A method, apparatus, equipment, and medium for multi-party collaborative scheduling of liquid energy logistics based on multi-stage robust optimization.

Abstract

This invention relates to the field of liquid energy technology and discloses a method, apparatus, equipment, and medium for multi-party collaborative scheduling of liquid energy logistics based on multi-stage robust optimization. It can construct a predicted demand value expression for various liquid energy sources and introduce it into the created material balance constraints to obtain uncertain material balance constraints. Through robust equivalence transformation, deterministic robust material balance constraints are obtained, creating a multi-party collaborative scheduling model for liquid energy logistics. Using simulation scenarios for each operating condition constructed based on preset disturbance intervals, the model is tested and optimized to obtain the final scheduling model. Based on the final scheduling model and a multi-stage rolling time-domain strategy, the supply and demand quantities of each product at each time point in each time window are determined by the supply and demand sides, as well as the logistics transportation volume of each product within each time window. This effectively enhances the anti-interference capability of multi-party collaborative scheduling of liquid energy logistics, including refined oil, methanol, liquid ammonia, and biofuels.

Description

Technical Field

[0001] This invention relates to the field of liquid energy technology, and in particular to a method, apparatus, equipment and medium for multi-party collaborative scheduling of liquid energy logistics based on multi-stage robust optimization. Background Technology

[0002] With the development of science and technology, technologies for the production, transportation, and storage of liquid energy are constantly improving.

[0003] Against the backdrop of the global low-carbon transition, refining and chemical enterprises are undergoing a structural shift from primarily producing refined oil products to producing diversified liquid energy sources, including methanol, liquid ammonia, and biofuels. Among these, the liquid energy logistics system is a core link in facilitating the spatial transfer of resources and achieving supply-demand matching.

[0004] Related technologies typically construct liquid energy logistics planning problems as static deterministic planning models, which assume that the refinery-side production plan and transportation-side allocation plan are deterministic throughout the entire planning period. However, this is out of touch with the real environment. When faced with sudden external disturbances, the initially seemingly optimal solution often fails midway through execution, and its resistance to interference is weak. Summary of the Invention

[0005] This invention provides a method, apparatus, equipment, and medium for multi-party collaborative scheduling of liquid energy logistics based on multi-stage robust optimization, which addresses the shortcomings of weak anti-interference capability of liquid energy logistics in related technologies, effectively enhances the anti-interference capability of collaborative scheduling of liquid energy logistics, and realizes efficient and dynamic collaborative configuration of liquid energy logistics under uncertain environments.

[0006] In a first aspect, the present invention provides a multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization, comprising: Based on the nominal demand variables, disturbance variables, and allowable prediction deviation variables of the multi-source liquid energy, a predicted demand value expression is generated; wherein, the multi-source liquid energy includes multiple liquid energy products; The predicted demand value expression is introduced into the created material balance constraint to obtain an uncertain material balance constraint; Based on strong duality theory, the uncertain material balance constraints are robustly transformed to obtain deterministic robust material balance constraints. Based on the robust material balance constraints, a multi-party collaborative scheduling model for liquid energy logistics is created. The liquid energy logistics multi-party collaborative scheduling model is tested and optimized using each working condition simulation scenario constructed based on a preset disturbance interval to obtain the final scheduling model. Based on the final scheduling model and the multi-stage rolling time-domain strategy, the supply volume of each liquid energy product by the supply side at each time endpoint of each time window, the demand volume of each liquid energy product by the demand side at each time endpoint, and the logistics transportation volume of each liquid energy product by the logistics side within each time window are determined.

[0007] Optionally, the step of testing and optimizing the multi-party collaborative scheduling model for liquid energy logistics using each operating condition simulation scenario constructed based on a preset disturbance interval to obtain the final scheduling model includes: Using the historical demand data of the diverse liquid energy sources from the demand side, the trained time-series forecasting model for demand values ​​is tested to determine the maximum allowable prediction deviation of the time-series forecasting model for demand values. Multiple random samplings are performed within the preset disturbance range. Based on each randomly sampled disturbance value, the maximum allowable prediction deviation value, and the set nominal demand value, each working condition simulation scenario is constructed. The multi-party collaborative scheduling model for liquid energy logistics is tested and optimized using each of the aforementioned operating condition simulation scenarios to obtain the final scheduling model.

[0008] Optionally, the historical demand data includes the actual demand value of the multi-source liquid energy products at each point in time within the historical period; The step of using historical demand data for the diverse liquid energy sources from the demand side to test the trained time-series demand forecasting model, in order to determine the maximum allowable prediction deviation of the time-series demand forecasting model, includes: The first P time points within the historical period are defined as the first historical time points, and the time points within the historical period other than the first historical time points are defined as the second historical time points; where P is an integer greater than 1. The actual demand values ​​of the multi-source liquid energy products at each first historical time point are arranged in chronological order to create a time series of actual demand for the multi-source liquid energy products. The actual demand time series of the multi-element liquid energy is input into the demand value time series prediction model for time series prediction, so as to obtain the predicted demand value of the multi-element liquid energy product at each second historical time point. Based on the predicted and actual demand values ​​of the multi-element liquid energy product at each of the second historical time points, the absolute value of the prediction deviation of the multi-element liquid energy at each of the second historical time points is determined. The maximum value among the absolute values ​​of each prediction deviation is determined and used as the maximum allowable prediction deviation value.

[0009] Optionally, the demand value time series prediction model includes a connected variational mode decomposition (VMD) model and a bidirectional long short-term memory (BiLSTM) network model. The step of inputting the actual demand time series of the multi-source liquid energy into the demand value time series prediction model for time series prediction, to obtain the predicted demand value of the multi-source liquid energy product at each second historical time point, includes: The actual demand time series of the multi-element liquid energy is input into the VMD model so that the VMD model performs variational mode decomposition on the actual demand time series of the multi-element liquid energy to obtain multiple intrinsic mode function components and residual components output by the VMD model. Each intrinsic mode function (IMF) component and the residual component are input into the BiLSTM model, respectively, so that the BiLSTM model extracts the forward and backward hidden states of each IMF component and the residual component, and concatenates the forward and backward hidden states of each IMF component and the residual component to obtain the bidirectional features of each IMF component and the residual component. A fully connected mapping is performed on the bidirectional features of each IMF component and the residual component to obtain the predicted value of each IMF component and the residual component. The predicted values ​​of each IMF component and the residual component are linearly superimposed to obtain the predicted demand value of the multi-element liquid energy product at each second historical time point.

[0010] Optionally, the expression for the predicted demand value is: ; in, For time windows t internal nodes n medium-sized products k The predicted demand value, For time windows t internal nodes n medium-sized products k The nominal demand variable, For time windows t internal nodes n medium-sized products k The perturbation variable, For time windows t internal nodes n medium-sized products k Allowable prediction bias variables, T , D and K These represent the time window set, the demand node set, and the product set, respectively.

[0011] Optionally, the robust material balance constraint is: ; ; ; in, For time points node medium-sized products Inventory levels, in tons; For time points node medium-sized products The amount of stock shortage, in tons; For time windows Internal transportation From node Send to node Products The transport volume, in tons; For time windows Internal transportation From node Send to node Products The amount of goods transported in transit, in tons; For time windows internal nodes n medium-sized products k The nominal demand variable, in tons; For time windows internal nodes medium-sized products The maximum permissible deviation, in tons; For nodes medium-sized products Uncertain budget parameters; and For robust equivalent variables; T , S , H , D, M and K These represent the time window set, supply node set, transit node set, demand node set, transportation mode set, and product set, respectively.

[0012] Optionally, the step of testing and optimizing the multi-party collaborative scheduling model for liquid energy logistics using each of the simulated operating scenarios to obtain the final scheduling model includes: The liquid energy logistics multi-party collaborative scheduling model was tested using at least one of the aforementioned operating condition simulation scenarios, and the resilience dimension index and cost dimension index of the liquid energy logistics multi-party collaborative scheduling model were obtained; wherein, the resilience dimension index includes at least one of average out-of-stock duration, out-of-stock quantity, and service level. If the resilience dimension index or the cost dimension index does not meet the requirements, the internal parameters of the liquid energy logistics multi-party collaborative scheduling model are optimized to obtain the optimized model. Continue to test the optimized model using at least one of the aforementioned operating condition simulation scenarios until the latest optimized model meets the requirements for both resilience and cost metrics.

[0013] Secondly, the present invention provides a multi-stage robust optimization-based multi-party collaborative scheduling device for liquid energy logistics, applied to the multi-stage robust optimization-based multi-party collaborative scheduling method for liquid energy logistics described in the first aspect above or any corresponding embodiment; the device includes: The generation unit is used to generate a predicted demand value expression based on the nominal demand variable, disturbance variable, and allowable prediction deviation variable of the multi-source liquid energy; wherein the multi-source liquid energy includes multiple liquid energy products; An introducing unit is used to introduce the predicted demand value expression into the created material balance constraint to obtain an uncertain material balance constraint; The transformation unit is used to perform a robust equivalent transformation on the uncertain material balance constraints based on strong duality theory, so as to obtain deterministic robust material balance constraints. A creation unit is used to create a multi-party collaborative scheduling model for liquid energy logistics based on the robust material balance constraints. The optimization unit is used to test and optimize the multi-party collaborative scheduling model of liquid energy logistics using each working condition simulation scenario constructed based on a preset disturbance interval, so as to obtain the final scheduling model. The determining unit is used to determine, based on the final scheduling model and the multi-stage rolling time-domain strategy, the supply quantity of each liquid energy product by the supply side at each time endpoint of each time window, the demand quantity of each liquid energy product by the demand side at each time endpoint, and the logistics transportation quantity of each liquid energy product by the logistics side within each time window.

[0014] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the multi-stage robust optimization-based multi-party collaborative scheduling method for liquid energy logistics described in the first aspect or any corresponding embodiment.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the multi-stage robust optimization-based multi-party collaborative scheduling method for liquid energy logistics described in the first aspect above or any corresponding embodiment.

[0016] This invention provides a multi-stage robust optimization-based method, apparatus, equipment, and medium for multi-party collaborative scheduling of liquid energy logistics. Addressing the issue of low actual execution rates caused by static models in related technologies, this invention introduces a rolling time-domain strategy during the robust optimization solution process. The long-cycle decision domain is divided into a series of sliding decision windows. At the current moment, the model is re-optimized only based on the updated state variables and the updated predicted requirements, and execution is sequentially pushed forward.

[0017] This invention constructs multiple types of physical constraints to achieve a mathematical representation of a practical liquid energy logistics system, including refinery production capacity constraints, refinery conversion constraints, receiving and dispatching capacity constraints, outbound capacity constraints, and material balance constraints. To enable the model to effectively cope with random fluctuations in demand, the deterministic material balance derivation is transformed into robust material balance constraints based on a cumulative flow perspective. This results in the optimal collaborative configuration solution for complex liquid energy logistics systems involving multiple products and multiple stakeholders, achieving efficient and flexible dynamic programming of multi-party collaborative scheduling plans for liquid energy in the face of nonlinear demand fluctuations and external disturbances.

[0018] This invention can effectively enhance the anti-interference capability of multi-party collaborative scheduling of liquid energy logistics, including refined oil, methanol, liquid ammonia, and biofuels, realize efficient and dynamic collaborative configuration of liquid energy logistics systems under uncertain environments, complete the optimal collaborative configuration solution of complex liquid energy logistics systems with multiple products and multiple entities, and realize efficient and flexible dynamic compilation of liquid energy multi-party collaborative scheduling plans in the face of nonlinear demand fluctuations and external disturbances. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of a multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization provided in this embodiment of the invention; Figure 2 A flowchart of another multi-stage robust optimization-based multi-party collaborative scheduling method for liquid energy logistics provided in an embodiment of the present invention; Figure 3 This invention provides a bar chart showing the monthly supply and demand situation at each node. Figure 4 A transportation cost table for various modes of transportation is provided for embodiments of the present invention; Figure 5 A schematic table illustrating the conversion rates between various liquid energy sources is provided for embodiments of the present invention. Figure 6 A schematic table illustrating the carbon emission factors and calorific values ​​of different products provided in this embodiment of the invention; Figure 7 This is a comparison chart of the effects of different models under different simulation scenarios provided by an embodiment of the present invention; Figure 8 This invention provides a heatmap visualization of the daily out-of-stock quantity distribution under a highly volatile scenario. Figure 9 This invention provides an analysis of the operating costs of different models under different fluctuation scenarios in an embodiment of the invention. Figure 10 A schematic diagram of a multi-party collaborative scheduling device for liquid energy logistics based on multi-stage robust optimization provided in an embodiment of the present invention; Figure 11 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] The following is combined Figures 1-9 This invention describes a multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization.

[0023] like Figure 1 As shown, this embodiment proposes a first multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization. This method may include the following steps: S101. Generate a forecast demand value expression based on the nominal demand variable, disturbance variable, and allowable forecast deviation variable of the multi-source liquid energy. The multi-source liquid energy includes multiple liquid energy products.

[0024] It should be noted that this embodiment can introduce perturbation variables and allowable prediction deviation variables to construct a prediction demand value expression for multi-element liquid energy.

[0025] The expression for the predicted demand value is as follows: ; in, For time windows t internal nodes n medium-sized products k The predicted demand value, For time windows t internal nodes n medium-sized products k The nominal demand variable, For time windows t internal nodes n medium-sized products k The perturbation variable, For time windows t internal nodes n medium-sized products k Allowable prediction bias variables, T , D and K These represent time period sets, node sets, and product sets, respectively.

[0026] S102. Introduce the predicted demand value expression into the created material balance constraint to obtain the uncertain material balance constraint.

[0027] It should be noted that this embodiment aims to reduce the total cost of the four links of transportation, conversion, storage and shortage in the liquid energy logistics system. An initial model for realizing multi-party collaborative scheduling of liquid energy logistics can be constructed first. This initial model includes constraints such as objective function, refinery production capacity constraint, refinery conversion constraint, receiving and dispatching capacity constraint and storage capacity constraint. The objective function includes transportation cost, conversion cost, storage cost and shortage penalty cost. The objective function, transportation cost, conversion cost, storage cost and shortage penalty cost can be referred to in equations (1) to (4) below.

[0028] ---------Formula (1); ----Equation (2); ------------Formula (3); ---------Form (4); ---------Form (5); In the formula: For transportation From node To the node Transport products The unit transportation cost, in yuan / ton; For time windows Internal transportation From node Send to node Products The transport volume, in tons; For refinery medium-sized products Transform into products The unit cost of the transferred-out volume, in yuan / ton; For time windows Domestic refinery Zhongcong Products Transform into products The amount transferred out, in tons; For nodes medium-sized products The unit inventory cost, in yuan / ton; For time points node medium-sized products Inventory quantity, in tons; For nodes medium-sized products The unit penalty cost for stockouts, in yuan / ton; For time points node medium-sized products The shortage amount is in tons.

[0029] Specifically, the refinery production capacity constraints are as follows: ---------Form (6); ---------Form (7); ---------Form (8); In the formula: For time windows internal nodes medium-sized products Production volume, in tons; For time windows internal nodes medium-sized products The planned production volume, in tons; \ For nodes medium-sized products Minimum or maximum inventory level, in tons; For nodes medium-sized products The maximum collaborative capacity is tons.

[0030] Equation (6) ensures that the production of refined petroleum products and other liquid products in these refineries equals the planned production. Equation (7) ensures that the inventory of refined petroleum products and other liquid products in these refineries does not exceed the maximum inventory level and is not lower than the minimum inventory level. Equation (8) ensures that the total inventory of refined petroleum products and other liquid products in these refineries does not exceed the maximum co-production capacity.

[0031] Specifically, the refinery conversion constraints are: ---------Form (9); ---------Formula (10); In the formula, For time windows internal nodes medium-sized products Supply volume, tons; For nodes medium-sized products Transform into products Product conversion rate.

[0032] Equation (9) represents the constraint on the transfer out of refined oil products, that is, the supply of refined oil products plus the cumulative transfer out of refined oil products is less than the planned production of refined oil products. Equation (10) represents the constraint on the transfer in of other liquid products, that is, the supply of liquid products is less than the cumulative transfer in of liquid products plus the planned production of liquid products.

[0033] Specifically, the send / receive capacity constraints are as follows: ---------Formula (11); ---------Formula (12); In the formula, For nodes Chinese transportation mode Maximum coordinated shipping capacity, tons; For time windows Internal transportation From node Send to node Products The amount of goods transported in transit, in tons; For nodes Chinese transportation mode The maximum collaborative receiving capacity is tons.

[0034] Equation (11) ensures that within a unit time window, the volume of shipments dispatched by the refinery or transit storage facility through various modes of transportation is less than the maximum coordinated dispatch capacity. Equation (12) ensures that the sum of the volume received from the refinery and transit storage facility and the volume of shipments in transit is less than the maximum coordinated receiving capacity of the storage facility.

[0035] Specifically, the storage capacity constraints are as follows: ---------Form (13); ---------Form (14); Equation (13) ensures that the inventory levels of these storage facilities do not exceed the maximum inventory level and are not lower than the minimum inventory level. Equation (14) ensures that the total inventory levels of refined oil products and other liquid products in these storage facilities do not exceed the maximum co-operation capacity.

[0036] Specifically, this embodiment can also create material balance constraints, which are as follows: ---------Form (14); ---------Form (15); ---------Form (16); In the formula, For nodes medium-sized products The initial inventory, in tons.

[0037] Equations (14) and (15) represent the material balance constraints for the refineries. At the initial moment, the inventory level of each refinery is equal to its initial inventory level. The inventory level in time window t+1 is equal to the inventory level in time window t, plus the supply in time window t, minus the transportation volume in time window t. Equation (16) represents the material balance constraints for transit warehouses and ordinary warehouses. At the initial moment, the inventory level of each warehouse is equal to its initial inventory level.

[0038] Specifically, in this embodiment, the predicted demand value expression can be introduced into the material balance constraint to obtain an uncertain material balance constraint. Wherein: The expression for the predicted demand value is: --------Formula (17); in, For time windows t internal nodes n medium-sized products k The predicted demand value, For time windows t internal nodes n medium-sized products k The nominal demand variable, For time windows t internal nodes n medium-sized products k The perturbation variable, For time windowst internal nodes n medium-sized products k Allowable prediction bias variables, T , D and K These represent the time window set, the demand node set, and the product set, respectively.

[0039] It is understandable that equations (14) to (17) constitute uncertain material balance constraints.

[0040] S103. Based on strong duality theory, perform robust equivalent transformation on uncertain material balance constraints to obtain deterministic robust material balance constraints.

[0041] It should be noted that equation (17) shows the expression for the predicted demand value, which is composed of the nominal demand variable, the disturbance variable, and the allowable forecast deviation variable. Substituting this formula into the cumulative material balance equation, the constraint equation includes the disturbance variable. This makes it impossible for conventional mathematical programming algorithms to solve it directly.

[0042] This embodiment can robustly transform uncertain material balance constraints based on strong duality theory. By converting model constraints containing perturbation variables and variables allowing for prediction bias into deterministic linear inequality constraints—that is, deterministic robust material balance constraints—randomly fluctuating demand is transformed into model constraints on inventory and transportation volumes. This constraint, from the perspective of cumulative flow, requires that at any point during the planning period, the cumulative supply of all types of products must be sufficient to cover all possible worst-case cumulative demand fluctuations. This ensures that the logistics execution plan can still maintain a high execution rate even when facing adverse external conditions.

[0043] Specifically, the robust material balance constraint is given by the following equations (18) to (20): --------Formula (18); --------Formula (19); --------Formula (20); In the formula: For time points node medium-sized products Inventory quantity, in tons; For time points node medium-sized products The amount out of stock, in tons; For time windows Internal transportation From node Send to node Products The transport volume, in tons; For time windows Internal transportation From node Send to node Products The amount of goods transported in transit, in tons; For time windows internal nodes n medium-sized products k The nominal demand variable is tons; For time windows internal nodes medium-sized products The maximum permissible deviation, in tons; For nodes medium-sized products Uncertain budget parameters; and For robust equivalent variables; T , S , H , D, M and K These represent the time window set, supply node set, transit node set, demand node set, transportation mode set, and product set, respectively.

[0044] Specifically, equations (19) to (20) further define the dual variables generated during the transformation process and their corresponding non-negative mathematical relationships.

[0045] S104. Based on robust material balance constraints, create a multi-party collaborative scheduling model for liquid energy logistics.

[0046] It should be noted that in this embodiment, the constructed objective function, refinery production capacity constraints, refinery conversion constraints, receiving and dispatching capacity constraints, storage capacity constraints, and deterministic robust material balance constraints can be used as a whole as a multi-party collaborative scheduling model for liquid energy logistics. Specifically, the multi-party collaborative scheduling model for liquid energy logistics includes the above equations (1) to (20).

[0047] S105. Using each working condition simulation scenario constructed based on a preset disturbance interval, test and optimize the multi-party collaborative scheduling model for liquid energy logistics to obtain the final scheduling model.

[0048] Specifically, this embodiment can use working condition simulation scenarios to test and optimize the multi-party collaborative scheduling model of liquid energy logistics until a final scheduling model that meets performance requirements is obtained.

[0049] Optionally, step S105 includes: By using historical demand data for various liquid energy sources from the demand side, the trained time series forecasting model for demand values ​​is tested to determine the maximum allowable forecast deviation of the time series forecasting model for demand values. Multiple random samplings are performed within a preset disturbance range. Based on each randomly sampled disturbance value, the maximum allowable prediction deviation value, and the set nominal demand value, a simulation scenario for each working condition is constructed. The multi-party collaborative scheduling model for liquid energy logistics was tested and optimized using each operating condition simulation scenario to obtain the final scheduling model.

[0050] Specifically, this embodiment can acquire historical demand data for each ordinary storage facility on the demand side for any liquid energy product, such as daily demand values ​​within a certain historical period. Then, the historical demand data for a particular liquid energy product from a specific ordinary storage facility is used to test the time-series demand forecasting model, determining the maximum permissible error (i.e., the maximum permissible forecast deviation) of the time-series demand forecasting model for that ordinary storage facility's daily predicted demand value for liquid energy products.

[0051] It is understandable that for different liquid energy products in different ordinary storage facilities, the time series forecasting model for demand values ​​has a corresponding maximum allowable forecast deviation value.

[0052] Specifically, in this embodiment, for a certain liquid energy product required by a general storage facility, the corresponding nominal demand value is first determined, and then random sampling is performed within a preset disturbance range. The disturbance value obtained by sampling is multiplied by the corresponding maximum allowable prediction deviation value, and the product is added to the nominal demand value to obtain the simulated daily demand value of the liquid energy product of the general storage facility under the condition of fluctuation.

[0053] Specifically, this embodiment can generate simulated demand values ​​for different liquid energy products from different ordinary storage facilities on different dates, and combine these values ​​to obtain a working condition simulation scenario. In this scenario, the working condition simulation scenario includes the simulated demand values ​​for each liquid energy product from each ordinary storage facility on a given day within a specified time period.

[0054] Optionally, historical demand data may include the actual demand for multi-functional liquid energy products at each point in time within a historical period.

[0055] The above-mentioned historical demand data of various liquid energy sources were used to test the trained time-series demand forecasting model to determine the maximum permissible prediction deviation of the time-series demand forecasting model, including: Using historical demand data for various liquid energy sources, a trained time-series demand forecasting model was tested to determine the maximum permissible forecast deviation, including: The first P time points within a historical period are defined as the first historical time points, and the time points within the historical period other than the first historical time points are defined as the second historical time points; where P is an integer greater than 1. The actual demand values ​​for multi-source liquid energy products at each first historical point in time are arranged in chronological order to create a time series of actual demand for multi-source liquid energy products. The actual demand time series of multi-source liquid energy is input into the demand value time series prediction model for time series prediction, so as to obtain the predicted demand value of multi-source liquid energy products at each second historical time point. Based on the predicted and actual demand values ​​for multi-source liquid energy products at each second historical time point, the absolute value of the prediction deviation of multi-source liquid energy at each second historical time point is determined. The maximum value among the absolute values ​​of each prediction deviation is determined and used as the maximum allowable prediction deviation value.

[0056] The time point can be a single day.

[0057] Specifically, the demand forecasting model can include interconnected Variational Mode Decomposition (VMD) and Bidirectional Long Short-Term Memory (BiLSTM) models. This embodiment can use known historical data (e.g., the past two years) as input for training the model, while reserving data for a specific period (e.g., the last month) as a test set. After obtaining the demand forecast through the model, the actual daily demand for each node and type of oil product on the test set is compared horizontally to calculate the actual forecast residual distribution during this period. The maximum positive or negative deviation value of the forecast is extracted and used as the error range of the forecast value, i.e., the maximum allowable forecast deviation value. .

[0058] Secondly, after obtaining the maximum permissible prediction deviation value, this invention compares it with the nominal demand value. By combining these parameters, a set of uncertainty parameters is constructed. Specifically, the model will predict the demand value. Defined as the sum of nominal demand and fluctuation deviation, its formula is expressed as follows: In the formula To take the value in [ A continuous disturbance variable between [1,1]. By adjusting the value of this disturbance variable, the error value can be added or subtracted from the nominal demand value, thereby representing different degrees of fluctuation.

[0059] Specifically, this embodiment takes into account that in actual operation, the daily demand will reach the maximum allowable forecast deviation value (i.e., the total demand for all days). The probability of all parameters being equal to 1 is extremely low. If the model is optimized to defend against this extreme case, it would lead to the system reserving too much safety stock for each storage depot, resulting in excessive conservatism and high logistics storage costs. Therefore, this embodiment further introduces an uncertainty budget parameter into the uncertainty parameter set. This is used to limit the sum of the absolute values ​​of all disturbance variables throughout the entire planning period (i.e., to satisfy the constraint). By providing this budget parameter, decision-makers can adjust the optimization model to defend against the worst-case extreme volatility based on the actual market risk appetite, thereby finding the optimal balance between reducing systemic shortage risk and controlling economic costs.

[0060] Specifically, this embodiment, after determining the maximum permissible forecast deviation, further constructs simulation test scenarios for model validation based on the daily demand of each node and type of oil product and the maximum permissible forecast deviation. This embodiment can, according to preset low, medium, and high-intensity disturbance fluctuation scenarios, test the disturbance variables (in the uncertainty set) A range of values ​​corresponding to the intensity of the disturbance variable is defined. For example, in low-fluctuation scenarios, the value of the disturbance variable is limited to a smaller range, while in high-fluctuation scenarios, its value is allowed to approach the upper limit of the extreme value. Subsequently, Monte Carlo sampling is used to randomly sample a large number of disturbance variables within these three different fluctuation value spaces. Finally, the sampled disturbance variables are substituted into the formula for predicting demand values ​​to construct multiple simulated test scenarios that closely resemble actual working conditions, thereby verifying the risk resistance and resilience of the logistics solution in the face of external disturbances of different degrees.

[0061] S106. Based on the final scheduling model and multi-stage rolling time domain strategy, determine the supply quantity of each liquid energy product at each time endpoint of the supply side, the demand quantity of each liquid energy product at each time endpoint of the demand side, and the logistics transportation quantity of each liquid energy product within each time window.

[0062] Specifically, in this embodiment, the model can adopt a multi-stage rolling time-domain strategy during execution, executing only the current time-domain decision in each decision window. The optimal instruction will be executed, and the system state after execution (such as the inventory in Equation 15) will be displayed. As feedback input, the updated prediction information is combined with the next round of optimization loop, thereby using the iterative logic between formulas to correct the execution deviation caused by external disturbances in real time.

[0063] The proposed method for multi-party collaborative scheduling of liquid energy logistics based on multi-stage robust optimization in this embodiment can generate a predicted demand value expression based on the nominal demand variable, disturbance variable, and allowable prediction deviation variable of multiple liquid energy sources. This predicted demand value expression is then introduced into the created material balance constraints to obtain uncertain material balance constraints. Based on strong duality theory, the uncertain material balance constraints are robustly transformed to obtain deterministic robust material balance constraints. Based on these robust material balance constraints, a multi-party collaborative scheduling model for liquid energy logistics is created. The model is tested and optimized using simulation scenarios for each operating condition constructed based on preset disturbance intervals to obtain the final scheduling model. Based on the final scheduling model and a multi-stage rolling time-domain strategy, the supply volume of each liquid energy product at each time endpoint of the supply side, the demand volume of each liquid energy product at each time endpoint of the demand side, and the logistics transportation volume of each liquid energy product within each time window are determined. This embodiment can effectively enhance the anti-interference capability of multi-party collaborative scheduling of liquid energy logistics, realize the efficient and dynamic collaborative configuration of liquid energy logistics system under uncertain environment, complete the optimal collaborative configuration solution of complex liquid energy logistics system with multiple products and multiple entities, and realize the efficient and flexible dynamic compilation of liquid energy multi-party collaborative scheduling plan in the face of nonlinear demand fluctuations and external disturbances.

[0064] based on Figure 1 This embodiment proposes a second method for multi-party collaborative scheduling of liquid energy logistics based on multi-stage robust optimization. In this method, the demand value time series prediction model includes a connected Variational Mode Decomposition (VMD) model and a Bidirectional Long Short-Term Memory (BiLSTM) model.

[0065] At this point, the actual demand time series of the aforementioned multi-source liquid energy is input into the demand value time series forecasting model for time series forecasting, to obtain the predicted demand value of multi-source liquid energy products at each second historical time point, including: The actual demand time series of multi-element liquid energy is input into the VMD model so that the VMD model can perform variational mode decomposition on the actual demand time series of multi-element liquid energy, and obtain multiple intrinsic mode function components and residual components output by the VMD model. Each intrinsic mode function (IMF) component and residual component is input into the BiLSTM model to extract the forward and backward hidden states of each IMF component and residual component. The forward and backward hidden states of each IMF component and residual component are then concatenated to obtain the bidirectional features of each IMF component and residual component. A fully connected mapping is then performed on the bidirectional features of each IMF component and residual component to obtain the predicted values ​​of each IMF component and residual component. The predicted values ​​of each IMF component and residual component are then linearly superimposed to obtain the predicted demand value of multi-element liquid energy products at each second historical time point.

[0066] Specifically, to address the issues of poor stability and limited robustness when directly applying raw demand time series to forecasting models, this invention employs VMD as a preprocessing technique. This technique decomposes the storage demand time series into multiple sub-components with higher stationarity and clearer temporal structure, specifically including a finite number of Intrinsic Mode Functions (IMFs) and a residual component, where each IMF can capture demand variation patterns at a specific center frequency.

[0067] ------------Formula (21); ------------Equation (22); ------------Formula (23); In the formula, For time windows node For the The original demand sequence value for this product; The first mode obtained by variational mode decomposition One intrinsic mode function component; These are the residual components; This represents the total number of modal components. For the first The center frequency of each modal component; It is the Dirac function; This is a convolution operation; Regarding time The partial derivative operator; It is a constant.

[0068] In the VMD solution process, an iterative solution using the alternating direction multiplier method is employed to effectively mitigate the mode aliasing problem. The extracted IMF components and residual terms are decomposed, significantly reducing the non-stationarity of the original data. Subsequently, all components are fed as input features into a BiLSTM. The BiLSTM utilizes its internal standard gating mechanism to update the neuron state and output the hidden state, thereby effectively capturing the implicit nonlinear time dependencies in each required component and enhancing prediction robustness.

[0069] ------------Formula (24); ------------Formula (25); ------------Formula (26); -----------------------Equation (27); ------------Formula (28); ------------Formula (29); In the formula, Output for the forget gate; For input gate output; Candidate cell state; For a moment The unit state; The state was hidden in the previous moment; For the first Each component at time... The input value; Output gate output; It is in a positive hidden state; , , and This is the corresponding weight matrix; , , and The input weight matrix; , , and For bias terms; Use the Sigmoid activation function; This is element-wise multiplication.

[0070] BiLSTM extracts the forward and backward hidden states separately and concatenates them to form bidirectional features. These features are then mapped through a fully connected layer to obtain the predicted values ​​of each component. This extraction and prediction process is applied independently to all IMF and residual components. Finally, the predicted values ​​of the storage demand are obtained by linearly superimposing the prediction results of each component.

[0071] ------------Formula (30); ------------Formula (31); Equation (32); In the formula, The feature representation is formed by splicing together bidirectional hidden states; For the first Predicted values ​​for each component; This is the weight matrix of the fully connected layer; For bias terms of fully connected layers; For a moment node For the The final predicted value for this product; These are the residual components.

[0072] Optionally, in other multi-stage robust optimization-based multi-party collaborative scheduling methods for liquid energy logistics proposed in this embodiment, the above-mentioned use of each operating condition simulation scenario to test and optimize the multi-party collaborative scheduling model for liquid energy logistics to obtain the final scheduling model includes: The multi-party collaborative scheduling model for liquid energy logistics was tested using at least one working condition simulation scenario, and the resilience and cost dimensions of the multi-party collaborative scheduling model for liquid energy logistics were obtained. Among them, the resilience dimension indicators include at least one of average out-of-stock duration, out-of-stock quantity, and service level. If the resilience or cost indicators do not meet the requirements, the internal parameters of the multi-party collaborative scheduling model for liquid energy logistics will be optimized to obtain the optimized model. Continue to test the optimized model using at least one working condition simulation scenario until the latest optimized model meets the requirements for both resilience and cost metrics.

[0073] It should be noted that, in order to evaluate the effectiveness of the operation scheme generated by the multi-stage robust optimization model, this embodiment can establish a comprehensive evaluation system covering both resilience and cost dimensions to achieve multi-dimensional assessment and effect analysis.

[0074] Specifically, this embodiment mainly includes three core evaluation indicators in terms of resilience: average stockout duration, stockout quantity, and service level. First, the average stockout duration aims to measure the overall impact of the system when facing supply disruptions by calculating the arithmetic mean of the number of stockout days of each oil depot in the logistics system within the planning period. As shown in Equation (33), the model uses a state indicator function to determine the stockout status of various products. When the total available resources of a certain demand node (i.e., the sum of planned receipts and existing inventory) is lower than its actual demand, a stockout is determined to have occurred, and the global average duration is calculated using Equation (34). Second, the stockout quantity is used to quantify the scale of the total demand that cannot be met within the planning period. As shown in Equation (35), when the available quantity of a demand node is lower than the actual demand, the unmet difference is calculated, reflecting the degree of shortage of various products at each stage. Finally, the service level is defined as the proportion of the demand that has been met to the total demand, as shown in Equation (36). This formula measures the degree to which the logistics system successfully achieves supply and demand matching at the macro level.

[0075] In terms of cost, this embodiment mainly uses operating costs for comprehensive quantification. The operating cost evaluation index aims to quantify the total expenditure of the entire logistics system during the planning period, as shown in formula (37). The total operating cost of the system consists of three core parts: the transportation costs of various products using different transportation methods; the conversion costs of refining refined oil into other energy products; and the storage costs of various products held at each node. By summing up the above-mentioned relevant operating costs throughout the planning period, the operational economy and resilience under the robust strategy are evaluated.

[0076] ------------Formula (33); ------------Formula (30); ------------Formula (34); ------------Formula (35); -------Formula (36); In the formula, For a moment node For the product The out-of-stock status indicator variable takes a value of 1 if an out-of-stock situation occurs, and 0 otherwise. This represents the average out-of-stock duration. This refers to the quantity out of stock; For service level; Operating costs; This represents the total number of product categories.

[0077] The proposed method for multi-party collaborative scheduling of liquid energy logistics based on multi-stage robust optimization can address nonlinear demand fluctuations by using variational mode decomposition to decompose the original time series into multiple intrinsic mode functions and residuals, and then performing prediction by combining a bidirectional long short-term memory network.

[0078] This embodiment addresses the issue of low actual execution rates caused by static models in related technologies by introducing a rolling time-domain strategy during the robust optimization process. The long-period decision domain is divided into a series of sliding decision windows. At the current moment, the model is re-optimized only based on the updated state variables and the updated prediction requirements, and this process is sequentially pushed forward for execution.

[0079] To achieve a mathematical representation of the actual liquid energy logistics system, this embodiment constructs multiple types of physical constraints, including refinery production capacity constraints, refinery conversion constraints, receiving and dispatching capacity constraints, outbound capacity constraints, and material balance constraints.

[0080] To enable the model to effectively cope with random fluctuations in demand, this embodiment derives a deterministic material balance derivation based on robust material balance constraints from a cumulative flow perspective. This achieves optimal collaborative configuration solutions for complex liquid energy logistics systems involving multiple products and stakeholders, enabling efficient and flexible dynamic programming of multi-party collaborative scheduling plans for liquid energy in the face of nonlinear demand fluctuations and external disturbances.

[0081] like Figure 2 As shown, in other multi-stage robust optimization-based multi-party collaborative scheduling methods for liquid energy logistics proposed in this embodiment, modules 1, 2, and 3 can be created.

[0082] Module 1 can be used for uncertainty quantification and scenario construction. Specifically, it is used to solve the prediction error range based on the historical demand data and demand value prediction algorithm of each ordinary storage facility, and then to construct simulation scenarios, including low volatility scenario, medium volatility scenario and high volatility scenario.

[0083] Specifically, Module 2 is a multi-stage robust decision-making and feedback compensation module. It can construct an objective function and model constraints, and then perform robust optimization on the model constraints to build a robust optimization model. Then, the rolling time domain strategy is introduced into the robust optimization model. The multi-stage robust optimization model is used to determine the supply volume of each product, the demand volume, and the logistics transportation volume within the corresponding time window at each time node.

[0084] Specifically, Module 3 can be used for multi-dimensional evaluation and effect analysis. Evaluation indicators can include two dimensions: resilience and cost. The effectiveness of the model is determined through testing.

[0085] This study uses the liquid energy logistics system of a petrochemical company as an example. The case involves 20 different liquid energy sources (O01-O20), including 15 refined petroleum products and 5 other liquid products: ethylene glycol, methanol, liquid ammonia, gray hydrogen, and blue hydrogen. Figure 3 The figure shows the monthly supply and demand for 16 supply nodes and 80 demand nodes. The supply-demand mismatches of various products across geographical areas necessitate the use of four modes of transportation: road, waterway, pipeline, and rail.

[0086] Figure 4 The unit transportation costs of various transportation modes were summarized. Figure 5 The conversion rate between various liquid energy sources, Figure 6 It provides a detailed list of the unit carbon emissions and calorific value of various liquid energy sources.

[0087] like Figure 7 As shown, this embodiment proposes a multi-stage robust optimization model (i.e. Figure 7 Compared to static deterministic simulation models (i.e., RRSM) in related technologies, Figure 7 In the context of TSDM (Service Level Management), under low volatility scenarios, service level improved from 95.49% to 97.63%, and the average out-of-stock duration decreased by 0.87 days. Under high volatility scenarios, service level improved from 91.48% to 94.66%, and the average out-of-stock duration decreased by 1.27 days. Figure 7 In this context, RAS and RSL represent the average out-of-stock duration and service level, respectively, in the resilience dimension evaluation indicators.

[0088] like Figure 8 As shown, the robust model proposed in this embodiment, compared with the existing static model, reduces the total shortage from 82.51 × 10⁻⁶ in high-fluctuation scenarios. 4 The tonnes decreased to 44.43 × 10 4 The volume decreased by 46.15% to tons. In terms of temporal and spatial distribution, the original continuous and concentrated shortages were effectively eliminated, and the peak of extreme shortages (taking the D40 oil depot on the 18th day as an example) decreased from 2.55 × 10⁻⁶ tons. 4 The volume of the substance dropped sharply to 3.7 tons.

[0089] Combination Figure 9 As shown, the total cost of the static model is constant at 64.49 × 10⁻⁶. 8 The total cost of the robust model in this embodiment increases slightly by 3.64% to 66.84 × 10⁻⁶ yuan under high-fluctuation scenarios. 8 Yuan. This additional cost is converted into upfront safety stock and direct supply capacity from refineries, enabling the system to avoid secondary allocation from transit warehouses during shortages, achieving the effect of converting products near refineries and directly reaching demand nodes.

[0090] This embodiment integrates VMD-BiLSTM hybrid prediction, multi-stage robust optimization, and rolling time-domain strategy to propose a multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization. Targeting large-scale logistics systems encompassing refined oil products and liquid new energy sources in related technologies, this method effectively solves the system shortage problem caused by the static lag of traditional planning when facing nonlinear demand fluctuations and external disturbances, laying a technical foundation for dynamic collaborative operation under long-term uncertain environments. The proposed scheme overcomes the shortcomings of high adjustment costs and passive response of traditional static models. Under the premise of strictly controlling operating cost premiums, it significantly suppresses the risk of supply-demand mismatch and improves system resilience. Verified by a real-world case study of a petrochemical company, under high demand fluctuation scenarios, compared with the static deterministic models in related technologies, this embodiment achieves significant optimization effects with only a slight increase in operating costs of approximately 3.64%, resulting in a 3.18% improvement in system service level, a 29.5% reduction in average shortage duration, and a 46.15% reduction in total shortage volume.

[0091] like Figure 10 As shown, this embodiment proposes a multi-stage robust optimization-based multi-party collaborative scheduling device for liquid energy logistics, applicable to any of the aforementioned multi-stage robust optimization-based multi-party collaborative scheduling methods for liquid energy logistics; the device includes: The generation unit 101 is used to generate a predicted demand value expression based on the nominal demand variable, disturbance variable, and allowable prediction deviation variable of the multi-source liquid energy; wherein, the multi-source liquid energy includes multiple liquid energy products; The unit 102 is used to introduce the predicted demand value expression into the created material balance constraint to obtain the uncertain material balance constraint; The transformation unit 103 is used to perform a robust equivalent transformation of uncertain material balance constraints based on strong duality theory to obtain deterministic robust material balance constraints. Create cell 104 to create a multi-party collaborative scheduling model for liquid energy logistics based on robust material balance constraints; The optimization unit 105 is used to test and optimize the multi-party collaborative scheduling model of liquid energy logistics using each working condition simulation scenario constructed based on a preset disturbance interval, so as to obtain the final scheduling model. The determining unit 106 is used to determine, based on the final scheduling model and the multi-stage rolling time domain strategy, the supply quantity of each liquid energy product at each time endpoint of the supply side, the demand quantity of each liquid energy product at each time endpoint of the demand side, and the logistics transportation quantity of each liquid energy product within each time window.

[0092] It should be noted that the processing procedures and beneficial effects of the generation unit 101, the introduction unit 102, the transformation unit 103, the creation unit 104, the optimization unit 105, and the determination unit 106 can be referred to respectively. Figure 1 Steps S101 to S106 are not described in detail here.

[0093] Optionally, the optimization unit 105 is also used for: By using historical demand data for various liquid energy sources from the demand side, the trained time series forecasting model for demand values ​​is tested to determine the maximum allowable forecast deviation of the time series forecasting model for demand values. Multiple random samplings are performed within a preset disturbance range. Based on each randomly sampled disturbance value, the maximum allowable prediction deviation value, and the set nominal demand value, a simulation scenario for each working condition is constructed. The multi-party collaborative scheduling model for liquid energy logistics was tested and optimized using each operating condition simulation scenario to obtain the final scheduling model.

[0094] Optionally, the historical demand data includes the actual demand for diversified liquid energy products at each point in time within the historical period. Optimization unit 105 is also used for: The first P time points within a historical period are defined as the first historical time points, and the time points within the historical period other than the first historical time points are defined as the second historical time points; where P is an integer greater than 1. The actual demand values ​​for multi-source liquid energy products at each first historical point in time are arranged in chronological order to create a time series of actual demand for multi-source liquid energy products. The actual demand time series of multi-source liquid energy is input into the demand value time series prediction model for time series prediction, so as to obtain the predicted demand value of multi-source liquid energy products at each second historical time point. Based on the predicted and actual demand values ​​for multi-source liquid energy products at each second historical time point, the absolute value of the prediction deviation of multi-source liquid energy at each second historical time point is determined. The maximum value among the absolute values ​​of each prediction deviation is determined and used as the maximum allowable prediction deviation value.

[0095] Optionally, the demand value time series prediction model includes an interconnected variational mode decomposition (VMD) model and a bidirectional long short-term memory (BiLSTM) network model. Optimization unit 105 is also used for: The actual demand time series of multi-element liquid energy is input into the VMD model so that the VMD model can perform variational mode decomposition on the actual demand time series of multi-element liquid energy, and obtain multiple intrinsic mode function components and residual components output by the VMD model. Each intrinsic mode function (IMF) component and residual component is input into the BiLSTM model to extract the forward and backward hidden states of each IMF component and residual component. The forward and backward hidden states of each IMF component and residual component are then concatenated to obtain the bidirectional features of each IMF component and residual component. A fully connected mapping is then performed on the bidirectional features of each IMF component and residual component to obtain the predicted values ​​of each IMF component and residual component. The predicted values ​​of each IMF component and residual component are then linearly superimposed to obtain the predicted demand value of multi-element liquid energy products at each second historical time point.

[0096] Optionally, the expression for the predicted demand value is: ; in, For time windows t internal nodes n medium-sized products k The predicted demand value, For time windows t internal nodes n medium-sized products k The nominal demand variable, For time windows t internal nodes n medium-sized products k The perturbation variable, For time windows t internal nodes n medium-sized products k Allowable prediction bias variables, T , D and K These represent the time window set, the demand node set, and the product set, respectively.

[0097] Optionally, robust material balance constraints are: ; ; ; in, For time points node medium-sized products Inventory levels, in tons; For time points node medium-sized products The amount of stock shortage, in tons; For time windows Internal transportation From node Send to node Products The transport volume, in tons; For time windows Internal transportation From node Send to node Products The amount of goods transported in transit, in tons; For time windows internal nodes n medium-sized products k The nominal demand variable, in tons; For time windows internal nodes medium-sized products The maximum permissible deviation, in tons; For nodes medium-sized products Uncertain budget parameters; and For robust equivalent variables; T , S , H , D, M and K These represent the time window set, supply node set, transit node set, demand node set, transportation mode set, and product set, respectively.

[0098] Optionally, optimization unit 105 is also used for: The multi-party collaborative scheduling model for liquid energy logistics was tested using at least one working condition simulation scenario, and the resilience and cost dimensions of the multi-party collaborative scheduling model for liquid energy logistics were obtained. Among them, the resilience dimension indicators include at least one of average out-of-stock duration, out-of-stock quantity, and service level. If the resilience or cost indicators do not meet the requirements, the internal parameters of the multi-party collaborative scheduling model for liquid energy logistics will be optimized to obtain the optimized model. Continue to test the optimized model using at least one working condition simulation scenario until the latest optimized model meets the requirements for both resilience and cost metrics.

[0099] The liquid energy logistics multi-party collaborative scheduling device proposed in this embodiment, based on multi-stage robust optimization, can generate a predicted demand value expression based on the nominal demand variable, disturbance variable, and allowable prediction deviation variable of multiple liquid energy sources. This predicted demand value expression is then introduced into the created material balance constraints to obtain uncertain material balance constraints. Based on strong duality theory, the uncertain material balance constraints are robustly transformed to obtain deterministic robust material balance constraints. Based on these robust material balance constraints, a liquid energy logistics multi-party collaborative scheduling model is created. Using simulation scenarios for each operating condition constructed based on preset disturbance intervals, the liquid energy logistics multi-party collaborative scheduling model is tested and optimized to obtain the final scheduling model. Based on the final scheduling model and the multi-stage rolling time-domain strategy, the supply volume on the supply side, the demand volume on the demand side, and the logistics transportation volume within the corresponding time window for each product are determined at each time node. This embodiment can effectively enhance the anti-interference capability of multi-party collaborative scheduling of liquid energy logistics, realize the efficient and dynamic collaborative configuration of liquid energy logistics system under uncertain environment, complete the optimal collaborative configuration solution of complex liquid energy logistics system with multiple products and multiple entities, and realize the efficient and flexible dynamic compilation of liquid energy multi-party collaborative scheduling plan in the face of nonlinear demand fluctuations and external disturbances.

[0100] In this embodiment, the liquid energy logistics multi-party collaborative scheduling device based on multi-stage robust optimization is presented in the form of functional units. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0101] This invention also provides a computer device having the above-described features. Figure 10 The device shown is a multi-party collaborative scheduling device for liquid energy logistics based on multi-stage robust optimization.

[0102] Please see Figure 11The present invention provides a schematic diagram of the structure of a computer device according to an optional embodiment. The computer device includes one or more processors 10, a memory 20, and interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The various components are interconnected via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, multiple processors and / or multiple buses can be used with multiple memories, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 11 Take a processor 10 as an example.

[0103] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0104] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0105] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function. The data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, which can be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0106] Memory 20 may include volatile memory, such as random access memory. Memory may also include non-volatile memory, such as flash memory, hard disk, or solid-state drive. Memory 20 may also include combinations of the above types of memory.

[0107] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0108] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-stage robust optimization based multi-party collaborative scheduling method for liquid energy logistics, characterized in that, include: Based on the nominal demand variables, disturbance variables, and allowable prediction deviation variables of the multi-source liquid energy, a predicted demand value expression is generated; wherein, the multi-source liquid energy includes multiple liquid energy products; The predicted demand value expression is introduced into the created material balance constraint to obtain an uncertain material balance constraint; Based on strong duality theory, the uncertain material balance constraints are robustly transformed to obtain deterministic robust material balance constraints. Based on the robust material balance constraints, a multi-party collaborative scheduling model for liquid energy logistics is created. The liquid energy logistics multi-party collaborative scheduling model is tested and optimized using each working condition simulation scenario constructed based on a preset disturbance interval to obtain the final scheduling model. Based on the final scheduling model and the multi-stage rolling time-domain strategy, the supply volume of each liquid energy product by the supply side at each time endpoint of each time window, the demand volume of each liquid energy product by the demand side at each time endpoint, and the logistics transportation volume of each liquid energy product by the logistics side within each time window are determined.

2. The method according to claim 1, characterized in that, The method of testing and optimizing the multi-party collaborative scheduling model for liquid energy logistics using each operating condition simulation scenario constructed based on a preset disturbance interval to obtain the final scheduling model includes: Using the historical demand data of the diverse liquid energy sources from the demand side, the trained time-series forecasting model for demand values ​​is tested to determine the maximum allowable prediction deviation of the time-series forecasting model for demand values. Multiple random samplings are performed within the preset disturbance range. Based on each randomly sampled disturbance value, the maximum allowable prediction deviation value, and the set nominal demand value, each working condition simulation scenario is constructed. The multi-party collaborative scheduling model for liquid energy logistics is tested and optimized using each of the aforementioned operating condition simulation scenarios to obtain the final scheduling model.

3. The method according to claim 2, characterized in that, The historical demand data includes the actual demand value of the multi-source liquid energy products at each point in time within the historical period. The step of using historical demand data for the diverse liquid energy sources from the demand side to test the trained time-series demand forecasting model, in order to determine the maximum allowable prediction deviation of the time-series demand forecasting model, includes: The first P time points within the historical period are defined as the first historical time points, and the time points within the historical period other than the first historical time points are defined as the second historical time points; where P is an integer greater than 1. The actual demand values ​​of the multi-source liquid energy products at each first historical time point are arranged in chronological order to create a time series of actual demand for the multi-source liquid energy products. The actual demand time series of the multi-element liquid energy is input into the demand value time series prediction model for time series prediction, so as to obtain the predicted demand value of the multi-element liquid energy product at each second historical time point. Based on the predicted and actual demand values ​​of the multi-element liquid energy product at each of the second historical time points, the absolute value of the prediction deviation of the multi-element liquid energy at each of the second historical time points is determined. The maximum value among the absolute values ​​of each prediction deviation is determined and used as the maximum allowable prediction deviation value.

4. The method according to claim 3, characterized in that, The demand value time series prediction model includes a connected variational mode decomposition (VMD) model and a bidirectional long short-term memory (BiLSTM) network model. The step of inputting the actual demand time series of the multi-source liquid energy into the demand value time series prediction model for time series prediction, to obtain the predicted demand value of the multi-source liquid energy product at each second historical time point, includes: The actual demand time series of the multi-element liquid energy is input into the VMD model so that the VMD model performs variational mode decomposition on the actual demand time series of the multi-element liquid energy to obtain multiple intrinsic mode function components and residual components output by the VMD model. Each intrinsic mode function (IMF) component and the residual component are input into the BiLSTM model, respectively, so that the BiLSTM model extracts the forward and backward hidden states of each IMF component and the residual component, and concatenates the forward and backward hidden states of each IMF component and the residual component to obtain the bidirectional features of each IMF component and the residual component. A fully connected mapping is performed on the bidirectional features of each IMF component and the residual component to obtain the predicted value of each IMF component and the residual component. The predicted values ​​of each IMF component and the residual component are linearly superimposed to obtain the predicted demand value of the multi-element liquid energy product at each second historical time point.

5. The method according to claim 1, characterized in that, The expression for the predicted demand value is: ； in, For time windows t internal nodes n medium-sized products k The predicted demand value, For time windows t internal nodes n medium-sized products k The nominal demand variable, For time windows t internal nodes n medium-sized products k The perturbation variable, For time windows t internal nodes n medium-sized products k Allowable prediction bias variables, T , D and K These represent the time window set, the demand node set, and the product set, respectively.

6. The method according to claim 1, characterized in that, The robust material balance constraint is: ； ； ； in, For time points node medium-sized products Inventory levels, in tons; For time points node medium-sized products The amount of stock shortage, in tons; For time windows Internal transportation From node Send to node Products The transport volume, in tons; For time windows Internal transportation From node Send to node Products The amount of goods transported in transit, in tons; For time windows internal nodes n medium-sized products k The nominal demand variable, in tons; For time windows internal nodes medium-sized products The maximum permissible deviation, in tons; For nodes medium-sized products Uncertainty in budget parameters; and For robust equivalent variables; T , S , H , D, M and K These represent the time window set, supply node set, transit node set, demand node set, transportation mode set, and product set, respectively.

7. The method according to claim 2, characterized in that, The process of testing and optimizing the multi-party collaborative scheduling model for liquid energy logistics using each of the aforementioned operational simulation scenarios to obtain the final scheduling model includes: The liquid energy logistics multi-party collaborative scheduling model was tested using at least one of the aforementioned operating condition simulation scenarios, and the resilience dimension index and cost dimension index of the liquid energy logistics multi-party collaborative scheduling model were obtained; wherein, the resilience dimension index includes at least one of average out-of-stock duration, out-of-stock quantity, and service level. If the resilience dimension index or the cost dimension index does not meet the requirements, the internal parameters of the liquid energy logistics multi-party collaborative scheduling model are optimized to obtain the optimized model. Continue to test the optimized model using at least one of the aforementioned operating condition simulation scenarios until the latest optimized model meets the requirements for both resilience and cost metrics.

8. A multi-party collaborative scheduling device for liquid energy logistics based on multi-stage robust optimization, characterized in that, The apparatus is applied to the multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization as described in any one of claims 1 to 7; the apparatus comprises: The generation unit is used to generate a forecast demand value expression based on the nominal demand variable, disturbance variable, and allowable forecast deviation variable of multiple liquid energy sources. An introducing unit is used to introduce the predicted demand value expression into the created material balance constraint to obtain an uncertain material balance constraint; The transformation unit is used to perform a robust equivalent transformation on the uncertain material balance constraints based on strong duality theory, so as to obtain deterministic robust material balance constraints. A creation unit is used to create a multi-party collaborative scheduling model for liquid energy logistics based on the robust material balance constraints. The optimization unit is used to test and optimize the multi-party collaborative scheduling model of liquid energy logistics using each working condition simulation scenario constructed based on a preset disturbance interval, so as to obtain the final scheduling model. The determining unit is used to determine, based on the final scheduling model and the multi-stage rolling time-domain strategy, the supply quantity of each liquid energy product by the supply side at each time endpoint of each time window, the demand quantity of each liquid energy product by the demand side at each time endpoint, and the logistics transportation quantity of each liquid energy product by the logistics side within each time window.

9. A computer device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the multi-party collaborative scheduling method for liquid energy logistics based on multi-stage robust optimization as described in any one of claims 1 to 7.

Images