Multi-period collaborative scheduling method for refinery production and utility systems under uncertain conditions

Abstract

The present invention relates to a multi-period collaborative scheduling method for oil refining and utility systems under uncertain conditions. The method comprises the following steps: Step S1: acquiring operating data and price data for purchased thermal coal and electricity; Step S2: predicting the boundaries of an uncertainty set based on the price data, thereby constructing a dynamic uncertainty set; Step S3: establishing a multi-period collaborative scheduling model for the refinery production system and utility systems under the dynamic uncertainty set; Step S4: introducing the model using a distributionally robust optimization method based on the dynamic uncertainty set; Step S5: solving the multi-period collaborative scheduling model for oil refining and utility systems under uncertain conditions using a robust dual reconstruction approach, and outputting a production plan. Compared to existing technologies, the present invention has advantages such as improving the robustness and conservatism of the production plan output by the model.

Description

Technical Field

[0001] The present invention relates to, in particular to, a multi-period collaborative production scheduling method for oil refining production and public systems under uncertain conditions. Background Art

[0002] The core function of a refinery's production system is to convert crude oil into various petroleum products, while the utility system provides the necessary electricity, steam, and water resources to support this process. Effective coordination between these two systems is crucial for ensuring efficient refinery operations. However, due to market uncertainties such as fluctuations in the prices of purchased thermal coal and electricity, nonlinear characteristics of equipment operation, and cyclical fluctuations in market demand, the coordinated optimization of refinery production and utility systems is extremely complex. These uncertainties directly impact resource allocation, the rationality of production planning, and overall production efficiency, significantly impacting the refinery's economic performance and environmental compliance.

[0003] In recent years, with the rapid development of data-driven technologies such as machine learning, predictive analysis methods based on historical data have become increasingly widely used in the refining industry. By analyzing and modeling historical data on purchased thermal coal and electricity prices, it is possible to effectively identify the cyclical characteristics of price fluctuations and construct a dynamic uncertainty set, providing a more accurate reference for optimization decisions. However, existing methods generally ignore the dynamic nature of uncertainties in multi-period environments and fail to fully address the impact of cyclical fluctuations in raw material prices on production plans. Furthermore, the recovery and utilization of waste heat resources in refineries has not been fully studied and optimized, resulting in wasted resources and reduced economic benefits. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-period collaborative production scheduling method for oil refining and utility systems under uncertain conditions in order to improve the robustness and conservatism of the production plan output by the model.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A multi-period collaborative scheduling method for oil refining production and utility systems under uncertain conditions, the method comprising the following steps:

[0007] Step S1: Obtaining the operating data of the entire refinery and the price data of purchased thermal coal and electricity in different periods;

[0008] Step S2: predicting the boundary of the uncertainty set based on the price data, and then constructing a dynamic uncertainty set;

[0009] Step S3: Under the dynamic uncertainty set, a multi-period collaborative scheduling model of the refinery production system and the utility system is established;

[0010] Step S4: using a distributed robust optimization method based on dynamic uncertainty sets to convert the model into a multi-period collaborative scheduling model for refinery production and utility systems under uncertain conditions;

[0011] Step S5: Use the robust dual reconstruction method to solve the multi-period collaborative scheduling model of the refinery production and the utility system under uncertain conditions, and output the production plan.

[0012] Furthermore, the dynamic uncertainty set is:

[0013]

[0014] Among them, ξ t,p represents the deviation of period t, is the deviation in the historical period, ρ i,p represents the regression coefficient, Π represents the length of the historical time period, Indicates some offset or fixed influence on the uncertainty at the current moment, U 1,t Indicates a dynamic boundary.

[0015] Furthermore, the dynamic boundary is predicted based on the XGBoost model.

[0016] Furthermore, the objective function of the multi-period collaborative scheduling model of the refinery production system and the utility system of S3 is to maximize the economic benefit, which is:

[0017]

[0018] Among them, profit means economic benefit, Price means p,t Cost m,t Cost i,t Cost u,t 、Price outfuel,t ,Price e,t , and Represent the purchase price of product p, the purchase cost of raw material m, the inventory cost of storage tank i, the operating cost of unit u, the price of purchased coal, the industrial electricity price, the treatment cost of CO2 and the treatment cost of SO2, respectively. V p,t 、V m,t 、Inv i,t 、V u,t 、V outfuel,b,t , V outE,t , and They represent the production of product p, the purchase quantity of crude oil m, the inventory level of storage tank i, the total flow of unit u, the amount of purchased coal, the amount of purchased electricity, the flow of CO2, and the flow of SO2.

[0019] Furthermore, the constraints of the multi-period collaborative scheduling model for the refinery production system and the utility system in S3 are the production planning model, the utility system model, and the connection constraints between the two systems. The connection constraints between the two systems include fuel balance, steam balance, and waste heat recovery steam balance. The fuel balance is specifically:

[0020] V infuel,b,t +V outfuel,b,t =V fuel,b,t b∈B,t∈T

[0021] Where V infuel,b,t Generate gas for production system, V outfuel,b,t For purchased coal, V fuel,b,t It represents the total fuel demand under boiler b.

[0022] Furthermore, the steam balance is specifically:

[0023]

[0024] Wherein, et represents extraction steam turbine, ET represents extraction steam turbine set, BT represents back-pressure steam turbine set, lv represents pressure reducing valve, LV represents pressure reducing valve set, u represents steam production unit, and U represents the set of processing equipment in production system and utility system;

[0025] represents the amount of high-pressure steam flowing out of the extraction turbine during period t, V hs,lv,t Indicates the high-pressure steam flow rate from the pressure reducing valve lv, V hs,t Indicates the total flow rate of high-pressure steam in the system, Indicates the amount of high-pressure steam flowing into the downstream turbine, DS hs,u,t Indicates the amount of high-pressure steam required by processing unit u in the production system; V ms,lv,t Indicates the medium pressure steam flow rate flowing out of the pressure reducing valve lv, represents the amount of medium-pressure steam flowing out of the extraction turbine during period t, V ms,lv,t Indicates the medium pressure steam flow rate flowing out of the pressure reducing valve lv, V ms,t Indicates the total flow of medium-pressure steam in the system, Indicates the amount of medium-pressure steam flowing into the downstream turbine, DS ms,u,t It represents the amount of medium-pressure steam required by process unit u in the production system; represents the amount of low-pressure steam flowing out of the extraction turbine et during period t, V ls,lv,t Indicates the low-pressure steam flow rate flowing out of the pressure reducing valve lv, V ls,t Indicates the total flow of low-pressure steam in the system, DS ls,u,t Indicates the amount of low-pressure steam required by processing unit u in the production system.

[0026] Furthermore, the waste heat recovery steam balance is:

[0027]

[0028]

[0029] Where: the total steam generated by the steam production unit u at steam level mls in period t Unit production flow V u,t , coefficient C psu,mls , PSU represents the set of self-generated steam units, MLS represents the set of steam levels, and steam demand DS mls,u,t , steam demand DS of non-steam producing units at steam level mls mls,u,t Externally produced steam supply and the steam consumed by the unit V u,mls,t The sum of the steam demand DS for the steam production unit mls,u,t is the unit's own steam usage External self-generated steam supply and the steam consumed in the unit V u,mls,t The sum of They represent the total amount of self-produced high-pressure steam, the total amount of self-produced medium-pressure steam and the total amount of self-produced low-pressure steam respectively.

[0030] Furthermore, the objective function of the multi-period collaborative scheduling model of refinery production and utility systems under uncertain conditions is:

[0031]

[0032] Among them, c T x t represents direct cost, x t represents the one-stage decision variable of period t, and c represents the decision variable x t The relevant cost coefficient, It represents the expected operation, that is, the expected value calculation of the function under uncertainty conditions. represents the probability distribution of uncertainty parameters, D represents the probability distribution set, and f represents the probability distribution of the decision variable x. t and uncertainty parameter ξ t The relevant objective function, ξ t represents the uncertainty parameter, which is a vector describing some uncertainty factors in the system under the time period t, b represents the right-hand constant vector of the constraint, W represents the constraint matrix, and defines the auxiliary variable y t The constraints that are satisfied, y t represents the two-stage decision variable adjusted according to the uncertainty condition, and h represents the decision variable x tRelated constraint functions, Y represents the set of auxiliary decision variables, that is, all possible y t P is a matrix that converts the uncertainty parameter ξ into t and the decision variable y t Combine and weight them to affect the value of the objective function.

[0033] Furthermore, the specific steps of solving the multi-period collaborative scheduling model of refinery production and utility systems under uncertain conditions using the robust dual reconstruction method are as follows:

[0034] The multi-period collaborative scheduling model of oil refining production and utility systems under uncertain conditions is transformed into a robust dual model and then solved. The robust dual model is:

[0035]

[0036]

[0037] Among them, λ t represents the uncertainty budget variable under period t, which controls the robustness of the model to uncertainty, s n,t Represents auxiliary variables, y n,t represents the two-stage decision variable for scenario n in period t, used to adjust the plan under specific circumstances, γ n,t The dual variable that represents the upper limit of the control uncertainty deviation, η n,t The dual variable θ represents the lower bound of the control uncertainty deviation. t represents the weight coefficient of the robustness level, N represents the number of uncertainty scenarios, n represents the index of the uncertainty scenario, q represents the weight vector, represents the actual uncertainty parameter under scenario n with period t, represents the upper bound of the uncertainty set, Denotes the lower bound of the uncertainty set.

[0038] Furthermore, the operating data and price data of thermal coal and electricity purchased at different times include process parameters of the production equipment, material balance data, energy consumption indicators and price data of thermal coal and electricity purchased at different times.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention integrates the refinery utility system model into the multi-period production planning model, while taking into account the waste heat recovery of the refinery production equipment, thereby achieving cost reduction and improved production efficiency. At the same time, the data-driven dynamic uncertainty set method is applied to the multi-period collaborative scheduling model, which can more accurately reflect the price fluctuations in the actual market and improve the robustness and flexibility of the production plan. The introduction of uncertainty sets enables the solution model to show significant advantages in optimizing solution quality, model efficiency and economic benefits, and a dynamic trade-off between robustness and conservatism can be achieved by adjusting the Wasserstein radius. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a step diagram of the multi-period collaborative scheduling method for oil refining production and public systems under uncertain conditions of the present invention;

[0042] Figure 2 To simplify the refinery production system flow chart.

[0043] Figure 3 To simplify the refinery utility system flow chart.

[0044] Figure 4 Optimize flow diagrams for the integration of refinery production systems and utility systems. DETAILED DESCRIPTION

[0045] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0046] The present invention proposes a multi-period collaborative scheduling method for oil refining production and public systems under uncertain conditions, which aims to cope with the impact of uncertain factors such as raw material price fluctuations on the operating efficiency of oil refineries in complex market environments. This method takes multi-period optimization as the core, comprehensively considers the cyclical fluctuation characteristics of thermal coal and electricity prices, and accurately identifies the dynamic range of price fluctuations by introducing a prediction model based on XGBoost, constructs a dynamic uncertainty set, and provides a scientific basis for the collaborative optimization of oil refining and public systems. Based on a data-driven dynamic optimization framework, the invention combines the operating data and historical price data of the refinery to achieve accurate prediction and dynamic adjustment of resource supply and demand. At the same time, waste heat recovery technology is integrated into the optimization model to effectively integrate the waste heat resources generated in the refining process, improve the utilization efficiency of high, medium and low pressure steam, and reduce energy consumption. By constructing a multi-period optimization model with strong robustness and high adaptability, the present invention meets the requirements of strict environmental protection regulations while improving the economic benefits of production, providing technical support for the transformation, upgrading and sustainable development of the oil refining industry.

[0047] The method of the present invention comprises the following steps, and the flow chart is as follows: Figure 1 As shown:

[0048] Step S1: Obtaining the operating data of the entire refinery and the price data of purchased thermal coal and electricity in different periods;

[0049] Step S2: predicting the boundary of the uncertainty set based on the price data, and then constructing a dynamic uncertainty set;

[0050] Step S3: Under the dynamic uncertainty set, a multi-period collaborative scheduling model of the refinery production system and the utility system is established;

[0051] Step S4: using a distributed robust optimization method based on dynamic uncertainty sets to convert the model into a multi-period collaborative scheduling model for refinery production and utility systems under uncertain conditions;

[0052] Step S5: Use the robust dual reconstruction method to solve the multi-period collaborative scheduling model of the refinery production and the utility system under uncertain conditions, and output the production plan.

[0053] Figure 2 The simplified refinery production system flow chart according to the present invention is as follows: Figure 2 As shown, in the refinery process in this embodiment, the main raw materials of the refinery are six types of crude oil (SAM, OMN, SHL, IRL, SAX, CBA) and purchased wax oil (BVG), purchased heavy aromatics (BHR), purchased methanol (BME), purchased hydrogen (BH2), purchased reforming raw material (BRF), purchased MTBE (MTB), and purchased ethylbenzene (BEB).

[0054] The processing units include two atmospheric and vacuum units (CDU1 and CDU2), a light hydrocarbon recovery unit (SCUT), a reforming pre-hydrogenation unit (SR1P), two catalytic cracking units (FCC1 and FCC2), an oil slurry decapping unit (OPOH), a high-pressure hydrocracking unit (SHCU), a delayed coking unit (DCFC), a solvent deasphalting unit (SDFC), an S-ZORB unit (S-ZORB), two gas separation units (GS1 and GS2), a polypropylene unit (SPPU), an MTBE unit, a PSA unit, an ethylbenzene unit (SEBN), a styrene unit (SSTR), an alkylation unit (SALK), a benzene extraction combined unit (CBEU), and seven hydrogenation units (HF1-HF8), totaling 27 units.

[0055] Storage tanks include six types of crude oil storage tanks, gasoline storage tanks, aviation kerosene storage tanks, diesel storage tanks, ethylene naphtha storage tanks, residual oil storage tanks, etc.

[0056] A total of 16 final products are produced, namely No. 92 gasoline (W92), No. 95 gasoline (W95), jet fuel (JET), vinyl naphtha (L10), xylene (EEN), automotive diesel (XYL), road asphalt (A60), commercial dry gas (PGS), polypropylene (PPP), petroleum coke (CCK), sulfur (SUL), styrene (STR), benzene (BNZ), commercial liquefied gas (PLG), partial tertiary feedstock (PUL) and liquid ammonia (NH3).

[0057] A total of 13 kinds of raw materials are purchased, namely:

[0058] Six types of crude oil (SAM, OMN, SHL, IRL, SAX, CBA) as well as purchased wax oil (BVG), purchased heavy aromatics (BHR), purchased methanol (BME), purchased hydrogen (BH2), purchased reforming feedstock (BRF), purchased MTBE (MTB), and purchased ethylbenzene (BEB).

[0059] Figure 3 This is a simplified refinery utility system flow chart of the present invention, as shown in FIG. Figure 3 As shown, in this embodiment, the refinery process includes three boilers responsible for providing steam for production. These boilers use fuel oil, natural gas, or coal as fuel to ensure a stable supply of heat. Four extraction steam turbines extract steam to meet varying pressure requirements while also providing power to support key equipment. Fifteen back-pressure steam turbines utilize the low-pressure steam generated by high-pressure steam during production to power low-pressure steam-consuming equipment, improving the system's energy efficiency. Three pressure reducing valves regulate steam pressure to meet the specific steam pressure requirements of different production units. These utilities together constitute the core of the refinery's energy and power system, ensuring efficient energy utilization and stable operation throughout the refinery's production processes.

[0060] The units that are considered to be able to produce their own steam include two atmospheric and vacuum units (CDU1, CDU2), two catalytic cracking units (FCC1, FCC2), one reforming pre-hydrogenation unit (SR1P) and one delayed coking unit (DCFC).

[0061] Figure 4 Optimize flow charts for integration of refinery production systems and utility systems, such as Figure 4 As shown, the production system includes atmospheric and vacuum distillation units, secondary processing units, and blending units, ultimately delivering the finished product. The utility system, comprised of boilers and steam turbines, provides energy for the production system.

[0062] For production systems, the atmospheric and vacuum distillation unit processes crude oil, separating light and heavy components; the secondary processing unit further processes the heavy component to improve product quality; and the blending unit blends the processed components into a final product that meets standards. For utility systems, the boiler produces steam to supply the production system and steam turbine; the steam turbine uses steam to generate electricity or drive equipment, while also outputting waste steam for use in various stages of the production system. Production systems not only consume energy from the utility system to produce gasoline and diesel but also generate some fuel, providing fuel resources for the utility system.

[0063] At the same time, the utility system must provide stable steam and electricity to provide operating power and a suitable production environment for the refinery's production units.

[0064] The steps of the present invention are specifically:

[0065] Step S1: Collect the operation data of the entire refinery and the price data of purchased thermal coal and electricity in different periods.

[0066] First, obtain the refinery's operating data, including process parameters of production units, material balance data, energy consumption indicators, etc., through real-time monitoring equipment, data acquisition systems or historical records; second, collect price data of purchased thermal coal and electricity in different periods from the market or supply chain management system; finally, combine data cleaning and preprocessing technologies to ensure the integrity, accuracy and consistency of the collected data.

[0067] Step S2: Predict the boundary of the uncertainty set based on the historical price data, and then construct a dynamic uncertainty set.

[0068] Dynamic Uncertain Set To describe the periodic fluctuation of coal prices, the uncertainty set has a time-varying boundary. The traditional static uncertainty set A fixed fluctuation range is defined, while the dynamic uncertainty set updates the boundary in real time by combining time series data and historical deviation estimates. This can more accurately reflect the actual fluctuations. The dynamic uncertainty set is defined as follows:

[0069]

[0070] Among them, ξ t,p represents the deviation at time t, is the deviation in the historical period, ρ i,p represents the regression coefficient, and the dynamic boundary U 1,t It is predicted by the XGBoost model.

[0071] The XGBoost model continuously updates historical data to predict uncertainty bounds for future periods. The model is trained using a supervised learning algorithm, with the deviation range of coal price fluctuations as the target variable. XGBoost uses gradient boosting decision trees to learn and predict uncertainty bounds, thereby generating uncertainty set bounds for a given time period. The following pseudocode shows how the model estimates the deviations using new input data at the end of each period: Algorithm: XGBoost Predicts Dynamic Uncertainty Sets

[0072] 1. Input:

[0073] I: Instance set of the current node (year, period).

[0074] m: Deviation dimension.

[0075] 2. Initialization:

[0076] Calculate the gain: gain←0.

[0077] Define global parameters: G←∑ i g i ,H←∑ i h i .

[0078] 3. Loop through the nodes: For k = 1 to m, perform the following operations:

[0079] Initialize local parameters: G L ←0, H L ←0.

[0080] According to the feature x k , sort the instances by their values.

[0081] 4. Feature splitting calculation: traverse the sorted instance j:

[0082] Update the accumulated value of the left child node: G L ←G L +g j , H L ←H L +h j .

[0083] Update the accumulated value of the right child node: G R ←GG L ,H R ←HH L .

[0084] Calculate the split score:

[0085] 5. Output:

[0086] The bias forecast of the uncertainty set for the next period.

[0087] Step S3: Under the condition of uncertainty in the prices of purchased coal and electricity, a multi-period collaborative scheduling model for the refinery production system and the public system is established.

[0088] This invention systematically optimizes the synergistic relationship between the refinery production system and the utility system by integrating the refinery utility system model into the multi-period production planning model. The refinery production system and the utility system each have distinct functions: the production system is responsible for crude oil processing and product output, while the utility system provides necessary support such as energy and steam. However, the two do not operate independently, but are highly coupled through energy and material flows. Therefore, integrating and optimizing the two in the model is particularly necessary, which can not only improve overall energy efficiency but also significantly reduce energy consumption and costs.

[0089] In addition, as a key link in optimizing synergy, the necessity of waste heat recovery is reflected in reducing energy waste, improving energy utilization efficiency and environmental protection. Refinery production equipment releases a large amount of heat energy during the processing process. If it is discharged directly, it will not only cause waste of resources, but also may cause thermal pollution. Through the waste heat recovery system, this part of heat energy can be efficiently utilized, such as for boiler water preheating or low-pressure steam production, thereby reducing dependence on external energy sources such as coal and electricity. This not only helps to achieve the low-carbon goal of refinery operation, but also improves production efficiency while reducing costs. Therefore, the present invention combines waste heat recovery with multi-cycle optimization, which not only improves the completeness of the model in theory, but also achieves the win-win goal of efficient energy utilization and cost minimization in practice.

[0090] In the production system, each processing unit requires steam and electricity to support its operation. These energies are provided by the utility system. At the same time, the gas produced by the production system can be used as fuel for the boiler.

[0091] Fuel balance between the production system and the utility system. The boiler fuel comes from gas generated by the production system and purchased coal. The function expression is:

[0092] V infuel,b,t +V outfuel,b,t =V fuel,b,t b∈B,t∈T

[0093] Where V infuel,b,t Generate gas for production system, V outfuel,b,t To purchase coal from outside.

[0094] Steam balance between production system and utility system. Three levels of steam are considered, including high pressure, medium pressure and low pressure steam. The generation and consumption of high, medium and low pressure steam are key energy balance factors. The functional expression is:

[0095]

[0096] In the formula, the two terms V hs,t They are the generation and consumption of high-pressure steam, specifically, they include the amount of steam provided by the high-pressure steam generating equipment (such as boilers or thermal equipment) in the refinery and the amount of high-pressure steam consumed for driving equipment, thermal demand or distribution to the next level of pressure (such as decompression process); the two V ms,t They are the generation and consumption of medium-pressure steam, specifically including the decompression of high-pressure steam or the output of medium-pressure steam generation equipment, as well as the direct application of medium-pressure steam in process equipment or the further decompression to low pressure; the two V ls,t They are the generation and consumption of low-pressure steam, specifically including the output of medium-pressure steam reduction or low-pressure steam generation equipment and the demand for low-pressure steam for process heating, pipeline insulation or other low-pressure application scenarios.

[0097] In the process of coordinated optimization of oil refining production and utility systems, the waste heat recovery system is integrated into the steam network to further reduce dependence on the utility system and raw material consumption. The functional expression is:

[0098]

[0099] Where, (1) represents the total steam generated by the steam production unit u at steam level mls in period t is obtained by increasing the unit's production flow V u,t Multiply by coefficient C psu,mls The latter is calculated to represent the steam output per unit of production flow. (2) represents the steam demand DS at the steam level mls for non-steam producing units (i.e. units that do not generate steam themselves). mls,u,t Externally produced steam supply and the steam consumed by the unit V u,mls,t The sum of ; (3) formula shows that for the steam production unit, the steam demand DS mls,u,t is the unit's own steam usage External self-generated steam supply and the steam consumed in the unit V u,mls,tThe total amount of high-pressure steam generated by all steam production units (excluding their own steam consumption) must be equal to the total amount of external high-pressure steam supplied to other units. Formula (5) shows that the total amount of medium-pressure steam generated by all steam production units, after deducting their own consumption and adding the medium-pressure steam obtained by heat exchange with high-pressure steam, must be equal to the total amount of medium-pressure steam supplied to all units as external steam. Formula (6) shows that the total amount of low-pressure steam generated by all steam production units, after deducting their own consumption and adding the low-pressure steam obtained by heat exchange with medium-pressure steam, must be equal to the total amount of low-pressure steam supplied to all units as external steam.

[0100] The objective function is to maximize economic benefits;

[0101] The constraints are the production planning model, the public system model and the connection constraints between the two systems.

[0102] The corresponding constraints of the production planning model include supply and demand constraints, processing equipment constraints, blending constraints and storage tank conditions;

[0103] The corresponding constraints of the utility system model include boiler model, steam turbine model, pressure reducing valve model and power constraints;

[0104] The corresponding constraints for the connection between the two systems include fuel balance, steam balance and waste heat recovery steam balance.

[0105] For the multi-period production planning model, since the refinery utility system model is introduced, the utility system energy cost also needs to be included in the objective function. The corresponding expression of the economic benefit model is:

[0106]

[0107] Among them, in the objective function, Price p,t Cost m,t Cost i,t Cost u,t 、Price outfuel,t ,Price e,t , and Represent the purchase price of product p, the purchase cost of raw material m, the inventory cost of storage tank i, the operating cost of unit u, the price of purchased coal, the industrial electricity price, the treatment cost of CO2, and the treatment cost of SO2. p,t 、V m,t 、Inv i,t 、V u,t 、V outfuel,b,t , V outE,t , and They represent the production of product p, the purchase quantity of crude oil m, the inventory level of storage tank i, the total flow of unit u, the amount of purchased coal, the amount of purchased electricity, the flow of CO2, and the flow of SO2.

[0108] The multi-period collaborative scheduling model includes a multi-period refinery production planning model and a refinery utility system model;

[0109] The production system and utility system of an oil refinery are closely related. Fluctuations in thermal coal and electricity prices will affect the cost of the utility system, which in turn affects the energy consumption and by-product utilization of the production system, and has a significant impact on production planning and economic benefits. Therefore, studying the optimization of collaborative systems under uncertain conditions of coal and electricity prices is crucial to improving the energy efficiency and economic benefits of oil refineries.

[0110] In step S4, the distributed robust optimization method based on dynamic uncertainty sets is applied to the current multi-period collaborative scheduling model to adapt to the cyclical fluctuations and random changes of market prices. The objective function is rewritten as follows:

[0111]

[0112] The main goal of the model is to minimize the cost while taking into account the uncertainty of the future state ξ t Among them, c T x t represents direct costs, and It summarizes the expected losses due to uncertainty.

[0113] At the same time, using dynamic uncertain set To describe the periodic fluctuation of coal prices, the uncertainty set has a time-varying boundary. The traditional static uncertainty set A fixed fluctuation range is defined, while the dynamic uncertainty set updates the boundary in real time by combining time series data and historical deviation estimates. This can more accurately reflect the actual fluctuations. The dynamic uncertainty set is defined as follows:

[0114]

[0115] Among them, ξ t,p represents the deviation at time t, is the deviation in the historical period, ρ i,p represents the regression coefficient, and the dynamic boundary U 1,t It is predicted by the XGBoost model.

[0116] Step S5, using a robust dual reconstruction method to solve the multi-period collaborative scheduling model of the oil refining production and utility system under uncertain conditions as described in claim 8, the function expression is:

[0117]

[0118]

[0119] The robust dual model integrates all variables and constraints, ensuring comprehensive management of uncertainties and achieving overall optimization by balancing the needs of different stakeholders.

[0120] In the previous collaborative model between the refinery production system and the utility system, the system operation focused on meeting production and energy needs, but did not fully consider the role of waste heat recovery, resulting in difficulty in further improving energy utilization efficiency. The multi-cycle integrated model proposed in the present invention achieves a closer coupling and efficient operation between the two by incorporating the waste heat recovery system into the collaborative optimization of the production and utility systems. Specifically, a large amount of waste heat generated during the operation of the processing unit can be converted into low-pressure or medium-pressure steam through the waste heat recovery system and directly supplied to other units for use, thereby reducing dependence on the utility system and reducing the operating burden of the utility system. The core of this collaborative model is to make full use of the waste heat generated by the production unit, realize the recycling of internal energy, and dynamically optimize the energy distribution between the production system and the utility system. Through the introduction of waste heat recovery, the model can significantly reduce the consumption of external energy such as coal and electricity, and improve the overall economic benefits and environmental friendliness of the system.

[0121] To verify this, two oil refining industry cases were compared, examining whether excess heat recovery from the refining process was permitted for steam production. The results are shown in Table 1. Case 1 included a self-generated steam module in the model, allowing for excess heat recovery from the refining process for steam production, without accounting for real-world fluctuations in coal and electricity prices. Case 2 did not include this steam recovery module, nor did it account for real-world fluctuations in coal and electricity prices.

[0122] The experimental results show that Case 1 consistently achieves higher economic benefits than Case 2, regardless of whether the model is run in a single or multiple cycles. This demonstrates that incorporating self-generated steam into the model significantly improves economic performance. Specifically, although Case 1 involves more continuous variables and constraints than Case 2, this increased complexity does not significantly impact computational efficiency. For example, the solution time for Case 1 increases only slightly across all cycles. The percentage difference between the two cases remains low, indicating that both models perform well in terms of solution accuracy. Furthermore, the target value representing economic benefits increases significantly with increasing planning cycles in Case 1. For example, in the 48-cycle scenario, the economic benefits of Case 1 reach 1.385618e+10, while those of Case 2 reach 1.309030e+10, a significant difference between the two. The economic benefits of Case 1 continue to increase with increasing cycles, with target values ​​increasing by 0.13182e+08, 0.51948e+08, and 1.05158e+09 for 12, 36, and 48 cycles, respectively. This indicates that the introduction of a self-generated steam recovery module can improve the overall system efficiency and economic benefits, especially in long-term operation.

[0123] Table 1 Experimental results of different production planning model methods under two cases

[0124]

[0125] Currently, various methods for managing uncertainty have been widely used, with robust optimization and stochastic programming being two of the most commonly used approaches. While the effectiveness of robust optimization has been demonstrated in various fields, it is noteworthy that these achievements have been achieved without relying on probabilistic information. In reality, the factors that cause uncertainty often follow a certain probability distribution. This observation highlights the potential limitations of robust optimization and has prompted further exploration of methods that incorporate probabilistic information to achieve more comprehensive uncertainty management approaches.

[0126] Dynamic uncertainty sets can adjust uncertainty descriptions in real time based on cyclical fluctuations in the environment (such as cyclical changes in coal and electricity prices), allowing the optimization model to dynamically adapt to changes in the external environment. This flexibility avoids the problem of traditional static uncertainty sets being overly conservative or inaccurate, thereby improving the practicality and accuracy of production scheduling. Conventional methods usually use static uncertainty sets to model the uncertainty of the external environment, but they cannot reflect cyclical fluctuations (such as coal and electricity prices) and dynamic changes, which may lead to overly conservative or inflexible results. Dynamic uncertainty sets can adjust the uncertainty range based on historical data and real-time environmental changes, more accurately describing complex uncertainties and improving the model's predictive and adaptable capabilities.

[0127] The analysis principle of the model in S5 is:

[0128] To do this, rewrite the objective function as follows:

[0129]

[0130] The main goal of the model is to minimize the cost while taking into account the uncertainty of the future state ξ t Among them, c T x t represents direct costs, and It summarizes the expected losses due to uncertainty.

[0131] By introducing the Lagrange multiplier λ t , Equation (8) reformulates the expectation part of the objective function as a minimization problem. This transformation not only facilitates the handling of uncertainty, but also enables the balance between the expected loss and the objective function during the optimization process.

[0132]

[0133] Here, a reasonable choice of λ t It can help decision makers deal with relevant trade-offs, thereby enhancing the robustness and reliability of the optimization process.

[0134] By introducing the slack variable s n,t , the above framework is further constructed as follows:

[0135]

[0136] Among them, the slack variable s n,t For modeling each sample point This enhances the flexibility of the optimization process and enables it to adapt to a variety of scenarios represented by the sample space.

[0137] Based on the duality principle, the dual relationship between maximization and minimization is demonstrated. This step simplifies the treatment of the original problem and enables the optimization process to utilize the known optimal solution characteristics.

[0138]

[0139] Next, further constraints are introduced into the optimization model:

[0140]

[0141] Here, formula (12) ensures that the maximum expected loss does not exceed the slack variable s n,t At the same time, it also imposes constraints on the Lagrange multiplier |z n,t | * ≤λ t, thus limiting the impact of the multiplier on the overall optimization.

[0142] By introducing the auxiliary variable y n,t ,γ n,t ,η n,t , the formula is further constructed as follows:

[0143]

[0144]

[0145] By integrating the results of all the previous formulas, an optimization model that includes all variables and constraints is finally formed:

[0146]

[0147] The robust dual model integrates all variables and constraints, ensuring comprehensive management of uncertainties and achieving overall optimization by balancing the needs of different stakeholders.

[0148] The present invention considers the uncertainty caused by fluctuations in coal and electricity prices and evaluates the performance of different optimization methods in dealing with uncertainty in the refining production system. The experiment includes Case 1 and Case 3. In Case 1, Case 1 includes a self-generated steam module in the model, allowing excess heat to be recovered from the refining process for steam production, without considering the actual fluctuations in coal and electricity prices; in Case 3, a self-generated steam module is included, considering the actual fluctuations in coal and electricity prices. Four different solutions are applied and compared to these two cases: deterministic methods, robust optimization methods, distributionally robust optimization methods, and dynamic distributionally robust optimization methods based on dynamic uncertainty sets. The experimental results are shown in Table 2. Compared with the performance of other uncertain methods, the dynamic distributionally robust optimization method based on dynamic uncertainty sets has the best solution results. Since the dynamic uncertainty set predicts the boundaries based on historical data, the robustness of the model is increased, thereby improving the economic benefit value. Compared with the distributionally robust optimization method, the dynamic distributionally robust optimization method based on dynamic uncertainty sets adds corresponding auxiliary variables in each cycle. The auxiliary variables can decompose complex constraints and reduce model complexity, which is more conducive to solving the model. Therefore, the solution efficiency is improved to a certain extent.

[0149] Table 2 Comparison of the effects of different optimization methods

[0150]

[0151]

[0152] The present invention proposes a multi-period collaborative scheduling method for refinery production and public systems under uncertain conditions. Taking into account the uncertainty of market thermal coal and electricity prices, the refinery public system model is integrated into the multi-period production planning model to achieve adaptive maintenance of the constant pressure reduction unit. At the same time, the data-driven dynamic uncertainty method is applied to the current multi-period model to more accurately reflect the price fluctuations in the actual market. Finally, the dynamic uncertainty set is introduced on the basis of the existing distributed robust optimization model, and a new solution model is proposed to improve the robustness and flexibility of the existing production plan and obtain the optimal production plan for the refinery production plan.

[0153] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A multi-period collaborative scheduling method for oil refining production and utility systems under uncertain conditions, characterized in that: The method comprises the following steps: Step S1: Obtain the operation data of the entire refinery and the price data of purchased thermal coal and electricity at different times; the operation data and the price data of purchased thermal coal and electricity at different times include the process parameters of the production equipment, material balance data, energy consumption indicators and the price data of purchased thermal coal and electricity at different times Step S2: predicting the boundary of the uncertainty set based on the price data, and then constructing a dynamic uncertainty set; Step S3: Under the dynamic uncertainty set, a multi-period collaborative scheduling model of the refinery production system and the utility system is established; Step S4: using a distributed robust optimization method based on dynamic uncertainty sets to convert the model into a multi-period collaborative scheduling model for refinery production and utility systems under uncertain conditions; Step S5: Solve the multi-period collaborative scheduling model of the refinery production and utility system under uncertain conditions by using a robust dual reconstruction method, and output a production plan; The dynamic uncertainty set is: Among them, ξ t,p represents the deviation of period t, is the deviation in the historical period, ρ i,p represents the regression coefficient, Π represents the length of the historical time period, Indicates some offset or fixed influence on the uncertainty at the current moment, U 1,t Represents a dynamic boundary; The constraints of the multi-period collaborative scheduling model for the refinery production system and utility system in S3 are the production planning model, the utility system model, and the connection constraints between the two systems. The connection constraints between the two systems include fuel balance, steam balance, and waste heat recovery steam balance. The fuel balance is specifically: V infuel,b,t +V outfuel,b,t =V fuel,b,t b∈B,t∈T Where V infuel,b,t Generate gas for production system, V outfuel,b,t For purchased coal, V fuel,b,t It represents the total fuel demand under boiler b; The objective function of the multi-period collaborative scheduling model of refinery production and utility systems under uncertain conditions is: s.t.Ax t ≤b Among them, c T x t represents direct cost, x t represents the one-stage decision variable of period t, and c represents the decision variable x t The relevant cost coefficient, It represents the expected operation, that is, the expected value calculation of the function under uncertainty conditions. represents the probability distribution of uncertainty parameters, D represents the probability distribution set, and f represents the probability distribution of the decision variable x. t and uncertainty parameter ξ t The relevant objective function, ξ t represents the uncertainty parameter, which is a vector describing some uncertainty factors in the system under the time period t, b represents the right-hand constant vector of the constraint, W represents the constraint matrix, and defines the auxiliary variable y t The constraints that are satisfied, y t represents the two-stage decision variable adjusted according to the uncertainty condition, and h represents the decision variable x t Related constraint functions, Y represents the set of auxiliary decision variables, that is, all possible y t P is a matrix that converts the uncertainty parameter ξ into t and the decision variable y t Combine and weight them to affect the value of the objective function.

2. The multi-period collaborative scheduling method for oil refining production and utility systems under uncertain conditions according to claim 1, characterized in that: The dynamic boundary is predicted based on the XGBoost model.

3. The multi-period collaborative scheduling method for oil refining production and utility systems under uncertain conditions according to claim 1, characterized in that: The objective function of the multi-period collaborative scheduling model of the refinery production system and the utility system of S3 is to maximize the economic benefits, which are: Among them, profit means economic benefit, Price means p,t Cost m,t Cost i,t Cost u,t 、Price outfuel,t ,Price e,t , and Represent the purchase price of product p, the purchase cost of raw material m, the inventory cost of storage tank i, the operating cost of unit u, the price of purchased coal, the industrial electricity price, the treatment cost of CO2 and the treatment cost of SO2, respectively. V p,t 、V m,t 、Inv i,t 、V u,t 、V outfuel,b,t , V outE,t , and They represent the production of product p, the purchase quantity of crude oil m, the inventory level of storage tank i, the total flow of unit u, the amount of purchased coal, the amount of purchased electricity, the flow of CO2, and the flow of SO2.

4. The multi-period collaborative scheduling method for oil refining production and utility systems under uncertain conditions according to claim 1, characterized in that: The steam balance is specifically: Wherein, et represents extraction steam turbine, ET represents extraction steam turbine set, VT represents back-pressure steam turbine set, lv represents pressure reducing valve, LV represents pressure reducing valve set, u represents steam production unit, and U represents the set of processing equipment in production system and utility system; represents the amount of high-pressure steam flowing out of the extraction turbine during period t, V hs,lv,t Indicates the high-pressure steam flow rate from the pressure reducing valve lv, V hs,t Indicates the total flow rate of high-pressure steam in the system, Indicates the amount of high-pressure steam flowing into the downstream turbine, DS hs,u,t Indicates the amount of high-pressure steam required by processing unit u in the production system; V ms,lv,t Indicates the medium pressure steam flow rate flowing out of the pressure reducing valve lv, represents the amount of medium-pressure steam flowing out of the extraction turbine during period t, V ms,lv,t Indicates the medium pressure steam flow rate flowing out of the pressure reducing valve lv, V ms,t Indicates the total flow of medium-pressure steam in the system, Indicates the amount of medium-pressure steam flowing into the downstream turbine, DS ms,u,t It represents the amount of medium-pressure steam required by process unit u in the production system; represents the amount of low-pressure steam flowing out of the extraction turbine et during period t, V ls,lv,t Indicates the low-pressure steam flow rate flowing out of the pressure reducing valve lv, V ls,t Indicates the total flow of low-pressure steam in the system, DS ls,u,t Indicates the amount of low-pressure steam required by processing unit u in the production system.

5. The multi-period collaborative scheduling method for oil refining and utility systems under uncertain conditions according to claim 4, characterized in that: The waste heat recovery steam balance is: Where: the total steam generated by the steam production unit u at steam level mls in period t Unit production flow V u,t , coefficient C psu,mls , PSU represents the set of self-generated steam units, MLS represents the set of steam levels, and steam demand DS mls,u,t , steam demand DS of non-steam producing units at steam level mls mls,u,t Externally produced steam supply and the unit steam consumption V u,mls,t The sum of the steam demand DS for the steam production unit mls,u,t is the unit's own steam usage External self-generated steam supply and the steam consumed in the unit V u,mls,t The sum of They represent the total amount of self-produced high-pressure steam, the total amount of self-produced medium-pressure steam and the total amount of self-produced low-pressure steam respectively.

6. The multi-period collaborative scheduling method for oil refining production and utility systems under uncertain conditions according to claim 1, characterized in that: The specific steps of solving the multi-period collaborative scheduling model of refinery production and utility systems under uncertain conditions using the robust dual reconstruction method are as follows: The multi-period collaborative scheduling model of oil refining production and utility systems under uncertain conditions is transformed into a robust dual model and then solved. The robust dual model is: s.t.Ax t ≤b ||(P) T y n,t -c n,t +n n,t || * ≤λ t c n,t ≥0,η n,t ≥0 Wy n,t ≥h(x t ) Among them, λ t represents the uncertainty budget variable under period t, which controls the robustness of the model to uncertainty, s n,t Represents auxiliary variables, y n,t represents the two-stage decision variable for scenario n in period t, used to adjust the plan under specific circumstances, γ n,t The dual variable that represents the upper limit of the control uncertainty deviation, η n,t The dual variable θ represents the lower bound of the control uncertainty deviation. t represents the weight coefficient of the robustness level, N represents the number of uncertainty scenarios, n represents the index of the uncertainty scenario, q represents the weight vector, represents the actual uncertainty parameter under scenario n with period t, represents the upper bound of the uncertainty set, ξ t Denotes the lower bound of the uncertainty set.

Images