand forecast based on inventory fluctuation patterns. Based on the forecast results, the corresponding replenishment quantity and timing for each enterprise are determined, thereby generating a joint replenishment strategy corresponding to the pattern-driven forecasting model.

[0028] After completing the above analysis, each prediction model in the set of joint heterogeneous prediction models outputs a corresponding joint replenishment strategy. Each joint replenishment strategy describes the material replenishment arrangements of multiple hydropower companies in the future time period, including the replenishment quantity, replenishment time, and material allocation relationship between companies.

[0029] Furthermore, the method provided in the application embodiments also includes: The collaborative state vector is analyzed based on the set of joint heterogeneous prediction models. The set of joint heterogeneous prediction models includes at least an operation-driven prediction model that predicts the trend of material demand changes based on the physical state of hydropower operation, a health-driven prediction model that predicts the trend of material demand changes based on the equipment health index, and a pattern-driven prediction model that predicts the trend of material demand changes based on the inventory fluctuation pattern.

[0030] In this embodiment, a set of joint heterogeneous prediction models is first established for joint replenishment decision analysis. This set analyzes the collaborative operation status of multiple hydropower enterprises using different prediction logics to improve the stability and adaptability of material demand forecasting. The set includes at least three types of prediction models: operation-driven, health-driven, and pattern-driven. The operation-driven model predicts material demand trends based on the physical state of hydropower operation; the health-driven model predicts material demand trends based on equipment health indices; and the pattern-driven model predicts material demand trends based on inventory fluctuation patterns. During the model building phase, a model training dataset is first constructed. Specifically, collaborative state vector sequences corresponding to multiple hydropower enterprises are extracted from historical time periods. These sequences are then matched with the actual material consumption and replenishment quantities recorded within the corresponding time periods, forming a historical collaborative state sample set for model training. This establishes a learnable data correspondence between historical collaborative operation status changes and material demand changes.

[0031] After constructing the training data, the operation-driven prediction model was trained. This model is used to predict the changing trend of material demand based on the physical state of hydropower operation. Specifically, the inflow rate, water level, generator output, and load rate were extracted from the cooperative state vector as input features of the operation state. Normalization was performed on these input features to map the data values ​​of different operation indicators to a unified numerical range, thus eliminating the influence of different units on model training. Subsequently, the material consumption within the future time period was used as the prediction target variable, and a multi-layer feedforward neural network model was trained. This model includes an input layer, two hidden layers, and an output layer. The number of neurons in the hidden layers was set to 64 and 32, respectively, and the ReLU activation function was used. The model parameters were iteratively updated using the backpropagation algorithm, enabling the operation-driven prediction model to learn the mapping relationship between changes in the physical state of hydropower operation and changes in material demand, thereby forming a prediction model for material demand based on the physical state of operation.

[0032] After training the operation-driven prediction model, a health-driven prediction model is trained. This health-driven model is used to predict material demand trends based on equipment health indices. Specifically, equipment operating efficiency parameters, equipment operating time, and equipment operating status identifiers are extracted from the collaborative state vector as equipment status input data. The equipment operating efficiency parameters and equipment operating time are normalized to eliminate the influence of different dimensions on the calculation of the equipment health index. Then, the normalized equipment operating efficiency parameters and normalized equipment operating time are weighted according to preset weights to obtain the equipment health index. The weight of the equipment operating efficiency parameter is set to 0.6, and the weight of the equipment operating time is set to 0.4, ensuring that the equipment health index reflects the degree of equipment performance degradation. The equipment health index sequence is then used as the model input variable, and the corresponding equipment maintenance material consumption within future time periods is used as the prediction target variable. The model is trained through a regression prediction model, enabling the health-driven prediction model to learn the correspondence between changes in equipment health status and changes in material demand, thus forming a prediction model for forecasting material demand based on the changing trends of the equipment health index.

[0033] After training the health-driven prediction model, a pattern-driven prediction model is trained. This model is used to predict trends in material demand based on inventory fluctuation patterns. Specifically, inventory quantity and material consumption per unit time are extracted from the co-state vector, and these are normalized to eliminate the impact of different units on time series forecasting. Subsequently, a continuous time-period inventory change sequence is constructed and used as input data for the model. Future inventory changes are used as the target data for prediction, and the model is trained using a time series forecasting model with an input time window length of 12 periods. This allows the pattern-driven prediction model to learn the relationship between the rate of inventory decline and material replenishment demand, thus forming a prediction model capable of forecasting trends in material demand based on inventory fluctuation patterns.

[0034] After completing the training of the joint heterogeneous prediction model set, the collaborative state vector corresponding to the current time window is input into the operation-driven prediction model, health-driven prediction model, and pattern-driven prediction model in the joint heterogeneous prediction model set for analysis. Each prediction model predicts the trend of material demand changes in the future time period based on the operation rules it has learned, and generates corresponding replenishment decision results based on the predicted material demand quantity, demand time, and inter-enterprise allocation relationship, thereby obtaining multiple joint replenishment strategies generated by the operation-driven prediction model, health-driven prediction model, and pattern-driven prediction model, respectively.

[0035] Step S400: Evaluate the joint replenishment strategy and obtain the decision-leading score for each joint heterogeneous prediction model. The decision-leading score includes a risk mitigation contribution index, a synergistic consistency stability index, and an operational state adaptability index.

[0036] In this embodiment, the joint replenishment strategy is evaluated to obtain the decision-leading score for each joint heterogeneous prediction model. The decision-leading score includes a risk mitigation contribution index, a synergistic consistency stability index, and an operational state adaptability index. Specifically, firstly, a risk mitigation assessment model is introduced for each joint replenishment strategy to evaluate the system risk level before and after the implementation of the replenishment strategy. The risk mitigation contribution index is obtained by calculating the reduction in system risk before and after the implementation of the joint replenishment strategy, where the reduction in system risk is used to characterize the degree of contribution of the joint replenishment strategy in reducing material supply risk. Subsequently, a distribution difference analysis is performed on the joint replenishment strategy and the set of joint replenishment strategies. The synergistic consistency stability index is obtained by calculating the average distribution difference between the joint replenishment strategy and the overall strategy distribution of the set of joint replenishment strategies, where the average distribution difference is used to reflect the synergistic stability of the joint replenishment strategy in the multi-model strategy set. Then, a model adaptation analysis is performed on the joint replenishment strategy and the synergistic state vector corresponding to the current time window. The operational state fit index is obtained by calculating the model fit between the joint replenishment strategy and the synergistic state vector, where the model fit is used to reflect the degree of matching between the joint replenishment strategy and the current synergistic operation state of the hydropower enterprise.

[0037] After obtaining the risk mitigation contribution index, the synergy consistency stability index, and the operational status adaptability index, each index is normalized to eliminate the differences in the dimensions of different indices, and then weighted according to the preset weights to obtain the decision-making dominance score corresponding to each joint heterogeneous prediction model.

[0038] Furthermore, the method provided in the application embodiments also includes: The decision-driven scoring includes a risk mitigation contribution index, a coordination consistency stability index, and an operational state adaptability index. The risk mitigation contribution index is obtained by calculating the reduction in system risk before and after implementing the joint replenishment strategy using a risk mitigation assessment model. The coordination consistency stability index is obtained by calculating the difference in the average distribution between the joint replenishment strategy and the set of joint replenishment strategies. The operational state adaptability index is obtained by calculating the model adaptability with the coordinated state vector.

[0039] In this embodiment, a risk mitigation contribution index is first calculated, which characterizes the contribution of the joint replenishment strategy to reducing the risk of system material supply. Specifically, a risk mitigation assessment model is introduced to quantify the level of material supply risk. This model is trained using historical operating data. During training, data on inventory levels, material consumption, power generation load changes, and equipment maintenance needs of multiple hydropower enterprises are extracted from historical time periods. A risk label dataset is then constructed based on material shortage events that occurred in historical records. Subsequently, inventory coverage time, inventory safety factor, material consumption rate, and inter-enterprise material allocation time are used as input features to the model, and the probability of material shortage occurrence is used as the model output. The model is trained using a logistic regression model, where the logistic regression model parameters are solved using the maximum likelihood estimation method and iteratively updated using the gradient descent algorithm. This enables the risk mitigation assessment model to establish a mapping relationship between operating status and material supply risk. After obtaining the risk mitigation assessment model, the system risk feature vector is constructed from the inventory quantity, material consumption rate, and inter-enterprise material allocation relationship before the joint replenishment strategy is implemented. This vector is then input into the risk mitigation assessment model to obtain the system risk value before replenishment is implemented. Subsequently, the inventory quantity is updated according to the replenishment quantity and replenishment time in the joint replenishment strategy to obtain the inventory quantity after replenishment is implemented. The updated inventory quantity, material consumption rate, and inter-enterprise material allocation relationship are then reconstructed into a system risk feature vector and input into the risk mitigation assessment model to obtain the system risk value after replenishment is implemented.

[0040] After obtaining the risk mitigation contribution index, a synergistic consistency stability index is calculated. This index characterizes the degree of synergistic stability of the joint replenishment strategy across a set of strategies generated by multiple prediction models. Specifically, a joint replenishment strategy set is formed by combining multiple joint replenishment strategies generated by a set of heterogeneous prediction models. Each joint replenishment strategy in the set is then vectorized, with replenishment quantity, replenishment time, and inter-enterprise material allocation relationships arranged according to a unified index order to form a corresponding strategy vector. Next, the Euclidean distance between the target joint replenishment strategy and other strategy vectors in the set is calculated, resulting in a set of strategy distribution distance values. These distance values ​​are then averaged to obtain the average distribution difference value. This average distribution difference value is then normalized to map to a unified numerical range. Finally, the synergistic consistency stability index is obtained by calculating the reciprocal of the normalized average distribution difference value. A smaller average distribution difference value corresponds to a higher degree of synergistic consistency stability, thus completing the acquisition of the synergistic consistency stability index.

[0041] The operational status fit index is obtained by calculating the model fit between the joint replenishment strategy and the current collaborative state vector. The collaborative state vector represents the joint operational status of multiple hydropower enterprises within the current time window and includes the hydrological fluctuation index, power generation load tension index, inventory risk margin index, maintenance concentration index, material sharing accessibility coefficient, and dispatch importance level coefficient. The model feature vector corresponding to each joint heterogeneous prediction model in the set of joint heterogeneous prediction models is obtained. The model feature vector represents the decision preference of the corresponding prediction model for different operational status characteristics during the replenishment decision process. Then, the similarity between the collaborative state vector and the model feature vector is calculated to obtain the operational status fit index of the corresponding joint heterogeneous prediction model.

[0042] Furthermore, in the method provided in the application embodiments, the running state adaptability index is obtained by calculating the model adaptability with the collaborative state vector, and further includes: The collaborative state vector includes the water situation fluctuation index, the power generation load tension index, the inventory risk margin index, the maintenance concentration index, the material sharing accessibility coefficient, and the scheduling importance level coefficient; the model feature vector of each joint heterogeneous prediction model in the set of joint heterogeneous prediction models is obtained to characterize the model decision preference to adapt to the operating state; the similarity between the collaborative state vector and the model feature vector is calculated to obtain the operating state adaptability index.

[0043] In this embodiment, the collaborative state vector includes a hydrological fluctuation index, a power generation load tension index, an inventory risk margin index, a maintenance concentration index, a material sharing accessibility coefficient, and a scheduling importance level coefficient. After the joint operation data matrix of hydropower enterprises is constructed, hydrological data, power generation load data, inventory data, maintenance plan data, and inter-enterprise material sharing accessibility data within the corresponding time window are read from the joint operation data matrix of hydropower enterprises. Based on the read data, index calculations are performed. Specifically, the change in inflow is obtained by differential calculation of the inflow sequence within a continuous time window, and the hydrological fluctuation index is calculated by comparing the change in inflow with the average inflow within the corresponding time window. The power generation output value of the generating unit is read, and the rated installed capacity of the corresponding unit is obtained. The current power generation output value is compared with the rated installed capacity... The power generation load tension index is obtained by calculating the ratio of the unit capacity; the inventory risk margin index is obtained by reading the inventory quantity and obtaining the safety stock threshold recorded in the material management system, and dividing the difference between the current inventory quantity and the safety stock threshold by the safety stock threshold; the maintenance concentration index is obtained by calculating the ratio of the number of equipment under maintenance to the total number of equipment by counting the number of equipment under maintenance in the current time window; and the material sharing accessibility coefficient is obtained by reading the inter-enterprise material transportation time and distance data, and calculating the ratio between time and distance after normalizing the transportation time and distance.

[0044] After calculating the above indicators, the scheduling importance level coefficient is calculated. Specifically, firstly, the power grid assessment weight corresponding to each hydropower enterprise is obtained from the power grid dispatch management system. This power grid assessment weight is a weight coefficient published by the power grid dispatching agency in the annual power station operation evaluation system, used to characterize the importance of each power station in the power grid operation stability evaluation. Then, the online status of each hydropower enterprise's units within the current time window is read from the power station production and operation management system, and the total capacity of units in operation is calculated to obtain the online unit capacity of each power station. Next, the total online capacity of regional hydropower enterprises is obtained from the regional power grid operation database. The online capacity ratio of each power station is calculated by comparing the online unit capacity of each power station with the total online capacity of regional hydropower. Finally, the power supply guarantee management system... The system reads the supply guarantee level corresponding to each power station and converts the supply guarantee level into the corresponding numerical level according to the level mapping rules stipulated in the power supply guarantee management system. Then, it normalizes the power grid assessment weight, the power station online capacity ratio, and the supply guarantee level value, and performs weighted summation calculation according to preset weights. The weight coefficients of the power grid assessment weight, the power station online capacity ratio, and the supply guarantee level are all set to 0.4, thus obtaining the dispatch importance level coefficient. The dispatch importance level coefficient is then arranged in a unified order with the water situation fluctuation index, the power generation load tension index, the inventory risk margin index, the maintenance concentration index, and the material sharing accessibility coefficient to form a collaborative state vector.

[0045] After obtaining the collaborative state vector, the model feature vector corresponding to each collaborative heterogeneous prediction model is obtained from the set of collaborative heterogeneous prediction models. Specifically, after the collaborative heterogeneous prediction model is trained, the model parameter weights corresponding to each input feature during the model training process are read, and the model parameter weights are arranged in the same feature order as the collaborative state vector. The feature order includes the water level fluctuation index, the power generation load tension index, the inventory risk margin index, the maintenance concentration index, the material sharing accessibility coefficient, and the scheduling importance level coefficient, thereby constructing a model feature vector to characterize the degree of attention the model pays to different operating state features during the replenishment decision process.

[0046] After obtaining the collaborative state vector and the model feature vector, a similarity calculation is performed on the collaborative state vector and the model feature vector to obtain the running state fit index. Specifically, firstly, the vector dot product between the collaborative state vector and the model feature vector is calculated, which involves multiplying the corresponding elements of the two vectors and summing the results. Then, the vector magnitudes of the collaborative state vector and the model feature vector are calculated, where the vector magnitude is obtained by summing the squares of each element of the vector and then taking the square root. Next, the dot product result is divided by the product of the two vector magnitudes to obtain the cosine similarity between the collaborative state vector and the model feature vector. Finally, the cosine similarity is used as the running state fit index, where a larger cosine similarity value indicates a higher degree of matching between the decision preferences of the joint heterogeneous prediction model and the current collaborative running state.

[0047] Step S500: Based on the decision-making dominant score, perform pattern determination, and optimize the joint replenishment strategy set output by the joint heterogeneous prediction model set according to the pattern determination result to obtain the joint replenishment optimization strategy.

[0048] In this embodiment, when determining the mode based on the decision-dominant score, the decision-dominant score set corresponding to the set of joint heterogeneous prediction models is first obtained. The decision-dominant score set is used to characterize the comprehensive evaluation results of the joint replenishment strategy generated by different joint heterogeneous prediction models under the current operating state. Then, difference analysis is performed on the decision-dominant score set by calculating the difference between the largest and second largest decision-dominant scores in the set. After that, the difference between the decision-dominant scores is compared with a first preset threshold. When the difference between the decision-dominant scores is greater than or equal to the first preset threshold, the current decision mode is determined to be a strong dominant mode. When the difference between the decision-dominant scores is less than the first preset threshold, the current decision mode is determined to be a collaborative fusion mode, thereby obtaining the corresponding mode determination result.

[0049] Next, the set of joint replenishment strategies output by the set of joint heterogeneous prediction models is optimized according to the mode determination results. In this process, when the mode determination result is a strong dominant mode, the first joint heterogeneous prediction model corresponding to the first dominant decision score is obtained, and the first joint replenishment strategy output by the first joint heterogeneous prediction model is used as a reference strategy to perform targeted optimization on the remaining joint replenishment strategies in the set of joint replenishment strategies, thereby obtaining the optimized joint replenishment strategy. When the mode determination result is a collaborative fusion mode, multiple joint heterogeneous prediction models with dominant decision scores greater than a second preset threshold are selected, and multiple joint replenishment strategies output by these multiple joint heterogeneous prediction models are fused and optimized, thereby obtaining the optimized joint replenishment strategy.

[0050] Furthermore, in the method provided in the application embodiments, the pattern determination based on the decision-driven scoring further includes: Obtain the decision-dominant score set corresponding to the joint heterogeneous prediction model set; analyze the dominant score difference of the decision-dominant score set, and obtain the mode determination result based on the dominant score difference; wherein, when the dominant score difference is greater than or equal to a first preset threshold, the mode determination result is a strong dominant mode, and when the dominant score difference is less than the first preset threshold, the mode determination result is a collaborative fusion mode.

[0051] In this embodiment, the decision-leading score set corresponding to the joint heterogeneous prediction model set is first obtained. Specifically, after the joint replenishment strategy evaluation is completed, each joint heterogeneous prediction model obtains a corresponding decision-leading score. Subsequently, the decision-leading scores corresponding to each prediction model are read from the joint heterogeneous prediction model set and summarized and recorded in the order of prediction model identification numbers to form a decision-leading score set. The decision-leading score set is used to centrally represent the comprehensive decision-making capability of each prediction model in the joint heterogeneous prediction model set under the current collaborative operation state.

[0052] After obtaining the decision-dominant score set, a ranking analysis is performed on the score set to calculate the dominant score difference. Specifically, firstly, the score values ​​in the decision-dominant score set are sorted in descending order to obtain a score sequence arranged from high to low. Then, the highest-scoring first decision-dominant score and the second-highest-scoring second decision-dominant score are read from the score sequence. Next, a difference calculation operation is performed to subtract the second decision-dominant score from the first decision-dominant score to obtain the dominant score difference. The dominant score difference is used to quantify the score advantage of the joint heterogeneous prediction model with the highest score relative to other prediction models.

[0053] After obtaining the dominant score difference, a threshold comparison process is performed on the dominant score difference to obtain the pattern determination result. Specifically, firstly, a first preset threshold is set to distinguish different decision modes. The first preset threshold is obtained statistically based on the historical joint replenishment strategy evaluation results and is used to characterize the threshold boundary when the difference in decision scores reaches a significant advantage. Then, the dominant score difference is numerically compared with the first preset threshold. When the dominant score difference is greater than or equal to the first preset threshold, the joint heterogeneous prediction model with the highest score is determined to have a significant advantage in the current collaborative operation state, and the pattern determination result is determined to be a strong dominant mode. When the dominant score difference is less than the first preset threshold, the difference in decision scores among multiple joint heterogeneous prediction models is determined to be small, and the pattern determination result is determined to be a collaborative fusion mode, thus completing the pattern determination process based on the dominant score difference.

[0054] Furthermore, in the method provided in the application embodiments, optimizing the joint replenishment strategy set output by the joint heterogeneous prediction model set according to the pattern determination result further includes: If the mode determination result is a strong dominant mode, obtain the first joint heterogeneous prediction model corresponding to the first dominant decision score, and use the first joint replenishment strategy output by the first joint heterogeneous prediction model as a reference to perform targeted optimization on the remaining joint replenishment strategies to obtain a joint replenishment optimization strategy; if the mode determination result is a collaborative fusion mode, select multiple joint heterogeneous prediction models with dominant decision scores greater than the second preset threshold, and perform fusion optimization on the multiple joint replenishment strategies output by the multiple joint heterogeneous prediction models to obtain a joint replenishment optimization strategy.

[0055] In this embodiment, when the mode determination result is a strong dominant mode, the decision-dominant scores corresponding to each joint heterogeneous prediction model are first read from the decision-dominant score set, and the decision-dominant score set is sorted in descending order. The highest-scoring first dominant decision score is obtained through the sorting result. Then, based on the model identifier corresponding to the first dominant decision score, the corresponding first joint heterogeneous prediction model is located from the joint heterogeneous prediction model set. Next, the first joint replenishment strategy generated by the first joint heterogeneous prediction model is read. The first joint replenishment strategy describes the material replenishment arrangements of multiple hydropower enterprises in the future time period, including replenishment quantity, replenishment time, and material allocation relationship between enterprises. Then, the first joint replenishment strategy is vectorized. The replenishment quantity, replenishment time, and material allocation relationship between enterprises are arranged in a unified index order to form a reference strategy vector. The remaining joint replenishment strategies generated by the other joint heterogeneous prediction models in the joint replenishment strategy set are vectorized in the same way to form a set of strategy vectors to be optimized.

[0056] After obtaining the sets of reference policy vectors and policy vectors to be optimized, targeted optimization is performed on each policy vector to be optimized. Specifically, firstly, the difference vector between each policy vector to be optimized and the reference policy vector is calculated, where the difference vector is obtained by subtracting the corresponding elements of the two vectors one by one; then, the deviation ratio of each policy dimension is calculated based on the difference vector, where the deviation ratio is obtained by dividing the difference value by the corresponding element value of the reference policy vector; next, the ratio of each policy element in the policy vector to be optimized is adjusted according to the deviation ratio, where when the deviation ratio exceeds a preset deviation threshold, the corresponding elements of the policy vector to be optimized are linearly converged according to the reference policy vector, specifically through formula S. opt =S cur +α(S ref -S cur ) is calculated, where S opt S represents the elements of the optimized policy vector. cur S represents the element of the policy vector to be optimized. refThe reference strategy vector element is represented by α, which represents the convergence adjustment coefficient and ranges from 0.3 to 0.7, thereby enabling the strategy to be optimized to gradually converge to the first joint replenishment strategy. After the adjustment of the strategy elements in each dimension is completed, the optimized strategy vector is re-analyzed into replenishment quantity, replenishment time, and inter-enterprise material allocation relationship, thereby obtaining a unified optimized joint replenishment strategy.

[0057] When the pattern determination result is a collaborative fusion mode, a threshold screening process is first performed on the decision-dominant score set to determine the set of prediction models participating in strategy fusion. Specifically, the decision-dominant score corresponding to each joint heterogeneous prediction model is read one by one from the decision-dominant score set, and the decision-dominant score is compared with a preset second threshold. When the decision-dominant score corresponding to a certain joint heterogeneous prediction model is greater than the second preset threshold, the joint heterogeneous prediction model is added to the candidate fusion model set, and the corresponding model identifier and decision-dominant score are recorded. Through the above comparison process, multiple joint heterogeneous prediction models that meet the conditions are obtained, thus forming a candidate fusion model set for strategy fusion calculation. Then, multiple joint replenishment strategies output by each joint heterogeneous prediction model in the candidate fusion model set are obtained. Specifically, based on the model identifiers in the candidate fusion model set, the strategy output interface of the corresponding joint heterogeneous prediction model is called one by one to read the joint replenishment strategy generated by each model in the current time window. The joint replenishment strategy is used to describe the replenishment decision information of multiple hydropower companies in the future time period, including the replenishment quantity, replenishment time, and material allocation relationship between companies. Then, the multiple joint replenishment strategies are organized according to a unified structure so that each joint replenishment strategy contains the same number of strategy elements, thereby obtaining a set of joint replenishment strategies for subsequent calculations.

[0058] After obtaining the set of joint replenishment strategies, strategy vectorization processing is performed on the set of joint replenishment strategies. Specifically, the replenishment quantity, replenishment time, and inter-enterprise material allocation relationship in each joint replenishment strategy are arranged according to a unified index order, and the values ​​of each index are converted into numerical expressions, thereby converting each joint replenishment strategy into a corresponding strategy vector. For example, the replenishment quantity of each hydropower company is arranged in order of company number to form a replenishment quantity vector, the replenishment time is converted into time code values ​​in order of time period, and the inter-enterprise material allocation relationship is encoded according to enterprise node pairs, so that each joint replenishment strategy is converted into a strategy vector with consistent dimensions, thus forming a set of strategy vectors.

[0059] After obtaining the policy vector set, a fusion weight calculation is performed on the policy vector set. Specifically, the decision-dominant scores corresponding to each joint heterogeneous prediction model in the candidate fusion model set are read, and each decision-dominant score is normalized to map each score value to the numerical range of 0 to 1. Then, the normalized score value is used as the fusion weight coefficient of the corresponding policy vector, so that the prediction model with a higher decision-dominant score has a higher weight contribution in the policy fusion process, thus obtaining a fusion weight set that corresponds one-to-one with the policy vector set. After obtaining the policy vector set and the fusion weight set, a weighted fusion calculation is performed on the policy vector set. For each policy element in the policy vector, a weighted sum is calculated according to the fusion weight coefficient of the corresponding policy vector. The weighted summation process is obtained by multiplying the corresponding element of each policy vector by the corresponding fusion weight and then summing the results, thus obtaining the fusion policy element value corresponding to each dimension. By performing the above weighted summation calculation on all policy elements dimension by dimension, the complete fusion policy vector is obtained.

[0060] After obtaining the fusion strategy vector, a strategy structure restoration process is performed on the fusion strategy vector. Specifically, based on the index order when the strategy vector was constructed, each element in the fusion strategy vector is parsed into the corresponding replenishment quantity, replenishment time, and inter-enterprise material allocation relationship. Then, the replenishment quantity is integerized to meet the constraint that the material replenishment quantity is an integer, and the replenishment time is aligned with the scheduling time window to ensure that the replenishment time meets the power dispatch cycle requirements. Finally, the parsed replenishment quantity, replenishment time, and inter-enterprise material allocation relationship are recombined to form the final joint replenishment optimization strategy.

[0061] Furthermore, in the method provided in the application embodiments, if the mode determination result is a strong dominant mode, it further includes: Collect the number of consecutive dominant periods of the first joint heterogeneous prediction model; if the number of consecutive dominant periods is greater than a preset period threshold, introduce a confidence decay factor to decay the dominant decision score of the first joint heterogeneous prediction model, and obtain the decayed dominant decision score; use the decayed dominant decision score to re-obtain the mode determination result.

[0062] In this embodiment, when the pattern determination result is a strong dominant pattern, to avoid a certain joint heterogeneous prediction model continuously dominating multiple decision cycles and causing path dependence in the replenishment decision results, continuous dominance state detection processing is performed on the first joint heterogeneous prediction model. First, the number of consecutive dominant cycles of the first joint heterogeneous prediction model is collected, where the number of consecutive dominant cycles represents the number of times the first joint heterogeneous prediction model obtains the highest dominant decision score in multiple consecutive replenishment decision cycles. During the collection process, by reading the decision dominance score set in the historical decision cycle records, the joint heterogeneous prediction model with the highest dominant decision score in each decision cycle is identified cycle by cycle, and it is determined whether the first joint heterogeneous prediction model has continuously obtained the highest dominant decision score in multiple adjacent cycles. When the dominant prediction model in the current cycle and the previous cycle are both the first joint heterogeneous prediction model, the continuous dominance count is accumulated. When other joint heterogeneous prediction models have higher scores than the first joint heterogeneous prediction model, the count is reset, thereby obtaining the number of consecutive dominant cycles.

[0063] After obtaining the number of consecutive dominant periods, this number is compared with a preset period threshold, where the period threshold represents the maximum number of decision periods that allow the same joint heterogeneous prediction model to maintain continuous dominance. The period threshold is determined through statistical analysis of historical joint replenishment decision period data, using the average number of consecutive dominant periods of the dominant model in historical periods as a reference. When the number of consecutive dominant periods exceeds the period threshold, a credibility attenuation factor is introduced to attenuate the dominant decision score corresponding to the first joint heterogeneous prediction model. This credibility attenuation factor reduces the scoring advantage of the long-term dominant prediction model in subsequent pattern determinations. The credibility attenuation factor is calculated based on the proportional relationship between the number of consecutive dominant periods and the period threshold, using the formula λ=T / N, where λ represents the credibility attenuation factor, T represents the period threshold, and N represents the number of consecutive dominant periods. The value of the credibility attenuation factor ranges from 0 to 1.

[0064] After obtaining the credibility decay factor, the decayed dominant decision score is obtained by multiplying the dominant decision score corresponding to the first joint heterogeneous prediction model with the credibility decay factor. Then, the decayed dominant decision score replaces the original dominant decision score to reconstruct the decision dominant score set. Following a predetermined pattern determination process, the dominant score difference is recalculated: the updated decision dominant score set is sorted in descending order to obtain new first and second decision dominant scores, and the difference between them is calculated to obtain a new dominant score difference. Finally, the new dominant score difference is compared with a first preset threshold to re-obtain the pattern determination result.

[0065] Furthermore, in the method provided in the application embodiments, after determining the pattern based on the decision-driven scoring, it further includes: Obtain the current joint replenishment optimization strategy based on the current mode determination result; obtain the previous replenishment optimization strategy corresponding to the previous replenishment cycle; if the replenishment fluctuation of the current joint replenishment optimization strategy is greater than the previous replenishment optimization strategy, adjust the current joint replenishment optimization strategy based on the fluctuation penalty factor.

[0066] In this embodiment, the current joint replenishment optimization strategy generated based on the current mode determination result is first obtained. Specifically, after completing the mode determination and optimizing the joint replenishment strategy set output by the joint heterogeneous prediction model set according to the mode determination result, the joint replenishment optimization strategy generated in the current replenishment decision cycle is read, and this joint replenishment optimization strategy is determined as the current joint replenishment optimization strategy.

[0067] Subsequently, the replenishment optimization strategy corresponding to the previous replenishment cycle is retrieved. Specifically, the replenishment decision records saved in the previous replenishment cycle are read from the replenishment strategy history database, and these records are identified as the replenishment optimization strategy for the previous cycle. During the retrieval process, the replenishment quantity, replenishment time, and inter-enterprise material allocation relationships in the previous replenishment optimization strategy are extracted in the same data arrangement order as the current joint replenishment optimization strategy, ensuring that the previous replenishment optimization strategy and the current joint replenishment optimization strategy maintain a consistent data structure across the three dimensions of replenishment quantity, replenishment time, and inter-enterprise material allocation relationships.

[0068] After obtaining the current joint replenishment optimization strategy and the previous cycle's replenishment optimization strategy, replenishment fluctuation calculations are performed on both strategies. Specifically, the replenishment quantity in the current joint replenishment optimization strategy is calculated as the difference between the replenishment quantity in the previous cycle's replenishment optimization strategy, the replenishment time in the current joint replenishment optimization strategy is calculated as the time difference between the replenishment time in the previous cycle's replenishment optimization strategy, and the changes in the material allocation relationship between enterprises are statistically analyzed to obtain the change values ​​between each strategy element. Then, the absolute values ​​of the change values ​​are processed, and all change values ​​are accumulated to obtain the overall strategy change. Finally, the ratio of the overall strategy change to the total replenishment quantity in the current joint replenishment optimization strategy is calculated to obtain the replenishment fluctuation.

[0069] After obtaining the replenishment fluctuation, it is compared with a preset fluctuation threshold, which limits the allowable range of replenishment strategy changes between consecutive replenishment cycles. When the replenishment fluctuation exceeds the fluctuation threshold, a fluctuation penalty factor is introduced to adjust the current joint replenishment optimization strategy. The fluctuation penalty factor is calculated based on the proportional relationship between the fluctuation threshold and the replenishment fluctuation, specifically using the formula β=θ / V, where β represents the fluctuation penalty factor, θ represents the fluctuation threshold, and V represents the replenishment fluctuation. The value of the fluctuation penalty factor is limited to a range of 0 to 1.

[0070] After obtaining the volatility penalty factor, the current joint replenishment optimization strategy is adjusted. Specifically, the replenishment quantity in the current joint replenishment optimization strategy is weighted and calculated with the replenishment quantity in the previous period's replenishment optimization strategy according to the volatility penalty factor, and then expressed using formula Q. adj =βQ cur +(1-β)Q pre The adjusted replenishment quantity is calculated, where Q cur Q represents the replenishment quantity in the current joint replenishment optimization strategy. pre Q represents the replenishment quantity in the previous replenishment optimization strategy. adj This indicates the adjusted replenishment quantity; then, the replenishment time is proportionally adjusted in the same way to converge the current replenishment time to the previous cycle's replenishment time; finally, while keeping the inter-enterprise material allocation relationship structure unchanged, the adjusted replenishment quantity, adjusted replenishment time, and inter-enterprise material allocation relationship are recombined to obtain the adjusted joint replenishment optimization strategy.

[0071] Furthermore, in the method provided in the application embodiments, after obtaining the joint replenishment optimization strategy, it further includes: Establish the hydraulic coupling relationship among the multiple hydropower enterprises; based on the hydraulic coupling relationship, perform replenishment priority scheduling on the obtained joint replenishment optimization strategy.

[0072] In this embodiment, the hydraulic coupling relationship between multiple hydropower enterprises is first established. Specifically, the inflow rate, reservoir water level, and inflow change sequence of each hydropower enterprise's power station are read from the hydrological monitoring system. The upstream and downstream connection relationships of the rivers where each power station is located are obtained from the watershed water resources scheduling database. These upstream and downstream connections represent the order of water flow transmission between different power stations in the river. Then, each power station is numbered according to the river flow direction; for example, the upstream power station is numbered 1, and the downstream power stations are numbered 2, 3, and 4 respectively, thus forming a power station sequence list. After obtaining the power station sequence list, the degree of influence of inflow changes between adjacent power stations is calculated. The hydraulic influence coefficient is obtained by calculating the proportional relationship between the inflow rates of adjacent power stations, specifically using the formula K.ij =Q j / Q i Perform the calculation, where K ij Q represents the hydraulic influence coefficient of power station i on power station j. i Q represents the average inflow value of upstream power station i. j The average inflow rate of downstream power station j is represented by the following: The hydraulic influence coefficients calculated between each power station are then filled into the matrix table according to the power station number. The rows of the matrix represent the upstream power station number, the columns of the matrix represent the downstream power station number, and the matrix elements represent the corresponding hydraulic influence coefficients. This forms a hydraulic coupling relationship matrix to represent the water resource transfer relationship between multiple hydropower enterprises, and the hydraulic coupling relationship matrix is ​​determined as the hydraulic coupling relationship between multiple hydropower enterprises.

[0073] After obtaining the hydraulic coupling relationship matrix, replenishment priority is calculated based on the obtained joint replenishment optimization strategy. Specifically, the replenishment quantity and replenishment time for each hydropower enterprise are first retrieved from the joint replenishment optimization strategy. Then, the hydraulic impact value of each enterprise is calculated based on the hydraulic coupling relationship matrix. The hydraulic impact value is obtained by summing all hydraulic impact coefficients in the row corresponding to the enterprise in the hydraulic coupling relationship matrix. The calculation formula is as follows: ,in K represents the hydraulic impact value of the i-th hydropower enterprise. ij The value represents the hydraulic influence coefficient in the i-th row and j-th column of the hydraulic coupling relationship matrix, and n represents the number of hydropower enterprises. The total hydraulic influence of each hydropower enterprise on downstream enterprises can be obtained by the above summation calculation.

[0074] After obtaining the hydraulic impact values ​​of each hydropower enterprise, a replenishment priority ranking is performed on each hydropower enterprise. Specifically, the hydraulic impact values ​​of all hydropower enterprises are sorted in descending order of numerical value, with hydropower enterprises having larger hydraulic impact values ​​indicating that they have a greater impact on the power generation operation of the basin, and therefore have higher priority in replenishment scheduling; then a replenishment priority sequence is generated based on the ranking results, such as priority first, priority second, and priority third.

[0075] After obtaining the replenishment priority sequence, the joint replenishment optimization strategy is implemented by prioritizing replenishment. Specifically, the replenishment quantity and time corresponding to each hydropower enterprise are read sequentially according to the replenishment priority sequence, and the replenishment execution order is arranged for hydropower enterprises with higher priority. After the replenishment quantity and time of the higher priority enterprise are determined, the replenishment tasks of the next priority enterprise are arranged in turn. After the replenishment order of all enterprises is arranged, the adjusted replenishment quantity, replenishment time, and material allocation relationship between enterprises are recombined to obtain the joint replenishment optimization strategy based on replenishment priority scheduling based on hydraulic coupling relationship.

[0076] In summary, the embodiments of this application have at least the following technical effects: This application collects operational data from multiple hydropower enterprises to construct a joint operation data matrix. It then extracts a collaborative state vector from this joint operation data matrix. A set of joint heterogeneous prediction models is established, and the collaborative state vector is analyzed to obtain a joint replenishment strategy generated by each model. The joint replenishment strategy is evaluated to obtain a decision-making dominance score for each model, including a risk mitigation contribution index, a collaborative consistency stability index, and an operational state adaptability index. Based on the decision-making dominance score, a pattern determination is performed, and the set of joint replenishment strategies output by the set of joint heterogeneous prediction models is optimized according to the pattern determination results to obtain an optimized joint replenishment strategy. This invention addresses the technical problem in existing technologies where multiple hydropower companies lack an effective collaborative decision-making mechanism during joint replenishment, resulting in low accuracy of replenishment decisions and low overall collaborative efficiency. By constructing a joint operation data matrix of hydropower companies and extracting collaborative state vectors, and using a set of joint heterogeneous prediction models to generate multiple joint replenishment strategies, these strategies are evaluated and their patterns determined. This collaborative optimization of the joint replenishment strategies improves the collaborativeness of joint replenishment decisions among hydropower companies and optimizes overall replenishment efficiency.

[0077] Example 2, based on the same inventive concept as the collaborative decision-making-based joint replenishment optimization method for hydropower enterprises in the previous examples, such as... Figure 2 As shown, this application provides a joint replenishment optimization system for hydropower enterprises based on collaborative decision-making. The system and method embodiments in this application are based on the same inventive concept. The system includes: The data matrix construction module 11 is used to collect operational data from multiple hydropower enterprises and construct a joint operation data matrix for hydropower enterprises; the vector extraction module 12 is used to extract the joint operation data matrix of hydropower enterprises and extract the collaborative state vector; the strategy generation module 13 is used to establish a set of joint heterogeneous prediction models, analyze the collaborative state vector for the set of joint heterogeneous prediction models, and obtain the joint replenishment strategy generated by each joint heterogeneous prediction model; the strategy evaluation module 14 is used to evaluate the joint replenishment strategy and obtain the decision-making dominance score of each joint heterogeneous prediction model, the decision-making dominance score including risk mitigation contribution index, collaborative consistency stability index, and operational state adaptability index; the strategy optimization module 15 is used to perform pattern determination based on the decision-making dominance score, and optimize the set of joint replenishment strategies output by the set of joint heterogeneous prediction models according to the pattern determination result to obtain the joint replenishment optimization strategy.

[0078] Furthermore, the system is also used to implement the following functions: The decision-driven scoring includes a risk mitigation contribution index, a coordination consistency stability index, and an operational state adaptability index. The risk mitigation contribution index is obtained by calculating the reduction in system risk before and after implementing the joint replenishment strategy using a risk mitigation assessment model. The coordination consistency stability index is obtained by calculating the difference in the average distribution between the joint replenishment strategy and the set of joint replenishment strategies. The operational state adaptability index is obtained by calculating the model adaptability with the coordinated state vector.

[0079] Furthermore, the system is also used to implement the following functions: The collaborative state vector includes the water situation fluctuation index, the power generation load tension index, the inventory risk margin index, the maintenance concentration index, the material sharing accessibility coefficient, and the scheduling importance level coefficient; the model feature vector of each joint heterogeneous prediction model in the set of joint heterogeneous prediction models is obtained to characterize the model decision preference to adapt to the operating state; the similarity between the collaborative state vector and the model feature vector is calculated to obtain the operating state adaptability index.

[0080] Furthermore, the system is also used to implement the following functions: Obtain the decision-dominant score set corresponding to the joint heterogeneous prediction model set; analyze the dominant score difference of the decision-dominant score set, and obtain the mode determination result based on the dominant score difference; wherein, when the dominant score difference is greater than or equal to a first preset threshold, the mode determination result is a strong dominant mode, and when the dominant score difference is less than the first preset threshold, the mode determination result is a collaborative fusion mode.

[0081] Furthermore, the system is also used to implement the following functions: If the mode determination result is a strong dominant mode, obtain the first joint heterogeneous prediction model corresponding to the first dominant decision score, and use the first joint replenishment strategy output by the first joint heterogeneous prediction model as a reference to perform targeted optimization on the remaining joint replenishment strategies to obtain a joint replenishment optimization strategy; if the mode determination result is a collaborative fusion mode, select multiple joint heterogeneous prediction models with dominant decision scores greater than the second preset threshold, and perform fusion optimization on the multiple joint replenishment strategies output by the multiple joint heterogeneous prediction models to obtain a joint replenishment optimization strategy.

[0082] Furthermore, the system is also used to implement the following functions: Collect the number of consecutive dominant periods of the first joint heterogeneous prediction model; if the number of consecutive dominant periods is greater than a preset period threshold, introduce a confidence decay factor to decay the dominant decision score of the first joint heterogeneous prediction model, and obtain the decayed dominant decision score; use the decayed dominant decision score to re-obtain the mode determination result.

[0083] Furthermore, the system is also used to implement the following functions: The collaborative state vector is analyzed based on the set of joint heterogeneous prediction models. The set of joint heterogeneous prediction models includes at least an operation-driven prediction model that predicts the trend of material demand changes based on the physical state of hydropower operation, a health-driven prediction model that predicts the trend of material demand changes based on the equipment health index, and a pattern-driven prediction model that predicts the trend of material demand changes based on the inventory fluctuation pattern.

[0084] Furthermore, the system is also used to implement the following functions: Obtain the current joint replenishment optimization strategy based on the current mode determination result; obtain the previous replenishment optimization strategy corresponding to the previous replenishment cycle; if the replenishment fluctuation of the current joint replenishment optimization strategy is greater than the previous replenishment optimization strategy, adjust the current joint replenishment optimization strategy based on the fluctuation penalty factor.

[0085] Furthermore, the system is also used to implement the following functions: Establish the hydraulic coupling relationship among the multiple hydropower enterprises; based on the hydraulic coupling relationship, perform replenishment priority scheduling on the obtained joint replenishment optimization strategy.

[0086] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A collaborative decision-making-based optimization method for joint replenishment of hydropower enterprises, characterized in that: The method includes: Collect operational data from multiple hydropower companies and construct a joint operation data matrix for hydropower companies; Extract the joint operation data matrix of the hydropower enterprises and extract the collaborative state vector; Establish a set of joint heterogeneous prediction models, analyze the collaborative state vector for the set of joint heterogeneous prediction models, and obtain the joint replenishment strategy generated by each joint heterogeneous prediction model; The joint replenishment strategy is evaluated to obtain the decision-leading score of each joint heterogeneous prediction model. The decision-leading score includes the risk mitigation contribution index, the synergy consistency stability index, and the operational status adaptability index. Based on the decision-making dominant score, a pattern is determined, and the joint replenishment strategy set output by the joint heterogeneous prediction model set is optimized according to the pattern determination result to obtain the joint replenishment optimization strategy.

2. The method as described in claim 1, characterized in that, The decision-making-led scoring includes risk mitigation contribution indicators, coordination and consistency stability indicators, and operational status adaptability indicators. The risk mitigation contribution index is obtained by calculating the reduction in system risk before and after implementing the joint replenishment strategy through a risk mitigation assessment model. The coordination consistency stability index is obtained by calculating the difference in the average distribution between the joint replenishment strategy and the set of joint replenishment strategies. The operational state adaptability index is obtained by calculating the model adaptability with the coordinated state vector.

3. The method as described in claim 2, characterized in that, The operational state adaptability index is obtained by calculating the model adaptability with the collaborative state vector. include: The collaborative state vector includes the water situation fluctuation index, the power generation load tension index, the inventory risk margin index, the maintenance concentration index, the material sharing accessibility coefficient, and the scheduling importance level coefficient. Obtain the model feature vector for each joint heterogeneous prediction model in the set of joint heterogeneous prediction models, which is used to characterize the model decision preference adaptation to the operating state; The similarity between the collaborative state vector and the model feature vector is calculated to obtain the running state fit index.

4. The method as described in claim 1, characterized in that, Pattern determination is performed based on the aforementioned decision-making dominance score, including the following methods: Obtain the decision-driven score set corresponding to the set of joint heterogeneous prediction models; Analyze the dominant score difference of the decision-making dominant score set, and obtain the determination result based on the dominant score difference; Specifically, when the difference in dominant scores is greater than or equal to a first preset threshold, the mode determination result is a strong dominant mode; when the difference in dominant scores is less than the first preset threshold, the mode determination result is a collaborative fusion mode.

5. The method as described in claim 4, characterized in that, The method for optimizing the set of joint replenishment strategies output by the set of joint heterogeneous prediction models based on the pattern determination results includes: If the mode determination result is a strong dominant mode, obtain the first joint heterogeneous prediction model corresponding to the first dominant decision score, and use the first joint replenishment strategy output by the first joint heterogeneous prediction model as a reference to optimize the remaining joint replenishment strategies in a targeted manner to obtain the joint replenishment optimization strategy. If the mode determination result is a collaborative fusion mode, select multiple joint heterogeneous prediction models with a dominant decision score greater than a second preset threshold, and perform fusion optimization on multiple joint replenishment strategies output by the multiple joint heterogeneous prediction models to obtain a joint replenishment optimization strategy.

6. The method as described in claim 5, characterized in that, If the pattern determination result is a strong dominant pattern, the method further includes: Collect the number of consecutive dominant periods of the first joint heterogeneous prediction model; If the number of consecutive dominant periods is greater than a preset period threshold, a reliable attenuation factor is introduced to attenuate the dominant decision score of the first joint heterogeneous prediction model, and the attenuated dominant decision score is obtained. The pattern determination result is re-obtained using the decayed dominant decision score.

7. The method as described in claim 1, characterized in that, The collaborative state vector is analyzed based on the set of joint heterogeneous prediction models. The set of joint heterogeneous prediction models includes at least an operation-driven prediction model that predicts the trend of material demand changes based on the physical state of hydropower operation, a health-driven prediction model that predicts the trend of material demand changes based on the equipment health index, and a pattern-driven prediction model that predicts the trend of material demand changes based on the inventory fluctuation pattern.

8. The method as described in claim 1, characterized in that, After determining the pattern based on the decision-driven scoring, the method further includes: Obtain the current joint replenishment optimization strategy based on the current mode determination result; Get the previous replenishment optimization strategy for the previous replenishment cycle; If the replenishment fluctuation of the current joint replenishment optimization strategy is greater than the preset fluctuation threshold of the replenishment optimization strategy in the previous cycle, the current joint replenishment optimization strategy is adjusted based on the fluctuation penalty factor.

9. The method as described in claim 1, characterized in that, After obtaining the joint replenishment optimization strategy, the methods include: Establish the hydraulic coupling relationship among the multiple hydropower enterprises; Based on the hydraulic coupling relationship, the obtained joint replenishment optimization strategy is used for replenishment priority scheduling.

10. A joint replenishment optimization system for hydropower enterprises based on collaborative decision-making, characterized in that: The system is used to execute the joint replenishment optimization method for hydropower enterprises based on collaborative decision-making as described in any one of claims 1-9, and the system includes: The data matrix construction module is used to collect operational data from multiple hydropower companies and construct a joint operation data matrix for hydropower companies. The vector extraction module is used to extract the joint operation data matrix of the hydropower enterprises and extract the collaborative state vector; The strategy generation module is used to establish a set of joint heterogeneous prediction models, analyze the collaborative state vector for the set of joint heterogeneous prediction models, and obtain the joint replenishment strategy generated by each joint heterogeneous prediction model. The strategy evaluation module is used to evaluate the joint replenishment strategy and obtain the decision-making dominance score of each joint heterogeneous prediction model. The decision-making dominance score includes the risk mitigation contribution index, the synergy consistency stability index, and the operational status adaptability index. The strategy optimization module is used to determine the pattern based on the decision-making dominant score, and optimize the joint replenishment strategy set output by the joint heterogeneous prediction model set according to the pattern determination result to obtain the joint replenishment optimization strategy.

Images